KR20250121074A - A machine learning method for predicting cell phenotypes using holographic imaging. - Google Patents

Abstract

Provided herein are methods for determining the cellular phenotype, such as activation status, of a T cell population. In some aspects, the provided methods are label-free. In some aspects, the provided methods can be used to monitor the cellular phenotype, such as activation status, of a T cell culture. Also provided herein are computing devices for use in performing the provided methods.

Description

A machine learning method for predicting cell phenotypes using holographic imaging.

Cross-reference to related applications

This application claims priority to U.S. Provisional Application No. 63/431,635, filed December 9, 2022, entitled "MACHINE LEARNING METHOD FOR PREDICTING CELLULAR PHENOTYPES USING HOLOGICAL IMAGING," and U.S. Provisional Application No. 63/537,467, filed September 8, 2023, entitled "MACHINE LEARNING METHOD FOR PREDICTING CELLULAR PHENOTYPES USING HOLOGICAL IMAGING," the contents of which are incorporated by reference in their entirety for all purposes.

Include references in the sequence list

This application is filed with an electronic sequence listing. The sequence listing is provided in a file named 735042024740SeqList.xml, created on December 7, 2023, and is 38,941 bytes in size. The electronic information in the sequence listing is incorporated by reference in its entirety.

field

The present disclosure relates to a method for determining the cellular phenotype, such as the activation state, of a T cell population. In some aspects, the provided method is a label-free method. In some aspects, the provided method can be used to monitor the cellular phenotype, such as the activation state, of a T cell culture. Also provided herein is a computing device for use in performing the provided method.

Existing methods for determining the cellular phenotype, such as activation status, of T cells, for example, during in vitro or ex vivo culture of T cells, can be time-consuming or require specialized reagents, equipment, or trained personnel. In some instances, existing methods involve directly manipulating or labeling T cells, for example, by incubating them with immunoaffinity-based reagents that may interfere with the quality or function of the T cells. Such methods can pose a risk of cell contamination or damage prior to any downstream processing. What is needed are improved methods for accurately determining the cellular phenotype, such as activation status, of T cells. Such methods are useful, for example, in the production of cell products containing T cells, such as cell therapy products. Methods and computing devices that address these needs are provided herein.

In some embodiments, provided herein is a method of determining a cell phenotype of a T cell population, comprising: determining, for a cell population comprising T cells, a population-level output measurement of expression of a marker expressed by T cells having a cell phenotype of interest, wherein the population-level output measurement is determined based on a plurality of population-level statistics, wherein: each population-level statistic is for one or more input measurements for one of a plurality of cell characteristics derived from holographic information obtained for the cell population, the plurality of population-level statistics comprising one or more population-level statistics for each of the plurality of cell characteristics; and wherein each input measurement is from an individual cell of the cell population.

In some of the embodiments, the method is for determining the activation status of a T cell population, wherein the marker is expressed by the activated T cells. In some of the embodiments, the marker is CD137 (4-1BB).

In some of the embodiments, the method is for determining a memory phenotype of a T cell population, wherein a marker, e.g., CCR7, is expressed by T cells having a central memory phenotype or a stem cell memory phenotype.

Also provided herein is a method of determining an activation state of a T cell population, comprising: determining, for a cell population comprising T cells, a population-level output measurement of expression of a marker expressed by activated T cells, wherein the population-level output measurement is determined based on a plurality of population-level statistics, wherein: each population-level statistic is for one or more input measurements of a cell characteristic among a plurality of cell characteristics derived from holographic information obtained for the cell population, the plurality of population-level statistics comprising one or more population-level statistics for each of the plurality of cell characteristics; and wherein each input measurement is from an individual cell of the cell population.

In some of the embodiments, the method is for determining expression of a recombinant receptor in a T cell population, wherein the marker is a recombinant receptor introduced into T cells of the cell population prior to obtaining holographic information. In some of the embodiments, the recombinant receptor is a chimeric antigen receptor (CAR) or an engineered T cell receptor (TCR).

Also provided herein is a method of determining recombinant receptor expression in a T cell population, comprising: determining, for a cell population comprising T cells, a population-level output measurement of expression of a recombinant receptor introduced into T cells of the T cell population, wherein the population-level output measurement is determined based on a plurality of population-level statistics, wherein: each population-level statistic is for one or more input measurements for one of a plurality of cell characteristics derived from holographic information obtained for the cell population, the plurality of population-level statistics comprising one or more population-level statistics for each of the plurality of cell characteristics; and wherein each input measurement is from an individual cell of the cell population.

In some of the embodiments, the method comprises determining a plurality of population-level statistics from one or more input measurements for each of the plurality of cell characteristics.

In some of the embodiments, the method comprises determining one or more input measurements for each of the plurality of cell features from the holographic information.

In some of the embodiments, the method comprises obtaining holographic information.

Also provided herein, in some embodiments, is a method of determining a cellular phenotype of a T cell population, comprising: (a) obtaining holographic information for a cell population comprising T cells; (b) determining one or more input measurements for each of a plurality of cell characteristics derived from the holographic information, wherein each input measurement is from an individual cell within the cell population; (c) determining a plurality of population-level statistics, wherein each population-level statistic is for one or more input measurements for one of the plurality of cell characteristics, and wherein the plurality of population-level statistics comprises one or more population-level statistics for each of the plurality of cell characteristics; and (d) determining, based on the plurality of population-level statistics, a population-level output measurement of expression of a marker expressed by T cells having a cell phenotype of interest for the cell population.

In some of the embodiments, the method is for determining the activation status of a T cell population, wherein the marker is expressed by the activated T cells. In some of the embodiments, the marker is CD137 (4-1BB).

Also provided herein, in some embodiments, is a method of determining an activation state of a T cell population, comprising: (a) obtaining holographic information for a cell population comprising T cells; (b) determining one or more input measurements for each of a plurality of cell characteristics derived from the holographic information, wherein each input measurement is from an individual cell within the cell population; (c) determining a plurality of population-level statistics, wherein each population-level statistic is for one or more input measurements for one of the plurality of cell characteristics, and wherein the plurality of population-level statistics comprises one or more population-level statistics for each of the plurality of cell characteristics; and (d) determining, based on the plurality of population-level statistics, a population-level output measurement of expression of a marker expressed by activated T cells for the cell population.

In some of the embodiments, the method is for determining expression of a recombinant receptor in a T cell population, wherein the marker is a recombinant receptor introduced into T cells of the cell population prior to obtaining holographic information. In some of the embodiments, the recombinant receptor is a chimeric antigen receptor (CAR) or an engineered T cell receptor (TCR).

Also provided herein, in some embodiments, is a method of determining recombinant receptor expression in a T cell population, comprising: (a) obtaining holographic information for a cell population comprising T cells; (b) determining one or more input measurements for each of a plurality of cell characteristics derived from the holographic information, wherein each input measurement is from an individual cell within the cell population; (c) determining a plurality of population-level statistics, wherein each population-level statistic is for one or more input measurements for one of the plurality of cell characteristics, and wherein the plurality of population-level statistics comprises one or more population-level statistics for each of the plurality of cell characteristics; and (d) determining, based on the plurality of population-level statistics, a population-level output measurement of expression for the cell population of a recombinant receptor introduced into T cells of the cell population.

In some of the embodiments, the method further comprises manipulating the T cell population after determining the population-level output measurement.

In some of the embodiments, the holographic information is obtained by differential digital holographic microscopy (DDHM). In some of the embodiments, obtaining the holographic information comprises imaging a cell population using DDHM.

In some of any of the embodiments, the one or more population-level statistics for at least one, and optionally each, of the plurality of cell characteristics comprises one or more quantiles of one or more input measurements of the cell characteristic. In some of any of the embodiments, the one or more population-level statistics for each of the plurality of cell characteristics comprises one or more quantiles of one or more input measurements of the cell characteristic. In some of any of the embodiments, the one or more quantiles are selected from (including one or more of) the 0.01, 0.1, 0.5, 0.9, and 0.99 quantiles of one or more input measurements of the cell characteristic.

In some of the embodiments, at least one, and optionally each, of the plurality of cell characteristics is determined by applying a distribution-based pooling filter to one or more input measurements of the cell characteristic. In some of the embodiments, at least one, and optionally each, of the plurality of cell characteristics is determined by applying a distribution-based pooling filter to one or more input measurements of the cell characteristic.

In some of the embodiments, the population-level output measurement is a number of cells, a percentage of cells, a proportion of cells, or a density of cells within a cell population expressing a marker.

In some of the embodiments, the cell population is from a cell culture grown in vitro or ex vivo.

In some of the embodiments, the holographic information is obtained during in vitro or ex vivo culture of the cell culture.

In some of the embodiments, the population-level output measurement is for expression of a marker during in vitro or ex vivo cultivation of a cell culture.

In some of the embodiments, the in vitro or ex vivo culture is performed under conditions that expand T cells in the cell culture.

In some of the embodiments, the in vitro or ex vivo culture is performed within a bioreactor.

In some of the embodiments, the cell population is incubated under T cell stimulating conditions prior to obtaining the holographic information. In some of the embodiments, the method comprises incubating the cell population under T cell stimulating conditions prior to obtaining the holographic information.

In some of the embodiments, the incubation occurs prior to in vitro or ex vivo culturing.

In some of the embodiments, the T cell stimulating conditions comprise incubation in the presence of a T cell stimulator that induces a primary activation signal and a co-stimulatory signal in the T cell. In some of the embodiments, the T cell stimulator comprises an anti-CD3 antibody or antibody fragment. In some of the embodiments, the T cell stimulator comprises an anti-CD28 antibody or antibody fragment. In some of the embodiments, the T cell stimulator is immobilized on a bead.

In some of the embodiments, the T cell stimulator is immobilized on an oligomeric streptavidin mutein reagent.

In some of the embodiments, the recombinant receptor is introduced into T cells of the cell population before obtaining holographic information. In some of the embodiments, the method comprises introducing the recombinant receptor into T cells of the cell population before obtaining holographic information.

In some of the embodiments, the introduction comprises contacting the cell population with an agent comprising a polynucleotide encoding a recombinant receptor.

In some of the embodiments, the recombinant receptor is a chimeric antigen receptor (CAR) or an engineered T cell receptor (TCR).

In some of the embodiments, the cell population is enriched for T cells. In some of the embodiments, at least 50, 55, 60, 65, 70, 75, 80, 85, 90, 91, 92, 93, 94, 95, 96, 97, 98, 99, or 100% of the cell population are T cells.

In some of the embodiments, the population-level output measurement is determined by providing the plurality of population-level statistics as input to a machine learning model trained to predict the population-level output measurement of marker expression based on the population-level statistics of the plurality of cell characteristics.

In some of the embodiments, the machine learning model is trained using a dataset of reference population-level statistics, wherein: for each of a first plurality of reference cell populations comprising T cells, the dataset of reference population-level statistics comprises one or more reference population-level statistics for each of a plurality of cell characteristics, wherein: each reference population-level statistic is for one or more reference input measurements for one of the plurality of cell characteristics derived from holographic information obtained for the reference cell population; and each reference input measurement is from an individual cell of the reference cell population.

In some of the embodiments, the machine learning model is trained using a dataset of reference population-level output measurements, wherein for each of a second plurality of reference cell populations, the dataset of reference population-level output measurements comprises reference population-level output measurements of expression of a marker for the reference cell population, wherein: the first and second plurality of reference cell populations are from reference cell cultures grown in vitro or ex vivo; and a reference cell population from the first plurality of reference cell populations is from the same reference cell culture as a reference cell population from the second plurality of reference cell populations.

Also provided in some embodiments is a method of training a machine learning model to predict a cell phenotype of a T cell population, comprising training the machine learning model using datasets: (i) a dataset of reference population-level statistics, wherein for each of a first plurality of reference cell populations comprising T cells, the dataset of reference population-level statistics comprises one or more reference population-level statistics for each of a plurality of cell features, wherein: each reference population-level statistic is for one or more reference input measurements for one of the plurality of cell features derived from holographic information obtained for the reference cell population; and each reference input measurement is from an individual cell of the reference cell population; and (ii) a dataset of reference population-level output measurements, wherein for each of a second plurality of reference cell populations, the dataset of reference population-level output measurements comprises reference population-level output measurements of expression of a marker expressed by T cells having a cell phenotype of interest for the reference cell population, wherein: the first and second plurality of reference cell populations are from reference cell cultures that are cultured in vitro or ex vivo; A method is provided herein wherein a reference cell population from a first plurality of reference cell populations is from the same reference cell culture as a reference cell population from a second plurality of reference cell populations; whereby a machine learning model is trained to predict a population-level output measure of marker expression based on population-level statistics of the plurality of cell characteristics.

In some of the embodiments, the method is for determining the activation status of a T cell population, wherein the marker is expressed by the activated T cells. In some of the embodiments, the marker is CD137 (4-1BB).

Also provided in some embodiments is a method of training a machine learning model to predict the activation state of a T cell population, comprising training the machine learning model using datasets: (i) a dataset of reference population-level statistics, wherein for each of a first plurality of reference cell populations comprising T cells, the dataset of reference population-level statistics comprises one or more reference population-level statistics for each of a plurality of cell characteristics, wherein: each reference population-level statistic is for one or more reference input measurements for one of the plurality of cell characteristics derived from holographic information obtained for the reference cell population; and each reference input measurement is from an individual cell of the reference cell population; and (ii) a dataset of reference population-level output measurements, wherein for each of a second plurality of reference cell populations, the dataset of reference population-level output measurements comprises reference population-level output measurements of expression of a marker expressed by activated T cells for the reference cell population, wherein: the first and second plurality of reference cell populations are from reference cell cultures that are cultured in vitro or ex vivo; A method is provided herein wherein a reference cell population from a first plurality of reference cell populations is from the same reference cell culture as a reference cell population from a second plurality of reference cell populations; whereby a machine learning model is trained to predict a population-level output measure of marker expression based on population-level statistics of the plurality of cell characteristics.

In some of the embodiments, the method is for determining expression of a recombinant receptor in a T cell population, wherein the marker is a recombinant receptor introduced into T cells of the cell population prior to obtaining holographic information. In some of the embodiments, the recombinant receptor is a chimeric antigen receptor (CAR) or an engineered T cell receptor (TCR).

Also provided in some embodiments is a method of training a machine learning model to predict recombinant receptor expression in a T cell population, comprising training the machine learning model using datasets: (i) a dataset of reference population-level statistics, wherein for each of a first plurality of reference cell populations comprising T cells, the dataset of reference population-level statistics comprises one or more reference population-level statistics for each of a plurality of cell features, wherein: each reference population-level statistic is for one or more reference input measurements for one of the plurality of cell features derived from holographic information obtained for the reference cell population; and each reference input measurement is from an individual cell of the reference cell population; and (ii) a dataset of reference population-level output measurements, wherein for each of a second plurality of reference cell populations, the dataset of reference population-level output measurements comprises reference population-level output measurements of expression for the reference cell population of a recombinant receptor introduced into T cells of the reference cell population, wherein: the first and second plurality of reference cell populations are from reference cell cultures that are cultured in vitro or ex vivo; A method is provided herein wherein a reference cell population from a first plurality of reference cell populations is from the same reference cell culture as a reference cell population from a second plurality of reference cell populations; whereby a machine learning model is trained to predict a population-level output measure of expression of a recombinant receptor based on population-level statistics of the plurality of cell characteristics.

In some of the embodiments, the plurality of cell characteristics comprises one or more of intensity skewness, intensity correlation, intensity homogeneity, intensity maximum, intensity minimum, intensity entropy, intensity contrast, phase entropy, cell area, and radius mean.

In some of the embodiments, the plurality of cell features are selected from (including one or more of) intensity maxima, intensity minima, intensity entropy, intensity contrast, phase entropy, cell area, and radius average. Also provided herein is a method of determining an activation state of a T cell population, the method comprising: determining, for a cell population comprising T cells, expression of a marker expressed by activated T cells, wherein the marker is CD137, and wherein the determination is based on one or more input measurements for each of the plurality of cell features derived from holographic information obtained for the cell population, wherein: the plurality of cell features are selected from (including one or more of) intensity maxima, intensity minima, intensity entropy, intensity contrast, phase entropy, cell area, and radius average; wherein each input measurement is from an individual cell of the cell population.

In some of the embodiments, the plurality of cell characteristics include intensity maxima, intensity minima, intensity entropy, intensity contrast, phase entropy, cell area, and radius mean.

In some embodiments, provided herein is a method of determining an activation state of a T cell population, comprising: determining, for a cell population comprising T cells, expression of a marker expressed by activated T cells, wherein the marker is CD137 (4-1BB), wherein the determination is based on one or more input measurements for each of a plurality of cell characteristics derived from holographic information obtained for the cell population, wherein: the plurality of cell characteristics comprise an intensity maximum, an intensity minimum, an intensity entropy, an intensity contrast, a phase entropy, a cell area, and a radius mean; and wherein each input measurement is from an individual cell of the cell population.

Also provided herein is a method of determining an activation state of a T cell population, comprising: determining, for a cell population comprising T cells, a population-level output measurement of expression of a marker expressed by activated T cells, wherein the marker is CD137 (4-1BB), wherein the population-level output measurement is determined based on one or more input measurements for each of a plurality of cell features derived from holographic information obtained for the cell population, wherein: the plurality of cell features comprise an intensity maximum, an intensity minimum, an intensity entropy, an intensity contrast, a phase entropy, a cell area, and a radius mean; and wherein each input measurement is from an individual cell of the cell population.

In some of the embodiments, the plurality of cell characteristics comprises one or more of intensity skewness, intensity correlation, and intensity homogeneity. In some of the embodiments, the plurality of cell characteristics comprises intensity skewness, intensity correlation, and intensity homogeneity.

Also provided herein is a method of determining an activation state of a T cell population, comprising: determining, for a cell population comprising T cells, expression of a marker expressed by activated T cells, wherein the marker is CD137 (4-1BB), wherein the determination is based on one or more input measurements for each of a plurality of cell characteristics derived from holographic information obtained for the cell population, wherein: the plurality of cell characteristics comprise intensity skewness, intensity correlation, and intensity homogeneity; and wherein each input measurement is from an individual cell of the cell population.

Also provided herein is a method of determining an activation state of a T cell population, comprising: determining, for a cell population comprising T cells, a population-level output measurement of expression of a marker expressed by activated T cells, wherein the marker is CD137 (4-1BB), wherein the population-level output measurement is determined based on one or more input measurements for each of a plurality of cell characteristics derived from holographic information obtained for the cell population, wherein: the plurality of cell characteristics comprise intensity skewness, intensity correlation, and intensity homogeneity; and wherein each input measurement is from an individual cell of the cell population.

In some of the embodiments, the plurality of cell characteristics comprises one or more of an intensity maximum, an intensity minimum, an intensity entropy, an intensity contrast, a phase entropy, a cell area, and a radius average. In some of the embodiments, the plurality of cell characteristics comprises an intensity maximum, an intensity minimum, an intensity entropy, an intensity contrast, a phase entropy, a cell area, and a radius average.

In some of the embodiments, the plurality of cell characteristics comprises one or more of a peak area, a phase average uniformity, an intensity geometric mean, a minimum optical height, and a normalized optical height. In some of the embodiments, the plurality of cell characteristics comprises a peak area, a phase average uniformity, an intensity geometric mean, a minimum optical height, and a normalized optical height.

In some of the embodiments, the plurality of cell characteristics comprises one or more of cell area, perimeter length, average intensity, normalized peak area, and equivalent peak diameter. In some of the embodiments, the plurality of cell characteristics comprises cell area, perimeter length, average intensity, normalized peak area, and equivalent peak diameter.

Also provided herein is a method of determining a memory phenotype of a T cell population, comprising determining, for a cell population comprising T cells, expression of a marker expressed by central memory T cells or stem cell memory T cells, wherein the marker is CCR7, and wherein the determination is based on one or more input measurements for each of a plurality of cell characteristics derived from holographic information obtained for the cell population, wherein: the plurality of cell characteristics comprise one or more of cell area, perimeter length, mean intensity, normalized peak area, and equivalent peak diameter; and wherein each input measurement is from an individual cell of the cell population.

In some of the embodiments, the plurality of cell characteristics comprises an intensity maximum, an intensity minimum, an intensity entropy, an intensity contrast, a phase entropy, a cell area, and a radial mean. In some of the embodiments, the plurality of cell characteristics comprises one or more of a peak area, a phase mean uniformity, an intensity geometric mean, a minimum optical height, and a normalized optical height.

In some of the embodiments, the plurality of cell characteristics include peak area, phase average uniformity, intensity geometric mean, minimum optical height, and normalized optical height.

In some of the embodiments, the plurality of cell characteristics comprises one or more of cell area, perimeter length, average intensity, normalized peak area, and equivalent peak diameter. In some of the embodiments, the plurality of cell characteristics comprises cell area, perimeter length, average intensity, normalized peak area, and equivalent peak diameter.

Also provided herein is a method of determining a memory phenotype of a T cell population, comprising: determining, for a cell population comprising T cells, a population-level output measurement of expression of a marker expressed by central memory T cells or stem cell memory T cells, wherein the marker is CCR7, and wherein the population-level output measurement is determined based on one or more input measurements for each of a plurality of cell characteristics derived from holographic information obtained for the cell population, wherein: the plurality of cell characteristics comprise one or more of cell area, perimeter length, mean intensity, normalized peak area, and equivalent peak diameter; and wherein each input measurement is from an individual cell of the cell population.

In some embodiments, the plurality of features include cell area, perimeter length, mean intensity, normalized peak area, and equivalent peak diameter.

Also provided herein are methods for determining recombinant receptor expression in a T cell population, comprising: determining, for a cell population comprising T cells, expression of a recombinant receptor introduced into T cells of the cell population, wherein the determination is based on one or more input measurements for each of a plurality of cell characteristics derived from holographic information obtained for the cell population, wherein: the plurality of cell characteristics comprise a peak area, a phase average uniformity, an intensity geometric mean, a minimum optical height, and a normalized optical height; and wherein each input measurement is from an individual cell of the cell population.

In some of any of the embodiments, at least one, and optionally one or more, input measurements for each of the plurality of cell features are determined by providing holographic information about individual cells of the cell population to a convolutional neural network, wherein: the plurality of cell features are cell features extracted by the convolutional neural network; and the one or more input measurements are determined from the convolutional neural network. In some of any of the embodiments, at least one input measurement for each of the plurality of cell features is determined by providing holographic information about individual cells of the cell population to a convolutional neural network, wherein: the plurality of cell features are cell features extracted by the convolutional neural network; and the one or more input measurements are determined from the convolutional neural network.

In some of any of the embodiments, at least one, and optionally one or more, reference input measurements for each of the plurality of cell features are determined by providing holographic information about individual cells in the reference cell population to a convolutional neural network, wherein: the plurality of cell features are cell features extracted by the convolutional neural network; and the one or more reference input measurements are determined from the convolutional neural network. In some of any of the embodiments, at least one reference input measurement for each of the plurality of cell features is determined by providing holographic information about individual cells in the reference cell population to a convolutional neural network, wherein: the plurality of cell features are cell features extracted by the convolutional neural network; and the one or more reference input measurements are determined from the convolutional neural network.

In some of the embodiments, the convolutional neural network is trained using a dataset of reference holographic information, wherein for each of a first plurality of reference cell populations comprising T cells, the dataset of reference holographic information comprises holographic information obtained for individual cells of the reference cell population.

In some of any of the embodiments, the one or more reference population-level statistics for at least one, and optionally each, of the plurality of cell characteristics comprises one or more quantiles of one or more reference input measurements for the cell characteristic. In some of any of the embodiments, the one or more reference population-level statistics for each of the plurality of cell characteristics comprises one or more quantiles of one or more reference input measurements for the cell characteristic. In some of any of the embodiments, the one or more quantiles are selected from the 0.01, 0.1, 0.5, 0.9, and 0.99 quantiles of one or more reference input measurements for the cell characteristic.

In some of the embodiments, at least one, and optionally each, of the plurality of cell characteristics is determined by applying a distribution-based pooling filter to one or more reference input measurements for the cell characteristic. In some of the embodiments, at least one reference population-level statistic for each of the plurality of cell characteristics is determined by applying a distribution-based pooling filter to one or more reference input measurements for the cell characteristic.

Also provided in some embodiments is a method of training a machine learning model to predict a cell phenotype of a T cell population, comprising: (a) training a convolutional neural network using a dataset of reference holographic information, wherein for each of a first plurality of reference cell populations comprising T cells, the dataset of reference holographic information comprises holographic information obtained for individual cells of the reference cell population; (b) determining from the convolutional neural network one or more reference input measurements for each cell feature among a plurality of cell features derived from the holographic information, wherein the plurality of cell features are cell features extracted by the convolutional neural network, and each reference input measurement is from an individual cell of the reference cell population; (c) determining a dataset of reference population-level statistics, wherein the dataset of reference population-level statistics comprises one or more reference population-level statistics for each of a plurality of cell characteristics, each reference population-level statistic determined by applying a distribution-based pooling filter to one or more reference input measurements for one of the plurality of cell characteristics; and (d) training a machine learning model using the dataset of reference population-level statistics and the dataset of reference population-level output measurements, wherein for each of a second plurality of reference cell populations, the dataset of reference population-level output measurements comprises reference population-level output measurements of expression of a marker expressed by T cells having a cell phenotype of interest for the reference cell population, wherein: the first and second plurality of reference cell populations are from reference cell cultures that are being cultured in vitro or ex vivo; and the reference cell population from the first plurality of reference cell populations is from the same reference cell culture as the reference cell population from the second plurality of reference cell populations; A method is provided herein wherein a machine learning model is trained to predict a population-level output measure of marker expression based on population-level statistics of multiple cell characteristics.

In some of the embodiments, the method is for determining the activation status of a T cell population, wherein the marker is expressed by the activated T cells. In some of the embodiments, the marker is CD137 (4-1BB).

Also provided in some embodiments is a method of training a machine learning model for predicting an activation state of a T cell population, comprising: (a) training a convolutional neural network using a dataset of reference holographic information, wherein for each of a first plurality of reference cell populations comprising T cells, the dataset of reference holographic information comprises holographic information obtained for individual cells of the reference cell population; (b) determining from the convolutional neural network one or more reference input measurements for each cell feature among a plurality of cell features derived from the holographic information, wherein the plurality of cell features are cell features extracted by the convolutional neural network, and each reference input measurement is from an individual cell of the reference cell population; (c) determining a dataset of reference population-level statistics, wherein the dataset of reference population-level statistics comprises one or more reference population-level statistics for each of a plurality of cell characteristics, each reference population-level statistic determined by applying a distribution-based pooling filter to one or more reference input measurements for one of the plurality of cell characteristics; and (d) training a machine learning model using the dataset of reference population-level statistics and the dataset of reference population-level output measurements, wherein for each of a second plurality of reference cell populations, the dataset of reference population-level output measurements comprises reference population-level output measurements of expression of a marker expressed by activated T cells for the reference cell population, wherein: the first and second plurality of reference cell populations are from reference cell cultures that are being cultured in vitro or ex vivo; and the reference cell population from the first plurality of reference cell populations is from the same reference cell culture as the reference cell population from the second plurality of reference cell populations; A method is provided herein wherein a machine learning model is trained to predict a population-level output measure of marker expression based on population-level statistics of multiple cell characteristics.

In some of the embodiments, the method is for determining expression of a recombinant receptor in a T cell population, wherein the marker is a recombinant receptor introduced into T cells of the cell population prior to obtaining holographic information. In some of the embodiments, the recombinant receptor is a chimeric antigen receptor (CAR) or an engineered T cell receptor (TCR).

Also provided in some embodiments is a method of training a machine learning model to predict a recombination receptor of a T cell population, comprising: (a) training a convolutional neural network using a dataset of reference holographic information, wherein for each of a first plurality of reference cell populations comprising T cells, the dataset of reference holographic information comprises holographic information obtained for individual cells of the reference cell population; (b) determining from the convolutional neural network one or more reference input measurements for each cell feature among a plurality of cell features derived from the holographic information, wherein the plurality of cell features are cell features extracted by the convolutional neural network, and each reference input measurement is from an individual cell of the reference cell population; (c) determining a dataset of reference population-level statistics, wherein the dataset of reference population-level statistics comprises one or more reference population-level statistics for each of a plurality of cell characteristics, each reference population-level statistic being determined by applying a distribution-based pooling filter to one or more reference input measurements for one of the plurality of cell characteristics; and (d) training a machine learning model using the dataset of reference population-level statistics and the dataset of reference population-level output measurements, wherein for each of the second plurality of reference cell populations, the dataset of reference population-level output measurements comprises reference population-level output measurements for the reference cell population of expression of a recombinant receptor introduced into T cells of the reference cell population, wherein: the first and second plurality of reference cell populations are from reference cell cultures that are cultured in vitro or ex vivo; and the reference cell population from the first plurality of reference cell populations is from the same reference cell culture as the reference cell population from the second plurality of reference cell populations; A method is provided herein wherein a machine learning model is trained to predict a population-level output measure of expression of a recombinant receptor based on population-level statistics of multiple cell characteristics.

In some embodiments, at least one, and optionally one or more, input measurements for each of the plurality of cell characteristics are obtained from a fully connected layer of a convolutional neural network. In some embodiments, at least one, and optionally one or more, input measurements for each of the plurality of cell characteristics are obtained from a fully connected layer of a convolutional neural network.

In some of the embodiments, the holographic information for the first plurality of reference cell populations is obtained by DDHM.

In some of the embodiments, the holographic information includes phase information and intensity information.

In some of the embodiments, the reference population-level output measurement is a number of cells, a percentage of cells, a proportion of cells, or a density of cells within the reference cell population that express the marker.

In some embodiments, the dataset of reference population-level statistics and the dataset of reference population-level output measures are time-matched.

In some of the embodiments, the in vitro or ex vivo cultivation of the reference cell culture is performed under the same or similar conditions as the in vitro or ex vivo cultivation of the cell culture.

In some of the embodiments, the holographic information for the first plurality of reference cell populations is obtained during in vitro or ex vivo culture of the reference cell culture.

In some of the embodiments, the dataset of reference population-level output measurements relates to expression of a marker during or after in vitro or ex vivo culturing of a reference cell culture.

In some of the embodiments, the dataset of reference population-level output measurements is determined using fluorescence imaging of a second plurality of reference cell populations. In some of the embodiments, the fluorescence imaging is by flow cytometry.

In some of the embodiments, the method is for determining the activation state of a T cell population, wherein the marker is expressed by activated T cells.

In some of the embodiments, the marker is CD137 (4-1BB).

In some of the embodiments, the method is for determining the memory phenotype of a T cell population, and the marker is expressed by T cells having a central memory phenotype or a stem cell memory phenotype. In some of the embodiments, the marker is expressed by T cells having a central memory phenotype. In some of the embodiments, the marker is expressed by T cells having a stem cell memory phenotype. In some of the embodiments, the marker is CCR7.

In some of the embodiments, the method is for determining expression of a recombinant receptor in a T cell population, wherein the marker is a recombinant receptor introduced into T cells of the cell population prior to obtaining holographic information. In some of the embodiments, the recombinant receptor is a chimeric antigen receptor (CAR) or an engineered T cell receptor (TCR).

In some of the embodiments, the first plurality of reference cell populations are incubated under T cell stimulating conditions before holographic information is obtained for the first plurality of reference cell populations.

In some of the embodiments, the incubation of the first plurality of reference cell populations is performed under the same or similar conditions as the incubation of the cell populations.

In some of the embodiments, the incubation of the first plurality of reference cell populations occurs prior to in vitro or ex vivo culturing of the reference cell culture.

In some of the embodiments, the first and/or second plurality of reference cell populations are enriched for T cells. In some of the embodiments, the first and second plurality of reference cell populations are each enriched for T cells.

In some of any of the embodiments, at least 50, 55, 60, 65, 70, 75, 80, 85, 90, 91, 92, 93, 94, 95, 96, 97, 98, 99 or 100% of the first and/or second plurality of reference cell populations are T cells. In some of any of the embodiments, at least 50, 55, 60, 65, 70, 75, 80, 85, 90, 91, 92, 93, 94, 95, 96, 97, 98, 99 or 100% of each of the first and second plurality of reference cell populations are T cells.

Also provided herein in some embodiments is a method of monitoring a cell phenotype of a T cell culture, comprising determining a population-level output measurement of the expression of a marker expressed by T cells having a cell phenotype of interest for a population of cells from the cell culture comprising T cells cultured in vitro or ex vivo, wherein the population-level output measurement is determined according to any of the methods provided.

In some of the embodiments, the method is for monitoring the activation status of a T cell population, wherein the marker is expressed by activated T cells. In some of the embodiments, the marker is CD137 (4-1BB).

Also provided herein in some embodiments is a method of monitoring the activation state of a T cell culture, comprising determining a population-level output measurement of the expression of a marker expressed by activated T cells for a population of cells from a cell culture comprising T cells cultured in vitro or ex vivo, wherein the population-level output measurement is determined according to any one of the provided methods.

In some of the embodiments, the method is for monitoring expression of a recombinant receptor in a T cell population, wherein the marker is a recombinant receptor introduced into T cells of the cell population before obtaining holographic information. In some of the embodiments, the recombinant receptor is a chimeric antigen receptor (CAR) or an engineered T cell receptor (TCR).

Also provided herein in some embodiments is a method of monitoring recombinant receptor expression in a T cell culture, comprising determining a population-level output measurement of expression of a recombinant receptor introduced into T cells of a cell culture, for a population of cells from the cell culture comprising T cells cultured in vitro or ex vivo, wherein the population-level output measurement is determined according to any one of the methods provided.

Also provided herein in some embodiments is a computing device comprising instructions in a memory for performing any of the provided methods, the instructions comprising: (a) instructions for receiving holographic information about an individual cell of a cell population comprising T cells, one or more input measurements for each of a plurality of cell characteristics of an individual cell of the cell population comprising T cells, or a plurality of population-level statistics for the cell population comprising T cells; and (b) instructions for determining, according to the method, from the holographic information, a population-level output measurement for the cell population, one or more input measurements for each of the plurality of cell characteristics, or a plurality of population-level statistics.

In some of the embodiments, the computing device further comprises in its memory a machine learning model trained according to any of the provided methods, wherein the population-level output measurement for the cell population is determined using the machine learning model.

Figures 1 and 2 show validation results for a machine learning method for determining the overall activation state (e.g., CD137 positivity) of a T cell population using holographic imaging. In Figures 1 and 2, the T cell populations were subjected to different stimulation processes.
Figure 3 shows validation results for a machine learning method for determining recombinant receptor expression in a T cell population using holographic imaging.
Figure 4 shows validation results for a machine learning method for determining the memory phenotype of a T cell population using holographic imaging.

Provided herein are methods for determining the cellular phenotype, e.g., activation state, of a cell population containing T cells. In some aspects, the provided methods do not require labeling of the cells. In some aspects, the provided methods do not include labeling of the T cells. In some aspects, the provided methods are label-free. In some aspects, the provided methods can be used to monitor the cellular phenotype, e.g., activation state, of a cell culture containing T cells, for example, during in vitro or ex vivo culture and prior to downstream processing steps for the T cells. In some embodiments, the determination of the cellular phenotype, e.g., activation state, is performed using machine learning. Also provided herein are methods for training a machine learning model to predict the cellular phenotype, e.g., activation state, of a cell population containing T cells, as well as a computing device for use in performing any of the provided methods.

Existing methods for determining the cellular phenotype, such as activation status, of T cells (e.g., during in vitro or ex vivo culture of T cells) can be time-consuming or require specialized reagents, equipment, or trained personnel. In some instances, existing methods involve directly manipulating or labeling T cells, for example, by incubating them with immunoaffinity-based reagents that may interfere with the quality or function of the T cells. Such methods may pose a risk of cell contamination or damage prior to any downstream processing.

The provided methods allow for the label-free determination and monitoring of the cellular phenotype, such as activation status, of T cells. In some aspects, the provided methods involve determining activation status based on the expression of CD137 (4-1BB) in T cells. The results demonstrated herein demonstrate that CD137 is a particularly useful marker for predicting overall activation status, as these results are consistent with the finding that CD137 expression is associated with changes in characteristics that can be monitored by holographic imaging (e.g., size of activated T cells). In some aspects, without wishing to be bound by theory, the extent to which CD137 expression dynamically changes over time during activation contributes to its utility as a marker for predicting activation status while monitoring cells by holographic imaging. In some such embodiments, the variation in CD137 expression over time is associated with changes in cell size. In some such embodiments, the variation in CD137 expression is associated with one or more of cell size, stiffness, texture (or smoothness), and circularity. In some embodiments, the increase in CD137 expression is associated with one or more of (e.g., at least one, at least two, at least three, or all of) increased intensity, increased size (e.g., increased cell area and/or increased cell radius), increased texture (decreased cell smoothness), and lower circularity. In some embodiments, the increase in CD137 expression is associated with one or more of (e.g., at least one, at least two, at least three, or all of) decreased intensity, decreased size (e.g., cell area and/or cell radius), greater cell smoothness, and greater circularity.

The provided methods also allow for the label-free determination and monitoring of another cell phenotype, such as a memory phenotype, of T cells. In some aspects, the provided methods involve determining the memory phenotype of a T cell population. In some embodiments, the determination is based on the expression of CCR7 on the T cells, wherein CCR7-positive cells are typically considered to have an early memory phenotype, such as a central memory or stem cell memory phenotype, while CCR7-negative cells are typically considered to have an effector memory phenotype. See Blaeschke et al., Cancer Immunol Immunother, 67: 1053-1066 (2018). Early memory phenotypes, such as stem cell memory or central memory phenotypes, have been reported to result in durable in vivo responses of CAR T cell therapy, given their proliferative capacity and effector capacity in both hematological and solid tumor settings. See Gargett et al., Cytotherapy, 21: 593-602 (2019). This makes CCR7 a particularly relevant marker for therapeutic populations of T cells. The results demonstrated herein demonstrate that CCR7 expression is associated with changes in characteristics that can be monitored by holographic imaging. In some such embodiments, changes in CCR7 expression are associated with one or more of cell size, intensity, texture (or smoothness), and circularity. In some embodiments, increases in CCR7 expression are associated with one or more of the following features: cell area, intensity mean, perimeter length, equivalent peak diameter, and normalized peak area (e.g., at least two, at least three, at least four, or all of these). In some embodiments, increases in CCR7 expression are associated with one or more of the following features: (i) larger cell size, (ii) smaller mean intensity, (iii) larger phase correlation, and (iv) smaller normalized radial variance (e.g., at least two, at least three, at least four, or all of these). In some embodiments, larger cell size is indicated by larger area, longer perimeter length, and/or larger diameter.

In some aspects, the provided methods further comprise manipulating T cells whose phenotypes are determined and/or monitored to produce a cell therapy product.

In some aspects, the provided methods involve determining a population-level output measurement of the expression of a marker expressed by T cells, such as a marker expressed by activated T cells, for a cell population containing T cells. For example, in some embodiments, the population-level output measurement is the number of cells, the percentage of cells, the proportion of cells, or the density of cells within the cell population expressing the marker.

In some aspects, the population-level output measurement is determined based on holographic information obtained for individual cells in the cell population. In some aspects, the population-level output measurement is determined based on cell characteristics derived from the holographic information for the individual cells. In some aspects, the population-level output measurement is determined based on population-level statistics of the cell characteristics. In some aspects, the population-level statistics summarize input measurements of the cell characteristics across individual cells in the cell population. For example, in some embodiments, the population-level statistics include quantile values of input measurements across individual cells in the cell population for the cell characteristics.

In some aspects, population-level statistics of cell characteristics derived from holographic information obtained for individual cells of a reference T cell population are used to train a machine learning model that predicts population-level output measurements of marker expression. In some embodiments, the machine learning model is trained using reference population-level output measurements of marker expression for the reference cell population. In some embodiments, the reference population-level output measurements are determined using fluorescence imaging, such as immunoaffinity-based fluorescent labeling of the marker.

As demonstrated herein, marker expression of individual cells, such as activation marker expression, was not required or utilized in the development of the provided methods, thereby obviating the need for any specialized equipment or reagents (e.g., for capturing matched holographic and fluorescent images of individual cells). Furthermore, since the holographically imaged cells did not need to be fluorescently labeled as part of the provided methods, any random or systematic morphometric changes that may occur in labeled cells compared to unlabeled cells were avoided, thereby avoiding the impact on cell feature measurements and model prediction accuracy.

Furthermore, the same holographic imaging system used in the provided method to obtain holographic information for training a machine learning model can also be used to obtain holographic information for prediction using the trained machine learning model. This contrasts with other methods that may use different imaging systems to obtain holographic information for model training and use, such as a dual fluorescence-holographic imaging system for model training and a dedicated holographic imaging system for model use. Therefore, in some aspects, the machine learning model of the provided method is well suited for use in subsequent model applications after model training, as compared to other methods that involve the use of different imaging systems before and after model training.

In some embodiments, holographic information is provided as input to a convolutional neural network. In some embodiments, selection or design of cell features derived from the holographic information is not required prior to model training. Instead, in some embodiments, cell features can be automatically extracted from the holographic information as part of model training, for example, by a convolutional neural network. In some embodiments, automatically extracted cell features include those that may not be detectable by humans or are not known to predict marker expression. In some embodiments, the automatically extracted cell features are not detectable by humans based on visual inspection of the holographic image. In some embodiments, one or more of the automatically extracted cell features are not known to predict marker expression (e.g., activation markers). Therefore, in some aspects, the provided method is an accurate, efficient, objective, and unbiased method for predicting cell phenotypes, such as activation states.

In some aspects, the methods described herein can be used to monitor and characterize T cell phenotypes, such as activation state or memory status, using a non-destructive and non-damaging approach that allows imaging of T cells (e.g., unlabeled T cells) without damaging them (e.g., cells can be removed from a bioreactor and returned to the bioreactor intact). The methods provided can be used to monitor the dynamics of T cell phenotypes, such as activation state or memory status, during cultivation without frequent cell sampling or demanding analytical techniques. In some aspects, the methods provided can be used to identify relationships between cell phenotypes (e.g., activation state or memory status) and outcomes (e.g., how cells evolve over time).

In some embodiments, the provided methods can be used to predict the quality of T cells applied to a manufacturing process, e.g., the quality of T cells during or after the manufacturing process. In some aspects, cell phenotypes (e.g., activation status or memory status) as predicted by the provided methods can be used as a readout during manufacturing, e.g., as a readout of the success of manufacturing or the success of a manufacturing step (e.g., cell stimulation). In some embodiments, the provided methods can be used to monitor whether T cells are sufficiently activated, e.g., for expansion of T cells to a desired threshold number, e.g., a number required for a clinical dose of T cells for T cell therapy.

In some aspects, T cell activation can lead to T cell differentiation. In T cell therapy, a higher proportion of early memory T cells, such as naive-like T cells, can improve patient outcomes (see, e.g., Jiang et al., Journal of Pharmaceutical Sciences (2021) 110:1871-1876). In some embodiments, the provided methods can be used to monitor the memory status of T cells, either directly or by monitoring the activation status of T cells. In some embodiments, the activation status of T cells is monitored to predict the memory status of T cells.

In some aspects, the cell phenotypic information obtained by the provided methods can be used to optimize the duration or other conditions of a manufacturing process or steps thereof to improve the quality of the processed T cells during process development. In some instances, this information can be used to develop a process control strategy, wherein, for example, if the predicted cell phenotype, e.g., activation state, falls outside a determined range, the conditions of one or more (e.g., the current or subsequent) manufacturing steps can be altered, such as the duration of the current or subsequent manufacturing steps, to improve the final quality of the T cells produced. For example, if the predicted cell phenotype, e.g., activation state, falls outside a determined range during cultivation, subsequent cultivation can be performed, in some instances, under perfusion conditions and/or in the presence of small molecules, e.g., to adjust the T cell phenotype toward a desired profile.

In some aspects, the cell phenotypic information obtained by the provided methods can be used to assess or reduce batch-to-batch variability of T cells applied to a manufacturing process. In some aspects, the cell phenotypic information can be used to assess or reduce batch-to-batch variability of a drug product produced using the manufacturing process. For example, by ensuring that T cells are in a similar activation state across different cell therapy manufacturing runs, the differentiation and memory status of the T cells can be maintained consistently. This can reduce variability (e.g., inter-patient variability) of the resulting T cell therapy (see, e.g., Jiang et al., Journal of Pharmaceutical Sciences (2021) 110:1871-1876).

In some embodiments, the provided methods can be used to monitor the activation status of T cells before or after manipulation of the T cells. In some aspects, transgene expression may be higher in activated versus non-activated T cells, such as after viral transduction of T cells (see, e.g., Ghassemi et al., Nature Biomedical Engineering (2022) 6:118-128). In some aspects, electroporation efficiency for manipulation may be higher in activated versus non-activated T cells (see, e.g., Zhang et al., BMC Biotechnology (2018) 18:4). In some embodiments, the T cells are monitored according to the provided methods before manipulation, e.g., manipulation may be initiated if the provided methods predict that the T cells are sufficiently activated for improved transgene expression. In some embodiments, the T cells are monitored according to the provided methods after manipulation, e.g., to determine whether the T cells are sufficiently activated to improve transgene expression or remain activated after manipulation.

All publications, including patent documents, scientific papers, and databases, referred to in this application are incorporated by reference in their entirety for all purposes to the same extent as if each individual publication were incorporated by reference individually. If a definition provided herein conflicts with or is inconsistent with a definition provided in a patent, application, published application, or other publication incorporated by reference herein, the definition provided herein shall take precedence over the definition incorporated by reference herein.

The section headings used herein are for organizational purposes only and should not be construed as limiting the subject matter covered.

How to determine I. T cell phenotype

In some embodiments, the provided method involves determining the cellular phenotype, such as the activation state, of a cell population containing T cells. In some embodiments, the provided method is for determining the cellular phenotype, such as the activation state, of a cell population containing T cells. Exemplary cell populations are described in Section II. In some embodiments, the provided method involves performing any of the cell processing steps described in Section II on the cell population.

In some embodiments, provided herein is a method of determining a cell phenotype of a T cell population, comprising: determining, for a cell population comprising T cells, a population-level output measurement of expression of a marker expressed by T cells having a cell phenotype of interest, wherein the population-level output measurement is determined based on a plurality of population-level statistics, wherein: each population-level statistic is for one or more input measurements for one of a plurality of cell characteristics derived from holographic information obtained for the cell population, the plurality of population-level statistics comprising one or more population-level statistics for each of the plurality of cell characteristics; and wherein each input measurement is from an individual cell of the cell population.

Also provided herein, in some embodiments, is a method of determining a cellular phenotype of a T cell population, comprising: (a) obtaining holographic information for a cell population comprising T cells; (b) determining one or more input measurements for each of a plurality of cell characteristics derived from the holographic information, wherein each input measurement is from an individual cell within the cell population; (c) determining a plurality of population-level statistics, wherein each population-level statistic is for one or more input measurements for one of the plurality of cell characteristics, and wherein the plurality of population-level statistics comprises one or more population-level statistics for each of the plurality of cell characteristics; and (d) determining, based on the plurality of population-level statistics, a population-level output measurement of expression of a marker expressed by T cells having a cell phenotype of interest for the cell population. In some embodiments, the term "determining" means making a prediction.

Accordingly, in some embodiments, provided herein is a method of predicting a cellular phenotype of a T cell population, comprising: (a) obtaining holographic information for a cell population comprising T cells; (b) predicting one or more input measurements for each of a plurality of cell characteristics derived from the holographic information, wherein each input measurement is from an individual cell within the cell population; (c) predicting a plurality of population-level statistics, wherein each population-level statistic is for one or more input measurements for one of the plurality of cell characteristics, and wherein the plurality of population-level statistics comprises one or more population-level statistics for each of the plurality of cell characteristics; and (d) predicting, based on the plurality of population-level statistics, a population-level output measurement of expression of a marker expressed by T cells having a cell phenotype of interest for the cell population.

Also provided in some embodiments is a method of training a machine learning model to predict a cell phenotype of a T cell population, comprising training the machine learning model using datasets: (i) a dataset of reference population-level statistics, wherein for each of a first plurality of reference cell populations comprising T cells, the dataset of reference population-level statistics comprises one or more reference population-level statistics for each of a plurality of cell features, wherein: each reference population-level statistic is for one or more reference input measurements for one of the plurality of cell features derived from holographic information obtained for the reference cell population; and each reference input measurement is from an individual cell of the reference cell population; and (ii) a dataset of reference population-level output measurements, wherein for each of a second plurality of reference cell populations, the dataset of reference population-level output measurements comprises reference population-level output measurements of expression of a marker expressed by T cells having a cell phenotype of interest for the reference cell population, wherein: the first and second plurality of reference cell populations are from reference cell cultures that are cultured in vitro or ex vivo; A method is provided herein wherein a reference cell population from a first plurality of reference cell populations is from the same reference cell culture as a reference cell population from a second plurality of reference cell populations; whereby a machine learning model is trained to predict a population-level output measure of marker expression based on population-level statistics of the plurality of cell characteristics.

Also provided herein is a method of determining an activation state of a T cell population, comprising: determining, for a cell population comprising T cells, a population-level output measurement of expression of a marker expressed by activated T cells, wherein the marker is CD137 (4-1BB), wherein the population-level output measurement is determined based on one or more input measurements for each of a plurality of cell features derived from holographic information obtained for the cell population, wherein: the plurality of cell features comprise an intensity maximum, an intensity minimum, an intensity entropy, an intensity contrast, a phase entropy, a cell area, and a radius mean; and wherein each input measurement is from an individual cell of the cell population.

Also provided herein is a method of determining an activation state of a T cell population, comprising: determining, for a cell population comprising T cells, a population-level output measurement of expression of a marker expressed by activated T cells, wherein the marker is CD137 (4-1BB), wherein the population-level output measurement is determined based on one or more input measurements for each of a plurality of cell characteristics derived from holographic information obtained for the cell population, wherein: the plurality of cell characteristics comprise intensity skewness, intensity correlation, and intensity homogeneity; and wherein each input measurement is from an individual cell of the cell population.

In some embodiments, the term “decision” means prediction.

Accordingly, in some embodiments, provided herein is a method of predicting an activation state of a T cell population, comprising: predicting, for a cell population comprising T cells, a population-level output measurement of expression of a marker expressed by activated T cells, wherein the marker is CD137 (4-1BB), wherein the population-level output measurement is predicted based on one or more input measurements for each of a plurality of cell features derived from holographic information obtained for the cell population, wherein: the plurality of cell features comprise an intensity maximum, an intensity minimum, an intensity entropy, an intensity contrast, a phase entropy, a cell area, and a radius mean; and wherein each input measurement is from an individual cell of the cell population.

Also provided herein is a method of predicting an activation state of a T cell population, comprising predicting, for a cell population comprising T cells, a population-level output measurement of expression of a marker expressed by activated T cells, wherein the marker is CD137 (4-1BB), wherein the population-level output measurement is predicted based on one or more input measurements for each of a plurality of cell characteristics derived from holographic information obtained for the cell population, wherein: the plurality of cell characteristics comprise intensity skewness, intensity correlation, and intensity homogeneity; and wherein each input measurement is from an individual cell of the cell population.

Also provided herein, in some embodiments, is a method of determining a memory phenotype of a T cell population, comprising: determining, for a cell population comprising T cells, a population-level output measurement of expression of a marker expressed by central memory T cells or stem cell memory T cells, wherein the marker is CCR7, and wherein the population-level output measurement is determined based on one or more input measurements for each of a plurality of cell characteristics derived from holographic information obtained for the cell population, wherein: the plurality of cell characteristics comprise one or more of cell area, perimeter length, mean intensity, normalized peak area, and equivalent peak diameter; and wherein each input measurement is from an individual cell of the cell population. In some embodiments, the term "determining" means making a prediction.

Accordingly, in some embodiments, provided herein is a method of predicting a memory phenotype of a T cell population, comprising predicting, for a cell population comprising T cells, a population-level output measurement of expression of a marker expressed by central memory T cells or stem cell memory T cells, wherein the marker is CCR7, and wherein the population-level output measurement is predicted based on one or more input measurements for each of a plurality of cell characteristics derived from holographic information obtained for the cell population, wherein: the plurality of cell characteristics comprise one or more of cell area, perimeter length, mean intensity, normalized peak area, and equivalent peak diameter; and wherein each input measurement is from an individual cell of the cell population.

Also provided in some embodiments is a method of training a machine learning model for predicting a cell phenotype of a T cell population, comprising: (a) training a convolutional neural network using a dataset of reference holographic information, wherein for each of a first plurality of reference cell populations comprising T cells, the dataset of reference holographic information comprises holographic information obtained for individual cells of the reference cell population; (b) determining from the convolutional neural network one or more reference input measurements for each cell feature among a plurality of cell features derived from the holographic information, wherein the plurality of cell features are cell features extracted by the convolutional neural network, and each reference input measurement is from an individual cell of the reference cell population; (c) determining a dataset of reference population-level statistics, wherein the dataset of reference population-level statistics comprises one or more reference population-level statistics for each of a plurality of cell characteristics, each reference population-level statistic determined by applying a distribution-based pooling filter to one or more reference input measurements for one of the plurality of cell characteristics; and (d) training a machine learning model using the dataset of reference population-level statistics and the dataset of reference population-level output measurements, wherein for each of a second plurality of reference cell populations, the dataset of reference population-level output measurements comprises reference population-level output measurements of expression of a marker expressed by T cells having a cell phenotype of interest for the reference cell population, wherein: the first and second plurality of reference cell populations are from reference cell cultures that are being cultured in vitro or ex vivo; and the reference cell population from the first plurality of reference cell populations is from the same reference cell culture as the reference cell population from the second plurality of reference cell populations; A method is provided herein wherein a machine learning model is trained to predict a population-level output measure of marker expression based on population-level statistics of multiple cell characteristics.

Also provided herein are methods for determining recombinant receptor expression in a T cell population, comprising: determining, for a cell population comprising T cells, expression of a recombinant receptor introduced into T cells of the cell population, wherein the determination is based on one or more input measurements for each of a plurality of cell features derived from holographic information obtained for the cell population, wherein: the plurality of cell features comprise a peak area, a phase average uniformity, an intensity geometric mean, a minimum optical height, and a normalized optical height; and wherein each input measurement is from an individual cell of the cell population. In some embodiments, the term "determining" means making a prediction.

Accordingly, in some embodiments, provided herein is a method of predicting recombinant receptor expression in a T cell population, comprising: predicting, for a cell population comprising T cells, expression of a recombinant receptor introduced into T cells of the cell population, wherein the determination is based on one or more input measurements for each of a plurality of cell features derived from holographic information obtained for the cell population, wherein: the plurality of cell features comprise a peak area, a phase average uniformity, an intensity geometric mean, a minimum optical height, and a normalized optical height; and wherein each input measurement is from an individual cell of the cell population.

In some embodiments, the cell phenotype is the expression of a marker by cells in the cell population. In some embodiments, the marker is expressed on the surface of T cells. In some embodiments, the provided methods involve determining the presence or absence of marker-expressing cells in the cell population, the extent to which marker-expressing cells are present in the cell population, or the extent to which marker-expressing cells are present in T cells in the cell population. In some embodiments, the presence or absence of marker-expressing cells in the cell population is determined. In some embodiments, the extent to which marker-expressing cells are present in the cell population is determined. In some embodiments, the extent to which marker-expressing cells are present in T cells in the cell population is determined.

In some embodiments, the cell phenotype is based on the expression of a combination of one or more markers by cells of the cell population. In some embodiments, the cell phenotype is based on a population-level output measure of the expression of a combination of one or more markers by cells of the cell population. In some embodiments, the markers are expressed on the surface of T cells. In some embodiments, the cell phenotype is the expression of some markers of the combination of markers and the non-expression of the remaining markers of the combination of markers. In some embodiments, the cell phenotype is a population-level output measure of the expression of some markers of the combination of markers and the non-expression of the remaining markers of the combination of markers. In some embodiments, the cell phenotype is the expression of each of the combination of markers. In some embodiments, the cell phenotype is a population-level output measure of the expression of each of the combination of markers.

In some embodiments, references herein to a "marker" also refer to a combination of markers, such as any combination of the markers described herein. In some embodiments, references herein to "marker expression," "marker-expressing cell," "cell expressing a marker," or similar terms also refer to combined marker expression and/or non-expression consistent with a cell phenotype of interest. For example, for a cell phenotype of CCR7+CD45RA-, a "marker-expressing cell" may refer to a CCR7+CD45RA- cell. Similarly, for a cell phenotype of CCR7+CD27+, a "marker-expressing cell" may refer to a CCR7+CD27+ cell.

In some embodiments, the provided methods involve determining the activation status of a cell population. In some embodiments, "activation status" refers to the presence or absence of activated T cells within a cell population, the extent to which activated T cells are present in a cell population, or the extent to which activated T cells are present among T cells in a cell population. In some embodiments, the presence or absence of activated T cells within a cell population is determined. In some embodiments, the extent to which activated T cells are present in a cell population is determined. In some embodiments, the extent to which activated T cells are present among T cells in a cell population is determined. In some embodiments, such determination is a prediction.

In some embodiments, the provided methods involve determining the memory status of a cell population. In some embodiments, "memory status" refers to the presence or absence of T cells of a particular memory phenotype within a cell population, the extent to which T cells of a particular memory phenotype are present in a cell population, or the extent to which T cells of a particular memory phenotype are present in T cells of a cell population. In some embodiments, the presence or absence of T cells of a particular memory phenotype within a cell population is determined. In some embodiments, the extent to which T cells of a particular memory phenotype are present in a cell population is determined. In some embodiments, the extent to which T cells of a particular memory phenotype are present in T cells of a cell population is determined.

In some embodiments, the memory phenotype is for naive-like T cells. In some embodiments, the presence or absence of naive-like T cells in a cell population is determined. In some embodiments, the extent to which naive-like T cells are present in the cell population is determined. In some embodiments, the extent to which naive-like T cells are present in the T cells of the cell population is determined. In some embodiments, the memory phenotype is a central memory phenotype or a stem cell memory phenotype. In some embodiments, the presence or absence of T cells having a central memory phenotype or a stem cell memory phenotype in a cell population is determined. In some embodiments, the extent to which T cells having a central memory phenotype or a stem cell memory phenotype are present in the cell population is determined. In some embodiments, the extent to which T cells having a central memory phenotype or a stem cell memory phenotype are present in the T cells of the cell population is determined. In some embodiments, the determination is a prediction.

In some embodiments, the memory phenotype is for central memory T cells. In some embodiments, the presence or absence of central memory T cells in a cell population is determined. In some embodiments, the extent to which central memory T cells are present in a cell population is determined. In some embodiments, the extent to which central memory T cells are present in T cells in a cell population is determined.

In some embodiments, the memory phenotype is for effector memory T cells. In some embodiments, the presence or absence of effector memory T cells in a cell population is determined. In some embodiments, the extent to which effector memory T cells are present in a cell population is determined. In some embodiments, the extent to which effector memory T cells are present in T cells in a cell population is determined.

In some embodiments, the memory phenotype is for terminally differentiated T cells. In some embodiments, the presence or absence of terminally differentiated T cells in a cell population is determined. In some embodiments, the extent to which terminally differentiated T cells are present in the cell population is determined. In some embodiments, the extent to which terminally differentiated T cells are present in T cells of the cell population is determined. In some embodiments, the provided methods involve determining, for a cell population, a population-level output measurement indicative of a cell phenotype, e.g., activation state, of the cell population. In some embodiments, the population-level output measurement indicates the presence or absence of marker-expressing cells in the cell population. In some embodiments, the population-level output measurement indicates the extent to which marker-expressing cells are present in the cell population. In some embodiments, the population-level output measurement indicates the extent to which marker-expressing cells are present in T cells of the cell population.

In some embodiments, the population-level output measurement is the number of marker-expressing cells, the percentage of marker-expressing cells, the proportion of marker-expressing cells, or the density of marker-expressing cells within the cell population. In some embodiments, the population-level output measurement is the number of marker-expressing cells, the percentage of marker-expressing cells, the proportion of marker-expressing cells, or the density of marker-expressing cells within T cells of the cell population. In some embodiments, the population-level output measurement is the percentage of marker-expressing cells within T cells of the cell population.

In some embodiments, the marker is expressed on the surface of the T cell.

In some embodiments, the marker is CD3. In some embodiments, the marker is CD4. In some embodiments, the marker is CD8.

In some embodiments, the marker is a recombinant protein. In some embodiments, the recombinant protein is a recombinant receptor. In some embodiments, the recombinant receptor is a chimeric antigen receptor (CAR). Exemplary CARs are described in Section II-C-2.

In some embodiments, the recombinant receptor is a T cell receptor (TCR). Exemplary TCRs are described in Section II-C-3.

In some embodiments, the provided method involves determining, for a cell population, a population-level output measurement indicative of an activation state of the cell population. In some embodiments, the population-level output measurement indicates the presence or absence of activated T cells within the cell population. In some embodiments, the population-level output measurement indicates the extent to which activated T cells are present in the cell population. In some embodiments, the population-level output measurement indicates the extent to which activated T cells are present in the T cells of the cell population.

In some embodiments, the population-level output measurement is the number of activated T cells, the percentage of activated T cells, the proportion of activated T cells, or the density of activated T cells within the cell population. In some embodiments, the population-level output measurement is the number of activated T cells, the percentage of activated T cells, the proportion of activated T cells, or the density of activated T cells within the T cells of the cell population. In some embodiments, the population-level output measurement is the percentage of activated T cells within the T cells of the cell population.

In some embodiments, the provided methods involve determining a population-level output measurement of the expression of a marker indicative of T cell activation for a cell population. In some embodiments, an "activation state" refers to the presence of T cells expressing the marker within a cell population, the extent to which T cells expressing the marker are present in the cell population, or the extent to which T cells expressing the marker are present in T cells of the cell population.

In some embodiments, the marker is a marker expressed by activated T cells (an "activation marker"). In some embodiments, an "activated T cell" refers to a T cell that expresses such a marker. In some embodiments, the marker is CD137 (4-1BB). In some embodiments, the marker is CD25. In some embodiments, the marker is CD69. In some embodiments, the marker is a combination of two or more activation markers, e.g., two or more of CD137, CD25, and CD69. In some embodiments, the marker comprises CD137. In some embodiments, the marker comprises CD137 and one or more additional activation markers. In some embodiments, the one or more additional activation markers comprise CD25 and/or CD69.

In other embodiments, the marker is a marker expressed by non-activated T cells. In some embodiments, "activated T cells" refers to T cells that do not express such markers.

In some embodiments, the provided methods involve determining, for a cell population, a population-level output measurement indicative of the memory status of the cell population. In some embodiments, the population-level output measurement indicates the presence or absence of T cells of a particular memory phenotype within the cell population. In some embodiments, the population-level output measurement indicates the extent to which T cells of a particular memory phenotype are present in the cell population. In some embodiments, the population-level output measurement indicates the extent to which T cells of a particular memory phenotype are present in the T cells of the cell population. Early memory phenotypes, such as stem cell memory or central memory phenotypes, have been reported to result in durable in vivo responses of CAR T cell therapies, given their proliferative capacity and effector capacity in both hematological and solid tumor settings. See Gargett et al., Cytotherapy, 21: 593-602 (2019). CCR7-positive cells are typically considered to be of the early memory phenotype.

In some embodiments, the population-level output measurement is the number of T cells of a particular memory phenotype, the percentage of T cells of a particular memory phenotype, the proportion of T cells of a particular memory phenotype, or the density of T cells of a particular memory phenotype within a cell population. In some embodiments, the population-level output measurement is the number of T cells of a particular memory phenotype, the percentage of T cells of a particular memory phenotype, the proportion of T cells of a particular memory phenotype, or the density of T cells of a particular memory phenotype within a cell population. In some embodiments, the population-level output measurement is the percentage of T cells of a particular memory phenotype within a cell population.

In some embodiments, the memory phenotype is for naive-like T cells. In some embodiments, the population-level output measurement is the number of naive-like T cells, the percentage of naive-like T cells, the proportion of naive-like T cells, or the density of naive-like T cells within a cell population. In some embodiments, the population-level output measurement is the number of naive-like T cells, the percentage of naive-like T cells, the proportion of naive-like T cells, or the density of naive-like T cells within T cells of the cell population. In some embodiments, the population-level output measurement is the percentage of naive-like T cells within T cells of the cell population.

In some embodiments, the memory phenotype is a central memory phenotype or a stem cell memory phenotype.

In some embodiments, the memory phenotype is for central memory T cells. In some embodiments, the population-level output measurement is the number of central memory T cells, the percentage of central memory T cells, the proportion of central memory T cells, or the density of central memory T cells within a cell population. In some embodiments, the population-level output measurement is the number of central memory T cells, the percentage of central memory T cells, the proportion of central memory T cells, or the density of central memory T cells within T cells of a cell population. In some embodiments, the population-level output measurement is the percentage of central memory T cells within T cells of a cell population.

In some embodiments, the memory phenotype is for stem cell memory T cells. In some embodiments, the population-level output measurement is the number of stem cell memory T cells, the percentage of stem cell memory T cells, the proportion of stem cell memory T cells, or the density of stem cell memory T cells within the cell population. In some embodiments, the population-level output measurement is the number of stem cell memory T cells, the percentage of stem cell memory T cells, the proportion of stem cell memory T cells, or the density of stem cell memory T cells within the T cells of the cell population. In some embodiments, the population-level output measurement is the percentage of stem cell memory T cells within the T cells of the cell population.

In some embodiments, the memory phenotype is for effector memory T cells. In some embodiments, the population-level output measurement is the number of effector memory T cells, the percentage of effector memory T cells, the proportion of effector memory T cells, or the density of effector memory T cells within a cell population. In some embodiments, the population-level output measurement is the number of effector memory T cells, the percentage of effector memory T cells, the proportion of effector memory T cells, or the density of effector memory T cells within T cells of a cell population. In some embodiments, the population-level output measurement is the percentage of effector memory T cells within T cells of a cell population.

In some embodiments, the memory phenotype is for terminally differentiated T cells. In some embodiments, the population-level output measurement is the number of terminally differentiated T cells, the percentage of terminally differentiated T cells, the proportion of terminally differentiated T cells, or the density of terminally differentiated T cells within the cell population. In some embodiments, the population-level output measurement is the number of terminally differentiated T cells, the percentage of terminally differentiated T cells, the proportion of terminally differentiated T cells, or the density of terminally differentiated T cells within the T cells of the cell population. In some embodiments, the population-level output measurement is the percentage of terminally differentiated T cells within the T cells of the cell population.

In some embodiments, the provided methods involve determining a population-level output measurement of the expression of a marker indicative of T cell memory status for a cell population. In some embodiments, "memory status" refers to the presence of T cells expressing the marker within a cell population, the extent to which T cells expressing the marker are present in the cell population, or the extent to which T cells expressing the marker are present in T cells of the cell population.

In some embodiments, the marker is a marker expressed by naive T cells (a "naive-like marker"). In some embodiments, "naive-like T cells" refers to T cells that express such a marker. In some embodiments, the marker is CCR7. In some embodiments, the marker is CD27. In some embodiments, the marker is CD45RA.

In other embodiments, the marker is a marker expressed by non-naive T cells. In some embodiments, "naive T cells" refers to T cells that do not express such markers.

In some embodiments, the combination of markers is any combination of markers described herein.

In some embodiments, the combination of markers is a combination of markers indicative of a memory phenotype of T cells. In some embodiments, the combination of markers is a combination of markers expressed by T cells of a specific memory phenotype.

In some embodiments, the memory phenotype is an early memory phenotype. In some embodiments, the early memory phenotype is a central memory phenotype or a stem cell memory phenotype. In some embodiments, the early memory phenotype is a central memory phenotype. In some embodiments, the early memory phenotype is a stem cell memory phenotype. In some embodiments, the marker expressed by the early memory phenotype, e.g., the central memory phenotype or the stem cell memory phenotype, is CCR7.

In some embodiments, the memory phenotype is for naive-like T cells. In some embodiments, the naive-like T cells are CCR7+CD27+. In some embodiments, the naive-like T cells are CCR7+CD45RA+. In some embodiments, the naive-like T cells are CD27+CD45RA+.

In some embodiments, the memory phenotype is for central memory T cells. In some embodiments, the central memory T cells are CCR7+CD45RA-.

In some embodiments, the memory phenotype is for effector memory T cells. In some embodiments, the effector memory T cells are CCR7-CD45RA-.

In some embodiments, the memory phenotype is for terminally differentiated T cells. In some embodiments, the effector memory T cells are CCR7-CD45RA+.

In some embodiments, the population-level output measurement indicates the presence of T cells expressing the marker within a cell population. In some embodiments, the population-level output measurement indicates the extent to which T cells expressing the marker are present within a cell population. In some embodiments, the population-level output measurement indicates the extent to which T cells expressing the marker are present within a cell population.

In some embodiments, the population-level output measurement is the number, percentage, proportion, or density of cells within a cell population that express the marker. In some embodiments, the population-level output measurement is the number, percentage, proportion, or density of cells within a T cell population that express the marker. In some embodiments, the population-level output measurement is the percentage of cells within a T cell population that express the marker.

In some embodiments, a cell phenotype, e.g., an activation state, is determined based on holographic information obtained for a cell population, e.g., an individual cell of the cell population. In some embodiments, a population-level output measurement is determined based on holographic information obtained for a cell population, e.g., an individual cell of the cell population. In some embodiments, the provided methods involve obtaining holographic information for a cell population, e.g., an individual cell of the cell population. In some embodiments, the holographic information comprises holographic information for one or more individual cells of the cell population. In some embodiments, the holographic information comprises holographic information for each of a plurality of individual cells of the cell population. Exemplary holographic information and methods for obtaining the same are described in Section I-A.

In some embodiments, the cell phenotype, e.g., activation state, is also determined based on non-holographic information obtained for the cell population. In some embodiments, the population-level output measurement is also determined based on non-holographic information obtained for the cell population. In some embodiments, the non-holographic information comprises non-holographic information for one or more individual cells of the cell population. In some embodiments, the non-holographic information comprises non-holographic information for each of a plurality of individual cells of the cell population.

In some embodiments, a cell phenotype, e.g., activation state, is determined based on one or more input measurements for each of one or more cell characteristics. In some embodiments, a population-level output measurement is determined based on one or more input measurements for each of one or more cell characteristics. In some embodiments, a "cell characteristic" refers to a characteristic of an individual cell. In some embodiments, one or more cell characteristics are derived from holographic information. In some embodiments, each cell characteristic is derived from holographic information obtained for an individual cell.

In some embodiments, an "input measurement" refers to a determined, e.g., measured or calculated, value for a feature, e.g., a cell feature. In some embodiments, the provided methods involve determining one or more input measurements for each of one or more cell features. In some embodiments, each input measurement is determined from holographic information obtained for a cell population, e.g., an individual cell of the cell population. In some embodiments, each input measurement is from an individual cell of the cell population. In some embodiments, each input measurement is determined from holographic information obtained for an individual cell of the cell population. In some embodiments, each input measurement is determined from holographic information. In some embodiments, each input measurement is calculated using holographic information. Exemplary cell features and methods for determining input measurements therefor are described in Section I-B.

In some embodiments, the cell phenotype, e.g., activation state, is also determined based on one or more other input measurements for each of one or more other features derived from non-holographic information. In some embodiments, the population-level output measurements are also determined based on one or more other input measurements for each of one or more other features derived from non-holographic information. In some embodiments, each of the other features is a characteristic of an individual cell. In some embodiments, each of the other features is derived from non-holographic information obtained for an individual cell.

In some embodiments, a cell phenotype, e.g., activation state, is determined based on at least one population-level statistic. In some embodiments, a population-level output measurement is determined based on at least one population-level statistic. In some embodiments, each population-level statistic is for one or more input measurements of one or more cell characteristics. In some embodiments, each population-level statistic describes one or more input measurements. In some embodiments, each population-level statistic describes a distribution of one or more input measurements. In some embodiments, the provided methods involve determining each population-level statistic from one or more input measurements. In some embodiments, each population-level statistic is calculated using one or more input measurements. Exemplary population-level statistics are described in Section I-C. Exemplary methods for determining a population-level output measurement based on at least one population-level statistic are described in Section I-D.

In some embodiments, the cell phenotype, e.g., activation state, is determined based on one or more other population-level statistics. In some embodiments, the population-level output measurements are determined based on one or more other population-level statistics. In some embodiments, the one or more other population-level statistics are for one or more other input measurements for each of one or more other features derived from non-holographic information.

In some embodiments, any number of the steps of the provided method are performed in a closed system. In some embodiments, any number of the steps of the provided method are automated.

A. Holographic information

In some embodiments, the cellular phenotype, e.g., activation state, of a cell population is determined based on holographic information obtained for the cell population, e.g., for individual cells of the cell population. In some embodiments, a population-level output measurement is determined based on holographic information obtained for the cell population, e.g., for individual cells of the cell population. In some embodiments, the holographic information obtained for the cell population comprises holographic information for each of a plurality of individual cells of the cell population. In some embodiments, the holographic information for each of the plurality of individual cells is obtained simultaneously or nearly simultaneously.

In some embodiments, the holographic information is obtained using a method that does not involve recording a projected image of the cell population. In some embodiments, the holographic information comprises a obtained hologram of the cell population. In some embodiments, the holographic information is obtained using an optical wavefront of interest and a reference wavefront. In some embodiments, the holographic information comprises phase information and intensity information obtained for the cell population, as described, for example, in Alm et al. (2013), "Cells and Holograms - Holograms and Digital Holographic Microscopy as a Tool to Study the Morphology of Living Cells" in Holography: Basic Principles and Contemporary Applications. In some embodiments, the phase information and intensity information are obtained using an optical wavefront of interest and a reference wavefront.

In some embodiments, the image of the cell population can be reconstructed from holographic information. In some embodiments, the image of the cell population can be reconstructed from holograms. In some embodiments, the image of the cell population can be reconstructed from phase information and intensity information. In some embodiments, the reconstruction is performed using a computer and numerical algorithm.

In some embodiments, the provided methods involve obtaining holographic information about a population of cells, e.g., individual cells of the population of cells. In some embodiments, the holographic information is obtained by microscopy. In some embodiments, the holographic information is obtained by imaging the population of cells, e.g., individual cells of the population of cells, using microscopy. In some embodiments, the holographic information about individual cells of the population of cells is obtained by segmenting the holographic information obtained for the population of cells.

In some embodiments, holographic information is obtained by repetitive imaging of cells, e.g., individual cells or a sample of cells, removed from a bioreactor containing a cell population (e.g., a cell population in suspension).

In some embodiments, "holographic information" refers to a 2D image of a cell population or individual cell obtained using holographic imaging. In some embodiments, the holographic information comprises multiple 2D images of a cell population or individual cell obtained using holographic imaging. In some embodiments, the holographic information comprises a 2D phase image of a cell population or individual cell. In some embodiments, the holographic information comprises a 2D intensity image of a cell population or individual cell. In some embodiments, the holographic information comprises a 2D phase image and a 2D intensity image of a cell population or individual cell.

In some embodiments, the holographic information itself can be used to train a machine learning model, such as a convolutional neural network, and the holographic information can be provided as input to the machine learning model, such as a convolutional neural network, according to any provided method. In other embodiments, cellular features derived from the holographic information, such as any of the cellular features described in Section I-B, can be used to train a machine learning model and can be provided as input to the machine learning model according to any provided method.

Imaging techniques can utilize a digital device to capture and, for example, record the output (e.g., result) of an imaging process. In some embodiments, the digital device is a charge-coupled device (e.g., a CCD camera). In some embodiments, the digital device is a complementary metal oxide semiconductor device (e.g., a CMOS camera). Therefore, in some embodiments, image data is obtained using a digital device. In some embodiments, the digital device interfaces with a computer to store and/or analyze the result (e.g., output) of the imaging process.

To obtain holographic information associated with T cells in a non-damaging and non-destructive manner, the T cells may be contained in a liquid, such as a culture medium. In some embodiments, the T cells may be suspended in a liquid, such as a culture medium, for imaging. Thus, in some embodiments, imaging techniques can image T cells contained and/or suspended in a liquid.

In some embodiments, the holographic information is obtained by interference microscopy, optical coherence tomography, diffraction phase microscopy, or digital holographic microscopy (DHM). In some embodiments, the holographic information is obtained by imaging a cell population using interference microscopy, optical coherence tomography, diffraction phase microscopy, or DHM.

In some embodiments, the holographic information is obtained by DHM (see, e.g., Carl et al., Applied Optics (2004) 43(36):6536-6544; and Marquet et al., Optics Letters (2005) 30(5):468-470). In some embodiments, the holographic information is obtained by imaging a population of cells, e.g., individual cells in the population, using DHM. In some embodiments, the DHM is traditional DHM, in-line DHM, or differential DHM (DDHM).

DHM is a technology that allows for the recording of 3D samples or objects without the need to scan the sample layer by layer. In this respect, DHM may be superior to confocal microscopy. In DHM, holographic information can be recorded using a digital camera, such as a CCD or CMOS camera, and subsequently stored or processed on a computer. Different DHM techniques, including DDHM, as well as microscope configurations and elements are described in US-7362449, US-9684281-B2, US-10578541-B2, US-2015248109-A1, US-9846151-B2, US-2014193850-A1, US-2016184817-A1, US-11067379-B2, US-2014195568-A1 and US-2021142472-A1.

To create a holographic display, or hologram, a highly coherent light source, such as laser light, can traditionally be used to illuminate a sample. The light from the source can be split into two beams, an object beam and a reference beam. The object beam is transmitted through an optical system to the sample, where it interacts with the sample, thereby changing the phase and amplitude of the light depending on the optical properties and 3D shape of the object. The object beam, reflected or transmitted through the sample, can then interfere with the reference beam (e.g., by a set of mirrors and/or beam splitters), resulting in an interference pattern that can be digitally recorded. Since holograms are more accurate when the object and reference beams have similar amplitudes, an absorbing element can be introduced into the reference beam, which reduces the amplitude to that of the object beam without changing the phase of the reference beam, or at best, changes the phase entirely. The recorded interference pattern can contain information about the phase and amplitude changes that depend on the optical properties and 3D shape of the object.

An alternative method for creating holograms involves the use of in-line holographic technology. In some respects, in-line DHM is similar to traditional DHM, but without beam splitting, or at least without the use of a beam splitter or other external optical elements. In-line DHM can be used to observe particles in a fluid, such as a solution of cells. For example, at least a portion of the coherent light will pass through the sample without interacting with the particles (the reference beam), and the light that has interacted with the particles (the object beam) will interfere to create an interference pattern, which can be digitally recorded and processed. In-line DHM can be used in transmission mode.

Another DHM technique is DDHM. In some aspects, DDHM differs from other techniques in that it uses a reference beam and an object beam. In a preferred DDHM setup, the sample can be illuminated by an illumination means comprising at least partially coherent light, either in reflection or transmission mode. The reflected or transmitted sample beam can be transmitted through an objective lens and subsequently split into two beams by a beam splitter and transmitted along different paths in a differential interferometer (e.g., a Michelson or Mach-Zehnder type). A beam-bending element or tilting means (e.g., a transparent wedge) can be inserted in one of the paths. The two beams can then interfere with each other at the focal plane of a focusing lens, and the interference pattern in this focal plane can be digitally recorded and stored, for example, by a CCD camera connected to a computer. Due to the beam-bending element, the two beams can be slightly shifted in a controlled manner, and the interference pattern can vary depending on the amount of shift. Afterwards, the beam-bending element can be rotated, thereby changing the amount of displacement. A new interference pattern can also be recorded. This can be performed several times (N), and from these N interference patterns, the slope (or spatial derivative) of the phase in the focal plane of the focusing lens can be approximately computed. This is called the phase-stepping method, but other methods for obtaining the phase slope, such as Fourier transform data processing techniques, are also known. The phase slope can be integrated to provide the phase as a function of position. The amplitude of the light as a function of position can be computed from a weighted average (possibly, but not necessarily) of the amplitudes of the N recorded interference patterns. Therefore, since the phase and amplitude are known, the same information can be obtained as with a direct holographic method (using a reference beam and an object beam), and subsequent 3D reconstruction of the object can be performed.

In DDHM, the illumination means may comprise light that is partially coherent both spatially and temporally. In some respects, this contrasts with other DHM methods that may utilize highly correlated laser light. Spatially and temporally coherent light may be generated, for example, by LEDs. LEDs are less expensive than lasers and can produce light with a spectrum centered around a known wavelength, which may be partially coherent both spatially and temporally. While not as coherent as, for example, laser light, it may still be coherent enough to produce holographic images of sufficient quality for the application. In some respects, LEDs may also have the advantage of being available in many different wavelengths, being very compact, and being easy to use and replace as needed. Therefore, in some respects, a method of using spatially and temporally coherent light to obtain holographic images may provide a more cost-effective device for implementing such a method. In some embodiments, the illumination means is a red LED.

Holographic images can be subjected to object segmentation and further analysis to obtain multiple features that quantitatively describe the imaged object (e.g., T cells, cell debris). Thus, various features, such as the cellular features described in Section I-B, can be directly assessed or calculated from DDHM using, for example, the image acquisition, image processing, image segmentation, and feature extraction steps. In some embodiments, a digital recording device is used to record the holographic image. In some embodiments, a computer comprising an algorithm for analyzing the holographic image can be used. In some embodiments, a monitor and/or a computer can be used to display the results of the holographic image analysis. In some embodiments, the analysis is automated (e.g., can be performed without user input).

Any type of DHM may be used according to the provided method. In some embodiments, the DHM is a traditional DHM. In some embodiments, the DHM is an inline DHM. In some embodiments, the DHM is a differential DHM.

An exemplary DHM system for use in the provided methods, such as the Ovizio iLine F (Ovizio Imaging Systems NV/SA, Brussels, Belgium), can be used with a bioreactor for T cell incubation. In some embodiments, the cell population from which holographic information is obtained is derived from a cell culture being cultured in vitro or ex vivo, for example, as described in Section II-D. In some embodiments, the in vitro or ex vivo culture is performed within a bioreactor. In some embodiments, the cell population is removed from the bioreactor for imaging. In some embodiments, the imaging is performed using a digital holographic microscope connected to the bioreactor. The microscope can be any microscope capable of obtaining phase information of a fluid sample and having an illumination means. The microscope can be any microscope used for DHM, including any microscope for traditional DHM, in-line DHM, or DDHM.

In some embodiments, one or more fluidic systems capable of directing fluid from the bioreactor to the microscope are connected to the bioreactor and the microscope. In some embodiments, at least one fluidic system comprises one or more tubes capable of directly contacting the fluid from the bioreactor. In some embodiments, at least one tube comprises a portion that is at least partially transparent to the microscope's illumination means for obtaining holographic information of the fluid sample. In some embodiments, the tube comprises a portion that is at least partially transparent to the microscope's illumination means and comprises a flow cell and/or a microfluidic system. In some embodiments, the flow cell and/or microfluidic system comprises a cross-section that varies in height and/or width along the cross-section. In some aspects, this allows for obtaining clear holographic images of objects of varying concentrations suspended in the fluid. High concentrations of suspended objects can cause a large number of objects to overlap each other and can make obtaining holographic images difficult, especially when the microscope is operating in transmission mode. A low concentration may cause the microscope to obtain only a holographic image of the fluid medium and not an image of an object suspended in the medium. A high concentration may allow the microscope to obtain a holographic image at a location with a small height or width, thereby ensuring that too many objects are not stacked in the illumination beam. A low concentration may allow the microscope to obtain a holographic image at a location with a large height or width, thereby ensuring that at least one suspended object is in the illumination beam. In some embodiments, the microfluidic system comprises a structure in which the tube branches into multiple tubes of different cross-sections, diameters, heights, and/or widths. This arrangement may allow for obtaining clear holographic images of objects of varying concentrations suspended in the fluid. In some embodiments, the cross-section, diameter, height, and/or width of the flow cell and/or microfluidic system are selected as a function of the size of the suspended object and/or the size of the illumination beam of the microscope. In some embodiments, the narrowest dimension in the cross-section of the flow cell and/or microfluidic system is greater than 10 micrometers, more preferably greater than 30 micrometers, even more preferably greater than 50 micrometers, and/or the widest dimension in the cross-section of the flow cell and/or microfluidic system is less than 5000 micrometers, more preferably less than 3000 micrometers, even more preferably less than 2500 micrometers. In some embodiments, the microfluidic system is attached to a substrate, e.g., to facilitate fabrication and/or to provide stability to the microfluidic system.

Although not strictly required, it may be desirable to have fluid flow in at least one fluid system. This may allow for timely sampling of the contents of the bioreactor and monitoring of different samples to obtain a better understanding of the condition and/or reaction of the bioreactor. Fluid flow may be due to natural phenomena such as convection, conduction, or radiation; density or pressure differences induced by reactions or thermal gradients occurring in the bioreactor; gravity; etc. If fluid flow is desirable but does not occur spontaneously or if the flow must be controlled, one or more pumping systems may be connected to the fluid system to induce fluid flow in the system. Thus, in some embodiments, at least one pumping system is connected to one or more fluid systems and may induce fluid flow in the fluid system.

In some embodiments, at least one fluid system comprises a fluid-tight flexible member that, when compressed, stretched, and/or pushed, causes fluid flow within the fluid system. Thus, fluid flow can be induced within the fluid system without a high risk of leakage and without contaminating the flow actuator. In some embodiments, a pumping system is connected to the fluid system and can cause fluid flow within the fluid system by stretching and/or pushing the fluid-tight flexible member.

In some aspects, the contents of the bioreactor can be non-destructively monitored. This allows the cell population observed in the microscope to be reintroduced into the bioreactor. Thus, in some embodiments, at least one fluid system forms a closed circuit between the bioreactor and the microscope, returning the fluid to the bioreactor. For example, the fluid system can direct fluid from the bioreactor to the microscope and back to the bioreactor.

B. Cell characteristics

In some embodiments, the cellular phenotype, e.g., activation state, of a cell population is determined based on one or more input measurements for each of one or more cellular characteristics. In some embodiments, a population-level output measurement is determined based on one or more input measurements for each of one or more cellular characteristics. In some embodiments, one or more cellular characteristics are derived from holographic information. In some embodiments, each input measurement is determined from holographic information obtained for an individual cell of the cell population. In some embodiments, determining the one or more input measurements for each of one or more cellular characteristics involves one or more steps selected from holographic information acquisition, holographic information processing, holographic information segmentation, and cell characteristic extraction. In some embodiments, image segmentation is used to determine holographic information associated with individual cells. Segmentation processes, such as watershed processing, clustering-based image thresholding, and the use of neural networks, are known in the art. Various algorithms exist for watershed processing, including Meyer's flooding algorithm and the optimal spanning forest algorithm (watershed cut). Cluster-based image thresholding can use Otsu's method. In some embodiments, the holographic information includes segmented images.

In some embodiments, the one or more input measurements are multiple input measurements. In some embodiments, the activation state is determined based on multiple input measurements for each of one or more cellular characteristics. In some embodiments, the population-level output measurement is determined based on multiple input measurements for each of one or more cellular characteristics.

In some embodiments, the one or more cell characteristics are multiple cell characteristics. In some embodiments, the activation state is determined based on one or more input measurements for each of the multiple cell characteristics. In some embodiments, the population-level output measurement is determined based on one or more input measurements for each of the multiple cell characteristics.

In some embodiments, the activation state is determined based on multiple input measurements for each of the multiple cell characteristics. In some embodiments, the population-level output measurement is determined based on multiple input measurements for each of the multiple cell characteristics.

In some embodiments, input measurements for some or all of one or more cell characteristics are obtained for individual cells. In some embodiments, input measurements for each of one or more cell characteristics are obtained for individual cells.

In some embodiments, the one or more input measurements for one of the one or more cell characteristics comprise input measurements from one or more individual cells of the cell population. In some embodiments, the one or more input measurements for one of the one or more cell characteristics comprise input measurements from a plurality of individual cells of the cell population. In some embodiments, the input measurements for the first cell characteristic are for a first plurality of individual cells, and the input measurements for the second cell characteristic are for a second distinct plurality of individual cells that may or may not contain some of the first plurality of individual cells. In some embodiments, the input measurements for the first cell characteristic are for the first plurality of individual cells, and the input measurements for the second cell characteristic are also for the first plurality of individual cells. In some embodiments, the input measurements for all cell characteristics are from the same plurality of individual cells.

In some embodiments, the one or more input measurements for one of the one or more cell characteristics include input measurements from some or all individual cells in the cell population. In some embodiments, the one or more input measurements for one of the one or more cell characteristics include input measurements from only live cells. In some embodiments, live cells are classified as live using an automated method. In some embodiments, live cells are classified as live by analysis of holographic images of cells in the cell population, for example, as exemplified herein (e.g., using OsOne software for live cell classification). In some embodiments, the one or more input measurements for one of the one or more cell characteristics include input measurements from only cells of at least a particular size. In some embodiments, the size is determined by a radius-average cell characteristic. In some embodiments, the size is determined by a cell area cell characteristic. In some embodiments, the one or more input measurements for one of the one or more cell characteristics include input measurements from only cells of sufficient circularity (e.g., as determined using the circularity cell characteristic). In some embodiments, the one or more input measurements for one of the one or more cell characteristics comprise input measurements from only cells having input measurements for the intensity smoothness cell characteristic below a particular threshold. In some embodiments, the one or more input measurements for one of the one or more cell characteristics comprise input measurements from only cells that meet all of the above criteria.

In some embodiments, the one or more input measurements for a cell feature of the one or more cell features comprises one or more input measurements from only cells that meet one or more (e.g., one, two, three, four, or all) of the following criteria: (i) being classified as alive, (ii) having a radius mean cell feature measurement > 5, (iii) having an intensity smoothness cell feature measurement < 0.03, (iv) having a cell area cell feature measurement > 60, and (v) having a circularity cell feature measurement > 0.5. In some embodiments, the cells meet all of the above criteria.

In some embodiments, the one or more input measurements for one of the one or more cell characteristics comprises input measurements from each individual cell of the cell population.

In some embodiments, one of the one or more cell features is derived from phase information of the holographic information. In some embodiments, one of the one or more cell features is derived from intensity information of the holographic information. In some embodiments, one of the one or more cell features is derived from phase information and intensity information. In some embodiments, one of the one or more cell features is derived from an image of an individual cell reconstructed from the holographic information. Exemplary methods for determining cell features from holographic information are described in US-9684281, US-20140193850, US-20140195568-A1, US-10578541-B2, US-9904248-B2, US-11067379-B2, and US-2021142472-A1. Exemplary software for determining cell characteristics from holographic information includes Obigeo OsOne (Obigeo Imaging Systems NV/SA, Brussels, Belgium).

In some embodiments, the one or more cellular features comprise one or more morphological features, one or more optical features, one or more intensity texture features, one or more topological texture features, or a combination of any of the foregoing. In some embodiments, the one or more cellular features comprise one or more morphological features, one or more intensity texture features, and one or more topological texture features. Exemplary cellular features and descriptions thereof are set forth in Table E1.

In some embodiments, the one or more cellular features comprise one or more morphological features. In some embodiments, the one or more morphological features describe one or more characteristics of the physical shape of an individual cell. In some embodiments, the one or more morphological features are selected from aspect ratio, cell area, circularity, compactness, elongation, height, diameter, hu moment 1, hu moment 2, hu moment 3, hu moment 4, hu moment 5, hu moment 6, hu moment 7, perimeter, radius mean, radius variance, and normalized radius variance. In some embodiments, the one or more morphological features comprise aspect ratio. In some embodiments, the one or more morphological features comprise cell area. In some embodiments, the one or more morphological features comprise circularity. In some embodiments, the one or more morphological features comprise compactness. In some embodiments, the one or more morphological features comprise elongation. In some embodiments, the one or more morphological features comprise elongation. In some embodiments, the one or more morphological features comprise diameter. In some embodiments, the one or more morphological features comprise a hu moment 1. In some embodiments, the one or more morphological features comprise a hu moment 2. In some embodiments, the one or more morphological features comprise a hu moment 3. In some embodiments, the one or more morphological features comprise a hu moment 4. In some embodiments, the one or more morphological features comprise a hu moment 5. In some embodiments, the one or more morphological features comprise a hu moment 6. In some embodiments, the one or more morphological features comprise a hu moment 7. In some embodiments, the one or more morphological features comprise a perimeter. In some embodiments, the one or more morphological features comprise a radius mean. In some embodiments, the one or more morphological features comprise a radius variance. In some embodiments, the one or more morphological features comprise a normalized radius variance.

In some embodiments, the one or more morphological features include mean cell area and radius.

In some embodiments, the one or more cell features comprise one or more optical features. In some embodiments, the one or more optical features describe one or more optical characteristics of an individual cell image. In some embodiments, the one or more optical features are selected from: refractive peak diameter, intensity maximum, average intensity, intensity minimum, mass eccentricity, optical height maximum (radians), optical height maximum (μm), average optical height (radians), average optical height (μm), normalized optical height, optical height minimum (radians), optical height minimum (μm), optical volume, refractive peak surface, normalized peak area, aggregate size, refractive peak intensity, and normalized peak height. In some embodiments, the one or more optical features comprise a refractive peak diameter. In some embodiments, the one or more optical features comprise an intensity maximum. In some embodiments, the one or more optical features comprise an average intensity. In some embodiments, the one or more optical features comprise an intensity minimum. In some embodiments, the one or more optical features comprise mass eccentricity. In some embodiments, one or more optical features comprise an optical height maximum (in radians). In some embodiments, one or more optical features comprise an optical height maximum (in μm). In some embodiments, one or more optical features comprise an average optical height (in radians). In some embodiments, one or more optical features comprise an average optical height (in μm). In some embodiments, one or more optical features comprise a normalized optical height. In some embodiments, one or more optical features comprise an optical height minimum (in radians). In some embodiments, one or more optical features comprise an optical height minimum (in μm). In some embodiments, one or more optical features comprise an optical volume. In some embodiments, one or more optical features comprise a refractive peak surface. In some embodiments, one or more optical features comprise a normalized peak area. In some embodiments, one or more optical features comprise an aggregate size. In some embodiments, one or more optical features comprise a refractive peak intensity. In some embodiments, one or more optical features comprise a normalized peak height.

In some embodiments, the one or more cellular features comprise one or more intensity texture features. In some embodiments, the one or more intensity texture features describe one or more characteristics of the intensity information of an individual cell. In some embodiments, the one or more intensity texture features are selected from intensity variance, intensity mean contrast, intensity mean entropy, intensity mean, intensity mean uniformity, intensity contrast, intensity correlation, intensity entropy, intensity homogeneity, intensity uniformity, intensity skewness, and intensity smoothness. In some embodiments, the one or more intensity texture features comprise intensity variance. In some embodiments, the one or more intensity texture features comprise intensity mean contrast. In some embodiments, the one or more intensity texture features comprise intensity mean entropy. In some embodiments, the one or more intensity texture features comprise intensity mean. In some embodiments, the one or more intensity texture features comprise intensity mean uniformity. In some embodiments, the one or more intensity texture features comprise intensity contrast. In some embodiments, the one or more intensity texture features comprise intensity correlation. In some embodiments, the one or more intensity texture features comprise intensity entropy. In some embodiments, one or more intensity texture features comprise intensity homogeneity. In some embodiments, one or more intensity texture features comprise intensity uniformity. In some embodiments, one or more intensity texture features comprise intensity skewness. In some embodiments, one or more intensity texture features comprise intensity smoothness.

In some embodiments, the one or more intensity texture features include maximum intensity, minimum intensity, intensity entropy, and intensity contrast.

In some embodiments, the one or more cellular features comprise one or more phase texture features. In some embodiments, the one or more phase texture features describe one or more characteristics of the phase information of an individual cell. In some embodiments, the one or more phase texture features are selected from optical height variance (radians), optical height variance (μm), phase mean contrast, phase mean entropy, phase mean, phase mean uniformity, phase contrast, phase correlation, phase entropy, phase homogeneity, phase skewness, phase smoothness, and phase uniformity. In some embodiments, the one or more phase texture features comprise optical height variance (radians). In some embodiments, the one or more phase texture features comprise optical height variance (μm). In some embodiments, the one or more phase texture features comprise phase mean contrast. In some embodiments, the one or more phase texture features comprise phase mean entropy. In some embodiments, the one or more phase texture features comprise phase mean. In some embodiments, the one or more phase texture features comprise phase mean uniformity. In some embodiments, the one or more phase texture features comprise phase contrast. In some embodiments, the one or more phase texture features comprise phase correlation. In some embodiments, the one or more phase texture features comprise phase entropy. In some embodiments, the one or more phase texture features comprise phase homogeneity. In some embodiments, the one or more phase texture features comprise phase skewness. In some embodiments, the one or more phase texture features comprise phase smoothness. In some embodiments, the one or more phase texture features comprise phase uniformity.

In some embodiments, the cell phenotype is that of an activated T cell.

In some embodiments, the plurality of cell features comprises one or more (e.g., any combination of two, three, four, five, six, seven, eight, nine, or all of) intensity skewness, intensity correlation, intensity homogeneity, intensity maximum, intensity minimum, intensity entropy, intensity contrast, phase entropy, cell area, and radius mean. In some embodiments, the plurality of cell features comprises intensity skewness, intensity correlation, intensity homogeneity, intensity maximum, intensity minimum, intensity entropy, intensity contrast, phase entropy, cell area, and radius mean.

In some embodiments, the plurality of cell characteristics are selected from (or include one or more of) maximum intensity, minimum intensity, intensity entropy, intensity contrast, phase entropy, cell area, and radius average. In some embodiments, the plurality of cell characteristics include maximum intensity, minimum intensity, intensity entropy, intensity contrast, phase entropy, cell area, and radius average.

In some embodiments, the one or more cell features comprise one or more of intensity maxima, intensity minima, intensity entropy, intensity contrast, phase entropy, cell area, and radius mean (e.g., any combination of two, three, four, five, six, or all of these).

In some embodiments, the plurality of cell characteristics comprises one or more of intensity skewness, intensity correlation, and intensity homogeneity (e.g., any combination of two or all of them). In some embodiments, the plurality of cell characteristics comprises intensity skewness, intensity correlation, and intensity homogeneity.

In some embodiments, the cell phenotype is a memory phenotype. In some embodiments, the memory phenotype is an early memory phenotype, such as a stem cell memory phenotype or a central memory phenotype. In some embodiments, the cell phenotype is a central memory phenotype. In some embodiments, the cell phenotype is a stem cell memory phenotype. In some embodiments, the marker is CCR7. In some embodiments, the cell phenotype is a memory phenotype, such as a stem cell memory phenotype or a central memory phenotype, and the marker is CCR7.

Accordingly, in some embodiments, the disclosed method is for determining the memory phenotype of a T cell population, wherein the marker is expressed by T cells having a central memory phenotype or a stem cell memory phenotype. In some embodiments, the marker is expressed by T cells having a central memory phenotype. In some embodiments, the marker is expressed by T cells having a stem cell memory phenotype. In some embodiments, the marker is CCR7.

In some embodiments, the plurality of cell characteristics comprises one or more of cell area, perimeter length, average intensity, normalized peak area, and equivalent peak diameter. In some embodiments, the plurality of cell characteristics comprises cell area. In some embodiments, the plurality of cell characteristics comprises perimeter length. In some embodiments, the plurality of cell characteristics comprises average intensity. In some embodiments, the plurality of cell characteristics comprises normalized peak area. In some embodiments, the plurality of cell characteristics comprises equivalent peak diameter. In some embodiments, the plurality of cell characteristics comprises cell area, perimeter length, average intensity, normalized peak area, and equivalent peak diameter.

In some embodiments, the cell phenotype is recombinant receptor expression, e.g., CAR or TCR expression.

In some embodiments, the plurality of cell features comprises one or more (e.g., any combination of two, three, four, or all) of a peak area, a phase average uniformity, an intensity geometric mean, a minimum optical height, and a normalized optical height. In some embodiments, the plurality of cell features comprises one or more of a peak area, a phase average uniformity, an intensity geometric mean, a minimum optical height, and a normalized optical height. In some embodiments, the plurality of cell features comprises one or more of a larger feature reflecting cell size (area, perimeter, diameter), a smaller average intensity feature, a larger phase correlation feature, and a smaller radial variance normalized feature. In some embodiments, the plurality of cell features comprises one or more of a larger cell area, a larger cell perimeter, a larger cell diameter, a smaller intensity average, a larger phase correlation, and a smaller radial variance (normalized).

In some embodiments, the plurality of cell features are cell features extracted by a machine learning model. In some embodiments, the extraction is performed automatically, for example, without prior design or manipulation of the cell features. In some embodiments, one or more input measurements for at least one of the plurality of cell features are determined by providing holographic information about individual cells in the cell population to the machine learning model. In some embodiments, one or more input measurements for each of the plurality of cell features are determined by providing holographic information about individual cells in the cell population to the machine learning model. In some embodiments, one or more input measurements are determined from the machine learning model.

In some embodiments, the machine learning model is a deep learning model. In some embodiments, the machine learning model is a convolutional neural network. The advantages of convolutional neural networks include the ability to automatically detect and extract features important for classification or regression without human guidance, and to detect features that may not be detectable by humans.

In some embodiments, one or more input measurements for at least one of the plurality of cell characteristics are obtained from a fully connected layer of a convolutional neural network. In some embodiments, one or more input measurements for each of the plurality of cell characteristics are obtained from a fully connected layer of a convolutional neural network.

The architecture of a convolutional neural network suitable for feature extraction can be identified and designed by a person skilled in the art. In some embodiments, the convolutional neural network is a LeNet, AlexNet, ResNet, GoogleNet/InceptionNet, MobileNetV1, ZfNet, Depth-based, Highway Network, Wide ResNet, VGG, PolyNet, Inception v2, Inception v3, Inception v4, Inception-ResNet, DenseNet, Pyramidal Net, Xception, channel-boosting, residual attention neural network, attention-based, feature-map exploitation-based, compression and stimulation, or competitive-compression and stimulation convolutional neural network.

In some embodiments, a machine learning model, e.g., a convolutional neural network, is trained using a dataset of reference holographic information. In some embodiments, for each of the first plurality of reference cell populations containing T cells, the dataset of reference holographic information contains holographic information obtained for the reference cell population, e.g., individual cells of the reference cell population. Exemplary reference cell populations are described in Section I-D-1.

In some embodiments, the machine learning model, e.g., a convolutional neural network, is trained using non-holographic images. In some embodiments, the machine learning model, e.g., a convolutional neural network, is trained using both holographic and non-holographic images.

C. Group-level statistics

In some embodiments, a cell phenotype, e.g., an activation state, is determined based on at least one population-level statistic. In some embodiments, a population-level output measurement is determined based on at least one population-level statistic. In some embodiments, each population-level statistic is for one or more input measurements of one of the one or more cell characteristics. In some embodiments, each population-level statistic describes one or more input measurements of one of the one or more cell characteristics. In some embodiments, each population-level statistic describes a distribution of one or more input measurements of one of the one or more cell characteristics.

In some embodiments, the at least one population-level statistic comprises one or more population-level statistics for each of one or more cell characteristics. In some embodiments, the at least one population-level statistic comprises multiple population-level statistics. In some embodiments, the multiple population-level statistics comprise one or more population-level statistics for each of one or more cell characteristics. In some embodiments, the multiple population-level statistics comprise multiple population-level statistics for each of one or more cell characteristics.

In some embodiments, one or more population-level statistics for the first cell characteristic are different statistics than one or more population-level statistics for the second cell characteristic. For example, in some embodiments, one or more population-level statistics for the first cell characteristic include the mean of one or more input characteristics for the first cell characteristic, and one or more population-level statistics for the second cell characteristic do not include the mean of one or more input characteristics for the second cell characteristic. In some embodiments, one or more population-level statistics for the first cell characteristic are the same statistics as the population-level statistics for the second cell characteristic. In some embodiments, one or more population-level statistics are the same statistics for all cell characteristics.

Exemplary population-level statistics are described in this section and are known in the art. The one or more population-level statistics for each of the one or more cell characteristics can be independently selected from any of these population-level statistics, e.g., any of the exemplary population-level statistics described. In some embodiments, the one or more population-level statistics for each of the one or more cell characteristics are identical and are selected from any of the exemplary population-level statistics described.

In some embodiments, the one or more population-level statistics for one or more cell characteristics comprise a population-level statistic summarizing one or more input measurements of the cell characteristic. In some embodiments, the population-level statistic is a summary statistic.

In some embodiments, the one or more population-level statistics for one of the one or more cell characteristics comprise a measure of central tendency of one or more input measurements of the cell characteristic. In some embodiments, the one or more population-level statistics comprise one or more of the mean, median, and mode of one or more input measurements of the cell characteristic.

In some embodiments, the one or more population-level statistics for one of the one or more cell characteristics comprise a statistical variance measure of one or more input measurements of the cell characteristic. In some embodiments, the one or more population-level statistics comprise one or more of the standard deviation, interquartile range, range, mean absolute difference, median absolute deviation, mean absolute deviation, and distance standard deviation of the one or more input measurements of the cell characteristic.

In some embodiments, the one or more population-level statistics for one of the one or more cell characteristics comprise a measure of the distribution shape of one or more input measurements of the cell characteristic. In some embodiments, the one or more population-level statistics comprise one or both of skewness and kurtosis of the one or more input measurements of the cell characteristic.

In some embodiments, the one or more population-level statistics for one of the one or more cell characteristics comprise one or more quantiles of one or more input measurements of the cell characteristic. In some embodiments, the one or more quantiles are selected from (including one or more of) the 0.01, 0.1, 0.5, 0.9, and 0.99 quantiles of the one or more input measurements of the cell characteristic. In some embodiments, the one or more quantiles comprise the 0.01, 0.1, 0.5, 0.9, and 0.99 quantiles of the one or more input measurements of the cell characteristic. In some embodiments, the one or more population-level statistics for each cell characteristic of the one or more cell characteristics comprise the 0.01, 0.1, 0.5, 0.9, and 0.99 quantiles of the one or more input measurements of the cell characteristic.

In some embodiments, one or more population-level statistics for one of the one or more cell characteristics are determined by applying a pooling filter to one or more input measurements of the cell characteristic. In some embodiments, one or more population-level statistics for each cell characteristic of the one or more cell characteristics are determined by applying a pooling filter to one or more input measurements of the cell characteristic.

In some embodiments, the pooling filter summarizes one or more input measurements of cell characteristics. In some embodiments, the pooling filter is a point estimate-based pooling filter. In some embodiments, the pooling filter is a mean pooling filter. In some embodiments, the pooling filter is a max pooling filter.

In some embodiments, the pooling filter is a distribution-based pooling filter. In some embodiments, the distribution-based pooling filter is based on an estimated marginal distribution of one or more input measurements of cell features, such as described in [Oner et al., arXiv:2006.01561]. In some embodiments, the estimated marginal distribution is computed using a kernel density estimation, such as a Gaussian kernel.

D. Determining group-level output measures

In some embodiments, the cellular phenotype, e.g., activation state, of a cell population is determined based on a population-level output measurement determined for the cell population. In some embodiments, the population-level output measurement is determined based on one or more input measurements for each of one or more cell characteristics. In some embodiments, the population-level output measurement is determined based on at least one population-level statistic determined for the cell population.

In some embodiments, each population-level statistic is compared to a corresponding threshold value. For threshold values for some population-level statistics, a population-level statistic higher than the threshold value may indicate the presence of marker-expressing cells. For threshold values for other population-level statistics, a population-level statistic lower than the threshold value may indicate the presence of marker-expressing cells within the cell population. In some embodiments, the threshold value for one of the one or more cell characteristics is a value exhibited by at least 2%, 4%, 6%, 8%, 10%, 20%, 30%, 40%, 50%, 60%, 70%, 80%, or 90% of a plurality of reference cell populations containing T cells. In some embodiments, each of the plurality of reference cell populations contains marker-expressing cells. In some embodiments, each of the plurality of reference cell populations contains marker-expressing T cells. In some embodiments, at least 2%, 4%, 6%, 8%, 10%, 20%, 30%, 40%, 50%, 60%, 70%, 80%, or 90% of each of the plurality of reference cell populations are marker-expressing cells. In some embodiments, at least 2%, 4%, 6%, 8%, 10%, 20%, 30%, 40%, 50%, 60%, 70%, 80%, or 90% of each of the plurality of reference cell populations are marker-expressing T cells. In some embodiments, at least 2%, 4%, 6%, 8%, 10%, 20%, 30%, 40%, 50%, 60%, 70%, 80%, or 90% of the T cells of each of the plurality of reference cell populations are marker-expressing T cells. An exemplary reference cell population is described in Section I-D-1.

In some embodiments, the population-level output measurement indicates whether at least one population-level statistic indicates the presence of marker-expressing cells in the cell population. In some embodiments, the population-level output measurement indicates whether at least 10%, 20%, 30%, 40%, 50%, 60%, 70%, 80%, or 90% of the population-level statistics indicate the presence of marker-expressing cells in the cell population. In some embodiments, the population-level output measurement indicates whether a majority of the population-level statistics indicate the presence of marker-expressing cells in the cell population. In some embodiments, the population-level output measurement indicates whether each population-level statistic indicates the presence of marker-expressing cells in the cell population.

In some embodiments, each population-level statistic is compared to a corresponding threshold value. For threshold values for some population-level statistics, a population-level statistic higher than the threshold value may indicate the presence of activated T cells. For threshold values for other population-level statistics, a population-level statistic lower than the threshold value may indicate the presence of activated T cells within the cell population. In some embodiments, the threshold value for one of the one or more cell characteristics is a value exhibited by at least 2%, 4%, 6%, 8%, 10%, 20%, 30%, 40%, 50%, 60%, 70%, 80%, or 90% of a plurality of reference cell populations containing T cells. In some embodiments, each of the plurality of reference cell populations contains activated T cells. In some embodiments, each of the plurality of reference cell populations contains T cells expressing a marker indicative of T cell activation. In some embodiments, at least 2%, 4%, 6%, 8%, 10%, 20%, 30%, 40%, 50%, 60%, 70%, 80%, or 90% of each of the plurality of reference cell populations are activated T cells. In some embodiments, at least 2%, 4%, 6%, 8%, 10%, 20%, 30%, 40%, 50%, 60%, 70%, 80%, or 90% of each of the plurality of reference cell populations are T cells that express the marker. In some embodiments, at least 2%, 4%, 6%, 8%, 10%, 20%, 30%, 40%, 50%, 60%, 70%, 80%, or 90% of the T cells of each of the plurality of reference cell populations are T cells that express the marker. An exemplary reference cell population is described in Section I-D-1.

In some embodiments, the population-level output measurement indicates whether at least one population-level statistic indicates the presence of activated T cells in the cell population. In some embodiments, the population-level output measurement indicates whether at least 10%, 20%, 30%, 40%, 50%, 60%, 70%, 80%, or 90% of the population-level statistics indicate the presence of activated T cells in the cell population. In some embodiments, the population-level output measurement indicates whether a majority of the population-level statistics indicate the presence of activated T cells in the cell population. In some embodiments, the population-level output measurement indicates whether each population-level statistic indicates the presence of activated T cells in the cell population.

In some embodiments, population-level output measurements are determined using a trained machine learning model. Exemplary machine learning models and methods for training them are described in Hastie et al., The Elements of Statistical Learning (2016); and Abu-Mostafa et al., Learning from Data (2012). Exemplary machine learning models are also described in Hastie et al., The Elements of Statistical Learning (2016); and Abu-Mostafa et al., Learning from Data (2012).

In some embodiments, the population-level output measurements are output or derived from a trained machine learning model. In some embodiments, one or more input measurements for each of the one or more cellular characteristics are provided as inputs to the trained machine learning model or a process comprising the trained machine learning model. In some embodiments, at least one population-level statistic for each of the one or more cellular characteristics is provided as inputs to the trained machine learning model or a process comprising the trained machine learning model. In some embodiments, as part of providing the input measurements or population-level statistics as inputs to the trained machine learning model, the input measurements or population-level statistic are subjected to one or more preprocessing steps. In some embodiments, the preprocessing step comprises a normalization step. In some embodiments, the preprocessing step comprises a dimensionality reduction step.

In some embodiments, the machine learning model is an unsupervised machine learning model. In some embodiments, the machine learning model is a semi-supervised machine learning model. In some embodiments, the machine learning model is a supervised machine learning model.

In some embodiments, the machine learning model is trained to predict the presence or absence of a marker-expressing cell in a cell population containing T cells, the extent to which the marker-expressing cell is present in the cell population containing T cells, or the extent to which the marker-expressing cell is present in T cells of the cell population containing T cells. In some embodiments, the machine learning model is trained to predict the presence or absence of a marker-expressing T cell in a cell population containing T cells, the extent to which the marker-expressing T cell is present in the cell population containing T cells, or the extent to which the marker-expressing T cell is present in T cells of the cell population containing T cells.

In some embodiments, the machine learning model is trained to predict the presence or absence of an activated T cell within a cell population containing T cells, the extent to which an activated T cell is present in the cell population containing T cells, or the extent to which an activated T cell is present among T cells in the cell population containing T cells. In some embodiments, the machine learning model is trained to predict the presence or absence of a T cell expressing a marker indicative of T cell activation within a cell population containing T cells, the extent to which a T cell expressing a marker indicative of T cell activation is present in the cell population containing T cells, or the extent to which a T cell expressing a marker indicative of T cell activation is present among T cells in the cell population containing T cells.

In some embodiments, the machine learning model is trained to make predictions based on input measurements of one or more cellular features. In some embodiments, the machine learning model is trained to make predictions based on population-level statistics for one or more cellular features. In some embodiments, the machine learning model is trained to make predictions based on one or more cellular features as well as other features, such as features of a cell population that are not based on the characteristics of individual cells.

In some embodiments, the population-level output measurement is determined by providing multiple population-level statistics as inputs to a machine learning model. In some embodiments, the machine learning model is trained to predict the population-level output measurement of marker expression based on the population-level statistics of multiple cell characteristics.

In some embodiments, the machine learning model is a classification model. In some embodiments, the machine learning model is a regression model. For any of the exemplary machine learning models described herein, both classification and regression versions of the machine learning model are disclosed.

In some embodiments, the machine learning model is a linear model. In some embodiments, the machine learning model is a non-linear model. In some embodiments, the machine learning model is a Bayesian model.

In some embodiments, the machine learning model is a regularization model. In some embodiments, the machine learning model is a Lasso-regularization model. In some embodiments, the machine learning model is a Ridge-regularization model. In some embodiments, the machine learning model is an Elastic Net-regularization model.

In some embodiments, the machine learning model is an artificial neural network. In some embodiments, the machine learning model is a support vector machine. In some embodiments, the machine learning model is an ensemble model. In some embodiments, the machine learning model comprises a decision tree. In some embodiments, the machine learning model comprises a boosted decision tree. In some embodiments, the machine learning model is a random forest model. In some embodiments, the machine learning model is an ensemble model comprising any combination of the machine learning models described herein. In some embodiments, a first machine learning model of the ensemble model is trained to predict a population-level output measurement for a first marker (e.g., any described herein), and a second machine learning model of the ensemble model is trained to predict a population-level output measurement for a second marker (e.g., any described herein) that is different from the first marker. In some embodiments, the first machine learning model of the ensemble model is trained to predict a population-level output measurement for a marker (e.g., any described herein), and the second machine learning model of the ensemble model is trained to predict a population-level output measurement for the same marker.

In some embodiments, the machine learning model is a multi-output model. In some embodiments, the multi-output model is a deep learning model (e.g., any of the convolutional neural networks described herein), such as any of the convolutional neural networks described in Section I-B. In some embodiments, the multi-output model is trained to predict population-level output measurements for each of a plurality of different markers, each of which can be independently selected from any of the markers described herein.

1. Reference dataset

In some embodiments, the threshold value is determined based on a dataset of reference input measurements. In some embodiments, the machine learning model is trained using the dataset of reference input measurements. In some embodiments, for each of the first plurality of reference cell populations containing T cells, the dataset of reference input measurements includes one or more reference input measurements for each of one or more cell characteristics.

In some embodiments, the threshold value is determined based on a dataset of reference population-level statistics. In some embodiments, the machine learning model is trained using the dataset of reference population-level statistics. In some embodiments, for each of the first plurality of reference cell populations containing T cells, the dataset of reference population-level statistics includes one or more reference population-level statistics for each of one or more cell characteristics. In some embodiments, each reference population-level statistic is for one or more reference input measurements for one of the plurality of cell characteristics.

In some embodiments, the provided method is for training a machine learning model to predict a cellular phenotype, e.g., activation state, of a T cell population. In some embodiments, the provided method involves training the machine learning model using a dataset of reference input measurements. In some embodiments, for each of the first plurality of reference cell populations containing T cells, the dataset of reference input measurements includes one or more reference input measurements for each of one or more cell characteristics. Exemplary machine learning models are described in Sections I-D.

In some embodiments, the provided method involves training a machine learning model using a dataset of reference population-level statistics. In some embodiments, for each of the first plurality of reference cell populations containing T cells, the dataset of reference population-level statistics includes one or more reference population-level statistics for each of one or more cell characteristics. In some embodiments, each reference population-level statistic is for one or more reference input measurements for one of the one or more cell characteristics. Exemplary machine learning models are described in Sections I-D.

In some embodiments, the threshold value is determined based on a dataset of reference population-level output measurements. In some embodiments, the machine learning model is trained using the dataset of reference population-level output measurements. In some embodiments, the provided methods involve training the machine learning model using the dataset of reference population-level output measurements. In some embodiments, for each of the second plurality of reference cell populations, the dataset of reference population-level output measurements comprises reference population-level output measurements for the reference cell population.

In some embodiments, the machine learning model is a multi-instance learning model. In some embodiments, the machine learning model is a deep multi-instance learning model.

In some embodiments, the provided method comprises: (a) training a convolutional neural network using a dataset of reference holographic information, wherein for each of a first plurality of reference cell populations containing T cells, the dataset of reference holographic information contains holographic information obtained for individual cells of the reference cell population; (b) determining from the convolutional neural network one or more reference input measurements for each cell feature of a plurality of cell features derived from the holographic information, wherein the plurality of cell features are cell features extracted by the convolutional neural network, and each reference input measurement is from an individual cell of the reference cell population; (c) determining a dataset of reference population-level statistics, wherein the dataset of reference population-level statistics contains one or more reference population-level statistics for each of the plurality of cell features; And (d) training a machine learning model using a dataset of reference population-level statistics and a dataset of reference population-level output measurements, wherein for each of the second plurality of reference cell populations, the dataset of reference population-level output measurements contains reference population-level output measurements of expression of a marker for the reference cell population. In some embodiments, the marker is a marker expressed by activated T cells.

In some embodiments, a machine learning model is trained to predict a population-level output measure of marker expression based on population-level statistics of multiple cell characteristics.

In some embodiments, the first and second plurality of reference cell populations are the same. In some embodiments, the first and second plurality of reference cell populations are different. In some embodiments, the dataset of reference population-level statistics and the dataset of reference population-level output measurements are time-matched, e.g., comprise population-level statistics and population-level output measurements measured simultaneously or within 1 hour, 30 minutes, 20 minutes, 10 minutes, or 5 minutes of each other from the same or different reference cell populations.

In some embodiments, for each of the first plurality of reference cell populations, each reference input measurement is derived from holographic information obtained for the reference cell population, e.g., an individual cell of the cell population. In some embodiments, each reference input measurement is from an individual cell of the reference cell population.

In some embodiments, the first and/or second plurality of reference cell populations comprise at least 5, 10, 15, 20, 25, 30, 35, 40, 45, 50, 55, 60, 65, 70, 75, 80, 85, 90, 95, or 100 reference cell populations. In some embodiments, each of the first and/or second plurality of reference cell populations is any cell population described in Section II. In some embodiments, the provided methods involve performing any of the cell processing steps described in Section II on each of the first and/or second plurality of reference cell populations. In some embodiments, each of the first and/or second plurality of reference cell populations is enriched for T cells. In some embodiments, at least 50, 55, 60, 65, 70, 75, 80, 85, 90, 91, 92, 93, 94, 95, 96, 97, 98, 99, or 100% of each of the first and/or second plurality of reference cell populations are T cells.

In some embodiments, the first and/or second plurality of reference cell populations are from a reference cell culture that is being cultured in vitro or ex vivo. In some embodiments, the first and second plurality of reference cell populations comprise reference populations from the same reference cell culture. In some embodiments, the provided method involves culturing the reference cell culture. Exemplary culturing methods and conditions are described in Section II-D. In some embodiments, the in vitro or ex vivo culturing of the reference cell culture is performed under the same or similar conditions as the in vitro or ex vivo culturing of the cell culture.

In some embodiments, the first and/or second plurality of reference cell populations are incubated under T cell stimulating conditions. In some embodiments, the incubation occurs before holographic information is obtained for the first plurality of reference cell populations. In some embodiments, the provided methods involve incubating the first and/or second plurality of reference cell populations under T cell stimulating conditions. Exemplary methods and conditions for stimulating T cells are described in Section II-B. In some embodiments, the incubation of the first and/or second plurality of reference cell populations is performed under the same or similar conditions as the incubation of the cell populations.

In some embodiments, the first and/or second plurality of reference cell populations are genetically engineered to express a recombinant protein. In some embodiments, the provided methods involve genetically engineering the first and/or second plurality of reference cell populations to express a recombinant protein. Exemplary engineering methods are described in Section II-C. In some embodiments, engineering the first and/or second plurality of reference cell populations is performed under the same or similar conditions as engineering the cell populations. In some embodiments, the first and/or second plurality of reference cell populations are engineered to express the same recombinant protein as the cell populations are engineered to express.

In some embodiments, the holographic information for the first plurality of reference cell populations is obtained according to any of the methods described in Section I-A. In some embodiments, the provided methods involve obtaining holographic information for the first plurality of reference cell populations. In some embodiments, the holographic information for the first plurality of reference cell populations is obtained during in vitro or ex vivo cultivation of the reference cell culture. In some embodiments, multiple reference cell populations among the first plurality of reference cell populations are from the same reference cell culture. In some embodiments, the holographic information for the multiple reference cell populations is obtained at multiple time points during in vitro or ex vivo cultivation of the reference cell culture.

In some embodiments, the holographic information for the first plurality of reference cell populations is obtained according to a method used to obtain the holographic information for the cell populations. In some embodiments, the holographic information for the first plurality of reference cell populations is obtained by DHM.

In some embodiments, the reference input measurements are determined according to any of the methods described in Section I-B. In some embodiments, the provided methods involve determining a reference input measurement. In some embodiments, the dataset of reference input measurements or reference population-level statistics comprises only one or more cell characteristics for which input measurements or population-level statistics are obtained for a cell population. In some embodiments, the dataset of reference input measurements or reference population-level statistics comprises other characteristics in addition to the one or more cell characteristics. The other characteristics can be determined from holographic or non-holographic information obtained for the first plurality of reference cell populations. The other characteristics may or may not be based on characteristics of individual cells of the first plurality of reference cell populations.

In some embodiments, the one or more reference population-level statistics for each of the one or more cell features are any of the population-level statistics described in Section I-C. In some embodiments, the provided methods involve determining one or more reference population-level statistics for each of the one or more cell features. In some embodiments, for one of the one or more cell features, the one or more reference population-level statistics are the same as one or more population-level statistics for the cell population. In some embodiments, the one or more reference population-level statistics for one of the one or more cell features comprise one or more quantiles of one or more reference input measurements of the cell feature. In some embodiments, the one or more quantiles are selected from the 0.01, 0.1, 0.5, 0.9, and 0.99 quantiles of one or more reference input measurements of the cell feature. In some embodiments, the one or more quantiles comprise the 0.01, 0.1, 0.5, 0.9, and 0.99 quantiles of one or more reference input measurements of the cell feature. In some embodiments, the one or more reference population-level statistics for each cell characteristic among the one or more cell characteristics comprise the 0.01, 0.1, 0.5, 0.9, and 0.99 quantiles of one or more reference input measurements of the cell characteristic.

In some embodiments, one or more reference population-level statistics for one of the one or more cell characteristics are determined by applying a pooling filter to one or more reference input measurements for the cell characteristic. In some embodiments, each reference population-level statistic is determined by applying a distribution-based pooling filter to one or more reference input measurements for one of the plurality of cell characteristics. In some embodiments, the pooling filter is a point estimate-based pooling filter. In some embodiments, the pooling filter is a mean pooling filter. In some embodiments, the pooling filter is a max pooling filter. In some embodiments, the pooling filter is a distribution-based pooling filter.

In some embodiments, for each of the second plurality of reference cell populations, the reference population-level output measurement represents a cellular phenotype of the reference cell population. In some embodiments, the reference population-level output measurement represents the presence or absence of marker-expressing cells within the reference cell population. In some embodiments, the reference population-level output measurement represents the extent to which marker-expressing cells are present in the reference cell population. In some embodiments, the reference population-level output measurement represents the extent to which marker-expressing cells are present in T cells of the reference cell population.

In some embodiments, the reference population-level output measurement is the number of marker-expressing cells, the percentage of marker-expressing cells, the proportion of marker-expressing cells, or the density of marker-expressing cells within the reference cell population. In some embodiments, the reference population-level output measurement is the number of marker-expressing cells, the percentage of marker-expressing cells, the proportion of marker-expressing cells, or the density of marker-expressing cells within T cells of the reference cell population. In some embodiments, the reference population-level output measurement is the percentage of marker-expression within T cells of the reference cell population.

In some embodiments, for each of the second plurality of reference cell populations, the reference population-level output measurement indicates an activation state of the reference cell population. In some embodiments, the reference population-level output measurement indicates the presence or absence of activated T cells within the reference cell population. In some embodiments, the reference population-level output measurement indicates the extent to which activated T cells are present in the reference cell population. In some embodiments, the reference population-level output measurement indicates the extent to which activated T cells are present in the T cells of the reference cell population.

In some embodiments, the reference population-level output measurement is the number of activated T cells, the percentage of activated T cells, the proportion of activated T cells, or the density of activated T cells within the reference cell population. In some embodiments, the reference population-level output measurement is the number of activated T cells, the percentage of activated T cells, the proportion of activated T cells, or the density of activated T cells within the T cells of the reference cell population. In some embodiments, the reference population-level output measurement is the percentage of activated T cells within the T cells of the reference cell population.

In some embodiments, the reference population-level output measurement is for the expression of a marker indicative of T cell activation by T cells of the reference cell population. In some embodiments, the reference population-level output measurement indicates the presence of T cells expressing the marker within the reference cell population. In some embodiments, the reference population-level output measurement indicates the extent to which T cells expressing the marker are present in the reference cell population. In some embodiments, the reference population-level output measurement indicates the extent to which T cells expressing the marker are present in T cells of the reference cell population.

In some embodiments, the reference population-level output measurement is the number, percentage, proportion, or density of cells within the reference cell population that express the marker. In some embodiments, the reference population-level output measurement is the number, percentage, proportion, or density of cells within the T cells of the reference cell population that express the marker. In some embodiments, the reference population-level output measurement is the percentage of cells within the T cells of the reference cell population that express the marker.

In some embodiments, the dataset of reference population-level output measurements relates to the expression of a marker during or after in vitro or ex vivo culturing of a reference cell culture. In some embodiments, multiple reference cell populations among the second plurality of reference cell populations are from the same reference cell culture. In some embodiments, the reference population-level output measurements for the multiple reference cell populations relate to the expression of a marker at multiple time points during in vitro or ex vivo culturing of the reference cell culture. In some embodiments, for a reference cell population, the reference population-level output measurements relate to the expression of a marker at the same time point as when holographic information for the reference cell population among the first plurality of reference cell populations is obtained.

In some embodiments, the dataset of reference population-level output measurements is determined using fluorescence imaging of cells in a reference cell culture. In some embodiments, the provided method involves determining the dataset of reference population-level output measurements using fluorescence imaging of a second plurality of reference cell populations. In some embodiments, the fluorescence imaging is by flow cytometry.

In some embodiments, one or more preparative and/or non-affinity-based cell separation steps are performed prior to labeling the cells for flow cytometry. In some embodiments, the cells are washed, centrifuged, and/or incubated in the presence of one or more reagents, for example, to remove unwanted components, enrich for desired components, or lyse or remove cells sensitive to a particular reagent. In some embodiments, the cells are separated based on one or more characteristics, such as density, adhesion properties, size, sensitivity, and/or resistance to a particular component. In some embodiments, the method comprises a density-based cell separation method.

In some embodiments, the cells are labeled with one or more fluorescent markers (e.g., one or more fluorophores) that generate a fluorescent signal that can be measured by a flow cytometer. In some embodiments, the cells are labeled by incubating the cells with one or more staining reagents, each of which contains a fluorescent signal or marker (e.g., a fluorophore). The staining reagents can be any reagent for characterizing, selecting, or isolating a specific cell type or cell subtype. In some embodiments, the staining reagents contain immunoaffinity-based reagents, such as antibodies.

In some embodiments, the staining reagent stains cells based on the expression or level of expression of a marker, e.g., a marker indicative of T cell activation. In some embodiments, the staining reagent contains one or more fluorescent markers that can be attached, e.g., by chemical conjugation, to a binding agent capable of binding to the marker, e.g., specifically binding. In some embodiments, the binding agent is a protein. In some embodiments, the binding agent is an antibody or antigen-binding fragment. The fluorescent marker can be conjugated to a binding agent, e.g., an antibody, by any method known in the art.

As is well known in the art, an "antibody" is an immunoglobulin (Ig) molecule capable of specifically binding to a target, such as a carbohydrate, polynucleotide, lipid, or polypeptide, via at least one epitope recognition site located in the variable region of the Ig molecule. As used herein, the term encompasses intact polyclonal or monoclonal antibodies, as well as fragments thereof, such as dAb, Fab, Fab', F(ab') ₂ , Fv), single chain (scFv), synthetic variants thereof, naturally occurring variants, fusion proteins comprising an antibody portion having an antigen-binding fragment of the desired specificity, chimeric antibodies, nanobodies, and any other modified conformation of an immunoglobulin molecule comprising an antigen-binding site or fragment (epitope recognition site) of the desired specificity. Minibodies comprising a scFv linked to a CH3 domain are also encompassed herein.

A binding agent, such as an antibody, that "specifically binds" or "preferentially binds" (used interchangeably herein) to a marker is a term well understood in the art. A molecule is said to exhibit "specific binding" or "preferential binding" if it reacts or associates with a particular marker more frequently, more rapidly, with longer duration, and/or with greater affinity than with alternative markers. An antibody binds specifically or preferentially to a target if it binds with greater affinity, avidity, more readily, and/or with longer duration than it binds to other substances. It is also understood that specific or preferential binding does not necessarily require exclusive binding (although it can include exclusive binding). Methods for determining such specific or preferential binding, such as immunoassays, are also well known in the art.

Antibodies for flow cytometry can be selected based on markers detected in multiple reference cell populations, such as markers indicative of T cell activation. In some embodiments, the marker is CD137 (4-1BB). Exemplary CD137-binding agents for flow cytometry, such as anti-CD137 antibodies, include clone 4B4-1 available from Miltenyi Biotec and clone 17B5 available from ThermoFisher Scientific.

In some embodiments, the marker is CD3. In some embodiments, the marker is CD4. In some embodiments, the marker is CD8. Exemplary CD3-, CD4-, and CD8-binding agents, such as antibodies, are described in Section II-A.

In some embodiments, the binding agent, such as an antibody, is conjugated to a fluorescent marker, such as a fluorophore. For example, cells can be incubated with one or more fluorescently labeled antibodies. In some embodiments, any fluorescent marker or fluorophore suitable for use in flow cytometry analysis can be used. Exemplary fluorescent markers include fluorescent proteins (e.g., GFP, YFP, RFP), fluorescent moieties (e.g., fluorescein isothiocyanate (FITC), phycoerythrin (PE), allophycocyanin (APC), and Alexa Fluor (AF)), nucleic acid stains (e.g., 4',6-diamidino-2-phenylindole (DAPI), SYT016, and propidium iodide (PI)), cell membrane stains (e.g., FMI-43), cell functional dyes (e.g., Fluo-4 and Indo-1), and synthetic dyes (e.g., brilliant violet (BV)). Exemplary fluorophores include hydroxycoumarin, Cascade Blue, Dylight 405 Pacific Orange, Alexa Fluor 430, fluorescein, Oregon Green, Alexa Fluor 488, BODIPY 493, 2,7-dichlorofluorescein, ATTO 488, Chromeo 488, Dylight 488, HiLyte 488, Alexa Fluor 532, Alexa Fluor 555, ATTO 550, BODIPY TMR-X, CF 555, Chromeo 546, Cy3, TMR, TRITC, Dy547, Dy548, Dy549, HiLyte 555, DyLyte 550, BODIPY 564, Alexa Fluor 568, Alexa Fluor 594, Rhodamine, Texas Red, Alexa Fluor 610, Alexa Fluor 633, Dylite 633, Alexa Fluor 647, APC, ATTO 655, CF633, CF640R, Chromeo 642, Cy5, Dylite 650, Alexa Fluor 680, IRDye 680, Alexa Fluor 700 (AF700), Cy5.5, ICG, Alexa Fluor 750, Dylite 755, IRDye 750, Cy7, PE-Cy7, Cy7.5, Alexa Fluor 790, Dylite 800, IRDye 800, BV421, BV510, BV570, BV605, BV650, Includes BV711, BV750, BV785, Qdot® 525, Qdot® 565, Qdot® 605, Qdot® 655, Qdot® 705 and Qdot® 800.

In some embodiments, cell staining for flow cytometry involves incubation with a staining reagent, followed in some embodiments by a washing step and a step of separating cells bound to the staining reagent from cells not bound to the staining reagent. In some embodiments, a volume of cells is mixed with an amount of the desired staining reagent and incubated under cell staining conditions. In some embodiments, the staining is performed at a temperature between 0°C and 25°C, such as at or about 4°C. In some embodiments, the staining is performed for more than 5 minutes, typically more than 15 minutes. In some embodiments, the staining is performed for between 15 minutes and 6 hours, such as between 30 minutes and 2 hours. In some embodiments, the staining is performed for, for example, 15 minutes, 30 minutes, 1 hour, 1.5 hours, 2 hours, 2.5 hours, 3 hours, or about the foregoing, or any time period between any of the foregoing. In some embodiments, one or more washing steps are performed prior to introducing the cells into a flow cytometer for analysis. In some embodiments, the stained cells are introduced into a flow cytometer.

In some embodiments, the cells are prepared by suspending single cells at a density of between 1 x 10 ⁶ and 1 x 10 ⁷ cells/mL to allow the cells to pass through a flow cytometer for reading. In some embodiments, this cell concentration is called a fluid sheath. In some embodiments, the fluid sheath affects the flow sorting rate, which is typically about 2,000-20,000 cells per second. The fluid sheath of the cell sample can be made of a phosphate buffered saline solution, although other solutions are available as known and understood by those of ordinary skill in the art.

II. T cell populations

Exemplary cell populations containing T cells and cell processing steps involving the cell populations are described in this section. In some embodiments, each of the reference cell populations mentioned in Section I-D-1 is individually selected from any of the described cell populations containing T cells. In some embodiments, the provided methods involve performing any of the described cell processing steps for each of the plurality of reference cell populations. In some embodiments, the same or similar cell processing steps are performed for each of the plurality of reference cell populations. In some embodiments, the same or similar cell processing steps are performed for each of the plurality of reference cell populations and the cell populations.

In some embodiments, at least 2%, 4%, 6%, 8%, 10%, 20%, 30%, 40%, 50%, 60%, 70%, 80%, or 90% of the cell population are T cells. In some embodiments, the cell population is enriched for T cells. In some embodiments, at least 50, 55, 60, 65, 70, 75, 80, 85, 90, 91, 92, 93, 94, 95, 96, 97, 98, 99, or 100% of the cell population are T cells. Exemplary methods for selecting T cells from a mixed cell population, for example, are described in Section I-A.

In some embodiments, the cell population is incubated under T cell stimulating conditions. In some embodiments, the provided methods involve incubating the cell population under T cell stimulating conditions. Exemplary T cell stimulating conditions are described in Section II-B.

In some embodiments, the cell population is genetically engineered to express a recombinant protein. In some embodiments, the provided methods involve genetically engineering a cell population to express a recombinant protein. Exemplary manipulation methods are described in Section II-C.

In some embodiments, the cell population is from a cell culture that is being cultured in vitro or ex vivo. In some embodiments, the provided method involves culturing the cell culture. Exemplary conditions for in vitro or ex vivo culture are described in Section II-D.

In some embodiments, holographic information about individual cells in a cell population is obtained during in vitro or ex vivo culture of the cell culture. In some embodiments, the cellular phenotype, e.g., activation state, of the cell population is determined during the in vitro or ex vivo culture. In some embodiments, the population-level output measurement represents the cellular phenotype, e.g., activation state, of the cell population during the in vitro or ex vivo culture. In some embodiments, the expression of a marker, e.g., a marker indicative of T cell activation, by the cell population is determined during the in vitro or ex vivo culture. In some embodiments, the population-level output measurement represents the expression of a marker during the in vitro or ex vivo culture of the cell culture.

In some embodiments, the cell culture is incubated under T cell stimulating conditions. In some embodiments, the provided method involves incubating the cell culture under T cell stimulating conditions. Exemplary T cell stimulating conditions are described in Section II-B.

In some embodiments, incubation under T cell stimulating conditions occurs before holographic information is obtained for individual cells in the cell population. In some embodiments, incubation under T cell stimulating conditions occurs after holographic information is obtained for individual cells in the cell population. In some embodiments, incubation under T cell stimulating conditions occurs before and after holographic information is obtained for individual cells in the cell population.

Sections II-A through II-C each describe exemplary methods and reagents for selecting, stimulating, and culturing a cell population containing T cells. In some embodiments, the provided methods involve one or more of the described steps of selecting, stimulating, and culturing T cells. In some embodiments, the provided methods involve stimulating T cells. In some embodiments, the provided methods involve culturing T cells. In some embodiments, the provided methods involve stimulating and culturing T cells.

In some embodiments, the step of selecting T cells is performed prior to the step of stimulating or culturing the T cells. In some embodiments, the step of selecting T cells is performed after the step of stimulating or culturing the T cells.

In some embodiments, incubation under T cell stimulating conditions occurs prior to in vitro or ex vivo culture. In some embodiments, at least a portion of the in vitro or ex vivo culture occurs under T cell stimulating conditions. In some embodiments, all of the in vitro or ex vivo culture occurs under T cell stimulating conditions. In some embodiments, incubation under T cell stimulating conditions occurs after in vitro or ex vivo culture.

In some embodiments, some or all of the steps of selecting, stimulating, and culturing T cells are performed at a temperature greater than room temperature, e.g., physiological temperature. In some embodiments, the temperature is about 35°C to about 39°C, such as about 37°C or above.

In some embodiments, any number of the steps of the provided method are performed in a closed system. In some embodiments, any number of the steps of the provided method are automated.

A. Select

In some embodiments, the cell population is from a biological sample. In some embodiments, the cell population is primary cells. In some embodiments, the primary cells are primary cells from a human subject. Biological samples can include tissues, fluids, and other samples taken directly from a subject. Biological samples can be samples obtained directly from a biological source or processed samples. Exemplary biological samples include bodily fluids such as blood, plasma, serum, cerebrospinal fluid, synovial fluid, urine, and sweat; tissues; and organ samples (including processed samples derived therefrom). Exemplary biological samples also include whole blood, peripheral blood mononuclear cells (PBMCs), white blood cells, bone marrow, thymus, tissue biopsies, tumors, leukemias, lymphomas, lymph nodes, gut-associated lymphoid tissue, mucosa-associated lymphoid tissue, spleen, other lymphoid tissue, liver, lung, stomach, intestine, colon, kidney, pancreas, breast, bone, prostate, cervix, testis, ovary, tonsil, or other organs, or cells derived therefrom.

In some instances, cells from the subject's circulating blood are obtained, for example, by apheresis or leukapheresis. The biological sample may contain lymphocytes, including T cells, monocytes, granulocytes, B cells, other nucleated white blood cells, red blood cells, and/or platelets, and in some aspects, cells other than red blood cells and platelets.

In some embodiments, the biological sample is a sample containing T cells. In some embodiments, the biological sample is a whole blood sample, a buffy coat sample, a peripheral blood mononuclear cell (PBMC) sample, an unfractionated T cell sample, a lymphocyte sample, a leukocyte sample, an apheresis product, or a leukapheresis product. In some embodiments, the biological sample is an apheresis product. In some embodiments, the biological sample is a leukapheresis product.

In some embodiments, cells obtained from the subject are washed, for example, to remove the plasma fraction, and the cells are placed in an appropriate buffer or medium for subsequent processing steps. In some embodiments, the cells are washed with phosphate buffered saline (PBS). In some embodiments, the wash solution is free of calcium, magnesium, and/or many or all divalent cations. In some aspects, the washing step is accomplished using a semi-automated "flow-through" centrifuge (e.g., Cobe 2991 Cell Processor, Baxter) according to the manufacturer's instructions. In some aspects, the washing step is accomplished by tangential flow filtration (TFF) according to the manufacturer's instructions. In some embodiments, the cells are resuspended in various biocompatible buffers, such as Ca ²⁺ /Mg ²⁺ -free PBS, after washing. In some embodiments, components of the blood cell biological sample are removed, and the cells are resuspended directly in culture medium. In some embodiments, the biological sample (e.g., apheresis product or leukapheresis product) is washed to remove one or more anticoagulants, such as heparin, added during apheresis or leukapheresis.

In some embodiments, the selection of cells comprises one or more preparation and/or non-affinity-based cell separation steps. In some instances, the cells are washed, centrifuged, and/or incubated in the presence of one or more reagents, for example, to remove unwanted components, enrich for desired components, or lyse or remove cells sensitive to specific reagents. In some instances, the cells are separated based on one or more characteristics, such as density, adhesive properties, size, sensitivity, and/or resistance to specific components. In some embodiments, the method involves preparing leukocytes from peripheral blood using a density-based cell separation method, such as lysing red blood cells and centrifuging through a Percoll or Ficoll gradient.

In some embodiments, the cryopreserved and/or cryoprotected apheresis product or leukapheresis product is thawed. In some embodiments, the thawed cell composition is subjected to dilution (e.g., serum-free medium) and/or washing (e.g., serum-free medium), which may, in some cases, remove or reduce unwanted or undesirable components. In some cases, dilution and/or washing removes or reduces the presence of cryoprotectants, such as DMSO, contained in the thawed sample, which may otherwise negatively affect cell viability, yield, or recovery upon prolonged room temperature exposure. In some embodiments, dilution and/or washing allows for media exchange of the thawed cryopreserved product with serum-free media, such as media described in US-20210207080.

Exemplary methods and reagents for the selection (including positive or negative selection) of T cells to produce a cell population containing T cells are described in US-11400115, US-20190112576, US-10228312, US-20200354677, US-20200384025, US-20210163893, US-20220002669, US-5985658, US-9023604, US-20030175850, US-20040082012, US-20080255004, US-20130059288, and US-9678061.

In some embodiments, at least part of the selection step comprises incubating the cells with a selection reagent. In some embodiments, the incubation is performed with the selection reagent or reagents as part of a selection method that may be performed using one or more selection reagents for the selection of one or more different cell types based on the expression or presence within or on the cells of one or more specific molecules, such as surface markers, e.g., surface proteins, intracellular markers, or nucleic acids. In some embodiments, any known method using a selection reagent or reagents for separation based on such markers may be used. In some embodiments, the selection reagent or reagents result in a separation that is an affinity- or immunoaffinity-based separation. For example, in some aspects, selection involves incubating cells or cell populations based on the expression or level of expression of one or more markers, typically cell surface markers, with a reagent or reagents for separation of cells and cell populations, for example, with a binding partner, e.g., an antibody, that specifically binds to such markers, typically followed by a washing step and a step of separating cells bound to the binding partner from cells not bound to the binding partner.

In some aspects of this process, a volume of cells is mixed with a quantity of a desired affinity-based selection reagent. Immunoaffinity-based selection can be performed using any system or method that results in a favorable energetic interaction between the cells to be isolated and a marker on the cells, e.g., a molecule that specifically binds to a binding partner on a solid surface, e.g., a particle. In some embodiments, the method is performed using particles, such as beads, e.g., magnetic beads, coated with a selection agent (e.g., an antibody) specific for a marker on the cells. The particles (e.g., beads) can be incubated or mixed with the cells in a vessel, such as a tube or bag, while shaking or mixing at a constant cell density-to-particle (e.g., bead) ratio to help promote an energetically favorable interaction.

In some embodiments, the total duration of incubation with the selection reagent is from about 5 minutes to about 6 hours, such as from about 30 minutes to about 3 hours, for example at least or at least about 30 minutes, 60 minutes, 120 minutes or 180 minutes.

In some embodiments, after incubation and/or mixing of the cells and the selection reagent and/or reagents, the incubated cells are subjected to separation for cell selection based on the presence or absence of a particular selection reagent or reagents. In some embodiments, after incubation with the selection reagent, the incubated cells, including the cells to which the selection reagent is bound, are transferred to a system for immunoaffinity-based separation of cells. In some embodiments, the system for immunoaffinity-based separation is or contains a magnetic separation column.

This separation step may be based on positive selection, which retains cells bound to a selection reagent, e.g., an antibody, for further use, and/or negative selection, which retains cells not bound to a selection reagent, e.g., an antibody. In some instances, both fractions are retained for further use. In some embodiments, the process steps further comprise negative and/or positive selection of the incubated cells, such as using a system or device capable of performing affinity-based selection. In some embodiments, the isolation is performed by enrichment for a specific cell population by positive selection, or depletion of a specific cell population by negative selection. In some embodiments, the positive or negative selection is accomplished by incubating the cells with one or more antibodies or other binding agents that specifically bind to one or more surface markers expressed at a relatively higher level (marker ^high ) or expressed (marker+ ) on the positively or negatively selected cells, respectively.

Separation need not result in 100% enrichment or elimination of a specific cell population or cells expressing a specific selectable marker. For example, positive selection or enrichment of a specific type of cell, such as cells expressing a selectable marker, refers to an increase in the number or percentage of such cells, but does not necessarily result in the complete absence of cells that do not express the selectable marker. Similarly, negative selection, elimination, or depletion of a specific type of cell, such as cells expressing a selectable marker, refers to a decrease in the number or percentage of such cells, but does not necessarily result in the complete elimination of all such cells.

In some instances, multiple rounds of separation steps are performed, wherein the positively or negatively selected fraction from one step is subjected to another separation step, such as a subsequent positive or negative selection. In some instances, a single separation step can simultaneously deplete cells expressing multiple markers, such as by incubating cells with multiple antibodies or other binding partners, each specific for the marker targeted for negative selection. Similarly, multiple cell types can be positively selected simultaneously by incubating cells with multiple antibodies or other binding partners expressed on different cell types. In certain embodiments, the separation steps are repeated and/or performed more than once, wherein the positively or negatively selected fraction from one step is subjected to the same separation step, such as repeated positive or negative selection. In some instances, a single separation step is repeated and/or performed more than once, for example, to increase the purity of the selected cells and/or to further remove and/or deplete negatively selected cells from the negatively selected fraction. In certain embodiments, one or more of the separation steps are performed 2, 3, 4, 5, 6, 7, 8, 9, 10, or more than 10 times. In certain embodiments, one or more of the selection steps are performed and/or repeated 1 to 10 times, 1 to 5 times, or 3 to 5 times.

In some embodiments, T cells are isolated from a PBMC, apheresis, or leukapheresis sample by negative selection for a marker expressed on non-T cells, such as B cells, monocytes, or other leukocytes, such as CD14. In some aspects, a CD3+ selection step is used to generate a population enriched for CD3+ T cells from a starting sample, such as a PBMC, apheresis, or leukapheresis sample, wherein the starting sample has not been subjected to positive or negative selection based on another marker, such as CD4 and/or CD8. In some embodiments, the starting sample is selected for CD3 without any prior, concurrent, or subsequent selection for another marker (e.g., CD4 and/or CD8) to generate a population enriched for CD3+ T cells. In some embodiments, the cell population is enriched for CD3+ T cells, and the population is not subjected to further selection, such as CD4+ and/or CD8+ selection, prior to the step of stimulating the cell population. In certain embodiments, the CD3+ T cell enriched population has a ratio of CD4+ T cells to CD8+ T cells of 1:10 to 10:1, 1:5 to 5:1, 4:1 to 1:4, 1:3 to 3:1, 2:1 to 1:2, 1.5:1 to 1:1.5, 1.25:1 to 1:1.25, 1.2:1 to 1:1.2, 1.1:1 to 1:1.1, or about 1:1 or 1:1.

In some embodiments, T cells are isolated from PBMCs, apheresis, or leukapheresis samples by negative selection for markers expressed on non-T cells, such as B cells, monocytes, or other leukocytes, such as CD14. In some aspects, a CD4+ or CD8+ selection step is used to isolate CD4+ helper and CD8+ cytotoxic T cells. These CD4+ and CD8+ populations can be further divided into subpopulations by positive or negative selection for markers expressed on, or relatively higher in, one or more naïve-like, memory, and/or effector T cell subpopulations.

In certain embodiments, a biological sample, such as a PBMC, apheresis, or leukapheresis sample, is subjected to selection of CD4+ T cells, wherein both negative and positive fractions are retained. In certain embodiments, CD8+ T cells are selected from the negative fraction. In some embodiments, a biological sample is subjected to selection of CD8+ T cells, wherein both negative and positive fractions are retained. In certain embodiments, CD4+ T cells are selected from the negative fraction.

In some aspects, a CD8-based positive selection step is used to generate a CD8+ T cell enriched population from a starting sample, such as a PBMC, apheresis, or leukapheresis sample, wherein the starting sample has not been subjected to selection based on another marker, such as CD3+ and/or CD4+. In some embodiments, both the negative and positive fractions from the CD8 positive selection step are retained, and the CD8-negative fraction is further subjected to a CD4-based positive selection step to generate a CD4+ T cell enriched population. In some embodiments, cells from the CD8+ T cell enriched population and cells from the CD4+ T cell enriched population are mixed, combined, and/or pooled to generate a population containing CD4+ T cells and CD8+ T cells.

In some aspects, a CD4-based positive selection step is used to generate a CD4+ T cell enriched population from a starting sample, such as a PBMC, apheresis, or leukapheresis sample, wherein the starting sample has not been subjected to selection based on another marker, such as CD3+ and/or CD8+. In some embodiments, both negative and positive fractions from the CD4 positive selection step are retained, and the CD4-negative fraction is further subjected to a CD8-based positive selection step used to generate a CD8+ T cell enriched population. In some embodiments, cells from the CD4+ T cell enriched population and cells from the CD8+ T cell enriched population are mixed, combined, and/or pooled to generate a population containing CD8+ T cells and CD4+ T cells.

In certain embodiments, the CD4+ T cell enriched population and the CD8+ T cell enriched population are pooled, mixed, and/or combined prior to stimulating the cells, e.g., prior to culturing the cells under stimulating conditions as described in Section I-B. In certain embodiments, the pooled, mixed, and/or combined cells or populations have a ratio of CD4+ T cells to CD8+ T cells of from 1:10 to 10:1, from 1:5 to 5:1, from 4:1 to 1:4, from 1:3 to 3:1, from 2:1 to 1:2, from 1.5:1 to 1:1.5, from 1.25:1 to 1:1.25, from 1.2:1 to 1:1.2, from 1.1:1 to 1:1.1, or about 1:1, or 1:1. In certain embodiments, the cells or populations are pooled, mixed, and/or combined such that the pooled, mixed, and/or combined cell composition has a ratio of CD4+ T cells to CD8+ T cells of 1:1 or about such ratio.

In some embodiments, a selection agent that specifically binds CD4 and a selection agent that specifically binds CD8 are used to generate a population enriched for CD4+ T cells and a population enriched for CD8+ T cells, respectively. In some embodiments, the CD4-specific selection agent and the CD8-specific selection agent have the same or substantially the same capacity, for example, a unit volume or unit weight of the selection agent (e.g., ClinicMACS CD4 selection reagent and CD8 selection reagent) can be used to select CD4 or CD8 cells from the same number of total cells. In some embodiments, a greater amount of CD4-specific selection agent can be used than a CD8-specific selection agent with the same or substantially the same capacity. For example, the volume or weight ratio of the CD4-specific selection agent and the CD8-specific selection agent can be about 5:1, 4:1, 3:1, 2:1, or 1:5:1.

In some aspects, the incubated sample or cell population to be separated is incubated with a selection reagent containing small, magnetizable or magnetically responsive substances, such as magnetically responsive particles or microparticles, such as paramagnetic beads (e.g., Dynabeads or MACS® beads). The magnetically responsive substances, e.g., particles, are typically attached directly or indirectly to a binding partner, e.g., an antibody, that specifically binds to a molecule, e.g., a surface marker, present on the cell population from which separation, e.g., negative or positive selection, is desired.

In some embodiments, the magnetic particles, e.g., beads, contain a magnetically responsive material bound to a specific binding member, such as an antibody or other binding partner. Many well-known magnetically responsive materials for use in magnetic separation methods are known, such as those described in US-4452773 and EP-452342. Colloidal-sized particles, such as those described in US-4795698 and US-5200084, may also be used.

Incubation can be performed under conditions in which the antibody or other binding partner, such as a secondary antibody or other reagent, that specifically binds to the magnetic particles, e.g., beads, binds specifically to cell surface molecules present on cells in the sample.

In certain embodiments, the magnetically reactive particles are coated with a primary antibody or other binding partner, a secondary antibody, a lectin, an enzyme, or streptavidin. In certain embodiments, the magnetic particles attach to cells through coating with a primary antibody specific for one or more markers. In certain embodiments, cells, rather than beads, are labeled with the primary antibody or binding partner, followed by addition of cell-type-specific secondary antibodies or other binding partners (e.g., streptavidin)-coated magnetic particles. In certain embodiments, the streptavidin-coated magnetic particles are used in conjunction with biotinylated primary or secondary antibodies.

In some aspects, separation is achieved by placing the sample in a magnetic field, such that cells with magnetically responsive or magnetizable particles attached to them are attracted to the magnet and separated from unlabeled cells. In positive selection, cells attracted to the magnet are retained; in negative selection, cells not attracted (unlabeled cells) are retained. In some aspects, a combination of positive and negative selections is performed during the same selection step, wherein the positive and negative fractions are retained and further processed or subjected to additional separation steps.

In some embodiments, affinity-based selection is achieved via magnetic-activated cell sorting (MACS) (Miltenyi Biotec, Auburn, CA). Magnetic-activated cell sorting (MACS), such as the CliniMACS system, can select cells with high purity that adhere to magnetized particles. In certain embodiments, MACS operates by sequentially eluting non-target and target species after the application of an external magnetic field. That is, cells adhered to magnetized particles are fixed in place, while non-adherent species are eluted. After this first elution step is completed, species that were trapped in the magnetic field and thus prevented from being eluted are released in some way for elution and recovery. In certain embodiments, non-target cells are labeled and depleted from the heterogeneous cell population.

In some embodiments, the separation and/or isolation step is performed using magnetic beads to which an immunoaffinity reagent is reversibly bound, such as through the interaction of a streptavidin mutein with a peptide ligand, as described in US-20170037369. An example of such magnetic beads is Streptamers®. In some embodiments, the separation and/or isolation step is performed using magnetic beads, such as those commercially available from Miltenyi Biotec.

In some embodiments, T cells are isolated, selected, or enriched by chromatographic isolation, such as column chromatography, including affinity chromatography or gel permeation chromatography. Such methods may be described as (trace-free) cell affinity chromatography technology (CATCH) and may include any of the methods or techniques described in US-10228312 and US-20170037369. Generally, the chromatographic method is fluid chromatography, typically liquid chromatography. In some aspects, the chromatography may be performed in flow-through mode, where a fluid sample containing cells is applied to one end of a column containing a chromatography matrix, for example, by gravity flow or a pump, and the fluid sample is present at the other end of the column. Additionally, chromatography can be performed in an "up-and-down" mode, where a fluid sample containing cells to be isolated is applied by a pipette to one end of a column containing a chromatography matrix packed within, for example, a pipette tip, and the fluid sample is introduced into and removed from the chromatography matrix/pipette tip at the other end of the column. Alternatively, chromatography can also be performed in a batch mode, where a chromatography material (e.g., a stationary phase) is incubated with a sample containing cells, for example, under shaking, rotation, or repeated contact and removal of the fluid sample (e.g., using a pipette).

In some embodiments, the selection agent is contained in the chromatography column (e.g., directly or indirectly bound to the chromatography matrix (e.g., stationary phase)). In some embodiments, the selection agent is present on the chromatography matrix (e.g., stationary phase) when the sample is added to the column. In some embodiments, the selection agent can be indirectly bound to the chromatography matrix (e.g., stationary phase) via a reagent, e.g., a selection reagent. In some embodiments, the selection reagent is covalently or non-covalently bound to the stationary phase of the column. In some embodiments, the selection reagent is reversibly immobilized on the chromatography matrix (e.g., stationary phase). In some cases, the selection reagent is immobilized on the chromatography matrix (e.g., stationary phase) via a covalent bond. In some aspects, the selection reagent is reversibly immobilized on the chromatography matrix (e.g., stationary phase) non-covalently.

In some embodiments, the selection agent is present on the chromatography matrix (e.g., the stationary phase) when the sample is added to the chromatography column (e.g., the stationary phase), and may be bound, for example, directly (e.g., covalently or non-covalently) or indirectly via a selection reagent. Thus, upon addition of the sample, T cells may be bound by the selection agent and immobilized on the chromatography matrix (e.g., the stationary phase) of the column. Alternatively, in some embodiments, the selection agent may be added to the sample. In this manner, the selection agent binds to T cells in the sample, and the sample may then be added to a chromatography matrix (e.g., the stationary phase) containing the selection reagent, wherein the selection agent, which is already bound to the T cells, binds to the selection reagent, thereby immobilizing the target cells on the chromatography matrix (e.g., the stationary phase).

In some embodiments, one or both of the first and/or second selections can utilize a plurality of affinity chromatography matrices and/or antibodies, whereby the plurality of matrices and/or antibodies are connected in series. In some embodiments, the affinity chromatography matrix or matrices used for selection can adsorb, select, or enrich at least about 50 x 10 ⁶ cells/mL, 100 x 10 ⁶ cells/mL, 200 x 10 ⁶ cells/mL, or 400 x 10 ⁶ cells/mL. In some embodiments, the adsorption capacity can be adjusted based on the diameter and/or length of the matrix. In some embodiments, the culture-initiation ratio of the selected or enriched population is achieved by selecting a sufficient amount of matrix and/or a sufficient relative amount to achieve the culture-initiation ratio, which assumes, for example, the adsorption capacity of the matrix or matrices for cell selection.

In some aspects, the chromatography matrix/stationary phase is a non-magnetic or non-magnetizable material. Such materials may include derivatized silica or a cross-linked gel. Cross-linked gels (typically prepared in bead form) may be based on natural polymers, such as cross-linked polysaccharides. Suitable examples include agarose gels or cross-linked dextran gels. Cross-linked gels may also be based on synthetic polymers, such as polymer classes that do not occur in nature. Typically, such synthetic polymers that form the basis of stationary phases for cell separation are polymers that contain polar monomer units and are therefore inherently polar.

Illustrative examples of suitable synthetic polymers include polyacrylamides, styrene-divinylbenzene gels, and copolymers of acrylates and diols or acrylamides and diols. An illustrative example is a polymethacrylate gel, commercially available as Fractogel®. A further example is a copolymer of ethylene glycol and methacrylate, commercially available as Toyopearl®. In some embodiments, the stationary phase may also include a composite matrix or composites or copolymers of natural and synthetic polymer components, such as polysaccharides and agarose (e.g., polyacrylamide/agarose composites, or polysaccharides and N,N'-methylenebisacrylamide). Illustrative examples of copolymers of dextran and N,N'-methylenebisacrylamide are the aforementioned Sephacryl® series of materials. Derivatized silica may include silica particles coupled to synthetic or natural polymers. Examples of such embodiments include polysaccharide grafted silica, polyvinylpyrrolidone grafted silica, polyethylene oxide grafted silica, poly(2-hydroxyethylaspartamide) silica, and poly(N-isopropylacrylamide) grafted silica.

In some embodiments, the isolation and/or selection results in one or more enriched T cell populations, e.g., CD3+ T cells, CD4+ T cells, and/or CD8+ T cells. In some embodiments, two or more separate enriched T cell populations are isolated, selected, enriched, or obtained from a single biological sample. In some embodiments, the separate populations are isolated, selected, enriched, and/or obtained from separate biological samples collected, harvested, and/or obtained from the same subject.

In certain embodiments, the isolation and/or selection results in one or more enriched T cell populations comprising at least 60%, at least 65%, at least 70%, at least 75%, at least 80%, at least 85%, at least 90%, at least 95%, at least 98%, at least 99%, at least 99.5%, at least 99.9%, or 100% or about 100% CD3+ T cells. In certain embodiments, the enriched T cell population consists essentially of CD3+ T cells.

In certain embodiments, the isolation and/or enrichment results in an enriched CD4+ T cell population comprising at least 60%, at least 65%, at least 70%, at least 75%, at least 80%, at least 85%, at least 90%, at least 95%, at least 98%, at least 99%, at least 99.5%, at least 99.9%, or 100% or about 100% CD4+ T cells. In certain embodiments, the CD4+ T cell population comprises less than 40%, less than 35%, less than 30%, less than 25%, less than 20%, less than 15%, less than 10%, less than 5%, less than 1%, less than 0.1%, or less than 0.01% CD8+ T cells, and/or contains no CD8+ T cells and/or is free or substantially free of CD8+ T cells. In some embodiments, the enriched T cell population consists essentially of CD4+ T cells.

In certain embodiments, the isolation and/or enrichment results in an enriched CD8+ T cell population comprising at least 60%, at least 65%, at least 70%, at least 75%, at least 80%, at least 85%, at least 90%, at least 95%, at least 98%, at least 99%, at least 99.5%, at least 99.9%, or 100% or about 100% CD8+ T cells. In certain embodiments, the CD8+ T cell population contains less than 40%, less than 35%, less than 30%, less than 25%, less than 20%, less than 15%, less than 10%, less than 5%, less than 1%, less than 0.1%, or less than 0.01% CD4+ T cells, and/or contains no CD4+ T cells, and/or is free or substantially free of CD4+ T cells. In some embodiments, the enriched T cell population consists essentially of CD8+ T cells.

In some embodiments, the selection marker may be CD4, and the selection agent specifically binds to CD4. In some aspects, the selection agent specifically binding to CD4 may be selected from the group consisting of an anti-CD4 antibody, a bivalent antibody fragment of an anti-CD4 antibody, a monovalent antibody fragment of an anti-CD4 antibody, and a proteinaceous CD4 binding molecule having antibody-like binding properties. In some embodiments, the anti-CD4 antibody, bivalent antibody fragment, or monovalent antibody fragment (e.g., an anti-CD4 Fab fragment) may be derived from antibody 13B8.2 or a functionally active mutant of 13B8.2 that retains specific binding to CD4. For example, exemplary mutants of antibody 13B8.2 or m13B8.2 are described in US-7482000, US-20140295458, US-10228312 and in the literature [Bes et al., J Biol Chem (2003) 278:14265-14273]. The mutant Fab fragment designated "ml3B8.2" retains the variable domain of the CD4 binding murine antibody 13B8.2 and the constant domains comprising a constant human CH1 domain of type gamma and a constant human light chain domain of type kappa for the heavy chain, which is described in US-7482000. In some embodiments, the anti-CD4 antibody, bivalent antibody fragment, or monovalent antibody fragment, e.g., a mutant of antibody 13B8.2, contains an amino acid substitution H91A in the variable light chain, an amino acid substitution Y92A in the variable light chain, an amino acid substitution H35A in the variable heavy chain, and/or an amino acid substitution R53A in the variable heavy chain, each according to Kabat numbering. In some aspects, compared to the variable domain of the 13B8.2 Fab fragment in ml3B8.2, the His residue at position 91 (position 93 of SEQ ID NO: 1) of the light chain is mutated to Ala, and the Arg residue at position 53 (position 55 of SEQ ID NO: 2) of the heavy chain is mutated to Ala. In some embodiments, the reagent that reversibly binds to an anti-CD4 antibody or fragment thereof is commercially available or derived from a commercially available reagent (e.g., catalog numbers 6-8000-206, 6-8000-205, or 6-8002-100; IBA GmbH, Göttingen, Germany). In some embodiments, the selection agent comprises an anti-CD4 Fab fragment. In some embodiments, the anti-CD4 Fab fragment comprises a variable heavy chain having the sequence set forth in SEQ ID NO: 2 and a variable light chain having the sequence set forth in SEQ ID NO: 1. In some embodiments, the anti-CD4 Fab fragment comprises a CDR of the variable heavy chain having the sequence set forth in SEQ ID NO: 2 and a CDR of the variable light chain having the sequence set forth in SEQ ID NO: 1.

In some embodiments, the selection marker may be CD8, and the selection agent specifically binds to CD8. In some aspects, the selection agent that specifically binds to CD8 may be selected from the group consisting of an anti-CD8 antibody, a bivalent antibody fragment of an anti-CD8 antibody, a monovalent antibody fragment of an anti-CD8 antibody, and a proteinaceous CD8 binding molecule having antibody-like binding properties. In some embodiments, the anti-CD8 antibody, bivalent antibody fragment, or monovalent antibody fragment (e.g., an anti-CD8 Fab fragment) may be derived from the antibody OKT8 (e.g., ATCC CRL-8014) or a functionally active mutant thereof that retains specific binding to CD8. In some embodiments, the reagent that reversibly binds to anti-CD8 or a fragment thereof is commercially available or derived from a commercially available reagent (e.g., catalog number 6-8003 or 6-8000-201; IBA GmbH, Göttingen, Germany). In some embodiments, the selection agent comprises an anti-CD8 Fab fragment. In some embodiments, the anti-CD8 Fab fragment comprises a variable heavy chain having the sequence set forth in SEQ ID NO: 3 and a variable light chain having the sequence set forth in SEQ ID NO: 4. In some embodiments, the anti-CD8 Fab fragment comprises a CDR of the variable heavy chain having the sequence set forth in SEQ ID NO: 3 and a CDR of the variable light chain having the sequence set forth in SEQ ID NO: 4.

In some embodiments, the selection marker can be CD3, and the selection agent specifically binds to CD3. In some aspects, the selection agent that specifically binds to CD3 can be selected from the group consisting of an anti-CD3 antibody, a bivalent antibody fragment of an anti-CD3 antibody, a monovalent antibody fragment of an anti-CD3 antibody, and a proteinaceous CD3 binding molecule having antibody-like binding properties. In some embodiments, the anti-CD3 antibody, bivalent antibody fragment, or monovalent antibody fragment (e.g., an anti-CD3 Fab fragment) can be derived from the antibody OKT3 (e.g., ATCC CRL-8001; see, e.g., Stemberger et al., PLoS One (2012) 7(4):e35798) or a functionally active mutant thereof that retains specific binding to CD3. In some embodiments, the reagent that reversibly binds to an anti-CD3 antibody or fragment thereof is commercially available or derived from a commercially available reagent (e.g., catalog number 6-8000-201 or 6-8001-100; IBA GmbH, Göttingen, Germany). In some embodiments, the selection agent contains an anti-CD3 Fab fragment. In some embodiments, the anti-CD3 Fab fragment contains a variable heavy chain having the sequence set forth in SEQ ID NO: 5 and a variable light chain having the sequence set forth in SEQ ID NO: 6. In some embodiments, the anti-CD3 Fab fragment contains CDRs of the variable heavy chain having the sequence set forth in SEQ ID NO: 5 and CDRs of the variable light chain having the sequence set forth in SEQ ID NO: 6.

In any of the above examples, the bivalent antibody fragment can be a F(ab') ₂ fragment or a bivalent single-chain Fv fragment. In some embodiments, the monovalent antibody fragment can be selected from the group consisting of a Fab fragment, an Fv fragment, and a single-chain Fv fragment (scFv). In any of the above examples, the proteinaceous binding molecule having antibody-like binding properties can be an aptamer, a mutein based on a polypeptide of the lipocalin family, a glubody, a protein based on an ankyrin scaffold, a protein based on a crystalline scaffold, an adnectin, or an avimer.

B. Stimulation

In some embodiments, the cell population is incubated under T cell stimulating conditions. In some embodiments, the provided methods involve incubating the cell population under T cell stimulating conditions.

Exemplary methods and stimulating reagents for stimulation of T cells are described in US-6040177, US-6352694, US-11400115, US-20190112576, US-20190136186, US-11274278, US-20210032297, US-20200354677, US-20200384025, US-20210163893, and US-20220002669.

In some embodiments, conditions for T cell stimulation can include one or more of a particular medium, temperature, oxygen content, carbon dioxide content, time, agents, e.g., nutrients, amino acids, antibiotics, ions, and/or stimulatory factors, such as cytokines, chemokines, antigens, binding partners, fusion proteins, recombinant soluble receptors, and any other agent designed to stimulate T cells.

In some embodiments, incubation occurs in a basal medium. In some embodiments, the basal medium is serum-free. In some embodiments, the basal medium is free of human serum. In some embodiments, the basal medium contains a mixture of inorganic salts, sugars, amino acids, and optionally, vitamins, organic acids, and/or buffers or other well-known cell culture nutrients. In addition to nutrients, the basal medium may also help maintain pH and osmolality. Various commercially available basal media are well known to those skilled in the art, including Dulbecco's Modified Eagle's Medium (DMEM), Roswell Park Memorial Institute Medium (RPMI), Iscove's Modified Dulbecco's Medium, and Ham's Medium. In some embodiments, the basal medium is Iscove's Modified Dulbecco's Medium, RPMI-1640, or α-MEM.

In some embodiments, the basal medium is a balanced salt solution (e.g., PBS, DPBS, HBSS, or EBSS). In some embodiments, the basal medium is selected from Dulbecco's Modified Eagle's Medium (DMEM), Minimum Essential Medium (MEM), Basal Medium Eagle (BME), F-10, F-12, RPMI 1640, Glasgow's Minimum Essential Medium (GMEM), Alpha Minimum Essential Medium (Alpha MEM), Iscove's Modified Dulbecco's Medium, and M199. In some embodiments, the basal medium is a complex medium (e.g., RPMI-1640 or IMDM). In some embodiments, the basal medium is OpTmizer™ CTS™ T-cell Expansion Basal Medium (ThermoFisher).

In some embodiments, the basal medium is supplemented with additional additives. In some embodiments, the basal medium is not supplemented with any additional additives. Supplements to the cell culture medium include nutrients, sugars such as glucose, amino acids, vitamins, and additives such as ATP and NADH.

In some embodiments, incubation is performed in a serum-free medium such as described in US-20210207080.

In some embodiments, the incubation occurs in the presence of one or more recombinant cytokines. In some embodiments, the one or more recombinant cytokines are human recombinant cytokines. In certain embodiments, the one or more recombinant cytokines bind and/or are capable of binding to a receptor expressed by and/or endogenous to a T cell. In certain embodiments, the one or more recombinant cytokines comprise a member of the 4-alpha-helix bundle family of cytokines. Members of the 4-alpha-helix bundle family of cytokines include interleukin-2 (IL-2), interleukin-4 (IL-4), interleukin-7 (IL-7), interleukin-9 (IL-9), interleukin-12 (IL-12), interleukin-15 (IL-15), granulocyte colony-stimulating factor (G-CSF), and granulocyte-macrophage colony-stimulating factor (GM-CSF). In some embodiments, the one or more recombinant cytokines are selected from IL-2, IL-15, and IL-7. In some embodiments, the incubation is performed in the presence of IL-2, IL-15, and IL-7.

In certain embodiments, the amount or concentration of one or more recombinant cytokines is measured and/or quantified in International Units (IU). International Units can be used to quantify vitamins, hormones, cytokines, vaccines, blood products, and similar biologically active substances. In some embodiments, an IU is or includes a unit of measurement of the potency of a biological product compared to an international reference standard of a specific weight and strength, such as the WHO First International Standard for Human IL-2, 86/504. International Units are the only recognized and standardized method of reporting units of biological activity derived and published from international collaborative research efforts. In certain embodiments, IUs for a population, sample, or source of cytokines can be obtained through product comparison testing with a similar WHO standard product. For example, in some embodiments, the IU/mg of the population, sample, or source of human recombinant IL-2, IL-7, or IL-15 is compared to a WHO Standard IL-2 product (NIBSC code: 86/500), a WHO Standard IL-17 product (NIBSC code: 90/530), and a WHO Standard IL-15 product (NIBSC code: 95/554), respectively.

In some embodiments, the biological activity (IU/mg) is equivalent to (ED50 (ng/mL))-1 x 10 ⁶ . In certain embodiments, the ED50 of recombinant human IL-2 or IL-15 is equivalent to the concentration required for half-maximal stimulation (XTT cleavage) of cell proliferation using CTLL-2 cells. In certain embodiments, the ED50 of recombinant human IL-7 is equivalent to the concentration required for half-maximal stimulation of proliferation of PHA-activated human peripheral blood lymphocytes. Details regarding the assay and calculation of IU for IL-2 are discussed in the literature [Wadhwa et al., Journal of Immunological Methods (2013) 379 (1-2): 1-7] and [Gearing and Thorpe, Journal of Immunological Methods (1988) 114 (1-2): 3-9]. Details regarding the assay and calculation of IU for IL-15 are discussed in the literature [Soman et al., Journal of Immunological Methods (2009) 348 (1-2):83-94].

In some embodiments, the cells are stimulated or subjected to stimulation in the presence of a recombinant cytokine, e.g., a recombinant human cytokine, at a concentration of 1 IU/mL to 1,000 IU/mL, 10 IU/mL to 50 IU/mL, 50 IU/mL to 100 IU/mL, 100 IU/mL to 200 IU/mL, 100 IU/mL to 500 IU/mL, 250 IU/mL to 500 IU/mL, or 500 IU/mL to 1,000 IU/mL.

In some embodiments, the cells are stimulated or subjected to stimulation in the presence of recombinant IL-2, e.g., human recombinant IL-2, at a concentration of 1 IU/mL to 500 IU/mL, 10 IU/mL to 250 IU/mL, 50 IU/mL to 200 IU/mL, 50 IU/mL to 150 IU/mL, 75 IU/mL to 125 IU/mL, 100 IU/mL to 200 IU/mL, or 10 IU/mL to 100 IU/mL. In certain embodiments, the cells are stimulated or subjected to stimulation in the presence of recombinant IL-2 at a concentration of or about 50 IU/mL, 60 IU/mL, 70 IU/mL, 80 IU/mL, 90 IU/mL, 100 IU/mL, 110 IU/mL, 120 IU/mL, 130 IU/mL, 140 IU/mL, 150 IU/mL, 160 IU/mL, 170 IU/mL, 180 IU/mL, 190 IU/mL, or 100 IU/mL. In some embodiments, the cells are stimulated or subjected to stimulation in the presence of recombinant IL-2, e.g., human recombinant IL-2, at or about 100 IU/mL.

In some embodiments, the cells are stimulated or subjected to stimulation in the presence of recombinant IL-7, e.g., human recombinant IL-7, at a concentration of 100 IU/mL to 2,000 IU/mL, 500 IU/mL to 1,000 IU/mL, 100 IU/mL to 500 IU/mL, 500 IU/mL to 750 IU/mL, 750 IU/mL to 1,000 IU/mL, or 550 IU/mL to 650 IU/mL. In certain embodiments, the cells are stimulated or subjected to stimulation in the presence of IL-7 at a concentration of or about 50 IU/mL, 100 IU/mL, 150 IU/mL, 200 IU/mL, 250 IU/mL, 300 IU/mL, 350 IU/mL, 400 IU/mL, 450 IU/mL, 500 IU/mL, 550 IU/mL, 600 IU/mL, 650 IU/mL, 700 IU/mL, 750 IU/mL, 800 IU/mL, 750 IU/mL, 750 IU/mL, or 1,000 IU/mL. In certain embodiments, the cells are stimulated or subjected to stimulation in the presence of recombinant IL-7, e.g., human recombinant IL-7, at or about 600 IU/mL.

In some embodiments, the cells are stimulated or subjected to stimulation in the presence of recombinant IL-15, e.g., human recombinant IL-15, at a concentration of 1 IU/mL to 500 IU/mL, 10 IU/mL to 250 IU/mL, 50 IU/mL to 200 IU/mL, 50 IU/mL to 150 IU/mL, 75 IU/mL to 125 IU/mL, 100 IU/mL to 200 IU/mL, or 10 IU/mL to 100 IU/mL. In certain embodiments, the cells are stimulated or subjected to stimulation in the presence of recombinant IL-15 at a concentration of or about 50 IU/mL, 60 IU/mL, 70 IU/mL, 80 IU/mL, 90 IU/mL, 100 IU/mL, 110 IU/mL, 120 IU/mL, 130 IU/mL, 140 IU/mL, 150 IU/mL, 160 IU/mL, 170 IU/mL, 180 IU/mL, 190 IU/mL, or 200 IU/mL. In some embodiments, the cells are stimulated or subjected to stimulation in the presence of recombinant IL-15, e.g., human recombinant IL-15, at or about 100 IU/mL.

In some embodiments, incubation occurs in the absence of recombinant cytokines.

In certain embodiments, stimulation is performed under static conditions, such as without centrifugation, shaking, spinning, rocking, or perfusion of the medium, e.g., continuous or semi-continuous perfusion. In some embodiments, cells are transferred (e.g., transferred under sterile conditions) to a container, such as a bag or vial, prior to or immediately following the initiation of stimulation, e.g., within 5, 15, or 30 minutes, and placed in an incubator. In certain embodiments, the incubator is set to, or about, 16°C, 24°C, or 35°C. In some embodiments, the incubator is set to, or at least about, 37°C, about 37°C, or 37°C ±2°C, ±1°C, ±0.5°C, or ±0.1°C. In certain embodiments, stimulation under static conditions is performed in a cell culture bag placed in an incubator. In some embodiments, the culture bag is comprised of a single-web polyolefin gas-permeable film that allows monocytes to adhere to the bag surface when present.

In some embodiments, the T cell stimulating conditions involve incubation in the presence of a T cell stimulator. In some embodiments, the T cell stimulator binds to a molecule expressed on the surface of the T cell. In some embodiments, one of the T cell stimulators induces a primary activation signal in the T cell. In some embodiments, one of the T cell stimulators induces a costimulatory signal in the T cell. In some embodiments, the T cell stimulator induces both a primary activation signal and a costimulatory signal in the T cell.

In some embodiments, a T cell stimulator that induces a primary activation signal binds to a member of the TCR/CD3 complex on a T cell. In some embodiments, the T cell stimulator binds to CD3. In some embodiments, the T cell stimulator is an anti-CD3 antibody, a bivalent antibody fragment of an anti-CD3 antibody, a monovalent antibody fragment of an anti-CD3 antibody, or a proteinaceous CD3 binding molecule having antibody-like binding properties. In some embodiments, the T cell stimulator is an anti-CD3 antibody or antibody fragment. In some embodiments, the anti-CD3 antibody, a bivalent antibody fragment of an anti-CD3 antibody, or a monovalent antibody fragment of an anti-CD3 antibody (e.g., an anti-CD3 Fab fragment) is derived from the antibody OKT3 (e.g., ATCC CRL-8001; see, e.g., Stemberger et al., PLoS One (2012) 7(4):e35798) or a functionally active mutant thereof that retains specific binding to CD3. In some embodiments, the T cell stimulator is an anti-CD3 Fab. In some embodiments, the anti-CD3 Fab comprises a variable heavy chain having the sequence set forth in SEQ ID NO: 5 and a variable light chain having the sequence set forth in SEQ ID NO: 6. In some embodiments, the anti-CD3 Fab comprises CDRs of the variable heavy chain having the sequence set forth in SEQ ID NO: 5 and CDRs of the variable light chain having the sequence set forth in SEQ ID NO: 6.

In some embodiments, the T cell stimulatory agent that induces a costimulatory signal binds to a costimulatory molecule on the T cell. In some embodiments, the costimulatory molecule is CD28, CD90 (Thy-1), CD95 (Apo-/Fas), CD137 (4-1BB), CD154 (CD40L), ICOS, LAT, CD27, OX40, or HVEM.

In some embodiments, the costimulatory molecule is CD28. In some embodiments, the T cell stimulator is an anti-CD28 antibody, a bivalent antibody fragment of an anti-CD28 antibody, a monovalent antibody fragment of an anti-CD28 antibody, or a proteinaceous CD28 binding molecule with antibody-like binding properties. In some embodiments, the T cell stimulator is an anti-CD28 antibody or antibody fragment. In some embodiments, the anti-CD28 antibody, the bivalent antibody fragment of the anti-CD28 antibody, or the monovalent antibody fragment of the anti-CD28 antibody (e.g., an anti-CD28 Fab fragment) is derived from antibody CD28.3 (deposited as a synthetic single chain Fv construct under GenBank Accession No. AF451974.1; see also Vanhove et al., Blood (2003) 102(2):564-570), the variable heavy and light chains of which contain the amino acid sequences set forth in SEQ ID NOs: 7 and 8, respectively. In some embodiments, the T cell stimulator is an anti-CD28 Fab. In some embodiments, the anti-CD28 Fab contains a variable heavy chain having the sequence set forth in SEQ ID NO: 7 and a variable light chain having the sequence set forth in SEQ ID NO: 8. In some embodiments, the anti-CD28 Fab contains a CDR of a variable heavy chain having the sequence set forth in SEQ ID NO: 7 and a CDR of a variable light chain having the sequence set forth in SEQ ID NO: 8.

In some embodiments, the costimulatory molecule is CD90. In some embodiments, the T cell stimulator is an anti-CD90 antibody, a bivalent antibody fragment of an anti-CD90 antibody, a monovalent antibody fragment of an anti-CD90 antibody, or a proteinaceous CD90 binding molecule having antibody-like binding properties. In some embodiments, the T cell stimulator is an anti-CD90 antibody or antibody fragment. In some embodiments, the T cell stimulator is an anti-CD90 Fab. In some embodiments, the anti-CD90 antibody, bivalent antibody fragment of an anti-CD90 antibody, or monovalent antibody fragment of an anti-CD90 antibody (e.g., an anti-CD90 Fab fragment) is derived from anti-CD90 antibody G7 (Biolegend, catalog no. 105201).

In some embodiments, the costimulatory molecule is CD95. In some embodiments, the T cell stimulator is an anti-CD95 antibody, a bivalent antibody fragment of an anti-CD95 antibody, a monovalent antibody fragment of an anti-CD95 antibody, or a proteinaceous CD95 binding molecule with antibody-like binding properties. In some embodiments, the T cell stimulator is an anti-CD28 antibody or antibody fragment. In some embodiments, the T cell stimulator is an anti-CD95 Fab. In some embodiments, the anti-CD95 antibody, bivalent antibody fragment of an anti-CD95 antibody, or monovalent antibody fragment of an anti-CD95 antibody (e.g., an anti-CD95 Fab fragment) is derived from monoclonal mouse anti-human CD95 CH11 (Upstate Biotechnology, Lake Placid, NY), anti-CD95 mAb 7C11, or anti-APO-1 (e.g., as described in Paulsen et al., Cell Death & Differentiation (2011) 18(4):619-631).

In some embodiments, the costimulatory molecule is CD137. In some embodiments, the T cell stimulator is an anti-CD137 antibody, a bivalent antibody fragment of an anti-CD137 antibody, a monovalent antibody fragment of an anti-CD137 antibody, or a proteinaceous CD137 binding molecule having antibody-like binding properties. In some embodiments, the T cell stimulator is an anti-CD137 antibody or antibody fragment. In some embodiments, the T cell stimulator is an anti-CD137 Fab. In some embodiments, the anti-CD137 antibody, a bivalent antibody fragment of an anti-CD137 antibody, or a monovalent antibody fragment of an anti-CD137 antibody (e.g., an anti-CD137 Fab fragment) is derived from LOB12, IgG2a or LOB12.3, IgG1 (as described in Taraban et al., Eur J Immunol. (2002) 32(12):3617-27). See also, for example, US-6569997, US-6303121 and the literature [Mittler et al., Immunol Res. (2004) 29(1-3):197-208].

In some embodiments, the costimulatory molecule is CD40. In some embodiments, the T cell stimulator is an anti-CD40 antibody, a bivalent antibody fragment of an anti-CD40 antibody, a monovalent antibody fragment of an anti-CD40 antibody, or a proteinaceous CD40 binding molecule with antibody-like binding properties. In some embodiments, the T cell stimulator is an anti-CD40 antibody or antibody fragment. In some embodiments, the T cell stimulator is an anti-CD40 Fab.

In some embodiments, the costimulatory molecule is CD40L. In some embodiments, the T cell stimulator is an anti-CD40L antibody, a bivalent antibody fragment of an anti-CD40L antibody, a monovalent antibody fragment of an anti-CD40L antibody, or a proteinaceous CD40L binding molecule having antibody-like binding properties. In some embodiments, the T cell stimulator is an anti-CD40L antibody or antibody fragment. In some embodiments, the T cell stimulator is an anti-CD40L Fab. In some embodiments, the anti-CD40L antibody, a bivalent antibody fragment of an anti-CD40L antibody, or a monovalent antibody fragment of an anti-CD40L antibody (e.g., an anti-CD40L Fab fragment) is derived from Hu5C8, as described in Blair et al., JEM (2000) 19(4):651-660. See also, for example, US-7563445, US20010026932, US7547438, and US-7172759.

In some embodiments, the costimulatory molecule is ICOS. In some embodiments, the T cell stimulator is an anti-ICOS antibody, a bivalent antibody fragment of an anti-ICOS antibody, a monovalent antibody fragment of an anti-ICOS antibody, or a proteinaceous ICOS binding molecule having antibody-like binding properties. In some embodiments, the T cell stimulator is an anti-ICOS antibody or antibody fragment. In some embodiments, the T cell stimulator is an anti-ICOS Fab. In some embodiments, the anti-ICOS antibody, a bivalent antibody fragment of an anti-ICOS antibody, or a monovalent antibody fragment of an anti-ICOS antibody (e.g., an anti-ICOS Fab fragment) is derived from any antibody described in US-20080279851 and in Deng et al., Hybrid Hybridomics (2004) 23(3):176-82.

In some embodiments, the co-stimulatory molecule is a linker (LAT) for T cell activation. In some embodiments, the T cell stimulator is an anti-LAT antibody, a bivalent antibody fragment of an anti-LAT antibody, a monovalent antibody fragment of an anti-LAT antibody, or a proteinaceous LAT binding molecule with antibody-like binding properties. In some embodiments, the T cell stimulator is an anti-LAT antibody or antibody fragment. In some embodiments, the T cell stimulator is an anti-LAT Fab.

In some embodiments, the costimulatory molecule is CD27. In some embodiments, the T cell stimulator is an anti-CD27 antibody, a bivalent antibody fragment of an anti-CD27 antibody, a monovalent antibody fragment of an anti-CD27 antibody, or a proteinaceous CD27 binding molecule having antibody-like binding properties. In some embodiments, the T cell stimulator is an anti-CD27 antibody or antibody fragment. In some embodiments, the T cell stimulator is an anti-CD27 Fab. In some embodiments, the anti-CD27 antibody, a bivalent antibody fragment of an anti-CD27 antibody, or a monovalent antibody fragment of an anti-CD27 antibody (e.g., an anti-CD27 Fab fragment) is derived from any antibody described in US-8481029.

In some embodiments, the costimulatory molecule is OX40. In some embodiments, the T cell stimulatory agent is an anti-OX40 antibody, a bivalent antibody fragment of an anti-OX40 antibody, a monovalent antibody fragment of an anti-OX40 antibody, or a proteinaceous OX40 binding molecule having antibody-like binding properties. In some embodiments, the T cell stimulatory agent is an anti-OX40 antibody or antibody fragment. In some embodiments, the T cell stimulatory agent is an anti-OX40 Fab. In some embodiments, the anti-OX40 antibody, a bivalent antibody fragment of an anti-OX40 antibody, or a monovalent antibody fragment of an anti-OX40 antibody (e.g., an anti-OX40 Fab fragment) is derived from any antibody described in US-9475880 and Melero et al., Clin Cancer Res. (2013) 19(5):1044-53.

In some embodiments, the costimulatory molecule is HVEM. In some embodiments, the T cell stimulator is an anti-HVEM antibody, a bivalent antibody fragment of an anti-HVEM antibody, a monovalent antibody fragment of an anti-HVEM antibody, or a proteinaceous HVEM binding molecule having antibody-like binding properties. In some embodiments, the T cell stimulator is an anti-HVEM antibody or antibody fragment. In some embodiments, the T cell stimulator is an anti-HVEM Fab. In some embodiments, the anti-HVEM antibody, a bivalent antibody fragment of an anti-HVEM antibody, or a monovalent antibody fragment of an anti-HVEM antibody (e.g., an anti-HVEM Fab fragment) is derived from any antibody described in WO-2006054961, US-8188232, and Park et al., Cancer Immunol Immunother. (2012) 61(2):203-14.

In some embodiments, the T cell stimulator is immobilized on a solid support. In some embodiments, the solid support is a bead. In some embodiments, the bead is biocompatible, for example, composed of a material suitable for biological applications. In some embodiments, the bead is non-toxic to T cells.

In some embodiments, the beads have a diameter greater than about 0.001 μm, greater than about 0.01 μm, greater than about 0.1 μm, greater than about 1.0 μm, greater than about 10 μm, greater than about 50 μm, greater than about 100 μm, or greater than about 1000 μm and less than or equal to about 1500 μm. In some embodiments, the beads have a diameter of from about 1.0 μm to about 500 μm, from about 1.0 μm to about 150 μm, from about 1.0 μm to about 30 μm, from about 1.0 μm to about 10 μm, from about 1.0 μm to about 5.0 μm, from about 2.0 μm to about 5.0 μm, or from about 3.0 μm to about 5.0 μm. In some embodiments, the beads have a diameter of from about 3 μm to about 5 μm. In some embodiments, the bead has a diameter of at least or at least about or about 0.001 μm, 0.01 μm, 0.1 μm, 0.5 μm, 1.0 μm, 1.5 μm, 2.0 μm, 2.5 μm, 3.0 μm, 3.5 μm, 4.0 μm, 4.5 μm, 5.0 μm, 5.5 μm, 6.0 μm, 6.5 μm, 7.0 μm, 7.5 μm, 8.0 μm, 8.5 μm, 9.0 μm, 9.5 μm, 10 μm, 12 μm, 14 μm, 16 μm, 18 μm, or 20 μm. In certain embodiments, the bead has a diameter of 4.5 μm. In certain embodiments, the bead has a diameter of 2.8 μm or about said value.

In some embodiments, the beads have a density greater than 0.001 g/cm3, greater than 0.01 g/cm3, greater than 0.05 g/cm3, greater than 0.1 g/cm3, greater than 0.5 g/cm3, greater than 0.6 g/cm3, greater than 0.7 g/cm3, greater than 0.8 g/cm3, greater than 0.9 g/cm3, greater than 1 g/cm3, greater than 1.1 g/cm3, greater than 1.2 g/cm3, greater than 1.3 g/cm3, greater than 1.4 g/cm3, greater than 1.5 g/cm3, greater than 2 g/cm3, greater than 3 g/cm3, greater than 4 g/cm3, or greater than 5 g/cm3. In some embodiments, the beads have a density of about 0.001 g/cm3 to about 100 g/cm3, about 0.01 g/cm3 to about 50 g/cm3, about 0.1 g/cm3 to about 10 g/cm3, about 0.1 g/cm3 to about 0.5 g/cm3, about 0.5 g/cm3 to about 1 g/cm3, about 0.5 g/cm3 to about 1.5 g/cm3, about 1 g/cm3 to about 1.5 g/cm3, about 1 g/cm3 to about 2 g/cm3, or about 1 g/cm3 to about 5 g/cm3. In some embodiments, the beads have a density of about 0.5 g/cm3, about 0.5 g/cm3, about 0.6 g/cm3, about 0.7 g/cm3, about 0.8 g/cm3, about 0.9 g/cm3, about 1.0 g/cm3, about 1.1 g/cm3, about 1.2 g/cm3, about 1.3 g/cm3, about 1.4 g/cm3, about 1.5 g/cm3, about 1.6 g/cm3, about 1.7 g/cm3, about 1.8 g/cm3, about 1.9 g/cm3, or about 2.0 g/cm3. In certain embodiments, the beads have a density of about 1.6 g/cm3. In certain embodiments, the beads have a density of about 1.5 g/cm3. In certain embodiments, the beads have a density of about 1.3 g/cm3.

In some embodiments, the plurality of beads have a uniform density. In some embodiments, the uniform density has a density standard deviation of less than 10%, less than 5%, or less than 1% of the average bead density.

In some embodiments, the beads respond to a magnetic field. In some embodiments, the beads are magnetic beads. In some embodiments, the beads are paramagnetic. In certain embodiments, the beads are superparamagnetic. In certain embodiments, the beads do not exhibit any magnetic properties unless exposed to a magnetic field.

In certain embodiments, the bead contains a magnetic core. In certain embodiments, the bead contains a paramagnetic core. In certain embodiments, the bead contains a superparamagnetic core. In some embodiments, the core contains a metal. In some embodiments, the metal can be iron, nickel, copper, cobalt, gadolinium, manganese, tantalum, zinc, zirconium, or any combination thereof. In certain embodiments, the core contains a metal oxide (e.g., iron oxide), a ferrite (e.g., manganese ferrite, cobalt ferrite, and nickel ferrite), hematite, and/or a metal alloy (e.g., CoTaZn). In some embodiments, the core contains one or more of a ferrite, a metal, a metal alloy, iron oxide, and chromium dioxide. In some embodiments, the core contains elemental iron or a compound thereof. In some embodiments, the core contains one or more of magnetite (Fe ₃ O ₄ ), maghemite (γFe ₂ O ₃ ), and greigeite (Fe ₃ S ₄ ). In some embodiments, the core contains iron oxide (e.g., Fe ₃ O ₄ ).

In some embodiments, the bead contains at least one substance on or near the surface of the bead that can be coupled, linked, or conjugated to an agent. In some embodiments, the T cell stimulator is immobilized on the bead via this substance. In some embodiments, the bead is surface-functionalized, for example, having a functional group capable of forming a covalent bond with a binding molecule, such as a polynucleotide or polypeptide. In certain embodiments, the bead has surface-exposed carboxyl, amino, hydroxyl, tosyl, epoxy, and/or chloromethyl groups. In certain embodiments, the bead has surface-exposed agarose and/or sepharose. In some embodiments, the bead has surface-exposed protein A, protein G, or biotin.

In certain embodiments, the bead contains a magnetic, paramagnetic, and/or superparamagnetic core covered with a surface-functionalized coat or coating. In some embodiments, the coat may contain a material that may include a polymer, a polysaccharide, silica, a fatty acid, a protein, carbon, agarose, sepharose, or a combination thereof. In some embodiments, the polymer may be polyethylene glycol, poly(lactic-co-glycolic acid), polyglutaraldehyde, polyurethane, polystyrene, or polyvinyl alcohol. In certain embodiments, the outer coat or coating comprises polystyrene. In certain embodiments, the outer coating is surface-functionalized.

In some embodiments, the cell population containing T cells is incubated with a stimulating reagent at a bead-to-cell ratio of 3:1, 2.5:1, 2:1, 1.5:1, 1.25:1, 1.2:1, 1.1:1, 1:1, 0.9:1, 0.8:1, 0.75:1, 0.67:1, 0.5:1, 0.3:1, or 0.2:1 or about such values. In particular embodiments, the bead-to-cell ratio is between 2.5:1 and 0.2:1, between 2:1 and 0.5:1, between 1.5:1 and 0.75:1, between 1.25:1 and 0.8:1, or between 1.1:1 and 0.9:1. In particular embodiments, the bead-to-cell ratio is about 1:1 or 1:1.

In some embodiments, the T cell stimulator is not immobilized on a solid support, such as a bead. In some embodiments, the T cell stimulator is part of a stimulating reagent in a soluble form. Exemplary T cell stimulating reagents in a soluble form are described in US2021/0032297 (see also Poltorak et al., Scientific Reports (2020)).

In some embodiments, the T cell stimulator is immobilized on an oligomeric streptavidin mutein reagent. In some embodiments, the T cell stimulator is reversibly immobilized on an oligomeric streptavidin mutein reagent. In some embodiments, the oligomeric streptavidin mutein reagent is an oligomer or polymer of a streptavidin mutein.

In some embodiments, the T cell stimulator comprises a binding partner that reversibly binds to a streptavidin mutein molecule of an oligomeric streptavidin mutein reagent. In some embodiments, the binding partner is fused to the C-terminus of the heavy chain of the T cell stimulator.

In some embodiments, the binding partner reversibly binds to the biotin-binding site of the streptavidin mutein. In some embodiments, binding of the binding partner to the streptavidin mutein is inhibited by the presence of biotin. In some embodiments, the biotin is D-biotin.

In some embodiments, one or both of the binding partners is biotin, a biotin analog, or a streptavidin-binding peptide. In some embodiments, one or both of the binding partners is a streptavidin-binding peptide. In some embodiments, the streptavidin-binding peptide comprises a sequence set forth in any of SEQ ID NOs: 9-15. In some embodiments, the sequence of the streptavidin-binding peptide is set forth in any of SEQ ID NOs: 9-15. In some embodiments, the streptavidin-binding peptide comprises a sequence set forth in SEQ ID NO: 15. In some embodiments, the sequence of the streptavidin-binding peptide is set forth in SEQ ID NO: 15.

In some embodiments, the oligomeric streptavidin mutein reagent comprises between 2,000 and 3,000 or about that number of streptavidin mutein tetramers. In some embodiments, the oligomeric streptavidin mutein reagent comprises about 2,400 streptavidin mutein tetramers.

In some embodiments, the oligomeric streptavidin mutein reagent has a radius of 90 to 110 nm or about that value. In some embodiments, the oligomeric streptavidin mutein reagent has a radius of about 100 nm. In some embodiments, the radius is the hydrodynamic radius.

In some embodiments, individual molecules of the oligomeric streptavidin mutein reagent are cross-linked by a bifunctional linker. In some embodiments, the bifunctional linker is a heterobifunctional linker. In some embodiments, the bifunctional linker is an amine-to-thiol linker.

In some embodiments, the streptavidin mutein reversibly binds to biotin, a biotin analog, or a streptavidin-binding peptide. In some embodiments, the streptavidin mutein reversibly binds to a streptavidin-binding peptide. In some embodiments, the streptavidin-binding peptide comprises a sequence set forth in any of SEQ ID NOs: 9-15. In some embodiments, the sequence of the streptavidin-binding peptide is set forth in any of SEQ ID NOs: 9-15.

In some embodiments, the streptavidin mutein comprises one or more mutations compared to wild-type streptavidin. In some embodiments, the streptavidin mutein comprises one or more mutations compared to the amino acid sequence set forth in SEQ ID NO: 16.

In some embodiments, the streptavidin mutein comprises one or more mutations compared to a minimal streptavidin. In some embodiments, the minimal streptavidin starts N-terminally in the region of amino acid positions 10 to 16 and ends C-terminally in the region of amino acid positions 133 to 142, compared to the sequence set forth in SEQ ID NO: 16. In some embodiments, the streptavidin mutein starts N-terminally in the region of amino acid positions 10 to 16 and ends C-terminally in the region of amino acid positions 133 to 142, compared to the sequence set forth in SEQ ID NO: 16.

In some embodiments, the sequence of the minimal streptavidin is from positions Ala13 to Ser139 of the amino acid sequence set forth in SEQ ID NO: 16. In some embodiments, the minimal streptavidin has an N-terminal methionine residue in place of Ala13.

In some embodiments, the sequence of the minimal streptavidin is set forth in SEQ ID NO: 17. In some embodiments, the streptavidin mutein comprises one or more mutations compared to the amino acid sequence set forth in SEQ ID NO: 17.

In some embodiments, the sequence of the minimal streptavidin is set forth in SEQ ID NO: 18. In some embodiments, the streptavidin mutein comprises one or more mutations compared to the amino acid sequence set forth in SEQ ID NO: 18.

In some embodiments, the streptavidin mutein comprises the amino acid sequence Val44-Thr45-Ala46-Arg47 (SEQ ID NO: 19) at sequence positions corresponding to positions 44 to 47 of the amino acid sequence set forth in SEQ ID NO: 16. In some embodiments, the streptavidin mutein comprises the amino acid sequence Ile44-Gly45-Ala46-Arg47 (SEQ ID NO: 20) at sequence positions corresponding to positions 44 to 47 of the amino acid sequence set forth in SEQ ID NO: 16.

In some embodiments, the streptavidin mutein further comprises amino acid substitutions Glu117, Gly120, and Try121 at sequence positions corresponding to the positions of the amino acid sequence set forth in SEQ ID NO: 16.

In some embodiments, the streptavidin mutein comprises an amino acid sequence set forth in any of SEQ ID NOs: 21-28. In some embodiments, the streptavidin mutein comprises an amino acid sequence set forth in SEQ ID NO: 21. In some embodiments, the sequence of the streptavidin mutein is set forth in SEQ ID NO: 21.

In some embodiments, the incubation under T cell stimulating conditions is performed for at least 12 hours. In some embodiments, the incubation under T cell stimulating conditions is performed for 48 hours, 42 hours, 36 hours, 30 hours, 24 hours, 22 hours, 20 hours, 18 hours, 16 hours, or 12 hours, about the time, or less than the time. In some embodiments, the incubation is performed for 16 hours to 24 hours, or about the time. In particular embodiments, the incubation is performed for 12 hours to 36 hours, 18 hours to 30 hours, or about the time, or for 24 hours, or about the time. In some embodiments, the incubation is performed for 2 days, about the time, or less than the time. In some embodiments, the incubation is performed for 1 day, about the time, or less than the time.

C. Genetic engineering

In some embodiments, the cell population is genetically engineered to express a recombinant protein. In some embodiments, the methods provided involve genetically engineering the cell population to express the recombinant protein. In some embodiments, a heterologous or recombinant polynucleotide encoding the recombinant protein is introduced into cells of the cell population. Any method for introducing a heterologous or recombinant polynucleotide that results in integration of the polynucleotide encoding the recombinant protein into the genome of a cell, such as a T cell, can be used, including viral and non-viral methods of genetic engineering. Introduction of a polynucleotide encoding a recombinant protein, e.g., a heterologous or recombinant polynucleotide, into a cell can be accomplished using any of a number of known vectors. Exemplary vectors are described in Section II-C-1. Such vectors include viral (including lentiviral and gammaretroviral) systems. Exemplary methods include methods for transferring a heterologous polynucleotide encoding the recombinant protein, including via viral (e.g., retroviral or lentiviral) transduction.

Exemplary methods for genetic manipulation of cells are described in US-11400115, US-20190112576, US-20190136186, US-11274278, US-20210032297, US-20200354677, US-20200384025, US-20210163893, and US-20220002669.

In some embodiments, the heterologous or recombinant polynucleotide encoding the recombinant protein is introduced using non-viral methods, such as electroporation, calcium phosphate transfection, protoplast fusion, cationic liposome-mediated transfection, nanoparticles such as lipid nanoparticles, tungsten particle-facilitated microparticle bombardment, strontium phosphate DNA co-precipitation, or other approaches (e.g., as described in US-10654928 and US-7446190). Transposon-based systems are also contemplated.

In certain embodiments, the cells are genetically engineered, transformed, or transduced after the cells have been stimulated, such as by any of the methods described herein, e.g., the methods described in Section II-B. In certain embodiments, the cells are genetically engineered, transformed, or transduced at, about, or within 72, 60, 48, 36, 24, or 12 hours after the initiation of stimulation (inclusive). In certain embodiments, the cells are genetically engineered, transformed, or transduced at, about, or within 3, 2, or 1 day after the initiation of stimulation (inclusive). In certain embodiments, the cells are genetically engineered, transformed, or transduced between 12 and 48 hours, between 16 and 36 hours, or between 18 and 30 hours after the initiation of stimulation (inclusive). In certain embodiments, the cells are genetically engineered, transformed, or transduced between 18 and 30 hours after the initiation of stimulation (inclusive). In certain embodiments, the cells are genetically engineered, transformed or transduced at or about 16, 18, 20, 22 or 24 hours after the initiation of stimulation.

In some embodiments, the manipulation, e.g., transduction, is performed for 24 to 48 hours, 36 to 12 hours, 18 to 30 hours, or 24 hours or about the time periods mentioned above. In some embodiments, the manipulation, e.g., transduction, is performed for 24 hours, 48 hours, or 72 hours or about the time periods mentioned above, or for 1 day, 2 days, or 3 days or about the time periods mentioned above. In particular embodiments, the manipulation, e.g., transduction, is performed for 24 hours ± 6 hours, 48 hours ± 6 hours, or 72 hours ± 6 hours or about the time periods mentioned above. In particular embodiments, the manipulation, e.g., transduction, is performed for 72 hours, 72 ± 4 hours, or for 3 days or about the time periods mentioned above.

In certain embodiments, the genetic engineering method is performed by contacting or introducing one or more cells of a population with a nucleic acid molecule or polynucleotide encoding a recombinant protein. In certain embodiments, the nucleic acid molecule or polynucleotide is heterologous to the cell. In certain embodiments, the heterologous nucleic acid molecule or polynucleotide is not native to the cell. In certain embodiments, the heterologous nucleic acid molecule or polynucleotide encodes a protein, e.g., a recombinant protein, that is not naturally expressed by the cell. In certain embodiments, the heterologous nucleic acid molecule or polynucleotide is or contains a nucleic acid sequence that is not found in the cell prior to the contact or introduction.

In some embodiments, the cells are manipulated, e.g., transduced, in the presence of a transduction adjuvant. Exemplary transduction adjuvants include polycations, fibronectin or fibronectin-derived fragments or variants, and retronectin. In certain embodiments, the cells are manipulated in the presence of polycations, fibronectin or fibronectin-derived fragments or variants, and/or retronectin. In certain embodiments, the cells are manipulated in the presence of a polycation that is polybrene, DEAE-dextran, protamine sulfate, poly-L-lysine, or cationic liposomes. In certain embodiments, the cells are manipulated in the presence of protamine sulfate.

In some embodiments, the genetic manipulation, e.g., transduction, is performed in any medium described in Section II-B. In some embodiments, the genetic manipulation, e.g., transduction, is performed in a serum-free medium, e.g., any medium described in US-20210207080.

In some embodiments, the genetic manipulation, e.g., transduction, is performed in the presence of one or more recombinant cytokines. In some embodiments, the genetic manipulation, e.g., transduction, is performed in the presence of any recombinant cytokine described in Section II-B and at any concentration described in Section II-B.

In some embodiments, the cells are genetically engineered, transformed, or transduced in the presence of the same or similar medium as that present during stimulation. In some embodiments, the cells are genetically engineered, transformed, or transduced in a medium containing the same cytokines as the medium present during stimulation. In certain embodiments, the cells are genetically engineered, transformed, or transduced in a medium containing the same cytokines at the same concentration as the medium present during stimulation.

In some embodiments, the genetic manipulation of a cell comprises or involves introducing a polynucleotide, e.g., a heterologous or recombinant polynucleotide, into the cell by transduction. In some embodiments, the cell is transduced or subjected to transduction using a viral vector. In a particular embodiment, the cell is transduced or subjected to transduction using a viral vector. In some embodiments, the virus is a retroviral vector, such as a gammaretroviral vector or a lentiviral vector. Methods for lentiviral transduction are well known. Exemplary methods are described, e.g., in Wang et al., J. Immunother. (2012) 35(9):689-701; Cooper et al., Blood (2003) 101:1637-1644; Verhoeyen et al., Methods Mol Biol. (2009) 506:97-114; and Cavalieri et al., Blood. (2003) 102(2):497-505].

In some embodiments, transduction is performed by contacting one or more cells of the population with a nucleic acid molecule encoding a recombinant protein. In some embodiments, contacting can be performed by centrifugation, such as spinoculation (e.g., centrifugal inoculation). Such methods include any of the methods described in US-10428351. Exemplary centrifugation chambers include those manufactured and sold by Biosafe SA, including those for use with the Sepax® and Sepax® 2 systems, including the A-200/F and A-200 centrifugation chambers and various kits for use with such systems. Exemplary chambers, systems, and processing equipment and cabinets are described, for example, in US-6123655, US-6733433, US-20080171951, and US-US6733433. Exemplary kits for use with these systems include the single-use kits sold by Biosafe SA under product names CS-430.1, CS-490.1, CS-600.1, and CS-900.2.

In certain embodiments, the genetic manipulation (e.g., by transforming (e.g., transducing) a cell with a viral vector) further comprises one or more steps of incubating the cell after introducing or contacting the cell with the viral vector. In some embodiments, the cell, e.g., a cell of a population of transformed cells (also referred to as a "transformed cell"), is incubated after a process to genetically manipulate, transform, transduce, or transfect the cell to introduce a viral vector into the cell.

In some embodiments, cells, e.g., transformed cells, are incubated after introduction of a heterologous or recombinant polynucleotide, e.g., a viral vector particle, without further processing of the cells. In certain embodiments, prior to incubation, the cells are washed to remove or substantially remove exogenous or residual polynucleotides encoding the heterologous or recombinant polynucleotide, e.g., viral vector particles (e.g., those remaining in the medium after a genetic engineering process following spinoculation).

In some embodiments, the additional incubation is performed under conditions that result in integration of the viral vector into the host genome of one or more cells. For example, the additional incubation provides time for the viral vector, which can bind to T cells, e.g., via spinoculation, after transduction to integrate into the cellular genome and deliver the gene of interest. In some aspects, the additional incubation is performed under conditions that allow the cells, e.g., the transformed cells, to rest or recover, wherein the cell culture supports or maintains the health of the cells during the incubation. In certain embodiments, the cells are incubated under static conditions, such as conditions that do not involve centrifugation, shaking, spinning, rocking, or perfusion of the medium, e.g., continuous or semi-continuous perfusion.

It is within the skill of one of ordinary skill in the art to assess or determine whether incubation has resulted in integration of the viral vector particle into the host genome, and thus empirically determine further incubation conditions. In some embodiments, integration of the viral vector into the host genome can be assessed by measuring the level of expression of a recombinant protein, such as a heterologous protein, encoded by the nucleic acid contained in the genome of the viral vector particle after incubation. Numerous well-known methods for assessing the level of expression of recombinant molecules can be used, such as affinity-based methods, such as immunoaffinity-based methods (e.g., in the context of cell surface proteins), or detection by flow cytometry. In some instances, expression is measured by detection of a transduction marker and/or reporter construct. In some embodiments, a nucleic acid encoding a truncated surface protein is included within the vector and used as a marker of expression and/or enhancement thereof.

In certain embodiments, the incubation is performed under static conditions, such as without centrifugation, shaking, spinning, rocking, or perfusion of the medium, e.g., continuous or semi-continuous perfusion. In some embodiments, prior to or immediately following the initiation of the incubation, e.g., within 5, 15, or 30 minutes, the cells are transferred (e.g., transferred under sterile conditions) to a container, such as a bag or vial, and placed in an incubator. In some embodiments, the cells are transferred to the container under closed or sterile conditions. In some embodiments, the container, e.g., a vial or bag, is then placed in an incubator for all or part of the incubation. In certain embodiments, the incubator is set to, or at least about, a temperature of 16°C, 24°C, or 35°C. In some embodiments, the incubator is set at 37°C, about 37°C, or 37°C ±2°C, ±1°C, ±0.5°C, or ±0.1°C.

In some embodiments, the incubation is performed in a serum-free medium. In some embodiments, the serum-free medium is a defined and/or well-defined cell culture medium. In certain embodiments, the serum-free medium is a controlled culture medium that has been processed, e.g., filtered, to remove inhibitors and/or growth factors. In some embodiments, the serum-free medium contains proteins. In certain embodiments, the serum-free medium may contain serum albumin, hydrolysates, growth factors, hormones, carrier proteins, and/or attachment factors.

In some embodiments, the additional incubation is performed in any medium described in Section II-B. In some embodiments, the additional incubation is performed in a serum-free medium, e.g., any medium described in US-20210207080.

In some embodiments, the additional incubation is performed in the presence of one or more recombinant cytokines. In some embodiments, the additional incubation is performed in the presence of any recombinant cytokine described in Section II-B and at any concentration described in Section II-B.

In some embodiments, the additional incubation is performed in the presence of the same or similar medium as that present during the manipulation. In some embodiments, the additional incubation is performed in a medium containing the same cytokines as the medium present during the manipulation. In certain embodiments, the additional incubation is performed in a medium containing the same cytokines at the same concentration as the medium present during the manipulation.

In certain embodiments, the cells are further incubated in the absence of cytokines. In certain embodiments, the cells are further incubated in the absence of any recombinant cytokines. In certain embodiments, the cells are further incubated in the absence of recombinant IL-2, IL-7, and IL-15.

In certain embodiments, the cells are further incubated for, about, or at least for 18 hours, 24 hours, 30 hours, 36 hours, 40 hours, 48 hours, 54 hours, 60 hours, 72 hours, 84 hours, 96 hours, or more than 96 hours after introduction of a polynucleotide encoding a heterologous or recombinant protein, e.g., a viral vector. In certain embodiments, the cells are further incubated for, about, or at least for 1 day, 2 days, 3 days, 4 days, or more than 4 days after introduction of a polynucleotide encoding a heterologous or recombinant protein, e.g., a viral vector. In some embodiments, the additional incubation is performed for an amount of time prior to genetic manipulation of 30 minutes to 2 hours, 1 hour to 8 hours, 6 hours to 12 hours, 12 hours to 18 hours, 16 hours to 24 hours, 18 hours to 30 hours, 24 hours to 48 hours, 24 hours to 72 hours, 42 hours to 54 hours, 60 hours to 120 hours, 96 hours to 120 hours, 90 hours to 1 day to 7 days, 3 days to 8 days, 1 day to 3 days, 4 days to 6 days, or 4 days to 5 days. In some embodiments, the additional incubation is for or about 18 hours to 30 hours. In particular embodiments, the additional incubation is for or about 24 hours, or for or about 1 day.

In certain embodiments, the total duration of the additional incubation is about, or at least about, 12 hours, 18 hours, 24 hours, 30 hours, 36 hours, 42 hours, 48 hours, 54 hours, 60 hours, 72 hours, 84 hours, 96 hours, 108 hours, or 120 hours. In certain embodiments, the total duration of the additional incubation is about, or at least about, 1 day, 2 days, 3 days, 4 days, or 5 days. In certain embodiments, the additional incubation is completed in, at, or within 120 hours, 108 hours, 96 hours, 84 hours, 72 hours, 60 hours, 54 hours, 48 hours, 42 hours, 36 hours, 30 hours, 24 hours, 18 hours, or 12 hours. In certain embodiments, the additional incubation is completed within or about 1 day, 2 days, 3 days, 4 days, or 5 days. In some embodiments, the total duration of the additional incubation is from 12 hours to 120 hours, from 18 hours to 96 hours, from 24 hours to 72 hours, or from 24 hours to 48 hours or about the above times (inclusive). In some embodiments, the total duration of the additional incubation is from 1 hour to 48 hours, from 4 hours to 36 hours, from 8 hours to 30 hours, or from 12 hours to 24 hours or about the above times (inclusive). In certain embodiments, the additional incubation is performed for or about 24 hours, 48 hours, or 72 hours, or for or about the above times, respectively, or for 1 day, 2 days, or 3 days. In certain embodiments, the additional incubation is performed for 24 hours ± 6 hours, 48 hours ± 6 hours, or 72 hours ± 6 hours. In certain embodiments, the additional incubation is performed for 72 hours or about the time, or for 3 days or about the time.

In some embodiments, the additional incubation is completed between 24 and 120 hours, between 36 and 108 hours, between 48 and 96 hours, or between 48 and 72 hours, or about these times (inclusive), after the initiation of stimulation. In some embodiments, the additional incubation is completed within, about, or about 120 hours, 108 hours, 96 hours, 72 hours, 48 hours, or 36 hours after the initiation of stimulation. In some embodiments, the additional incubation is completed within, about, or about 5 days, 4.5 days, 4 days, 3 days, 2 days, or 1.5 days after the initiation of stimulation. In particular embodiments, the additional incubation is completed 24 hours ± 6 hours, 48 hours ± 6 hours, or 72 hours ± 6 hours after the initiation of stimulation. In some embodiments, additional incubation is completed after 72 hours or about that time or after 3 days or about that time.

1. Virus particles

In some embodiments, the nucleic acid sequence encoding the recombinant protein is contained in a viral particle. In some embodiments, the viral particle is a recombinant infectious viral particle. In some embodiments, the viral particle is a viral vector, such as a vector derived from simian virus 40 (SV40), adenovirus, or adeno-associated virus (AAV). In some embodiments, the viral particle is a recombinant lentiviral vector or retroviral vector, such as a gamma-retroviral vector (see, e.g., Koste et al., Gene Therapy (2014) doi: 10.1038/gt.2014.25; Carlens et al., Exp Hematol (2000) 28(10): 1137-46; Alonso-Camino et al. (2013) Mol Ther Nucl Acids 2, e93; Park et al., Trends Biotechnol. 2011 November 29(11): 550-557). In some embodiments, the viral particle is a recombinant lentiviral vector.

In some embodiments, the retroviral vector has a long terminal repeat (LTR) sequence (e.g., a retroviral vector derived from Moloney murine leukemia virus (MoMLV), myeloproliferative sarcoma virus (MPSV), murine embryonic stem cell virus (MESV), murine stem cell virus (MSCV), spleen focus-forming virus (SFFV), or adeno-associated virus (AAV)). Many retroviral vectors are derived from murine retroviruses. In some embodiments, the retrovirus includes those derived from any avian or mammalian cell source. The retrovirus can be amphotropic, meaning that it can infect host cells of multiple species, including humans. In one embodiment, the gene to be expressed replaces the retroviral gag, pol, and/or env sequences. A number of exemplary retroviral systems have been described (e.g., U.S. Pat. Nos. 5,219,740; 6,207,453; 5,219,740; Miller and Rosman (1989) BioTechniques 7:980-990; Miller, A. D. (1990) Human Gene Therapy 1:5-14; Scarpa et al. (1991) Virology 180:849-852; Burns et al. (1993) Proc. Natl. Acad. Sci. USA 90:8033-8037; and Boris-Lawrie and Temin (1993) Cur. Opin. Genet. Develop. 3:102-109).

The viral vector genome can be constructed in the form of a plasmid that can be transfected into a packaging or producer cell line. In some embodiments, the nucleic acid encoding the recombinant protein, such as the recombinant receptor, is inserted or positioned in a region of the viral vector, such as a nonessential region of the viral genome. In some embodiments, the nucleic acid is inserted into the viral genome in place of a specific viral sequence, resulting in a replication-defective virus.

Any of a variety of known methods can be used to produce retroviral particles having a genome containing an RNA copy of the viral vector genome. In some embodiments, creating a virus-based gene delivery system involves at least two components: first, a packaging plasmid containing the structural proteins and enzymes necessary to produce the viral vector particle, and second, the viral vector itself, e.g., the genetic material to be transferred. Biosafety safeguards may be incorporated into the design of one or both of these components.

In some embodiments, the packaging plasmid may contain all retroviral (e.g., HIV-1) proteins other than the envelope protein (Naldini et al., 1998). In other embodiments, the viral vector may lack additional viral genes, such as genes associated with virulence, such as vpr, vif, vpu, and nef, and/or Tat, the primary transcriptional activator of HIV. In some embodiments, a lentiviral vector, such as an HIV-based lentiviral vector, contains only three genes of the parental virus: gag, pol, and rev, which reduces or eliminates the possibility of reconstitution of a wild-type virus through recombination.

In some embodiments, the viral vector genome is introduced into a packaging cell line that contains all the components necessary to package the viral genomic RNA transcribed from the viral vector genome into a viral particle. Alternatively, the viral vector genome may contain one or more genes encoding viral components in addition to one or more sequences of interest, e.g., a recombinant nucleic acid. However, in some aspects, to prevent genome replication in target cells, endogenous viral genes required for replication are removed and provided separately in the packaging cell line.

In some embodiments, the packaging cell line is transfected with one or more plasmid vectors containing the components necessary for particle production. In some embodiments, the packaging cell line is transfected with a plasmid containing a viral vector genome comprising an LTR, a cis-acting packaging sequence, and a nucleic acid encoding a sequence of interest, i.e., an antigen receptor, such as a CAR; and one or more helper plasmids encoding viral enzymatic and/or structural components (e.g., Gag, pol, and/or rev). In some embodiments, multiple vectors are utilized to separate the various genetic components that produce the retroviral vector particle. In some such embodiments, providing separate vectors to the packaging cell reduces the likelihood of recombination events that could otherwise produce replication competent viruses. In some embodiments, a single plasmid vector containing all retroviral components can be used.

In some embodiments, retroviral vector particles, such as lentiviral vector particles, are pseudotyped to increase the transduction efficiency of host cells. For example, retroviral vector particles, such as lentiviral vector particles, in some embodiments are pseudotyped with the VSV-G glycoprotein, which provides a broad cellular host range that extends the cell types that can be transduced. In some embodiments, the packaging cell line is transfected with a plasmid or polynucleotide encoding a non-native envelope glycoprotein (including, for example, a heterologous, polytropic, or amphotropic envelope), such as the Sindbis virus envelope, GALV, or VSV-G.

In some embodiments, the packaging cell line provides components, including viral regulatory and structural proteins, necessary for packaging viral genomic RNA into lentiviral vector particles. In some embodiments, the packaging cell line can be any cell line capable of expressing lentiviral proteins and producing functional lentiviral vector particles. In some aspects, suitable packaging cell lines include 293 (ATCC CCL X), 293T, HeLA (ATCC CCL 2), D17 (ATCC CCL 183), MDCK (ATCC CCL 34), BHK (ATCC CCL-10), and Cf2Th (ATCC CRL 1430) cells.

In some embodiments, the packaging cell line stably expresses viral protein(s). For example, in some aspects, a packaging cell line can be constructed that contains gag, pol, rev, and/or other structural genes, but lacks LTRs and packaging components. In some embodiments, the packaging cell line can be transiently transfected with a nucleic acid molecule encoding one or more viral proteins, together with a viral vector genome containing a nucleic acid molecule encoding a heterologous protein and/or a nucleic acid molecule encoding an envelope glycoprotein.

In some embodiments, the viral vector and packaging and/or helper plasmids are introduced into a packaging cell line via transfection or infection. The packaging cell line can produce viral vector particles containing the viral vector genome. Methods for transfection or infection are well known. Examples include calcium phosphate, DEAE-dextran, and lipofection methods, electroporation, and microinjection.

When a recombinant plasmid and a retroviral LTR and packaging sequence are introduced into a specialized cell line (e.g., by calcium phosphate precipitation), the packaging sequence can allow RNA transcripts of the recombinant plasmid to be packaged into viral particles, which can then be secreted into the culture medium. In some embodiments, the medium containing the recombinant retrovirus is then collected, optionally concentrated, and used for gene transfer. For example, in some aspects, after co-transfection of the packaging plasmid and the transfer vector into a packaging cell line, the viral vector particles are recovered from the culture medium and titered using standard methods used by those skilled in the art.

In some embodiments, retroviral vectors, such as lentiviral vectors, can be produced in a packaging cell line, such as the exemplary HEK 293T cell line, by introducing a plasmid to allow for the production of lentiviral particles. In some embodiments, the packaging cell is transfected with and/or contains polynucleotides encoding gag and pol, and a polynucleotide encoding a recombinant receptor, such as an antigen receptor, e.g., a CAR. In some embodiments, the packaging cell line is optionally and/or additionally transfected with and/or contains a polynucleotide encoding a rev protein. In some embodiments, the packaging cell line is optionally and/or additionally transfected with and/or contains a polynucleotide encoding a non-native envelope glycoprotein, such as VSV-G. In some such embodiments, approximately 2 days after transfection of cells, such as HEK 293T cells, the cell supernatant contains the recombinant lentiviral vector, which can be recovered and titered.

The recovered and/or produced retroviral vector particles can be used to transduce target cells using the methods described. Upon reaching the target cell, the viral RNA is reverse transcribed and exported to the nucleus, where it becomes stably integrated into the host genome. One or two days after viral RNA integration, expression of a recombinant protein, such as an antigen receptor such as a CAR, can be detected.

In some embodiments, the vector is a viral vector, such as a retroviral vector. In some embodiments, the polynucleotide encoding the recombinant receptor and/or additional polypeptide(s) is introduced into the cell via a retroviral or lentiviral vector or via a transposon (see, e.g., Baum et al. (2006) Molecular Therapy: The Journal of the American Society of Gene Therapy. 13:1050-1063; Frecha et al. (2010) Molecular Therapy 18:1748-1757; and Hackett et al. (2010) Molecular Therapy 18:674-683).

In some embodiments, one or more polynucleotide(s) or vector(s) encoding the recombinant receptor and/or additional polypeptide(s) are introduced into the cells, e.g., T cells, prior to isolation, cultivation, and/or expansion. Such introduction of the polynucleotide(s) or vector(s) can be accomplished using any suitable retroviral vector. In some embodiments, after manipulation, the resulting genetically engineered cells can be released from the initial stimulation (e.g., anti-CD3/anti-CD28 stimulation) and subsequently stimulated in the presence of a second type of stimulation (e.g., via the de novo introduced recombinant receptor). This second type of stimulation can include antigen stimulation in the form of peptide/MHC molecules, a cognate (cross-linking) ligand of the genetically engineered receptor (e.g., a native antigen and/or a ligand of a CAR), or any ligand (e.g., an antibody) that binds directly within the framework of the new receptor (e.g., by recognizing a constant region within the receptor). See, for example, Cheadle et al., "Chimeric antigen receptors for T-cell based therapy" Methods Mol Biol. 2012; 907:645-66 or Barrett et al., Chimeric Antigen Receptor Therapy for Cancer Annual Review of Medicine Vol. 65: 333-347 (2014).

In some cases, vectors may be used that do not require activation of cells, such as T cells. In some such examples, cells may be selected and/or transduced prior to activation.

In some embodiments, a polynucleotide encoding a recombinant receptor contains at least one promoter operably linked to control the expression of the recombinant receptor. In some examples, the polynucleotide contains two, three, or more promoters operably linked to control the expression of the recombinant receptor. In some embodiments, the polynucleotide may contain regulatory sequences, such as transcription and translation initiation and termination codons, which are specific to the type of host into which the polynucleotide is to be introduced (e.g., bacteria, fungi, plants, or animals), taking into account whether the polynucleotide is DNA- or RNA-based. In some embodiments, the polynucleotide may contain regulatory/control elements, such as a promoter, an enhancer, an intron, a polyadenylation signal, a Kozak consensus sequence, an internal ribosome entry site (IRES), a 2A sequence, and a splice acceptor or donor. In some embodiments, the polynucleotide may contain a non-natural promoter operably linked to a nucleotide sequence encoding a recombinant receptor and/or one or more additional polypeptide(s). In some embodiments, the promoter is selected from an RNA pol I, pol II, or pol III promoter. In some embodiments, the promoter is recognized by RNA polymerase II (e.g., the CMV, SV40 early region, or adenovirus major late promoter). In another embodiment, the promoter is recognized by RNA polymerase III (e.g., the U6 or H1 promoter). In some embodiments, the promoter may be a non-viral promoter or a viral promoter, such as the cytomegalovirus (CMV) promoter, the SV40 promoter, the RSV promoter, and the promoter found in the long terminal repeat of murine stem cell virus. Other known promoters are also contemplated.

In some embodiments, the promoter is or comprises a constitutive promoter. Exemplary constitutive promoters include the simian virus 40 early promoter (SV40), the cytomegalovirus immediate-early promoter (CMV), the human ubiquitin C promoter (UBC), the human elongation factor 1α promoter (EF1α), the mouse phosphoglycerate kinase 1 promoter (PGK), and the chicken β-actin promoter coupled with the CMV early enhancer (CAGG). In some embodiments, the constitutive promoter is a synthetic or modified promoter. In some embodiments, the promoter is or comprises the MND promoter, which is a synthetic promoter containing the U3 region of the modified MoMuLV LTR together with the myeloproliferative sarcoma virus enhancer (see Challita et al. (1995) J. Virol. 69(2):748-755). In some embodiments, the promoter is a tissue-specific promoter. In another embodiment, the promoter is a viral promoter. In another embodiment, the promoter is a non-viral promoter. In some embodiments, exemplary promoters include the human elongation factor 1 alpha (EF1α) promoter or a modified form thereof, or the MND promoter.

In another embodiment, the promoter is a regulated promoter (e.g., an inducible promoter). In some embodiments, the promoter is an inducible promoter or an inhibitory promoter. In some embodiments, the promoter comprises or is an analog of a Lac operator sequence, a tetracycline operator sequence, a galactose operator sequence, or a doxycycline operator sequence, or can be bound to or recognized by a Lac repressor or a tetracycline repressor or an analog thereof. In some embodiments, the polynucleotide does not comprise a regulatory element, e.g., a promoter.

In some cases, the nucleic acid sequence encoding the recombinant receptor contains a signal sequence encoding a signal peptide. In some aspects, the signal sequence may encode a signal peptide derived from a natural polypeptide. In other aspects, the signal sequence may encode a heterologous or non-natural signal peptide. In some cases, the nucleic acid sequence encoding the recombinant receptor, e.g., a chimeric antigen receptor (CAR), contains a signal sequence encoding a signal peptide. Exemplary signal peptides include, for example, the GMCSFR alpha chain signal peptide or the CD8 alpha signal peptide.

In some embodiments, the polynucleotide contains a nucleic acid sequence encoding one or more additional polypeptides, e.g., one or more tag(s) and/or one or more effector molecules. In some embodiments, the one or more tag(s) comprises a transduction tag, a surrogate tag, and/or a resistance tag or a selection tag. For example, the introduced additional nucleic acid sequence encoding one or more additional polypeptide(s) includes a nucleic acid sequence that can improve the efficacy of the therapy (e.g., by promoting the viability and/or function of transferred cells); a nucleic acid sequence that provides a genetic tag for selection and/or evaluation of cells (e.g., to assess survival or localization in vivo); a nucleic acid sequence that improves safety (e.g., by sensitizing cells to negative selection in vivo as described in Lupton S. D. et al., Mol. and Cell Biol., 11:6 (1991); and Riddell et al., Human Gene Therapy 3:319-338 (1992)); See also WO 1992008796 and WO 1994028143, and U.S. Patent No. 6,040,177, which describe the use of bifunctional selectable fusion genes derived from fusing a dominant positive selectable tag with a negative selectable tag.

In some embodiments, the tag is a transduction tag or a surrogate tag. The transduction tag or surrogate tag can be used to detect cells that have been introduced with a polynucleotide, e.g., a polynucleotide encoding a recombinant receptor. In some embodiments, the transduction tag can indicate or identify a cell modification. In some embodiments, the surrogate tag is a protein engineered to be co-expressed on the cell surface with the recombinant receptor, e.g., a CAR. In certain embodiments, such a surrogate tag is a surface protein that has been modified to have little or no activity. In certain embodiments, the surrogate tag is encoded on the same polynucleotide encoding the recombinant receptor. In some embodiments, the nucleic acid sequence encoding the recombinant receptor is operably linked to a nucleic acid sequence encoding a tag, optionally separated by an internal ribosome entry site (IRES), or a nucleic acid encoding a self-cleaving peptide or a peptide that induces ribosome skipping, such as a 2A sequence. The exogenous tag gene can, in some cases, be utilized to allow for the detection or selection of cells in relation to the engineered cells, and in some cases, can also be utilized to promote cell removal and/or apoptosis.

Exemplary surrogate tags can include truncated forms of cell surface polypeptides, such as truncated forms that are non-functional and do not or cannot transmit a signal or signals normally transmitted by the full-length form of the cell surface polypeptide, and/or truncated forms that do not or cannot be internalized. Exemplary truncated cell surface polypeptides include truncated forms of growth factors or other receptors, such as truncated human epidermal growth factor receptor 2 (tHER2), truncated epidermal growth factor receptor (tEGFR), or prostate-specific membrane antigen (PSMA) or modified forms thereof, such as truncated PSMA (tPSMA). In some aspects, the tEGFR may contain an epitope recognized by the antibody cetuximab (Erbitux®) or other therapeutic anti-EGFR antibodies or binding molecules, which may be used to identify or select cells engineered with the tEGFR construct and the encoded exogenous protein and/or to eliminate or isolate cells expressing the encoded exogenous protein. See U.S. Patent No. 8,802,374 and Liu et al., Nature Biotech. 2016 April; 34(4): 430-434. In some aspects, the tag, e.g., a surrogate tag, comprises all or a portion of CD34, NGFR, CD19, or a truncated CD19, e.g., a truncated form of a truncated non-human CD19.

In some embodiments, the tag is or comprises a detectable protein, such as a fluorescent protein, such as green fluorescent protein (GFP), enhanced green fluorescent protein (EGFP), such as super-fold GFP (sfGFP), red fluorescent protein (RFP), such as tdTomato, mCherry, mStrawberry, AsRed2, DsRed or DsRed2, cyan fluorescent protein (CFP), blue green fluorescent protein (BFP), enhanced blue fluorescent protein (EBFP) and yellow fluorescent protein ( YFP ), and variants thereof (including species variants, monomeric variants, codon-optimized, stabilized and/or enhanced variants of fluorescent proteins). In some embodiments, the tag is or comprises an enzyme, such as luciferase, the lacZ gene from E. coli, alkaline phosphatase, secreted embryonic alkaline phosphatase (SEAP), chloramphenicol acetyl transferase (CAT). Exemplary luminescent reporter genes include luciferase (luc), β-galactosidase, chloramphenicol acetyltransferase (CAT), β-glucuronidase (GUS), or variants thereof. In some aspects, expression of the enzyme can be detected by the addition of a substrate that can detect expression and functional activity of the enzyme.

In some embodiments, the tag is a resistance generator or a selection tag. In some embodiments, the resistance tag or selection tag is or comprises a polypeptide that confers resistance to an exogenous agent or drug. In some embodiments, the resistance tag or selection tag is an antibiotic resistance gene. In some embodiments, the resistance tag or selection tag is an antibiotic resistance gene that confers antibiotic resistance to a mammalian cell. In some embodiments, the resistance tag or selection tag is or comprises a puromycin resistance gene, a hygromycin resistance gene, a blasticidin resistance gene, a neomycin resistance gene, a geneticin resistance gene, or a zeocin resistance gene, or a modified form thereof.

Any of the recombinant receptors and/or additional polypeptide(s) described herein can be encoded by one or more polynucleotides containing one or more nucleic acid sequences encoding the recombinant receptor, in any combination, orientation, or arrangement. For example, one, two, three, or more polynucleotides can encode one, two, three, or more different polypeptides, e.g., the recombinant receptor or portions or components thereof, and/or one or more additional polypeptide(s), e.g., tags and/or effector molecules. In some embodiments, one polynucleotide contains a nucleic acid sequence encoding a recombinant receptor, e.g., a CAR, or portions or components thereof, and a nucleic acid sequence encoding one or more additional polypeptide(s). In some embodiments, one vector or construct contains a nucleic acid sequence encoding a recombinant receptor, e.g., a CAR, or portions or components thereof, and a separate vector or construct contains a nucleic acid sequence encoding one or more additional polypeptide(s). In some embodiments, the nucleic acid sequence encoding the recombinant receptor and the nucleic acid sequence encoding one or more additional polypeptide(s) are operably linked to two different promoters. In some embodiments, the nucleic acid encoding the recombinant receptor is upstream of the nucleic acid encoding the one or more additional polypeptide(s). In some embodiments, the nucleic acid encoding the recombinant receptor is downstream of the nucleic acid encoding the one or more additional polypeptide(s).

In certain instances, a single polynucleotide contains nucleic acid sequences encoding two or more different polypeptide chains, e.g., a recombinant receptor, and one or more additional polypeptide(s), e.g., a tag and/or an effector molecule. In some embodiments, the nucleic acid sequences encoding two or more different polypeptide chains, e.g., a recombinant receptor, and one or more additional polypeptide(s), are present in two separate polynucleotides. For example, two separate polynucleotides are provided, each of which can be individually transferred or introduced into a cell for expression in the cell. In some embodiments, the nucleic acid sequence encoding the tag and the nucleic acid sequence encoding the recombinant receptor are present in or inserted at different locations within the cellular genome. In some embodiments, the nucleic acid sequence encoding the tag and the nucleic acid sequence encoding the recombinant receptor are operably linked to two different promoters.

In some embodiments, such as when a polynucleotide comprises a first and a second nucleic acid sequence, the coding sequences encoding each of the different polypeptide chains may be operably linked to promoters, which may be the same or different. In some embodiments, the nucleic acid molecule may contain promoters that drive the expression of two or more different polypeptide chains. In some embodiments, such nucleic acid molecules may be multicistronic (bicistronic or tricistronic, see, e.g., U.S. Patent No. 6,060,273). In some embodiments, the nucleic acid sequence encoding the recombinant receptor and the nucleic acid sequence encoding one or more additional polypeptide(s) are operably linked to the same promoter and optionally separated by a nucleic acid encoding an internal ribosome entry site (IRES), or a self-cleaving peptide or a peptide that induces ribosome skipping, such as a 2A element. For example, the exemplary tag, and optionally the ribosome skipping sequence sequence, can be any of those disclosed in PCT Publication No. WO2014031687.

In some embodiments, the transcription unit can be engineered as a bicistronic unit containing an IRES, which allows co-expression of gene products (e.g., encoding a recombinant receptor and an additional polypeptide) driven by a message from a single promoter. Alternatively, in some cases, a single promoter can direct expression of RNA containing two or three genes (e.g., a gene encoding a tag and a gene encoding a recombinant receptor) separated from each other by sequences encoding a self-cleaving peptide (e.g., a 2A sequence) or a protease recognition site (e.g., furin) within a single open reading frame (ORF). The ORF thus encodes a single polypeptide, which is processed into individual proteins either during translation (in the case of 2A) or post-translationally. In some cases, a peptide, such as T2A, can cause the ribosome to skip peptide bond synthesis at the C-terminus of the 2A element (ribosome skipping), resulting in dissociation between the end of the 2A sequence and the next peptide downstream (see, e.g., de Felipe, Genetic Vaccines and Ther. 2:13 (2004) and de Felipe et al. Traffic 5:616-626 (2004)). Various 2A elements are known. Examples of 2A sequences that can be used in the methods and systems disclosed herein include 2A sequences from foot-and-mouth disease virus (F2A), equine rhinitis A virus (E2A), Thosea asigna virus (T2A), and porcine tescovirus-1 (P2A) (e.g., as described in U.S. Patent Publication No. 20070116690).

In some embodiments, the polynucleotide encoding the recombinant receptor and/or additional polypeptide may be contained in a vector or cloned into one or more vector(s). In some embodiments, one or more vector(s) may be used to transform or transfect a host cell, e.g., a cell for manipulation. Exemplary vectors include vectors designed for introduction, propagation, and expansion, or for expression, or both, such as plasmids and viral vectors. In some aspects, the vector is an expression vector, e.g., a recombinant expression vector. In some embodiments, the recombinant expression vector may be prepared using standard recombinant DNA techniques.

In some embodiments, the vector can be a vector from the pUC series (Fermentas Life Sciences), the pBluescript series (Stratagene, La Jolla, CA), the pET series (Novagen, Madison, WI), the pGEX series (Pharmacia Biotech, Uppsala, Sweden), or the pEX series (Clontech, Palo Alto, CA). In some cases, bacteriophage vectors such as λG10, λGT11, λZapII (Stratagene), λEMBL4, and λNM1149 can also be used. In some embodiments, plant expression vectors can be used, including pBI01, pBI101.2, pBI101.3, pBI121, and pBIN19 (Clontech). In some embodiments, the animal expression vectors include pEUK-Cl, pMAM, and pMAMneo (Clontech).

2. Chimeric antigen receptor

In some embodiments, the recombinant protein is a recombinant receptor. In some embodiments, the recombinant protein is a chimeric antigen receptor (CAR). In some embodiments, a chimeric receptor, such as a chimeric antigen receptor, contains one or more domains that combine a ligand-binding domain (e.g., an antibody or antibody fragment) that provides specificity for a desired antigen (e.g., a tumor antigen) with an intracellular signaling domain. In some embodiments, the intracellular signaling domain is a portion of an activating intracellular domain that provides a primary activation signal, such as a T cell activation domain. In some embodiments, the intracellular signaling domain contains or additionally contains a costimulatory signaling domain to promote effector function. In some embodiments, the chimeric receptor, when genetically engineered into an immune cell, can modulate T cell activity, and in some cases, modulate T cell differentiation or homeostasis, thereby generating genetically engineered cells with improved in vivo lifespan, survival, and/or persistence, such as for use in adoptive cell therapy methods.

Exemplary antigen receptors including CARs and methods for manipulating and introducing such receptors into cells are described, for example, in WO-200014257, WO-2013126726, WO-2012129514, WO-2014031687, WO-2013166321, WO-2013071154, WO-2013123061, WO-20160046724, WO-2016014789, WO-2016090320, WO-2016094304, WO-2017025038, WO-2017173256, US-2002131960, US-2013287748, US-20130149337, US-6451995, US-7446190, US-8252592, US-8339645, US-8398282, US-7446179, US-6410319, US-7070995, US-7265209, US-7354762, US-7446191, US-8324353, US-8479118, US-9765342, EP-2537416, Sadelain et al., Cancer Discov. (2013); 3(4): 388-398]; [Davila et al. (2013) PLoS ONE 8(4): e61338]; [Turtle et al., Curr. Opin. Immunol., October 2012; 24(5): 633-39]; And [Wu et al., Cancer, 2012 March 18(2): 160-75]. In some aspects, the antigen receptor comprises a CAR as described in U.S. Patent No.: 7,446,190 and those described in International Patent Application Publication No.: WO/2014055668 A1. Examples of CARs include any of the publications mentioned above, such as WO2014031687, US 8,339,645, US 7,446,179, US 2013/0149337, U.S. Patent No.: 7,446,190, U.S. Patent No.: 8,389,282, Kochenderfer et al., 2013, Nature Reviews Clinical Oncology, 10, 267-276 (2013)]; [Wang et al. (2012) J. Immunother. 35(9): 689-701]; and [Brentjens et al., Sci Transl Med. 2013 5(177)]. See also WO2014031687, US 8,339,645, US 7,446,179, US 2013/0149337, US Patent No. 7,446,190, and US Patent No. 8,389,282.

Exemplary antigen receptors, such as CARs, are also described in the literature [Marofi et al., Stem Cell Res Ther 12: 81 (2021)]; [Townsend et al., J Exp Clin Cancer Res 37: 163 (2018)]; [Ma et al., Int J Biol Sci 15(12): 2548-2560 (2019)]; [Zhao and Cao, Front Immunol 10: 2250 (2019)]; [Han et al., J Cancer 12(2): 326-334 (2021)]; [Specht et al., Cancer Res 79: 4 Supplement, Abstract P2-09-13]; [Byers et al., Journal of Clinical Oncology 37, no. 15_suppl (2019)]; [Panowski et al., Cancer Res 79 (13 Supplement) 2326 (2019)]; and [Sauer et al., Blood 134 (Supplement_1): 1932 (2019)]; or any of the antibodies or antigen-binding fragments described in U.S. Patent Nos. 8,153,765; 8,603477, 8,008,450; U.S. Publication No. US20120189622 or US20100260748; and International PCT Publication Nos. WO2006099875, WO2009080829, WO2012092612, WO2014210064.

Additional exemplary antigen receptors, e.g., CARs, such as anti-BCMA CARs, include idecaptagene bicleucel, Avecma©, BCMA02, JCARH125, JNJ-68284528 (LCAR-B38M; siltacaptagene autoleucel; CARBYKTI™) (Janssen/Legend), P-BCMA-101 (Poseida), PBCAR269A (Poseida), P-BCMA-Allo1 (Poseida), Allo-715 (Pfizer/Allogene), CT053 (Carsgen), Descartes-08 (Cartesian), PHE885 (Novartis), ARI-002 (Barcelona Clinic Hospital Clinic Barcelona), IDIBAPS), and CTX120 (CRISPR Therapeutics). In certain embodiments, the CAR is a CAR of idecaptagene bicleucel cells. In certain embodiments, the CAR is a CAR of ABECMA® cells (cells used in ABECMA® immunotherapy). In certain embodiments, the CAR is a CAR of siltacaptagene autoleucel cells. In certain embodiments, the CAR is a CAR of Cavicti™ cells (cells used in Cavicti™ immunotherapy).

Exemplary antigen receptors, e.g., CARs, also include the FDA-approved products BREYANZI® (lithocaptagene maraleucel), TECARTUS™ (brexucaptagene autoleucel), KYMRIAH™ (tisagenleucel), and CARs from YESCARTA™ (axicaptagene ciloleucel), AVECMA® (idecaptagene bicleucel), and CAVICTI™ (siltacaptagene autoleucel). In some of any of the provided embodiments, the CAR is a CAR from BREYANZI® (lithocaptagene maraleucel), TECARTUS™ (brexucaptagene autoleucel), KYMRIAH™ (tisagenleucel), KYMRIAH™ (axicaptagene ciloleucel), AVECMA® (idecaptagene bicleucel), or CAVICTI™ (siltacaptagene autoleucel). In some of any of the provided embodiments, the CAR is a CAR of Breyanzi® (lithocaptagene maraleucel, see Sehgal et al., 2020, Journal of Clinical Oncology 38:15_suppl, 8040; Teoh et al., 2019, Blood 134(Supplement_1):593; and Abramson et al., 2020, The Lancet 396(10254): 839-852). In some of any of the provided embodiments, the CAR is a CAR of Tecartus™ (brexucaptagene autoleucel, see Mian and Hill, 2021, Expert Opin Biol Ther; 21(4):435-441; and Wang et al., 2021, Blood 138(Supplement 1):744). In some of any of the provided embodiments, the CAR is the CAR of Kymriah™ (tisagenlecleucel, see Bishop et al., 2022, N Engl J Med 386:629:639; Schuster et al., 2019, N Engl J Med 380:45-56; Halford et al., 2021, Ann Pharmacother 55(4):466-479; Mueller et al., 2021, Blood Adv. 5(23):4980-4991; and Fowler et al., 2022, Nature Medicine 28:325-332). In some of any of the provided embodiments, the CAR is a CAR of Yescarta™ (axicaptagene ciloleucel, see Neelapu et al., 2017, N Engl J Med 377(26):2531-2544; Jacobson et al., 2021, The Lancet 23(1):P91-103; and Locke et al., 2022, N Engl J Med 386:640-654). In some of any of the provided embodiments, the CAR is a CAR of Avekma® (idecaptagene bicleucel, see Raje et al., 2019, N Engl J Med 380:1726-1737; and Munshi et al., 2021, N Engl J Med 384:705-716). In some of the embodiments provided, the CAR is a CAR of Cavicti™ (siltacaptagene autoleucel, see Berdeja et al., Lancet. 2021 Jul 24;398(10297):314-324; and Martin, Abstract #549 [Oral Presentation], Presented at the 2021 American Society of Hematology (ASH) Annual Meeting & Exposition).

In some embodiments, the antigen targeted by the receptor is a polypeptide. In some embodiments, it is a carbohydrate or other molecule. In some embodiments, the antigen is selectively expressed or overexpressed on cells of a disease or condition, such as tumor or pathogenic cells, compared to normal or non-targeted cells or tissues. In other embodiments, the antigen is expressed on normal cells and/or manipulated cells.

In some embodiments, the antigen is αvβ6 integrin (avb6 integrin), B cell maturation antigen (BCMA), B7-H3, B7-H6, carbonic anhydrase 9 (CA9, also known as CAIX or G250), cancer-testis antigen, cancer/testis antigen 1B (CTAG, also known as NY-ESO-1 and LAGE-2), carcinoembryonic antigen (CEA), cyclin, cyclin A2, C-C motif chemokine ligand 1 (CCL-1), CD19, CD20, CD22, CD23, CD24, CD30, CD33, CD38, CD44, CD44v6, CD44v7/8, CD123, CD133, CD138, CD171, chondroitin sulfate proteoglycan 4 (CSPG4), epidermal growth factor protein (EGFR), type III epidermal growth factor receptor mutant (EGFR vIII), epithelial glycoprotein 2 (EPG-2), epithelial glycoprotein 40 (EPG-40), ephrin B2, ephrin receptor A2 (EPHa2), estrogen receptor, Fc receptor-like 5 (FCRL5; also known as Fc receptor homolog 5 or FCRH5), fetal acetylcholine receptor (fetal AchR), folate binding protein (FBP), folate receptor alpha, ganglioside GD2, O-acetylated GD2 (OGD2), ganglioside GD3, glycoprotein 100 (gp100), glypican-3 (GPC3), G protein-coupled receptor 5D (GPRC5D), Her2/neu (receptor tyrosine kinase erb-B2), Her3 (erb-B3), Her4 (erb-B4), erbB dimer, human high-molecular-weight-melanoma-associated antigen (HMW-MAA), hepatitis B surface antigen, human leukocyte antigen A1 (HLA-A1), human leukocyte antigen A2 (HLA-A2), IL-22 receptor alpha (IL-22Rα), IL-13 receptor alpha 2 (IL-13Rα2), kinase insert domain receptor (kdr), kappa light chain, L1 cell adhesion molecule (L1-CAM), CE7 epitope of L1-CAM, leucine-rich repeat containing 8 family member A (LRRC8A), Lewis Y, melanoma-associated antigen (MAGE)-A1, MAGE-A3, MAGE-A6, MAGE-A10, mesothelin (MSLN), c-Met, murine cytomegalovirus (CMV), mucin 1 (MUC1), MUC16, natural killer cell group 2 member D (NKG2D) ligand, melan A (MART-1), neural cell adhesion molecule (NCAM), oncofetal antigen, preferentially expressed antigen of melanoma (PRAME), progesterone receptor, prostate specific antigen, prostate stem cell antigen (PSCA), prostate specific membrane antigen (PSMA), receptor tyrosine kinase-like orphan receptor 1 (ROR1), survivin, trophoblast glycoprotein (TPBG, also known as 5T4), tumor-associated glycoprotein 72 (TAG72), tyrosinase-related protein 1 (TRP1, also known as TYRP1 or gp75), tyrosinase-related protein 2 (TRP2, also known as dopachrome tautomerase, dopachrome delta-isomerase or DCT), vascular endothelial growth factor receptor (VEGFR), vascular endothelial growth factor receptor 2 (VEGFR2), Wilms' tumor 1 (WT-1), a pathogen-specific or pathogen-expressed antigen, or an antigen associated with a universal tag, and/or a biotinylated molecule, and/or A molecule expressed by HIV, HCV, HBV, or another pathogen, or comprises a molecule. In some embodiments, the antigen targeted by the receptor comprises an antigen associated with a B cell malignancy, such as any of a number of known B cell markers. In some embodiments, the antigen is or comprises CD20, CD19, CD22, ROR1, CD45, CD21, CD5, CD33, Ig kappa, Ig lambda, CD79a, CD79b, or CD30. In some embodiments, the antigen is or comprises a pathogen-specific or pathogen-expressed antigen. In some embodiments, the antigen is a viral antigen (e.g., a viral antigen from HIV, HCV, HBV, etc.), a bacterial antigen, and/or a parasitic antigen.

Chimeric receptors, such as CARs, typically comprise an extracellular antigen binding domain, such as a portion of an antibody molecule, typically the variable heavy (VH) region and/or the variable light (VL) region of an antibody, e.g., an scFv antibody fragment.

The term “antibody” herein is used in the broadest sense and includes polyclonal and monoclonal antibodies (including intact antibodies and functional (antigen-binding) antibody fragments), including fragments of antigen binding (Fab) fragments, F(ab') _fragments , Fab' fragments, Fv fragments, recombinant IgG (rIgG) fragments, heavy chain variable (VH) regions capable of specifically binding to an antigen, single-chain antibody fragments (including single-chain variable fragments (scFv)), and single-domain antibodies (e.g., sdAb, sdFv, nanobodies). The term encompasses genetically engineered and/or otherwise modified forms of immunoglobulins, such as intrabodies, peptibodies, chimeric antibodies, fully human antibodies, humanized antibodies and heteroconjugate antibodies, multispecific (e.g. bispecific or trispecific) antibodies, diabodies, triabodies and tetrabodies, tandem di-scFv, tandem tri-scFv. Unless otherwise stated, the term "antibody" should be understood to include functional antibody fragments thereof, also referred to herein as "antigen-binding fragments". The term also encompasses intact or full-length antibodies, including antibodies of any class or subclass, including IgG and its subclasses, IgM, IgE, IgA and IgD.

The terms "complementarity determining region" and "CDR" (synonymous with "hypervariable region" or "HVR") are known in the art to refer to noncontiguous amino acid sequences within the variable region of an antibody, which confer antigen specificity and/or binding affinity. Typically, each heavy chain variable region has three CDRs (CDR-H1, CDR-H2, CDR-H3), and each light chain variable region has three CDRs (CDR-L1, CDR-L2, CDR-L3). "Framework region" and "FR" are known in the art to refer to the non-CDR portions of the variable regions of the heavy and light chains. Typically, each full-length heavy chain variable region has four FRs (FR-H1, FR-H2, FR-H3, and FR-H4), and each full-length light chain variable region has four FRs (FR-L1, FR-L2, FR-L3, and FR-L4).

The precise amino acid sequence boundaries of a given CDR or FR are determined by references such as Kabat et al. (1991), "Sequences of Proteins of Immunological Interest," 5th Ed. Public Health Service, National Institutes of Health, Bethesda, MD ("Kabat" numbering system); Al-Lazikani et al., (1997) JMB 273,927-948 ("Chotia" numbering system); MacCallum et al., J. Mol. Biol. 262:732-745 (1996), "Antibody-antigen interactions: Contact analysis and binding site topography," J. Mol. Biol. 262, 732-745."] ("Contact" numbering scheme); Lefranc MP et al., "IMGT unique numbering for immunoglobulin and T cell receptor variable domains and Ig superfamily V-like domains," Dev Comp Immunol, 2003 Jan;27(1):55-77] ("IMGT" numbering scheme); Honegger A and Plueckthun A, "Yet another numbering scheme for immunoglobulin variable domains: an automatic modeling and analysis tool," J Mol Biol, 2001 Jun 8;309(3):657-70] ("Aho" numbering scheme); and Martin et al., "Modeling antibody hypervariable loops: a combined algorithm," PNAS, 1989, 86(23):9268-9272] ("AbM" numbering scheme). It can be easily determined using any of the many well-known systems.

The boundaries of a given CDR or FR can vary depending on the system used for identification. For example, the Kabat system is based on structural alignment, whereas the Chothia system is based on structural information. Numbering for both the Kabat and Chothia systems is based on the length of the most common antibody region sequence, with insertions designated by insertion characters, such as "30a," and deletions occurring in some antibodies. The two systems assign specific insertions and deletions ("indels") to different positions, resulting in differential numbering. The Contact system is based on complex crystal structure analysis and is similar in many respects to the Chothia numbering system. The AbM system is a compromise between the Kabat and Chothia definitions, based on that used by Oxford Molecular's AbM antibody modeling software.

Table 1 below lists exemplary position boundaries of CDR-L1, CDR-L2, CDR-L3, and CDR-H1, CDR-H2, CDR-H3 as identified by the Kabat, Chothia, AbM, and Contact systems, respectively. For CDR-H1, residue numbering is listed using both the Kabat and Chothia numbering systems. FRs are positioned between CDRs, e.g., FR-L1 is positioned before CDR-L1, FR-L2 is positioned between CDR-L1 and CDR-L2, FR-L3 is positioned between CDR-L2 and CDR-L3, etc. Note that since the indicated Kabat numbering system places the insertion at H35A and H35B, the terminus of the Chothia CDR-H1 loop when numbered using the indicated Kabat numbering notation varies between H32 and H34 depending on the loop length.

Table 1. CDR boundaries according to various numbering systems.

Therefore, unless otherwise specified, a "CDR" or "complementarity determining region" of a given antibody or region thereof, such as a variable region thereof, or an individual specified CDR (e.g., CDR-H1, CDR-H2, CDR-H3) should be understood to comprise (or be specific for) a complementarity determining region as defined by any of the aforementioned systems or other known systems. For example, if a particular CDR (e.g., CDR-H3) is stated to contain the amino acid sequence of a corresponding CDR in a given VH or VL region amino acid sequence, such CDR is understood to have the sequence of the corresponding CDR (e.g., CDR-H3) within the variable region as defined by any of the aforementioned systems or other known systems. In some embodiments, specific CDR sequences are specified. Although exemplary CDR sequences of antibodies are described using various numbering systems, it is understood that antibodies may include CDRs described according to any of the other aforementioned numbering systems or other numbering systems known to those of ordinary skill in the art.

Likewise, unless otherwise specified, it should be understood that a given antibody or region thereof, such as a FR or individually specified FR(s) (e.g., FR-H1, FR-H2, FR-H3, FR-H4) of a variable region thereof, comprises (or is specific to) a framework region as defined by any of the known systems. In some instances, a particular CDR, FR, or system for the identification of FRs or CDRs is specified, e.g., a CDR is as defined by the Kabat, Chothia, AbM or Contact method, or other known systems. In other cases, the particular amino acid sequence of a CDR or FR is given.

The term "variable region" or "variable domain" refers to the domain of an antibody heavy or light chain that is involved in binding the antibody to an antigen. The variable regions of the heavy and light chains of native antibodies (V _H and V _L , respectively) generally have similar structures, each domain comprising four conserved framework regions (FRs) and three CDRs (see, e.g., Kindt et al., Kuby, Immunology, 6th ed., WH Freeman and Co., page 91 (2007)). A single V _H or V _L domain may be sufficient to confer antigen-binding specificity. Furthermore, antibodies that bind to a particular antigen can be isolated using the V _H or V _L domain from an antibody that binds the antigen to screen a library of complementary V _L or V _H domains, respectively. See, e.g., Portolano et al., J. Immunol. 150:880-887 (1993); See [Clarkson et al., Nature 352:624-628 (1991)].

Among the antibodies included in the CAR are antibody fragments. An "antibody fragment" or "antigen-binding fragment" refers to a molecule other than an intact antibody that comprises a portion of an intact antibody that binds to an antigen to which the intact antibody binds. Examples of antibody fragments include Fv, Fab, Fab', Fab'-SH, F(ab') ₂ ; diabodies; linear antibodies; single-chain antibody molecules, such as scFv, and single-domain antibodies comprising only _{the V H region} ; and multispecific antibodies formed from antibody fragments. In some embodiments, the antigen-binding domain in the CAR is or comprises an antibody fragment comprising a variable heavy (V _H ) and a variable light (V _L ) region. In particular embodiments, the antibody is a single-chain antibody fragment, such as a scFv, comprising a heavy chain variable (V _H ) region and/or a light chain variable (V _L ) region.

In some embodiments, the chimeric antigen receptor comprises an extracellular portion containing an antibody or antibody fragment. In some aspects, the chimeric antigen receptor comprises an extracellular portion containing an antibody or fragment and an intracellular signaling domain. In some embodiments, the antibody or fragment comprises an scFv.

In some embodiments, the antibody portion of the recombinant receptor, e.g., the CAR, further comprises at least a portion of an immunoglobulin constant region, such as a hinge region, e.g., an IgG4 hinge region, and/or a CH1/CL and/or Fc region. In some embodiments, the constant region or portion is directed to a human IgG, such as an IgG4 or IgG1. In some aspects, a portion of the constant region serves as a spacer region between the antigen-recognition component (e.g., scFv) and the transmembrane domain. The spacer can have a length that provides increased responsiveness of the cell after antigen binding compared to the absence of the spacer. Exemplary spacers include those described in Hudecek et al. (2013) Clin. Cancer Res., 19:3153, International Patent Application Publication No. WO2014031687, U.S. Patent No. 8,822,647, or Published Application No. US2014/0271635.

In some embodiments, the antigen receptor comprises an intracellular domain directly or indirectly linked to an extracellular domain. In some embodiments, the chimeric antigen receptor comprises a transmembrane domain connecting the extracellular domain and the intracellular signaling domain. In some embodiments, the intracellular signaling domain comprises an ITAM. For example, in some aspects, the antigen recognition domain (e.g., the extracellular domain) is generally linked to one or more intracellular signaling components, such as a signaling component that mimics activation through an antigen receptor complex, such as a TCR complex in the case of a CAR, and/or a signal through another cell surface receptor. In some embodiments, the chimeric receptor comprises a transmembrane domain linked or fused between an extracellular domain (e.g., an scFv) and an intracellular signaling domain. Thus, in some embodiments, the antigen-binding component (e.g., an antibody) is linked to one or more transmembrane and intracellular signaling domains.

In one embodiment, a transmembrane domain that naturally associates with one of the domains of a receptor, e.g., a CAR, is used. In some instances, the transmembrane domain is selected or modified by amino acid substitutions to avoid binding to the transmembrane domain of the same or a different surface membrane protein, thereby minimizing interactions with other members of the receptor complex.

In some embodiments, the transmembrane domain is derived from a natural or synthetic source. If the source is natural, in some aspects, the domain is derived from any membrane-bound or transmembrane protein. The transmembrane domain comprises a region (i.e., a region comprising at least a transmembrane region(s) thereof) derived from the alpha, beta, or zeta chain of the T-cell receptor, CD28, CD3 epsilon, CD45, CD4, CD5, CD8, CD9, CD16, CD22, CD33, CD37, CD64, CD80, CD86, CD134, CD137, CD154. Alternatively, in some embodiments, the transmembrane domain is synthetic. In some aspects, the synthetic transmembrane domain predominantly comprises hydrophobic residues, such as leucine and valine. In some aspects, a triplet of phenylalanine, tryptophan, and valine will be found at each terminus of the synthetic transmembrane domain. In some embodiments, the connection is via a linker, spacer, and/or transmembrane domain(s). In some aspects, the transmembrane domain comprises a transmembrane portion of CD28.

In some embodiments, the extracellular domain and the transmembrane domain may be directly or indirectly linked. In some embodiments, the extracellular domain and the transmembrane domain are linked by a spacer (e.g., any of those described herein). In some embodiments, the receptor contains the extracellular portion of a molecule from which the transmembrane domain is derived, such as the extracellular portion of CD28.

The intracellular signaling domain includes domains that mimic or approximate signaling via native antigen receptors, signaling via such receptors in combination with costimulatory receptors, and/or signaling via costimulatory receptors alone. In some embodiments, a short oligo- or polypeptide linker, e.g., a linker between 2 and 10 amino acids in length, such as a linker containing glycine and serine, e.g., a glycine-serine doublet, is present, forming a linkage between the transmembrane domain of the CAR and the cytoplasmic signaling domain.

In some aspects, T cell activation is described as being mediated by two classes of cytoplasmic signaling sequences: sequences that initiate antigen-dependent primary activation via the TCR (primary cytoplasmic signaling sequences), and sequences that act in an antigen-independent manner to provide secondary or costimulatory signals (secondary cytoplasmic signaling sequences). In some aspects, CARs comprise one or both of these signaling components.

A receptor, such as a CAR, typically comprises at least one intracellular signaling component or components. In some aspects, the CAR comprises a primary cytoplasmic signaling sequence that regulates the primary activation of the TCR complex. A stimulatory primary cytoplasmic signaling sequence may contain a signaling motif known as an immunoreceptor tyrosine-based activation motif, or ITAM. Examples of ITAMs containing a primary cytoplasmic signaling sequence include those derived from the CD3 zeta chain, FcR gamma, CD3 gamma, CD3 delta, and CD3 epsilon. In some embodiments, the cytoplasmic signaling molecule(s) within the CAR comprises a cytoplasmic signaling domain, a portion thereof, or a sequence derived from CD3 zeta.

In some embodiments, the receptor comprises an intracellular component of a TCR complex, such as the TCR CD3 chain, e.g., the CD3 zeta chain, which mediates T-cell activation and cytotoxicity. Therefore, in some aspects, the antigen-binding portion is linked to one or more cell signaling modules. In some embodiments, the cell signaling module comprises a CD3 transmembrane domain, a CD3 intracellular signaling domain, and/or another CD3 transmembrane domain. In some embodiments, the receptor, e.g., a CAR, further comprises part of one or more additional molecules, such as Fc receptor γ, CD8, CD4, CD25, or CD16. For example, in some aspects, the CAR or other chimeric receptor comprises a chimeric molecule between CD3-zeta (CD3-ζ) or Fc receptor γ and CD8, CD4, CD25, or CD16.

In some embodiments, upon ligation of a CAR or other chimeric receptor, the cytoplasmic domain or intracellular signaling domain of the receptor activates at least one of the normal effector functions or responses of an immune cell, e.g., a T cell engineered to express the CAR. For example, in some contexts, the CAR induces a function of the T cell, such as cytolytic activity or T-helper activity, such as the secretion of cytokines or other factors. In some embodiments, a truncated portion of the intracellular signaling domain of an antigen receptor component or a co-stimulatory molecule is used in place of the intact immunostimulatory chain, e.g., when transducing an effector function signal. In some embodiments, the intracellular signaling domain or domains comprise a cytoplasmic sequence of a T cell receptor (TCR), and in some aspects also comprise a cytoplasmic sequence of a co-receptor that acts in concert with such receptor to initiate signal transduction following antigen receptor engagement in its natural environment.

In the context of natural TCRs, full activation generally requires signaling through the TCR, as well as a costimulatory signal. Therefore, in some embodiments, to facilitate full activation, the CAR also includes a component for generating a secondary or costimulatory signal. In other embodiments, the CAR does not include a component for generating a costimulatory signal. In some aspects, an additional CAR is expressed on the same cell and provides a component for generating a secondary or costimulatory signal.

In some embodiments, the chimeric antigen receptor contains the intracellular domain of a T cell costimulatory molecule. In some embodiments, the CAR comprises a signaling domain and/or a transmembrane portion of a costimulatory receptor, such as CD28, 4-1BB, OX40, DAP10, or ICOS. In some aspects, the same CAR comprises both activating and costimulatory components. In some embodiments, the chimeric antigen receptor contains an intracellular domain derived from a T cell costimulatory molecule or a functional variant thereof, such as between the transmembrane domain and the intracellular signaling domain. In some aspects, the T cell costimulatory molecule is CD28 or 4-1BB.

In some embodiments, the activation domain is contained within one CAR, while the costimulatory component is provided by another CAR that recognizes another antigen. In some embodiments, the CAR comprises an activating or stimulatory CAR and a costimulatory CAR, both of which are expressed on the same cell (see WO2014/055668). In some aspects, the cell comprises one or more stimulatory or activating CARs and/or costimulatory CARs. In some embodiments, the cell further comprises an inhibitory CAR (iCAR, see Fedorov et al., Sci. Transl. Medicine, 5(215) (December, 2013)), e.g., a CAR that recognizes an antigen other than an antigen associated with and/or specific for a disease or condition, whereby an activating signal delivered via the disease-targeting CAR is attenuated or inhibited by binding of the inhibitory CAR to its ligand, e.g., reducing an off-target effect.

In some embodiments, the two receptors induce activating and inhibitory signals in the cell, respectively, such that binding of one receptor to its antigen activates the cell or induces a response, while binding of the second inhibitory receptor to its antigen induces a signal that inhibits or attenuates that response. An example is the combination of an activating CAR and an inhibitory CAR (iCAR). This strategy can be used to reduce the likelihood of off-target effects, for example, in a context where an activating CAR binds to an antigen expressed on cells in a disease or condition but also on normal cells, and an inhibitory receptor binds to a separate antigen expressed on cells in a normal but not diseased or condition.

In some aspects, the chimeric receptor is or comprises an inhibitory CAR (e.g., an iCAR) and comprises an intracellular component that attenuates or inhibits an immune response in a cell, such as an ITAM- and/or costimulatory-promoted response. Examples of such intracellular signaling components are those found on immune checkpoint molecules, including PD-1, CTLA4, LAG3, BTLA, OX2R, TIM-3, TIGIT, LAIR-1, PGE2 receptor, EP2/4 adenosine receptor (including A2AR). In some aspects, the engineered cell comprises an inhibitory CAR comprising a signaling domain of such an inhibitory molecule or a signaling domain derived therefrom, thereby attenuating a cellular response induced by, for example, an activating and/or costimulatory CAR.

In certain embodiments, the intracellular signaling domain comprises a CD28 transmembrane and signaling domain linked to a CD3 (e.g., CD3-zeta) intracellular domain. In some embodiments, the intracellular signaling domain comprises a chimeric CD28 and CD137 (4-1BB, TNFRSF9) costimulatory domain linked to a CD3 zeta intracellular domain.

In some embodiments, the CAR comprises one or more, e.g., two or more, costimulatory domains and an activation domain, e.g., a primary activation domain, in the cytoplasmic portion. Exemplary CARs comprise intracellular components of CD3-zeta, CD28, and 4-1BB.

In some embodiments, the antigen receptor further comprises a tag, and/or the cell expressing the CAR or other antigen receptor further comprises a surrogate tag, such as a cell surface tag, which can be used to identify transduction or manipulation of the cell to express the receptor. In some aspects, the tag comprises all or a portion (e.g., a truncated form) of CD34, NGFR, or epidermal growth factor receptor, such as a truncated version of such cell surface receptor (e.g., tEGFR). In some embodiments, the nucleic acid encoding the tag is operably linked to a linker sequence, such as a cleavable linker sequence, e.g., a polynucleotide encoding T2A. For example, the tag and optionally the linker sequence can be any of those disclosed in published patent application No. WO2014031687. For example, the tag can be a truncated EGFR (tEGFR), optionally linked to a linker sequence, such as a T2A cleavable linker sequence.

In some embodiments, the tag is a molecule not naturally found on a T cell or not naturally found on the surface of a T cell, e.g., a cell surface protein, or a portion thereof. In some embodiments, the molecule is a non-self molecule, e.g., a non-self protein, i.e., not recognized as "self" by the immune system of the host to which the cell is to be adaptively transferred.

In some embodiments, the tag does not provide a therapeutic function and/or does not produce any effect other than being used as a tag for genetic manipulation, e.g., to select successfully manipulated cells. In other embodiments, the tag may be a therapeutic molecule or a molecule that otherwise exerts some desired effect, such as a ligand that interacts with a cell in vivo, such as a co-stimulatory or immune checkpoint molecule that enhances and/or dampens the cell's response upon encountering the ligand and/or adoptive transfer.

In some cases, CARs are referred to as first-generation, second-generation, and/or third-generation CARs. In some aspects, a first-generation CAR is a CAR that provides only a CD3-chain-induced signal upon antigen binding; in some aspects, a second-generation CAR is a CAR that provides such a signal and a costimulatory signal, such as a CAR comprising an intracellular signaling domain from a costimulatory receptor, such as CD28 or CD137; and in some aspects, a third-generation CAR is a CAR that comprises multiple costimulatory domains from different costimulatory receptors.

For example, in some embodiments, the CAR contains an antibody, e.g., an antibody fragment, such as a scFv, a transmembrane domain that is or comprises a transmembrane portion of CD28 or a functional variant thereof, and an intracellular signaling domain that comprises a signaling portion of CD28 or a functional variant thereof and a signaling portion of CD3 zeta or a functional variant thereof, which is specific for an antigen, including any antigen as described. In some embodiments, the CAR contains an antibody, e.g., an antibody fragment, such as a scFv, a transmembrane domain that is or comprises a transmembrane portion of CD28 or a functional variant thereof, which is specific for an antigen, including any antigen as described, and an intracellular signaling domain that comprises a signaling portion of 4-1BB or a functional variant thereof and a signaling portion of CD3 zeta or a functional variant thereof. In some such embodiments, the receptor further comprises a spacer containing a portion of an Ig molecule, such as a human Ig molecule, such as an Ig hinge, e.g., an IgG4 hinge, such as a hinge-only spacer.

For example, in some embodiments, the CAR comprises an antibody, such as an antibody fragment comprising a scFv, a spacer, such as a spacer containing a portion of an immunoglobulin molecule, such as a hinge region and/or one or more constant regions of a heavy chain molecule, such as an Ig-hinge containing spacer, a transmembrane domain containing all or a portion of a CD28-derived transmembrane domain, a CD28-derived intracellular signaling domain, and a CD3 zeta signaling domain. In some embodiments, the CAR comprises an antibody or fragment, such as a scFv, a spacer, such as any of an Ig-hinge containing spacer, a CD28-derived transmembrane domain, a 4-1BB-derived intracellular signaling domain, and a CD3 zeta-derived signaling domain.

Exemplary surrogate tags can include truncated forms of cell surface polypeptides, such as truncated forms that are non-functional and do not or cannot transmit a signal or signals normally transmitted by the full-length form of the cell surface polypeptide, and/or truncated forms that do not or cannot be internalized. Exemplary truncated cell surface polypeptides include truncated forms of growth factors or other receptors, such as truncated human epidermal growth factor receptor 2 (tHER2), truncated epidermal growth factor receptor (tEGFR), or prostate-specific membrane antigen (PSMA), or modified forms thereof. The tEGFR may contain an epitope recognized by the antibody cetuximab (Erbitux®) or other therapeutic anti-EGFR antibodies or binding molecules, which may be used to identify or select cells engineered to express the tEGFR construct and the encoded exogenous protein, and/or to eliminate or isolate cells expressing the encoded exogenous protein. See U.S. Patent No. 8,802,374 and Liu et al., Nature Biotech. 2016 April; 34(4): 430-434. In some aspects, the tag, e.g., a surrogate tag, comprises CD34, NGFR, CD19, or a truncated CD19, e.g., all or part of a truncated non-human CD19 (e.g., a truncated form), or an epidermal growth factor receptor (e.g., a tEGFR). In some embodiments, the tag is or comprises a fluorescent protein, such as green fluorescent protein (GFP), enhanced green fluorescent protein (EGFP), such as super-fold GFP (sfGFP), red fluorescent protein (RFP), such as tdTomato, mCherry, mStrawberry, AsRed2, DsRed or DsRed2, cyan fluorescent protein (CFP), blue green fluorescent protein (BFP), enhanced blue fluorescent protein (EBFP) and yellow fluorescent protein (YFP), and variants thereof (including species variants, monomeric variants, and codon-optimized and/or enhanced variants of the fluorescent proteins). In some embodiments, the tag is or comprises an enzyme, such as luciferase, the lacZ gene from E. coli, alkaline phosphatase, secreted embryonic alkaline phosphatase (SEAP), chloramphenicol acetyl transferase (CAT). Exemplary luminescent reporter genes include luciferase (luc), β-galactosidase, chloramphenicol acetyltransferase (CAT), β-glucuronidase (GUS) or variants thereof.

In some embodiments, the tag is a resistance tag or a selection tag. In some embodiments, the resistance tag or selection tag is or comprises a polypeptide that confers resistance to an exogenous agent or drug. In some embodiments, the resistance tag or selection tag is an antibiotic resistance gene. In some embodiments, the resistance tag or selection tag is an antibiotic resistance gene that confers antibiotic resistance to a mammalian cell. In some embodiments, the resistance tag or selection tag is or comprises a puromycin resistance gene, a hygromycin resistance gene, a blasticidin resistance gene, a neomycin resistance gene, a geneticin resistance gene, or a zeocin resistance gene, or a modified form thereof.

In some embodiments, the nucleic acid encoding the tag is operably linked to a polynucleotide encoding a linker sequence, such as a cleavable linker sequence, e.g., T2A. For example, the tag and optionally the linker sequence may be any of those disclosed in PCT Publication No. WO2014031687.

3. T cell receptor

In some embodiments, the recombinant protein is a recombinant receptor. In some embodiments, the recombinant protein is a T cell receptor (TCR) or an antigen-binding portion thereof that recognizes a peptide epitope or a T cell epitope of a target polypeptide, such as an antigen of a tumor, virus, or autoimmune protein. In some aspects, the TCR is or comprises a recombinant TCR.

In some embodiments, a "T cell receptor" or "TCR" is a molecule that contains variable α and β chains (also known as TCRα and TCRβ, respectively) or variable γ and δ chains (also known as TCRα and TCRβ, respectively) or antigen-binding portions thereof and that is capable of specifically binding a peptide bound to an MHC molecule. In some embodiments, the TCR is in the αβ form. Typically, TCRs that exist in the αβ and γδ forms are structurally similar, but the T cells that express them may have distinct anatomical locations or functions. TCRs can be found on the cell surface or in a soluble form. Typically, TCRs are found on the surface of T cells (or T lymphocytes), where they typically serve to recognize antigens bound to major histocompatibility complex (MHC) molecules.

Unless otherwise stated, the term "TCR" should be understood to include a complete TCR as well as an antigen-binding portion or fragment thereof. In some embodiments, the TCR is an intact or full-length TCR, including an αβ form or a γδ form of the TCR. In some embodiments, the TCR is an antigen-binding portion that is smaller than the full-length TCR but binds a specific peptide bound to an MHC molecule, such as an MHC-peptide complex. In some cases, the antigen-binding portion or fragment of a TCR may contain only a portion of the structural domains of a full-length or intact TCR, but may bind a peptide epitope that a complete TCR binds, such as an MHC-peptide complex. In some cases, the antigen-binding portion contains a variable domain of a TCR sufficient to form a binding site for binding to a specific MHC-peptide complex, such as the variable α chain and the variable β chain of the TCR. Typically, the variable chain of a TCR contains complementarity-determining regions involved in the recognition of peptides, MHC, and/or MHC-peptide complexes.

In some embodiments, the variable domain of a TCR contains hypervariable loops or complementarity determining regions (CDRs), which are generally major contributors to antigen recognition and binding ability and specificity. In some embodiments, the CDRs or combinations thereof of a TCR form all or substantially all of the antigen-binding site of a given TCR molecule. The various CDRs within the variable region of a TCR chain are generally separated by framework regions (FRs), which generally exhibit less variability between TCR molecules compared to the CDRs (see, e.g., Jores et al., Proc. Nat'l Acad. Sci. U.S.A. 87:9138, 1990; Chothia et al., EMBO J. 7:3745, 1988; see also Lefranc et al., Dev. Comp. Immunol. 27:55, 2003). In some embodiments, CDR3 is the major CDR responsible for antigen binding or specificity, or is the CDR most important among the three CDRs on a given TCR variable region for antigen recognition and/or interaction with the processed peptide portion of a peptide-MHC complex. In some contexts, CDR1 of the alpha chain may interact with the N-terminal portion of a particular antigenic peptide. In some contexts, CDR1 of the beta chain may interact with the C-terminal portion of the peptide. In some contexts, CDR2 is the major CDR most significantly responsible for or contributing to interaction with or recognition of the MHC portion of an MHC-peptide complex. In some embodiments, the variable region of the β chain may contain an additional hypervariable region (CDR4 or HVR4) that is generally involved in superantigen binding, but not antigen recognition (Kotb (1995) Clinical Microbiology Reviews, 8:411-426).

In some embodiments, the TCR may also contain a constant domain, a transmembrane domain, and/or a short cytoplasmic tail (see, e.g., Janeway et al., Immunobiology: The Immune System in Health and Disease, 3rd Ed., Current Biology Publications, p. 4:33, 1997). In some aspects, each chain of the TCR may have an N-terminal immunoglobulin variable domain, an immunoglobulin constant domain, a transmembrane region, and a short cytoplasmic tail at the C-terminal end. In some embodiments, the TCR associates with a non-variant protein of the CD3 complex, which is involved in signal transduction mediation.

In some embodiments, the TCR chain contains one or more constant domains. For example, the extracellular portion of a given TCR chain (e.g., an α-chain or a β-chain) may contain two immunoglobulin-like domains adjacent to the cell membrane, such as a variable domain (e.g., Vα or Vβ; typically amino acids 1 to 116 of the chain based on Kabat numbering, Kabat et al., "Sequences of Proteins of Immunological Interest, US Dept. Health and Human Services, Public Health Service National Institutes of Health, 1991, 5th ed.) and a constant domain (e.g., an α-chain constant domain or Cα, typically amino acids 117 to 259 of the chain based on Kabat numbering or a β-chain constant domain or Cβ, typically amino acids 117 to 295 of the chain based on Kabat numbering). For example, in some cases, the extracellular portion of a TCR formed by two chains may contain two membrane-proximal constant domains and It contains two membrane-distal variable domains, each variable domain containing a CDR. The constant domain of the TCR may contain a short connecting sequence in which cysteine residues form disulfide bonds, thereby connecting the two chains of the TCR. In some embodiments, the TCR may have an additional cysteine residue in each of the α and β chains, which allows the TCR to contain two disulfide bonds in the constant domain.

In some embodiments, the TCR chain contains a transmembrane domain. In some embodiments, the transmembrane domain is positively charged. In some cases, the TCR chain contains a cytoplasmic tail. In some cases, the structure allows the TCR to associate with other molecules, such as CD3 and its subunits. For example, a TCR containing a constant domain with a transmembrane region can anchor the protein to the cell membrane and associate with a non-variant subunit of the CD3 signaling apparatus or complex. The intracellular tail of the CD3 signaling subunit (e.g., CD3γ, CD3δ, CD3ε, and CD3ζ chains) contains one or more immunoreceptor tyrosine-based activation motifs, or ITAMs, that are involved in the signaling ability of the TCR complex.

In some embodiments, the TCR may be a heterodimer of two chains, α and β (or optionally γ and δ), or may be a single chain TCR construct. In some embodiments, the TCR is a heterodimer containing two separate chains (α and β chains or γ and δ chains) linked, e.g., by a disulfide bond or disulfide bonds.

In some embodiments, the TCR can be generated from known TCR sequence(s) for which substantially the full-length coding sequence is readily available, such as the sequence of the Vα,β chain. Methods for obtaining full-length TCR sequences, including V chain sequences, from cell sources are well known. In some embodiments, the nucleic acid encoding the TCR can be obtained from a variety of sources, such as by polymerase chain reaction (PCR) amplification of TCR-encoding nucleic acid within or isolated from a given cell or cells, or by synthesis of publicly available TCR DNA sequences.

In some embodiments, the recombinant receptor comprises a recombinant TCR and/or a TCR cloned from a naturally occurring T cell. In some embodiments, a high-affinity T cell clone for a target antigen (e.g., a cancer antigen) is identified, isolated from a patient, and introduced into a cell. In some embodiments, the TCR clone for the target antigen is generated in a transgenic mouse engineered with human immune system genes (e.g., the human leukocyte antigen system or HLA). See, e.g., tumor antigens (see, e.g., Parkhurst et al. (2009) Clin Cancer Res. 15:169-180 and Cohen et al. (2005) J Immunol. 175:5799-5808). In some embodiments, phage display is used to isolate TCRs against a target antigen (see, e.g., Varela-Rohena et al. (2008) Nat Med. 14:1390-1395 and Li (2005) Nat Biotechnol. 23:349-354).

In some embodiments, the TCR is obtained from a biological source, such as a cell, such as a T cell (e.g., a cytotoxic T cell), a T-cell hybridoma, or other publicly available source. In some embodiments, the T-cell can be obtained from cells isolated in vivo. In some embodiments, the TCR is a thymus-selected TCR. In some embodiments, the TCR is a neoepitope-restricted TCR. In some embodiments, the T-cell can be a cultured T-cell hybridoma or clone. In some embodiments, the TCR or antigen-binding portion thereof can be synthetically generated from knowledge of the TCR sequence.

In some embodiments, the TCR is generated from a TCR identified or selected from a screening of a candidate TCR library for a target polypeptide antigen or its target T cell epitopes. The TCR library can be generated by amplifying the Vα and Vβ repertoire from T cells isolated from a subject, including cells present in PBMCs, spleen, or other lymphoid organs. In some cases, the T cells can be amplified from tumor-infiltrating lymphocytes (TILs). In some embodiments, the TCR library can be generated from CD4+ or CD8+ cells. In some embodiments, the TCR can be amplified from a source of T cells from a normal healthy subject, i.e., a normal TCR library. In some embodiments, the TCR can be amplified from a source of T cells from a diseased subject, i.e., a diseased TCR library. In some embodiments, degenerate primers are used to amplify the Vα and Vβ gene repertoire from a sample, e.g., a human, by RT-PCR. In some embodiments, the scTv library can be assembled from naive Vα and Vβ libraries, where the amplified products are cloned or assembled such that they are separated by a linker. Depending on the source of the subject and cells, the library can be HLA allele-specific. Alternatively, in some embodiments, the TCR library can be generated by mutagenesis or diversification of parental or scaffold TCR molecules. In some aspects, the TCR is subjected to directed evolution, such as mutagenesis of the α or β chain. In some aspects, specific residues within the CDRs of the TCR are altered. In some embodiments, the selected TCR can be modified by affinity maturation. In some embodiments, antigen-specific T cells can be selected, for example, by screening to assess CTL activity against a peptide. In some aspects, for example, the TCR present on the antigen-specific T cell can be selected, for example, by binding activity, such as a specific affinity or avidity for the antigen.

In some embodiments, the TCR or antigen-binding portion thereof is modified or engineered. In some embodiments, directed evolution methods are used to generate TCRs with altered properties, such as higher affinity for specific MHC-peptide complexes. In some embodiments, directed evolution is achieved by display methods, including but not limited to yeast display (Holler et al., (2003) Nat Immunol, 4, 55-62; Holler et al., (2000) Proc Natl Acad Sci U S A, 97, 5387-92), phage display (Li et al., (2005) Nat Biotechnol, 23, 349-54), or T cell display (Chervin et al., (2008) J Immunol Methods, 339, 175-84). In some embodiments, the display approach involves engineering or modifying a known parental or reference TCR. For example, in some cases, a wild-type TCR can be used as a template to generate a mutagenized TCR in which one or more residues of the CDRs are mutated, and mutants with desired altered properties, such as higher affinity for a desired target antigen, are selected.

Exemplary TCR T cell therapies for use in accordance with the methods provided herein are known in the art. Suitable TCR T cell therapies for use in accordance with the methods provided herein include any of the therapies described in Zhao and Cao, Front Immunol 10: 2250 (2019); Ping et al., Protein Cell 9(3): 254-266 (2018); and Zhang and Wang, Technol Cancer Res Treat 18: 1533033819831068 (2019), each of which is incorporated herein by reference in its entirety.

Exemplary TCR T cell therapies targeting PRAME include those investigated or under investigation in clinical trials NCT03503968 and NCT02743611. Exemplary TCR T cell therapies targeting MAGE-A3/A6 include those investigated or under investigation in clinical trial NCT03139370. Exemplary TCR T cell therapies targeting CEA include those investigated or under investigation in clinical trial NCT00923806. Exemplary TCR T cell therapies targeting MAGE-A3/12 include those investigated or under investigation in clinical trial NCT01273181. Exemplary TCR T cell therapies targeting MAGE-A10 include those investigated or under investigation in clinical trials NCT02592577, NCT03391791, and NCT02989064. Exemplary TCR T cell therapies targeting NY-ESO-1 include those investigated or are investigational in clinical trials NCT01343043, NCT03029273, NCT03462316, NCT01892293, NCT01352286, NCT01567891, NCT01350401, NCT02588612, NCT03691376, and NCT03168438. Exemplary TCR T cell therapies targeting AFP include those investigated or are investigational in clinical trial NCT03132792. Exemplary TCR T cell therapies targeting HA-1 include those investigated or are investigational in clinical trial NCT03326921. Exemplary TCR T cell therapies targeting WT1 include those investigated or under investigation in clinical trials NCT02550535 and NCT02770820. Exemplary TCR T cell therapies targeting Gp100 include those investigated or under investigation in clinical trials NCT00923195 and NCT02889861. Exemplary TCR T cell therapies targeting CMV include those investigated or under investigation in clinical trial NCT02988258. Exemplary TCR T cell therapies targeting MART-1 include those investigated or under investigation in clinical trial NCT00091104. Exemplary TCR T cell therapies targeting HBV include those investigated or under investigation in clinical trial NCT02719782. Exemplary TCR T cell therapies targeting P53 include those investigated or under investigation in clinical trial NCT00393029. Exemplary TCR T cell therapies targeting HPV-16 E6 include those investigated or under investigation in clinical trials NCT03578406 and NCT02280811. Exemplary TCR T cell therapies targeting HPV-16 E7 include those investigated or under investigation in clinical trial NCT02858310. Exemplary TCR T cell therapies targeting SL9 include those investigated or under investigation in clinical trial NCT00991224. Exemplary TCR T cell therapies targeting TGFβII include those investigated or under investigation in clinical trial NCT03431311. Exemplary TCR T cell therapies targeting MCPyV include those investigated or under investigation in clinical trial NCT03747484. Exemplary TCR T cell therapies targeting TRAIL include those investigated or under investigation in clinical trial NCT00923390. Exemplary TCR T cell therapies targeting EBV include those investigated or are being investigated in clinical trial NCT03648697. Exemplary TCR T cell therapies targeting KRAS include those investigated or are being investigated in clinical trials NCT03190941 and NCT03745326.

In some embodiments, the peptides of the target polypeptide for use in producing or generating a TCR of interest are known or can be routinely readily identified. In some embodiments, suitable peptides for use in generating a TCR or antigen-binding portion can be determined based on the presence of HLA-restricted motifs in the target polypeptide of interest, such as those described below. In some embodiments, the peptides are routinely identified using computer prediction models. In some embodiments, such models for predicting MHC class I binding sites include, but are not limited to, ProPred1 (see Singh and Raghava (2001) Bioinformatics 17(12):1236-1237) and SYFPEITHI (see Schuler et al., (2007) Immunoinformatics Methods in Molecular Biology, 409(1): 75-93 2007). In some embodiments, the MHC-restricted epitope is HLA-A0201, which is expressed in approximately 39-46% of all Caucasians and therefore represents a suitable MHC antigen choice for use in constructing a TCR or other MHC-peptide binding molecule.

HLA-A0201-binding motifs and cleavage sites for the proteasome and immunoproteasome are known using computer prediction models. For predicting MHC class I binding sites, these models include, but are not limited to, ProPred1 (described in more detail in Singh and Raghava, ProPred: prediction of HLA-DR binding sites. BIOINFORMATICS 17(12):1236-1237 2001) and SYFPEITHI (see Schuler et al., SYFPEITHI, Database for Searching and T-Cell Epitope Prediction. in Immunoinformatics Methods in Molecular Biology, vol 409(1): 75-93 2007).

In some embodiments, the TCR or antigen-binding portion thereof may be a recombinantly produced native protein or a mutated form thereof with one or more altered properties, such as binding characteristics. In some embodiments, the TCR may be derived from one of various animal species, such as a human, mouse, rat, or other mammal. The TCR may be cell-associated or in a soluble form. In some embodiments, for purposes of the provided methods, the TCR is in a cell-associated form expressed on the cell surface.

In some embodiments, the TCR is a full-length TCR. In some embodiments, the TCR is an antigen-binding portion. In some embodiments, the TCR is a dimeric TCR (dTCR). In some embodiments, the TCR is a single-chain TCR (sc-TCR). In some embodiments, the dTCR or scTCR has a structure as described in WO 03/020763, WO 04/033685, or WO2011/044186.

In some embodiments, the TCR contains a sequence corresponding to a transmembrane sequence. In some embodiments, the TCR contains a sequence corresponding to a cytoplasmic sequence. In some embodiments, the TCR can form a TCR complex with CD3. In some embodiments, any TCR, including a dTCR or scTCR, can be linked to a signaling domain that generates an active TCR on the surface of a T cell. In some embodiments, the TCR is expressed on the surface of a cell.

In some embodiments, the dTCR comprises a first polypeptide in which a sequence corresponding to a TCR α chain variable region sequence is fused to the N-terminus of a sequence corresponding to a TCR α chain constant region extracellular sequence, and a second polypeptide in which a sequence corresponding to a TCR β chain variable region sequence is fused to the N-terminus of a sequence corresponding to a TCR β chain constant region extracellular sequence, wherein the first and second polypeptides are linked by a disulfide bond. In some embodiments, the bond may correspond to a native interchain disulfide bond present in a native dimeric αβ TCR. In some embodiments, the interchain disulfide bond is not present in a native TCR. For example, in some embodiments, one or more cysteines may be incorporated into the constant region extracellular sequence of a dTCR polypeptide pair. In some cases, both native and non-native disulfide bonds may be desirable. In some embodiments, the TCR contains a transmembrane sequence that anchors it to the membrane.

In some embodiments, the dTCR comprises a TCR α chain comprising a variable α domain, a constant α domain, and a first dimerization motif attached to the C-terminus of the constant α domain, and a TCR β chain comprising a variable β domain, a constant β domain, and a first dimerization motif attached to the C-terminus of the constant β domain, wherein the first and second dimerization motifs readily interact to form a covalent bond between an amino acid of the first dimerization motif and an amino acid of the second dimerization motif to link the TCR α chain and the TCR β chain together.

In some embodiments, the TCR is a scTCR. Typically, scTCRs can be generated using any suitable known method. See, e.g., Soo Hoo, W. F. et al., PNAS (USA) 89, 4759 (1992); Wulfing, C. and Plueckthun, A., J. Mol. Biol. 242, 655 (1994); Kurucz, I. et al., PNAS (USA) 90, 3830 (1993); International Publication Nos. WO 96/13593, WO 96/18105, WO99/60120, WO99/18129, WO 03/020763, WO2011/044186; and Schlueter, C. J. et al., J. Mol. Biol. 242, 655 (1994). 256, 859 (1996)]. In some embodiments, the scTCR contains a non-natural disulfide interchain bond introduced to promote TCR chain association (see, e.g., International Publication No. WO 03/020763). In some embodiments, the scTCR is a non-disulfide-linked truncated TCR, wherein a heterologous leucine zipper fused to its C-terminus promotes chain association (see, e.g., International Publication No. WO99/60120). In some embodiments, the scTCR contains a TCRα variable domain covalently linked to a TCRβ variable domain via a peptide linker (see, e.g., International Publication No. WO99/18129).

In some embodiments, the scTCR contains a first segment comprising an amino acid sequence corresponding to a TCR α chain variable region, a second segment comprising an amino acid sequence corresponding to a TCR β chain variable region sequence fused to the N-terminus of an amino acid sequence corresponding to a TCR β chain constant domain extracellular sequence, and a linker sequence connecting the C-terminus of the first segment to the N-terminus of the second segment.

In some embodiments, the scTCR contains a first segment consisting of an α chain variable region sequence fused to the N terminus of an α chain extracellular constant domain sequence, and a second segment consisting of a β chain variable region sequence fused to the N terminus of a β chain extracellular constant and transmembrane sequence, and optionally a linker sequence connecting the C terminus of the first segment to the N terminus of the second segment.

In some embodiments, the scTCR contains a first segment consisting of a TCR β chain variable region sequence fused to the N terminus of a β chain extracellular constant domain sequence, and a second segment consisting of an α chain variable region sequence fused to the N terminus of an α chain extracellular constant and transmembrane sequence, and optionally a linker sequence connecting the C terminus of the first segment to the N terminus of the second segment.

In some embodiments, the linker of the scTCR connecting the first and second TCR segments can be any linker capable of forming a single polypeptide strand while maintaining TCR binding specificity. In some embodiments, the linker sequence can have, for example, the formula -P-AA-P-, wherein P is proline and AA represents an amino acid sequence, wherein the amino acids are glycine and serine. In some embodiments, the first and second segments are paired such that their variable region sequences are oriented for such binding. Thus, in some cases, the linker is long enough to cover the distance between the C-terminus of the first segment and the N-terminus of the second segment, but not so long as to block or reduce binding of the scTCR to its target ligand, or vice versa. In some embodiments, the linker can contain 10 to 45 amino acids, such as 10 to 30 amino acids or 26 to 41 or about that number of amino acid residues, for example 29, 30, 31 or 32 amino acids.

In some embodiments, the scTCR contains a covalent disulfide bond linking an immunoglobulin region residue of the constant domain of the α chain to an immunoglobulin region residue of the constant domain of the β chain. In some embodiments, the interchain disulfide bond is not present in a native TCR. For example, in some embodiments, one or more cysteines may be incorporated into the extracellular sequence of the constant region of the first and second segments of the scTCR polypeptide. In some cases, both native and non-native disulfide bonds may be preferred.

In some embodiments of dTCRs or scTCRs containing introduced interchain disulfide bonds, the native disulfide bonds are absent. In some embodiments, one or more naturally occurring cysteines forming the native interchain disulfide bonds are substituted with another residue, such as serine or alanine. In some embodiments, the introduced disulfide bonds can be formed by mutating non-cysteine residues on the first and second segments to cysteine. Exemplary non-native disulfide bonds in TCRs are described in published international PCT No. WO2006/000830.

In some embodiments, the TCR or antigen-binding fragment thereof exhibits an affinity for a target antigen with an equilibrium binding constant of 10-5 to 10-12 M or about such values and all individual values and ranges therein. In some embodiments, the target antigen is an MHC-peptide complex or a ligand.

D. Culture

In some embodiments, the cell population is from a cell culture that is being cultured in vitro or ex vivo. In some embodiments, the provided method involves culturing the cell culture in vitro or ex vivo. In some embodiments, the cell population is removed from the cell culture to obtain holographic information. In some embodiments, the holographic information is obtained, and the in vitro or ex vivo culture occurs in a closed system.

Exemplary methods and systems for in vitro or ex vivo culture are described in US-11400115, US-20190112576, US-20190136186, US-11274278, US-20210032297, US-20200354677, US-20200384025, US-20210163893, and US-20220002669.

In some embodiments, the in vitro or ex vivo culture is conducted in a basal medium. In some embodiments, the basal medium is serum-free. In some embodiments, the basal medium is free of human serum. In some embodiments, the basal medium contains a mixture of inorganic salts, sugars, amino acids, and optionally, vitamins, organic acids, and/or buffers or other well-known cell culture nutrients. In addition to nutrients, the basal medium may also help maintain pH and osmolality. Various commercially available basal media are well known to those skilled in the art, including Dulbecco's Modified Eagle's Medium (DMEM), Roswell Park Memorial Institute Medium (RPMI), Iscove's Modified Dulbecco's Medium, and Ham's Medium. In some embodiments, the basal medium is Iscove's Modified Dulbecco's Medium, RPMI-1640, or α-MEM.

In some embodiments, the basal medium is a balanced salt solution (e.g., PBS, DPBS, HBSS, or EBSS). In some embodiments, the basal medium is selected from Dulbecco's Modified Eagle's Medium (DMEM), Minimum Essential Medium (MEM), Basal Medium Eagle (BME), F-10, F-12, RPMI 1640, Glasgow's Minimum Essential Medium (GMEM), Alpha Minimum Essential Medium (Alpha MEM), Iscove's Modified Dulbecco's Medium, and M199. In some embodiments, the basal medium is a complex medium (e.g., RPMI-1640 or IMDM). In some embodiments, the basal medium is OpTmizer™ CTS™ T-cell Expansion Basal Medium (ThermoFisher).

In some embodiments, the basal medium is supplemented with additional additives. In some embodiments, the basal medium is not supplemented with any additional additives. Supplements to the cell culture medium include nutrients, sugars such as glucose, amino acids, vitamins, and additives such as ATP and NADH.

In some embodiments, the in vitro or ex vivo culture is performed in the presence of one or more recombinant cytokines. In some embodiments, the culture is performed in the presence of any recombinant cytokine described in Section II-B, and at any concentration described in Section II-B. In some embodiments, the one or more recombinant cytokines are selected from IL-2, IL-15, and IL-7. In some embodiments, the in vitro or ex vivo culture is performed in the presence of IL-2, IL-15, and IL-7.

In some embodiments, the in vitro or ex vivo culture is performed in the absence of recombinant cytokines.

In some embodiments, the in vitro or ex vivo culture is performed under conditions that expand T cells in the cell culture. In some embodiments, the expansion conditions include one or more of a specific medium, temperature, oxygen content, carbon dioxide content, time, agents, such as nutrients, amino acids, antibiotics, ions, and/or stimulatory factors, such as cytokines, chemokines, antigens, binding partners, fusion proteins, recombinant soluble receptors, and any other agent designed to promote the growth, division, and/or expansion of T cells. In some embodiments, the in vitro or ex vivo culture is performed for a period of time until a desired or threshold density, concentration, or number of T cells is achieved.

In some embodiments, the culturing is generally performed under conditions that include a temperature suitable for the growth of primary immune cells, such as human T lymphocytes, for example, at least about 25°C, generally at least about 30°C, and generally at or about 37°C. In some embodiments, the cell culture is incubated at a temperature of 25 to 38°C, such as 30 to 37°C, for example, at or about 37°C ± 2°C. In some embodiments, the incubation is performed for a period of time until the culturing results in a desired or threshold density, concentration, or cell number. In some embodiments, the culturing is performed for more than or about 24 hours, 48 hours, 72 hours, 96 hours, 5 days, 6 days, 7 days, 8 days, 9 days, or more days.

In some embodiments, the in vitro or ex vivo culture is performed within a bioreactor. Examples of bioreactors suitable for in vitro or ex vivo culture include the GE Xuri W25, the GE Xuri W5, the Sartorius BioSTAT RM 20 | 50, the Finesse SmartRocker bioreactor system, and the Pall XRS bioreactor system. In some embodiments, the bioreactor is used to perfuse and/or mix the cell culture during at least a portion of the in vitro or ex vivo culture. In some embodiments, a cell population is removed from the cell culture within the bioreactor to obtain holographic information. In some embodiments, the bioreactor is connected to a microscope to obtain holographic information, for example, as described in Section I-A.

In some embodiments, cells cultured under the containment, connection, and/or control of a bioreactor undergo faster expansion during culture than cells cultured without a bioreactor, e.g., cells cultured under static conditions, such as without mixing, shaking, movement, and/or perfusion. In some embodiments, cells cultured under the containment, connection, and/or control of a bioreactor reach or achieve threshold expansion, cell count, and/or density within 14 days, 10 days, 9 days, 8 days, 7 days, 6 days, 5 days, 4 days, 3 days, 2 days, 60 hours, 48 hours, 36 hours, 24 hours, or 12 hours. In some embodiments, cells cultured under the sealing, connection and/or control of a bioreactor reach or achieve a threshold expansion, cell count and/or density that is at least 50%, at least 60%, at least 70%, at least 80%, at least 90%, at least 95%, at least 100%, at least 150%, at least 1-fold, at least 2-fold, at least 3-fold, at least 4-fold, at least 5-fold greater than cells cultured in exemplary and/or alternative processes that do not involve sealing, connection and/or control of a bioreactor.

In some embodiments, the mixing is or includes rocking and/or motion. In some cases, the bioreactor may be subjected to motion or rocking, which may, in some aspects, increase oxygen transfer. Moving the bioreactor may include rotation about a horizontal axis, rotation about a vertical axis, rocking along an inclined or inclined horizontal axis of the bioreactor, or any combination thereof. In some embodiments, at least a portion of the culturing is performed with rocking. The rocking speed and rocking angle can be adjusted to achieve the desired agitation. In some embodiments, the rocking angle is 20°, 19°, 18°, 17°, 16°, 15°, 14°, 13°, 12°, 11°, 10°, 9°, 8°, 7°, 6°, 5°, 4°, 3°, 2°, or 1°. In certain embodiments, the rocking angle is 6-16°. In other embodiments, the rocking angle is 7-16°. In other embodiments, the rocking angle is 8-12°. In some embodiments, the rocking speed is 1, 2, 3, 4, 5, 6, 7, 8, 9, 10, 11, 1 12, 13, 14 15, 16, 17, 18, 19, 20, 21, 22, 23, 24, 25, 26, 27, 28, 29, 30, 31, 32, 33, 34, 35, 36, 37, 38, 39, 40 rpm. In some embodiments, the rocking speed is 4 to 12 rpm, such as 4 to 6 rpm (inclusive).

In some embodiments, the bioreactor maintains a _constant air flow at a rate of about, or at least greater than, 0.01 L/min, 0.05 L/min, 0.1 L/min, 0.2 L/min, 0.3 L/min, 0.4 L/min, 0.5 L/min, 1.0 L/min, 1.5 L/min, or 2.0 L/min, or greater than, or at least about, 2.0 L/min, while maintaining a temperature at or near 37°C and a CO2 level at or near 5%. In certain embodiments, at least a portion of the culture is performed by perfusion, such as at a rate of 290 ml/day, 580 ml/day, and/or 1160 ml/day, depending, for example, on the time point relative to the start of the culture and/or the density of the cultured cells. In some embodiments, at least a portion of the cell culture expansion is performed with a rocking motion at an angle of, for example, 5° to 10°, for example, 6°, and at a constant rocking speed, for example, 5 to 15 RPM, for example, 6 RMP or 10 RPM.

In some embodiments, at least a portion of the culture step is performed under constant perfusion, e.g., slow, constant perfusion. In some embodiments, the perfusion comprises or involves the outflow of liquid, e.g., spent medium, and the inflow of fresh medium. In certain embodiments, the perfusion replaces spent medium with fresh medium. In some embodiments, at least a portion of the culture is 100 ml/day, 200 ml/day, 250 ml/day, 275 ml/day, 290 ml/day, 300 ml/day, 350 ml/day, 400 ml/day, 450 ml/day, 500 ml/day, 550 ml/day, 575 ml/day, 580 ml/day, 600 ml/day, 650 ml/day, 700 ml/day, 750 ml/day, 800 ml/day, 850 ml/day, 900 ml/day, 950 ml/day, 1000 ml/day, 1100 ml/day, 1160 ml/day, 1200 ml/day, 1400 ml/day, 1600 ml/day, 1800 ml/day, 2000 ml/day, 2200 ml/day or 2400 ml/day or at a constant rate of about or at least the above rate.

In some embodiments, the in vitro or ex vivo culture is performed under conditions that minimize or eliminate further expansion of the T cells in the cell culture. In some embodiments, the cell culture is not incubated under conditions that increase the amount of T cells during the in vitro or ex vivo culture. In some embodiments, the cell culture is incubated under conditions that may result in expansion, but are not performed for the purpose of expanding the T cells in the cell culture.

In some embodiments, the in vitro or ex vivo culture occurs in an incubator. In some embodiments, the cell culture is transferred to a vessel for in vitro or ex vivo culture. In some embodiments, the vessel is a vial. In particular embodiments, the vessel is a bag. In some embodiments, the cell culture is transferred to the vessel under closed or sterile conditions. In some embodiments, the vessel, e.g., a vial or bag, is then placed in an incubator for all or part of the in vitro or ex vivo culture. In particular embodiments, the incubator is set to, about, or at least at 16°C, 24°C, or 35°C. In some embodiments, the incubator is set to, about, or at 37°C, or at, about 37°C, or 37°C ±2°C, ±1°C, ±0.5°C, or ±0.1°C.

In certain embodiments, the in vitro or ex vivo culture is performed under static conditions, such as without centrifugation, shaking, rotation, rocking, or perfusion of the medium. In some embodiments, the in vitro or ex vivo culture is performed under gentle mixing conditions, such as with rocking.

In some embodiments, the in vitro or ex vivo culture is performed for at least 12 hours. In some embodiments, the culture is performed for 6 hours, 12 hours, 18 hours, 24 hours, 36 hours, 48 hours, 60 hours, 72 hours, 2 days, 3 days, 4 days, 5 days, 6 days, 7 days, 7 days, 8 days, 9 days, 10 days, 1 week, 2 weeks, 3 weeks, or 4 weeks, or about or for less than or equal to the time.

In some embodiments, the in vitro or ex vivo culture is performed for 14 days or less. In some embodiments, the in vitro or ex vivo culture is performed for 7 days or less. In some embodiments, the in vitro or ex vivo culture is for 0.5 to 12 days, 0.5 to 10 days, 0.5 to 8 days, 0.5 to 6 days, 0.5 to 4 days, 0.5 to 2 days, 1 to 12 days, 1 to 10 days, 1 to 8 days, 1 to 6 days, 1 to 4 days, 1 to 2 days, 2 to 12 days, 2 to 10 days, 2 to 8 days, 2 to 6 days, 2 to 4 days, 4 to 12 days, 4 to 10 days, 4 to 8 days, 4 to 6 days, 6 to 12 days, 6 to 10 days, 6 to 8 days, 8 to 12 days, 8 to 10 days, or 10 to 12 days or about such times (each inclusive).

In some embodiments, the in vitro or ex vivo culture is performed for 7 days or less. In some embodiments, the in vitro or ex vivo culture is performed for 0.5 to 7 days, 0.5 to 6 days, 0.5 to 5 days, 0.5 to 4 days, 0.5 to 3 days, 0.5 to 2 days, 0.5 to 1 day, 1 to 7 days, 1 to 6 days, 1 to 5 days, 1 to 4 days, 1 to 3 days, 1 to 2 days, 2 to 7 days, 2 to 6 days, 2 to 5 days, 2 to 4 days, 2 to 3 days, 3 to 7 days, 3 to 6 days, 3 to 5 days, 3 to 4 days, 4 to 7 days, 4 to 6 days, 4 to 5 days, 5 to 7 days, 5 to 6 days, or 6 to 7 days or about such times (each inclusive).

III. Computing Devices

Also provided herein is a computing device that can be used to perform any of the provided methods. In some embodiments, the computing device has instructions in its memory for performing any of the provided methods.

In some embodiments, the instructions comprise instructions for receiving holographic information about individual cells in a cell population containing T cells. In some embodiments, the instructions comprise instructions for receiving one or more input measurements for each of a plurality of cell characteristics for individual cells in the cell population containing T cells. In some embodiments, the instructions comprise instructions for determining one or more input measurements for each of a plurality of cell characteristics for individual cells in the cell population containing T cells. In some embodiments, the instructions comprise instructions for receiving a plurality of population-level statistics for a cell population containing T cells. In some embodiments, the instructions comprise instructions for determining a plurality of population-level statistics for a cell population containing T cells.

In some embodiments, the computing device has in memory a machine learning model trained according to any provided method.

In some embodiments, the computing device has instructions in memory for training a machine learning model according to any provided method.

In some embodiments, the instructions comprise instructions for receiving holographic information for individual cells of a plurality of reference cell populations comprising T cells. In some embodiments, the instructions comprise instructions for receiving a dataset of reference input measurements for the plurality of reference cell populations comprising T cells. In some embodiments, the instructions comprise instructions for determining a dataset of reference input measurements for the plurality of reference cell populations comprising T cells. In some embodiments, the instructions comprise instructions for receiving a dataset of reference population-level statistics for the plurality of reference cell populations comprising T cells. In some embodiments, the instructions comprise instructions for determining a dataset of reference population-level statistics for the plurality of reference cell populations comprising T cells. In some embodiments, the instructions comprise instructions for receiving a dataset of reference population-level output measurements for the plurality of reference cell populations comprising T cells.

In some embodiments, the instructions include instructions for training a machine learning model using a dataset of reference input measurements, reference population-level statistics, and/or reference population-level output measurements according to any provided method.

In some embodiments, the instructions comprise instructions for determining a population-level output measurement for a cell population, one or more input measurements for each of a plurality of cell characteristics, or a plurality of population-level statistics from holographic information, according to any provided method. In some embodiments, the population-level output measurement for the cell population is determined using a machine learning model. In some embodiments, the instructions comprise instructions for determining a population-level output measurement for the cell population using a machine learning model.

A typical system for the provided computing device may include a system processor comprising one or more processing elements in communication with a system data store (SDS) comprising one or more storage elements. The system processor may be programmed and/or adapted to perform the functionality described herein. The system may include one or more input devices for receiving input from a user and/or a software application. The system may include one or more output devices for presenting output to the user and/or the software application. In some embodiments, the output devices may include a monitor capable of displaying a graphical representation of the described analytical functionality to the user.

The described functionality may be supported using a computer including one or more processing elements, such as a CELERON, PENTIUM, XEON, CORE 2 DUO, or CORE 2 QUAD class microprocessor (Intel Corp., Santa Clara, Calif.) or a suitable system processor including a SEMPRON, PHENOM, OPTERON, ATHLON X2, or ATHLON 64 X2 (AMD Corp., Sunnyvale, Calif.), although other general purpose processors may be used. In some embodiments, as further described below, the functionality may be distributed across multiple processing elements. The term processing element may refer to (1) a process executing on or across a particular piece of hardware, (2) a particular piece of hardware, or (1) or (2), as the context requires. Some implementations may include one or more limited special purpose processors, such as a digital signal processor (DSP), an application specific integrated circuit (ASIC), or a field programmable gate array (FPGA). Additionally, some implementations may use a combination of general-purpose and special-purpose processors.

The environment may further include an SDS that may include various primary and secondary storage elements. In one implementation, the SDS will include registers and RAM as part of the primary storage. In some implementations, the primary storage may include other forms of memory, such as cache memory, non-volatile memory (e.g., FLASH, ROM, EPROM, etc.). The SDS may also include secondary storage that includes single, multiple, and/or various server and storage elements. For example, the SDS may utilize an internal storage device connected to the system processor. In implementations where a single processing element supports all functionality, a local hard disk drive may serve as secondary storage for the SDS, and a disk operating system running on this single processing element may act as a data server that receives and services data requests.

Those skilled in the art will appreciate that the different information used in the systems and methods provided herein may be logically or physically isolated within a single device that serves as secondary storage for the SDS; multiple related data repositories accessible through a unified management system (which together serve as the SDS); or multiple independent data repositories individually accessible through non-compliant management systems (which may in some implementations be collectively referred to as the SDS). The various storage elements comprising the physical architecture of the SDS may be centrally located or distributed across various different locations.

Additionally, or instead, the functionality and approaches discussed above may be stored within and/or on one or more computer-readable storage media. Such media may include primary storage and/or secondary storage integrated with and/or internal to a computer, such as RAM and/or magnetic disks, and/or removable from the computer, such as solid-state devices or removable magnetic or optical disks. The media may utilize any technology known to those skilled in the art, including ROM, RAM, magnetic, optical, paper, and/or solid-state media technologies.

IV. Definition

Unless otherwise defined, all technical terms, notations, and other technical and scientific terms or terminology used herein are intended to have the same meaning as commonly understood by one of ordinary skill in the art to which the claimed subject matter pertains. In some instances, terms having a commonly understood meaning are defined herein for clarity and/or ease of reference, and the inclusion of such definitions herein should not necessarily be construed as indicating a material difference from the commonly understood meaning in the relevant art.

As used herein, the singular forms include plural referents unless the context clearly dictates otherwise. For example, "one" means "at least one" or "one or more." It is understood that aspects and variations described herein include aspects "consisting of" and/or "consisting essentially of" aspects and variations.

Throughout this disclosure, various aspects of the claimed subject matter are presented in a range format. It should be understood that the description in range format is provided solely for convenience and brevity and should not be construed as an inflexible limitation on the scope of the claimed subject matter. Therefore, the description of a range should be considered to specifically disclose all possible sub-ranges within that range, as well as individual numerical values. For example, if a range of values is recited, the smaller ranges between each intermediate value between the upper and lower limits of the recited range and any other intermediate value within the recited range are understood to be encompassed within the claimed subject matter. The upper and lower limits of these smaller ranges may independently be included within the recited range and are also encompassed within the claimed subject matter subject to any specifically excluded limit in the recited range. If the recited range includes one or both limits, ranges excluding either or both of the included limits are also encompassed within the claimed subject matter. This applies regardless of the breadth of the recited range.

As used herein, the term "about" refers to a generally known error range for each value. References herein to "about" a value or parameter include and describe embodiments of that value or parameter itself. For example, a description referring to "about X" includes a description of "X."

As used herein, reference to a cell or population of cells being "positive" for a particular marker refers to the detectable presence of the particular marker, typically a surface marker, on or within the cell. When referring to a surface marker, the term refers to the presence of surface expression as detected by flow cytometry (e.g., by staining with an antibody that specifically binds to the marker and detecting the antibody), wherein the staining is detectable by flow cytometry at a level substantially greater than the level of staining detected by performing the same procedure using an isotype-matched control under otherwise identical conditions, and/or at a level substantially similar to that of cells known to be positive for the marker, and/or at a level substantially greater than that of cells known to be negative for the marker.

As used herein, reference to a cell or cell population being "negative" for a particular marker refers to the absence of substantial detectable presence of the particular marker, typically a surface marker, on or within the cell. When referring to a surface marker, the term refers to the absence of surface expression as detected by flow cytometry (e.g., by staining with an antibody that specifically binds to the marker and detecting the antibody), wherein the staining is not detected by flow cytometry at a level substantially higher than the level of staining detected by performing the same procedure using an isotype-matched control under otherwise identical conditions, and/or at a level substantially lower than cells known to be positive for the marker, and/or at a level substantially similar to cells known to be negative for the marker.

As used herein, a composition refers to any mixture of two or more products, substances, or compounds, including cells. It may be a solution, suspension, liquid, powder, paste, aqueous, non-aqueous, or any combination of the foregoing.

As used herein, a “subject” is a mammal, such as a human or other animal, and is typically a human.

The term "determining" as used herein means predicting. For example, in some embodiments, determining the cellular phenotype of a T cell population, e.g., determining the activation state of the T cell population, or determining the memory phenotype of the T cell population, means predicting the cellular phenotype of the T cell population, e.g., predicting the activation state of the T cell population, or predicting the memory phenotype of the T cell population; determining one or more input measurements means predicting one or more input measurements; determining a plurality of population-level statistics means predicting a plurality of population-level statistics; and determining a population-level output measurement of expression means predicting a population-level output measurement of expression.

V. Exemplary Embodiments

The embodiments provided are as follows:

1. A method for determining an activation state of a T cell population, comprising: determining a population-level output measurement of expression of a marker expressed by activated T cells for a cell population including T cells, wherein the population-level output measurement is determined based on a plurality of population-level statistics, wherein:

Each population-level statistic is for one or more input measurements of a cell characteristic among a plurality of cell characteristics derived from holographic information obtained for the cell population, wherein the plurality of population-level statistics includes one or more population-level statistics for each of the plurality of cell characteristics;

Each input measurement is from an individual cell in the cell population.

method.

2. A method according to embodiment 1, wherein the method comprises determining a plurality of population-level statistics from one or more input measurements for each of a plurality of cell characteristics.

3. A method according to embodiment 1 or embodiment 2, wherein the method comprises determining one or more input measurements for each of the plurality of cell characteristics from the holographic information.

4. A method according to any one of embodiments 1 to 3, wherein the method comprises obtaining holographic information.

5. A method for determining the activation status of a T cell population, comprising:

(a) obtaining holographic information about a cell population including T cells;

(b) determining one or more input measurements for each of a plurality of cell characteristics derived from the holographic information, wherein each input measurement is from an individual cell within the cell population;

(c) a step of determining a plurality of population-level statistics, wherein each population-level statistic is for one or more input measurements of one of the plurality of cell characteristics, and the plurality of population-level statistics includes one or more population-level statistics for each of the plurality of cell characteristics; and

(d) determining a population-level output measure of the expression of a marker expressed by activated T cells for a cell population based on multiple population-level statistics.

6. A method according to any one of embodiments 1-5, further comprising manipulating a T cell population after determining a population-level output measurement.

7. A method according to any one of embodiments 1 to 6, wherein the holographic information is obtained by differential digital holographic microscopy (DDHM).

8. A method according to any one of embodiments 4-7, wherein obtaining holographic information comprises imaging a cell population using DDHM.

9. A method according to any one of embodiments 1-8, wherein at least one, and optionally one or more population-level statistics for each of the plurality of cell characteristics comprises one or more quantiles of one or more input measurements of the cell characteristic.

10. A method according to embodiment 9, wherein one or more quantiles are selected from the 0.01, 0.1, 0.5, 0.9, and 0.99 quantiles of one or more input measurements of a cell characteristic.

11. A method according to any one of embodiments 1-8, wherein at least one, and optionally one or more population-level statistics for each of the plurality of cell characteristics are determined by applying a distribution-based pooling filter to one or more input measurements of the cell characteristic.

12. A method according to any one of embodiments 1-11, wherein the population-level output measurement is a number of cells, a percentage of cells, a proportion of cells, or a density of cells within a cell population expressing a marker.

13. A method according to any one of embodiments 1-12, wherein the cell population is from a cell culture being cultured in vitro or ex vivo.

14. A method according to embodiment 13, wherein the holographic information is obtained during in vitro or ex vivo cultivation of a cell culture.

15. A method according to embodiment 13 or embodiment 14, wherein the population-level output measurement is for expression of a marker during in vitro or ex vivo cultivation of a cell culture.

16. A method according to any one of embodiments 13-15, wherein the in vitro or ex vivo culture is performed under conditions that expand T cells in the cell culture.

17. A method according to any one of embodiments 13-16, wherein the in vitro or ex vivo culture is performed in a bioreactor.

18. A method according to any one of embodiments 1-17, wherein the cell population is incubated under T cell stimulating conditions before obtaining holographic information.

19. A method according to any one of embodiments 1-18, wherein the method comprises incubating the cell population under T cell stimulating conditions before obtaining holographic information.

20. A method according to embodiment 18 or embodiment 19, wherein the incubation is performed prior to in vitro or ex vivo culture.

21. A method according to any one of embodiments 18-20, wherein the T cell stimulating conditions comprise incubation in the presence of a T cell stimulator that induces a primary activation signal and a co-stimulatory signal in the T cell.

22. A method according to embodiment 21, wherein the T cell stimulator comprises an anti-CD3 antibody or antibody fragment.

23. A method according to embodiment 21 or embodiment 22, wherein the T cell stimulator comprises an anti-CD28 antibody or antibody fragment.

24. A method according to any one of embodiments 21-23, wherein the T cell stimulator is immobilized on a bead.

25. A method according to any one of embodiments 1-24, wherein the cell population is enriched for T cells.

26. A method according to any one of embodiments 1-25, wherein at least 50, 55, 60, 65, 70, 75, 80, 85, 90, 91, 92, 93, 94, 95, 96, 97, 98, 99 or 100% of the cell population are T cells.

27. A method according to any one of embodiments 1-26, wherein the population-level output measurement is determined by providing the plurality of population-level statistics as input to a machine learning model trained to predict the population-level output measurement of marker expression based on the population-level statistics of the plurality of cell characteristics.

28. In embodiment 27, the machine learning model is trained using a dataset of reference population-level statistics, wherein:

For each of the first plurality of reference cell populations comprising T cells, the dataset of reference population-level statistics comprises one or more reference population-level statistics for each of the plurality of cell characteristics, wherein:

Each reference population-level statistic is for one or more reference input measurements of one cell characteristic among a plurality of cell characteristics derived from holographic information obtained for the reference cell population;

Each reference input measurement is from an individual cell in the reference cell population.

method.

29. In embodiment 28, the machine learning model is trained using a dataset of reference population-level output measurements, wherein for each of the second plurality of reference cell populations, the dataset of reference population-level output measurements includes reference population-level output measurements of expression of a marker for the reference cell population, wherein:

The first and second plurality of reference cell populations are from reference cell cultures cultured in vitro or ex vivo;

The reference cell population from the first plurality of reference cell populations is from the same reference cell culture as the reference cell population from the second plurality of reference cell populations.

method.

30. A method for training a machine learning model to predict the activation state of a T cell population, comprising training the machine learning model using the following dataset:

(i) a dataset of reference population-level statistics, wherein for each of a first plurality of reference cell populations comprising T cells, the dataset of reference population-level statistics comprises one or more reference population-level statistics for each of a plurality of cell characteristics, wherein:

Each reference population-level statistic is for one or more reference input measurements of one cell characteristic among a plurality of cell characteristics derived from holographic information obtained for the reference cell population;

Each reference input measurement is from an individual cell in the reference cell population; and

(ii) a dataset of reference population-level output measurements, wherein for each of the second plurality of reference cell populations, the dataset of reference population-level output measurements comprises a reference population-level output measurement of the expression of a marker expressed by activated T cells for the reference cell population, wherein:

The first and second plurality of reference cell populations are from reference cell cultures cultured in vitro or ex vivo;

The reference cell population from the first plurality of reference cell populations is from the same reference cell culture as the reference cell population from the second plurality of reference cell populations;

Hereby, a machine learning model is trained to predict a population-level output measure of marker expression based on population-level statistics of multiple cell features.

method.

31. A method according to any one of embodiments 1 to 30, wherein the plurality of cell characteristics are selected from intensity maxima, intensity minima, intensity entropy, intensity contrast, phase entropy, cell area, and radius average.

32. A method according to any one of embodiments 1 to 31, wherein the plurality of cell characteristics include intensity maxima, intensity minima, intensity entropy, intensity contrast, phase entropy, cell area, and radius average.

33. A method for determining an activation state of a T cell population, comprising: determining, for a cell population comprising T cells, expression of a marker expressed by activated T cells, wherein the marker is CD137, and wherein the determination is based on one or more input measurements for each of a plurality of cell characteristics derived from holographic information obtained for the cell population, wherein:

Multiple cell characteristics include intensity maximum, intensity minimum, intensity entropy, intensity contrast, phase entropy, cell area and radius mean;

Each input measurement is from an individual cell in the cell population.

method.

34. In any one of embodiments 1-30, at least one, optionally one or more input measurements of each of the plurality of cell characteristics are determined by providing holographic information about individual cells of the cell population to a convolutional neural network, wherein:

The multiple cell features are cell features extracted by a convolutional neural network;

One or more input measurements are determined from a convolutional neural network.

method.

35. In any of embodiments 28-30 and 34, at least one, optionally one or more, reference input measurements for each of the plurality of cell characteristics are determined by providing holographic information about individual cells of the reference cell population to a convolutional neural network, wherein:

The multiple cell features are cell features extracted by a convolutional neural network;

One or more reference input measurements are determined from a convolutional neural network.

method.

36. A method according to embodiment 34 or embodiment 35, wherein the convolutional neural network is trained using a dataset of reference holographic information, wherein for each of the first plurality of reference cell populations including T cells, the dataset of reference holographic information includes holographic information obtained for individual cells of the reference cell population.

37. A method according to any one of embodiments 28-32 and 34-36, wherein at least one of the plurality of cell characteristics, and optionally one or more reference population-level statistics for each, comprises one or more quantiles of one or more reference input measurements for the cell characteristic.

38. A method according to embodiment 37, wherein one or more quantiles are selected from the 0.01, 0.1, 0.5, 0.9, and 0.99 quantiles of one or more reference input measurements for cell characteristics.

39. A method according to any one of embodiments 28-32 and 34-36, wherein at least one, and optionally one or more, reference population-level statistics for each of the plurality of cell characteristics are determined by applying a distribution-based pooling filter to one or more reference input measurements for the cell characteristic.

40. A method for training a machine learning model to predict the activation state of a T cell population, comprising:

(a) a step of training a convolutional neural network using a dataset of reference holographic information, wherein for each of a first plurality of reference cell populations including T cells, the dataset of reference holographic information includes holographic information obtained for individual cells of the reference cell population;

(b) a step of determining, from a convolutional neural network, one or more reference input measurements for each cell feature among a plurality of cell features derived from holographic information, wherein the plurality of cell features are cell features extracted by the convolutional neural network, and each reference input measurement is from an individual cell of the reference cell population;

(c) determining a dataset of reference population-level statistics, wherein the dataset of reference population-level statistics comprises one or more reference population-level statistics for each of a plurality of cell characteristics, each reference population-level statistic being determined by applying a distribution-based pooling filter to one or more reference input measurements for one of the plurality of cell characteristics; and

(d) training a machine learning model using a dataset of reference population-level statistics and a dataset of reference population-level output measurements, wherein for each of the second plurality of reference cell populations, the dataset of reference population-level output measurements comprises reference population-level output measurements of expression of a marker expressed by activated T cells for the reference cell population, wherein:

The first and second plurality of reference cell populations are from reference cell cultures cultured in vitro or ex vivo;

The reference cell population from the first plurality of reference cell populations is from the same reference cell culture as the reference cell population from the second plurality of reference cell populations;

Hereby, a machine learning model is trained to predict a population-level output measure of marker expression based on population-level statistics of multiple cell features.

method.

41. A method according to any one of embodiments 34-40, wherein at least one, and optionally one or more, input measurements for each of the plurality of cell characteristics are obtained from a fully connected layer of a convolutional neural network.

42. A method according to any one of embodiments 28-32 and 34-41, wherein holographic information for the first plurality of reference cell populations is obtained by DDHM.

43. A method according to any one of embodiments 1 to 42, wherein the holographic information includes phase information and intensity information.

44. A method according to any one of embodiments 29-32 and 34-43, wherein the reference population-level output measurement is a number of cells, a percentage of cells, a proportion of cells, or a density of cells within the reference cell population expressing the marker.

45. A method according to any one of embodiments 29-32 and 34-44, wherein the dataset of reference population-level statistics and the dataset of reference population-level output measurements are time-matched.

46. A method according to any one of embodiments 29-32 and 34-45, wherein the in vitro or ex vivo cultivation of the reference cell culture is performed under the same or similar conditions as the in vitro or ex vivo cultivation of the cell culture.

47. A method according to any one of embodiments 29-32 and 34-46, wherein the holographic information for the first plurality of reference cell populations is obtained during in vitro or ex vivo culture of the reference cell culture.

48. A method according to any one of embodiments 29-32 and 34-47, wherein the dataset of reference population-level output measurements relates to expression of a marker during or after in vitro or ex vivo culturing of a reference cell culture.

49. A method according to any one of embodiments 29-32 and 34-48, wherein the dataset of reference population-level output measurements is determined using fluorescence imaging of a second plurality of reference cell populations.

50. A method according to embodiment 49, wherein the fluorescence imaging is by flow cytometry.

51. A method according to any one of embodiments 1-32 and 34-50, wherein the marker is CD137 (4-1BB).

52. A method according to any one of embodiments 28-32 and 34-51, wherein the first plurality of reference cell populations are incubated under T cell stimulating conditions before holographic information is obtained for the first plurality of reference cell populations.

53. A method according to embodiment 52, wherein incubation of the first plurality of reference cell populations is performed under the same or similar conditions as incubation of the cell population.

54. A method according to embodiment 52 or embodiment 53, wherein the incubation of the first plurality of reference cell populations is performed prior to in vitro or ex vivo culturing of the reference cell culture.

55. A method according to any one of embodiments 28-32 and 34-54, wherein the first and/or second plurality of reference cell populations are enriched for T cells.

56. A method according to any one of embodiments 28-32 and 34-55, wherein at least 50, 55, 60, 65, 70, 75, 80, 85, 90, 91, 92, 93, 94, 95, 96, 97, 98, 99 or 100% of the first and/or second plurality of reference cell populations are T cells.

57. A method for monitoring the activation status of a T cell culture, comprising determining a population-level output measurement of the expression of a marker expressed by activated T cells for a cell population from a cell culture comprising T cells cultured in vitro or ex vivo, wherein the population-level output measurement is determined according to any one of the methods of embodiments 1-29, 31, 32, 34-39 and 41-56.

58. A computing device comprising instructions in memory for performing any one of the methods of embodiments 1-29, 31, 32, 34-39 and 41-56, wherein the instructions comprise the following instructions:

(a) instructions for receiving holographic information about an individual cell in a cell population comprising T cells, one or more input measurements for each of a plurality of cell characteristics for an individual cell in a cell population comprising T cells, or a plurality of population-level statistics for a cell population comprising T cells; and

(b) instructions for determining a population-level output measurement for a cell population, one or more input measurements for each of a plurality of cell characteristics, or a plurality of population-level statistics from the holographic information, according to the method.

59. In embodiment 58, the computing device further comprises in its memory a machine learning model trained according to any one of the methods of embodiments 30-32 and 34-56, wherein the population-level output measurement for the cell population is determined using the machine learning model.

Additionally, the provided embodiments are as follows:

1. A method for determining the cell phenotype of a T cell population, comprising:

(a) obtaining holographic information about a cell population including T cells;

(b) determining one or more input measurements for each of a plurality of cell characteristics derived from the holographic information, wherein each input measurement is from an individual cell within the cell population;

(c) a step of determining a plurality of population-level statistics, wherein each population-level statistic is for one or more input measurements of one of the plurality of cell characteristics, and the plurality of population-level statistics includes one or more population-level statistics for each of the plurality of cell characteristics; and

(d) determining a population-level output measure of the expression of a marker expressed by T cells having a cell phenotype of interest for the cell population based on multiple population-level statistics.

2. A method for determining a cell phenotype of a T cell population, comprising: determining a population-level output measurement of expression of a marker expressed by T cells having a cell phenotype of interest for a cell population comprising T cells, wherein the population-level output measurement is determined based on a plurality of population-level statistics, wherein:

Each population-level statistic is for one or more input measurements of a cell characteristic among a plurality of cell characteristics derived from holographic information obtained for the cell population, wherein the plurality of population-level statistics includes one or more population-level statistics for each of the plurality of cell characteristics;

Each input measurement is from an individual cell in the cell population.

method.

3. A method according to embodiment 2, wherein the method comprises determining a plurality of population-level statistics from one or more input measurements for each of a plurality of cell characteristics.

4. A method according to embodiment 2 or embodiment 3, wherein the method comprises determining one or more input measurements for each of the plurality of cell characteristics from the holographic information.

5. A method according to any one of embodiments 2 to 4, wherein the method comprises obtaining holographic information.

6. A method according to any one of embodiments 1 to 5, wherein the holographic information is obtained by differential digital holographic microscopy (DDHM).

7. A method according to any one of embodiments 4-6, wherein obtaining holographic information comprises imaging a cell population using DDHM.

8. A method according to any one of embodiments 1-7, wherein at least one, and optionally one or more population-level statistics for each of the plurality of cell characteristics comprises one or more quantiles of one or more input measurements of the cell characteristic.

9. A method according to embodiment 8, wherein the at least one quantile comprises at least one of the 0.01, 0.1, 0.5, 0.9, and 0.99 quantiles of at least one input measurement of a cell characteristic.

10. A method according to any one of embodiments 1-7, wherein at least one, and optionally one or more population-level statistics for each of the plurality of cell characteristics are determined by applying a distribution-based pooling filter to one or more input measurements of the cell characteristic.

11. A method according to any one of embodiments 1-10, wherein the population-level output measurement is a number of cells, a percentage of cells, a proportion of cells, or a density of cells within a cell population expressing a marker.

12. A method according to any one of embodiments 1-11, wherein the cell population is from a cell culture being cultured in vitro or ex vivo.

13. A method according to embodiment 12, wherein the holographic information is obtained during in vitro or ex vivo cultivation of a cell culture.

14. A method according to embodiment 12 or embodiment 13, wherein the population-level output measurement is for the expression of a marker during in vitro or ex vivo cultivation of a cell culture.

15. A method according to any one of embodiments 12-14, wherein the in vitro or ex vivo culture is performed under conditions that expand T cells in the cell culture.

16. A method according to any one of embodiments 12-15, wherein the in vitro or ex vivo culture is performed in a bioreactor.

17. A method according to any one of embodiments 1-16, wherein the cell population is incubated under T cell stimulating conditions before obtaining holographic information.

18. A method according to any one of embodiments 1-17, wherein the method comprises incubating the cell population under T cell stimulating conditions before obtaining holographic information.

19. A method according to embodiment 17 or embodiment 18, wherein the incubation is performed prior to in vitro or ex vivo culture.

20. A method according to any one of embodiments 17-19, wherein the T cell stimulating condition comprises incubation in the presence of a T cell stimulator that induces a primary activation signal and a co-stimulatory signal in the T cell.

21. A method according to embodiment 20, wherein the T cell stimulator comprises an anti-CD3 antibody or antibody fragment.

22. A method according to embodiment 20 or embodiment 21, wherein the T cell stimulator comprises an anti-CD28 antibody or antibody fragment.

23. A method according to any one of embodiments 20-22, wherein the T cell stimulator is immobilized on a bead.

24. A method according to any one of embodiments 20-22, wherein the T cell stimulator is immobilized on an oligomeric streptavidin mutein reagent.

25. A method according to any one of embodiments 1-24, wherein the recombinant receptor is introduced into T cells of the cell population before holographic information is obtained.

26. A method according to any one of embodiments 1-25, wherein the method comprises introducing a recombinant receptor into T cells of the cell population before obtaining holographic information.

27. A method according to embodiment 25 or embodiment 26, wherein the introduction comprises contacting the cell population with an agent comprising a polynucleotide encoding a recombinant receptor.

28. A method according to embodiment 26 or embodiment 27, wherein the recombinant receptor is a chimeric antigen receptor (CAR) or an engineered T cell receptor (TCR).

29. A method according to any one of embodiments 1-28, wherein at least 50, 55, 60, 65, 70, 75, 80, 85, 90, 91, 92, 93, 94, 95, 96, 97, 98, 99 or 100% of the cell population are T cells.

30. A method according to any one of embodiments 1-29, wherein the population-level output measurement is determined by providing the plurality of population-level statistics as input to a machine learning model trained to predict the population-level output measurement of marker expression based on the population-level statistics of the plurality of cell characteristics.

31. In embodiment 30, the machine learning model is trained using a dataset of reference population-level statistics, wherein:

For each of the first plurality of reference cell populations comprising T cells, the dataset of reference population-level statistics comprises one or more reference population-level statistics for each of the plurality of cell characteristics, wherein:

Each reference population-level statistic is for one or more reference input measurements of one cell characteristic among a plurality of cell characteristics derived from holographic information obtained for the reference cell population;

Each reference input measurement is from an individual cell in the reference cell population.

method.

32. In embodiment 31, the machine learning model is trained using a dataset of reference population-level output measurements, wherein for each of the second plurality of reference cell populations, the dataset of reference population-level output measurements includes reference population-level output measurements of expression of a marker for the reference cell population, wherein:

The first and second plurality of reference cell populations are from reference cell cultures cultured in vitro or ex vivo;

The reference cell population from the first plurality of reference cell populations is from the same reference cell culture as the reference cell population from the second plurality of reference cell populations.

method.

33. A method for training a machine learning model to predict a cell phenotype of a T cell population, comprising training the machine learning model using the following dataset:

(i) a dataset of reference population-level statistics, wherein for each of a first plurality of reference cell populations comprising T cells, the dataset of reference population-level statistics comprises one or more reference population-level statistics for each of a plurality of cell characteristics, wherein:

Each reference population-level statistic is for one or more reference input measurements of one cell characteristic among a plurality of cell characteristics derived from holographic information obtained for the reference cell population;

Each reference input measurement is from an individual cell in the reference cell population; and

(ii) a dataset of reference population-level output measurements, wherein for each of the second plurality of reference cell populations, the dataset of reference population-level output measurements comprises a reference population-level output measurement of the expression of a marker expressed by T cells having a cell phenotype of interest for the reference cell population, wherein:

The first and second plurality of reference cell populations are from reference cell cultures cultured in vitro or ex vivo;

The reference cell population from the first plurality of reference cell populations is from the same reference cell culture as the reference cell population from the second plurality of reference cell populations;

Hereby, a machine learning model is trained to predict a population-level output measure of marker expression based on population-level statistics of multiple cell features.

method.

34. A method according to any one of embodiments 1 to 33, wherein the plurality of cell characteristics include at least one of intensity skewness, intensity correlation, intensity homogeneity, intensity maximum, intensity minimum, intensity entropy, intensity contrast, phase entropy, cell area, and radius average.

35. A method according to any one of embodiments 1 to 34, wherein the plurality of cell characteristics include intensity maxima, intensity minima, intensity entropy, intensity contrast, phase entropy, cell area, and radius average.

36. A method for determining an activation state of a T cell population, comprising: determining, for a cell population comprising T cells, a population-level output measurement of expression of a marker expressed by activated T cells, wherein the marker is CD137 (4-1BB), and wherein the population-level output measurement is determined based on one or more input measurements for each of a plurality of cell characteristics derived from holographic information obtained for the cell population, wherein:

Multiple cell characteristics include intensity maximum, intensity minimum, intensity entropy, intensity contrast, phase entropy, cell area and radius mean;

Each input measurement is from an individual cell in the cell population.

method.

37. A method according to any one of embodiments 1-36, wherein the plurality of cell characteristics include intensity skewness, intensity correlation, and intensity homogeneity.

38. A method for determining an activation state of a T cell population, comprising: determining, for a cell population comprising T cells, a population-level output measurement of expression of a marker expressed by activated T cells, wherein the marker is CD137 (4-1BB), and wherein the population-level output measurement is determined based on one or more input measurements for each of a plurality of cell characteristics derived from holographic information obtained for the cell population, wherein:

Multiple cellular characteristics include intensity skewness, intensity correlation and intensity homogeneity;

Each input measurement is from an individual cell in the cell population.

method.

39. A method according to embodiment 38, wherein the plurality of cell characteristics include intensity maximum, intensity minimum, intensity entropy, intensity contrast, phase entropy, cell area, and radius average.

40. A method according to any one of embodiments 1-33, wherein the plurality of cell characteristics comprise at least one of peak area, phase average uniformity, intensity geometric mean, minimum optical height, and normalized optical height.

41. A method according to any one of embodiments 1-33 and 40, wherein the plurality of cell characteristics include peak area, phase average uniformity, intensity geometric mean, minimum optical height, and normalized optical height.

42. A method according to any one of embodiments 1-33, wherein the plurality of cell characteristics comprise at least one of cell area, perimeter length, average intensity, normalized peak area, and equivalent peak diameter.

43. A method according to any one of embodiments 1-33 and 42, wherein the plurality of cell characteristics include cell area, perimeter length, average intensity, normalized peak area, and equivalent peak diameter.

44. A method for determining a memory phenotype of a T cell population, comprising: determining, for a cell population comprising T cells, a population-level output measurement of expression of a marker expressed by central memory T cells or stem cell memory T cells, wherein the marker is CCR7, and wherein the population-level output measurement is determined based on one or more input measurements for each of a plurality of cell characteristics derived from holographic information obtained for the cell population, wherein:

The multiple cell characteristics include one or more of cell area, perimeter length, mean intensity, normalized peak area, and equivalent peak diameter;

Each input measurement is from an individual cell in the cell population.

method.

45. A method according to embodiment 44, wherein the plurality of characteristics include cell area, perimeter length, average intensity, normalized peak area, and equivalent peak diameter.

46. A method for determining expression of a recombinant receptor in a T cell population, comprising: determining, for a cell population comprising T cells, expression of a recombinant receptor introduced into T cells of the cell population, wherein the determination is based on one or more input measurements for each of a plurality of cell characteristics derived from holographic information obtained for the cell population, wherein:

Multiple cell characteristics include peak area, phase average uniformity, intensity geometric mean, minimum optical height, and normalized optical height;

Each input measurement is from an individual cell in the cell population.

method.

47. In any one of embodiments 1-33, at least one, optionally one or more input measurements for each of the plurality of cell characteristics are determined by providing holographic information about individual cells of the cell population to a convolutional neural network, wherein:

The multiple cell features are cell features extracted by a convolutional neural network;

One or more input measurements are determined from a convolutional neural network.

method.

48. In any one of embodiments 31-33 and 47, at least one, optionally one or more, reference input measurements for each of the plurality of cell characteristics are determined by providing holographic information about individual cells of the reference cell population to a convolutional neural network, wherein:

The multiple cell features are cell features extracted by a convolutional neural network;

One or more reference input measurements are determined from a convolutional neural network.

method.

49. A method according to embodiment 47 or embodiment 48, wherein the convolutional neural network is trained using a dataset of reference holographic information, wherein for each of the first plurality of reference cell populations including T cells, the dataset of reference holographic information includes holographic information obtained for individual cells of the reference cell population.

50. A method according to any one of embodiments 31-35, 37, 40-43 and 47-49, wherein at least one of the plurality of cell characteristics, and optionally one or more reference population-level statistics for each, comprises one or more quantiles of one or more reference input measurements for the cell characteristic.

51. A method according to embodiment 50, wherein one or more quantiles are selected from the 0.01, 0.1, 0.5, 0.9, and 0.99 quantiles of one or more reference input measurements for cell characteristics.

52. A method according to any one of embodiments 31-35, 37, 40-43 and 47-49, wherein at least one, and optionally one or more, reference population-level statistics for each of the plurality of cell characteristics are determined by applying a distribution-based pooling filter to one or more reference input measurements for the cell characteristic.

53. A method for training a machine learning model to predict a cell phenotype of a T cell population, comprising:

(a) a step of training a convolutional neural network using a dataset of reference holographic information, wherein for each of a first plurality of reference cell populations including T cells, the dataset of reference holographic information includes holographic information obtained for individual cells of the reference cell population;

(b) a step of determining, from a convolutional neural network, one or more reference input measurements for each cell feature among a plurality of cell features derived from holographic information, wherein the plurality of cell features are cell features extracted by the convolutional neural network, and each reference input measurement is from an individual cell of the reference cell population;

(c) determining a dataset of reference population-level statistics, wherein the dataset of reference population-level statistics comprises one or more reference population-level statistics for each of a plurality of cell characteristics, each reference population-level statistic being determined by applying a distribution-based pooling filter to one or more reference input measurements for one of the plurality of cell characteristics; and

(d) training a machine learning model using a dataset of reference population-level statistics and a dataset of reference population-level output measurements, wherein for each of the second plurality of reference cell populations, the dataset of reference population-level output measurements comprises reference population-level output measurements of expression of a marker expressed by T cells having a cell phenotype of interest for the reference cell population, wherein:

The first and second plurality of reference cell populations are from reference cell cultures cultured in vitro or ex vivo;

The reference cell population from the first plurality of reference cell populations is from the same reference cell culture as the reference cell population from the second plurality of reference cell populations;

Hereby, a machine learning model is trained to predict a population-level output measure of marker expression based on population-level statistics of multiple cell features.

method.

54. A method according to any one of embodiments 47-53, wherein at least one, and optionally one or more, input measurements for each of the plurality of cell characteristics are obtained from a fully connected layer of a convolutional neural network.

55. A method according to any one of embodiments 31-35, 37, 40-43 and 47-54, wherein holographic information for the first plurality of reference cell populations is obtained by DDHM.

56. A method according to any one of embodiments 1 to 55, wherein the holographic information includes phase information and intensity information.

57. A method according to any one of embodiments 32-35, 37, 40-43 and 47-56, wherein the reference population-level output measurement is the number of cells, the percentage of cells, the proportion of cells or the density of cells in the reference cell population expressing the marker.

58. A method according to any one of embodiments 32-35, 37, 40-43 and 47-57, wherein the dataset of reference population-level statistics and the dataset of reference population-level output measurements are time-matched.

59. A method according to any one of embodiments 32-35, 37, 40-43 and 47-58, wherein the in vitro or ex vivo cultivation of the reference cell culture is performed under the same or similar conditions as the in vitro or ex vivo cultivation of the cell culture.

60. A method according to any one of embodiments 32-35, 37, 40-43 and 47-59, wherein the holographic information for the first plurality of reference cell populations is obtained during in vitro or ex vivo culturing of the reference cell culture.

61. A method according to any one of embodiments 32-35, 37, 40-43 and 47-60, wherein the dataset of reference population-level output measurements relates to expression of a marker during or after in vitro or ex vivo culturing of a reference cell culture.

62. A method according to any one of embodiments 32-35, 37, 40-43 and 47-61, wherein the dataset of reference population-level output measurements is determined using fluorescence imaging of a second plurality of reference cell populations.

63. A method according to embodiment 62, wherein the fluorescence imaging is by flow cytometry.

64. A method according to any one of embodiments 1-35, 37 and 47-63, wherein the method is for determining the activation state of a T cell population, and the marker is expressed by the activated T cells.

65. A method according to any one of embodiments 1-35, 37 and 47-64, wherein the marker is CD137 (4-1BB).

66. In any one of embodiments 1-35 and 42-63, the method is for determining the memory phenotype of a T cell population, and the marker is expressed by T cells having a central memory phenotype or a stem cell memory phenotype.

67. A method according to embodiment 66, wherein the marker is expressed by a T cell having a central memory phenotype.

68. A method according to embodiment 66, wherein the marker is expressed by T cells having a stem cell memory phenotype.

69. A method according to any one of embodiments 1-35, 42-63 and 66-68, wherein the marker is CCR7.

70. In any one of embodiments 1-33 and 40-63, the method is for determining recombinant receptor expression in a T cell population, wherein the marker is a recombinant receptor introduced into T cells of the cell population before holographic information is obtained.

71. A method according to any one of embodiments 46-63 and 70, wherein the recombinant receptor is a chimeric antigen receptor (CAR) or an engineered T cell receptor (TCR).

72. A method according to any one of embodiments 31-35, 37, 40-43 and 47-71, wherein the first plurality of reference cell populations are incubated under T cell stimulating conditions before holographic information is obtained for the first plurality of reference cell populations.

73. A method according to embodiment 72, wherein the incubation of the first plurality of reference cell populations is performed under the same or similar conditions as the incubation of the cell population.

74. A method according to embodiment 72 or embodiment 73, wherein the incubation of the first plurality of reference cell populations is performed prior to in vitro or ex vivo culturing of the reference cell culture.

75. A method according to any one of embodiments 31-35, 37, 40-43 and 47-74, wherein the first and/or second plurality of reference cell populations are enriched for T cells.

76. A method according to any one of embodiments 31-35, 37, 40-43 and 47-75, wherein at least 50, 55, 60, 65, 70, 75, 80, 85, 90, 91, 92, 93, 94, 95, 96, 97, 98, 99 or 100% of the first and/or second plurality of reference cell populations are T cells.

77. A method for monitoring a cell phenotype of a T cell culture, comprising determining a population-level output measurement of the expression of a marker expressed by T cells having a cell phenotype of interest for a cell population from a cell culture comprising T cells cultured in vitro or ex vivo, wherein the population-level output measurement is determined according to any one of the methods of embodiments 1-32, 34, 35, 37, 40-43, 47-52, and 54-76.

78. A computing device comprising instructions in memory for performing any one of the methods of embodiments 1-32, 34, 35, 37, 40-43, 47-52, and 54-76, wherein the instructions comprise the following instructions:

(a) instructions for receiving holographic information about an individual cell in a cell population comprising T cells, one or more input measurements for each of a plurality of cell characteristics for an individual cell in a cell population comprising T cells, or a plurality of population-level statistics for a cell population comprising T cells; and

(b) instructions for determining a population-level output measurement for a cell population, one or more input measurements for each of a plurality of cell characteristics, or a plurality of population-level statistics from the holographic information, according to the method.

79. In embodiment 78, the computing device further comprises in a memory a machine learning model trained according to any one of the methods of embodiments 33-35, 37, 40-43, and 47-76, wherein the population-level output measurement for the cell population is determined using the machine learning model.

VI. Examples

The following examples are included for illustrative purposes only and are not intended to limit the scope of the present invention.

Example 1: Machine learning method for determining overall activation state of a T cell population using holographic imaging.

We developed a machine learning method to determine the overall activation state of a T cell population using holographic imaging. To assess overall activation state, T cell populations were stimulated to induce activation, and then periodically evaluated during cultivation in a bioreactor to determine the percentage of cells expressing surface expression of an exemplary T cell activation marker (in this case, CD137 (4-1BB)). CD137 activation marker expression was assessed using flow cytometry. To predict flow cytometry results of CD137 expression as a surrogate for overall activation state, a machine learning method was trained using population-level statistics summarizing cell characteristic measurements across individual cells in the stimulated T cell population. Cell characteristic measurements were derived from holographic information about the stimulated T cell population captured during cultivation.

In this approach, activation marker expression in individual cells was not required or utilized for model training, thereby eliminating the need for any specialized equipment or reagents to capture matched holographic and fluorescent images of individual cells. Furthermore, because fluorescently labeled cells were not imaged holographically, any random or systematic morphometric changes in labeled cells compared to unlabeled cells were avoided, thereby avoiding the impact of these changes on cell feature measurements and model prediction accuracy.

Furthermore, using this approach, the same holographic imaging system used to obtain images for training a machine learning model can be used to obtain images for predicting the overall T cell activation state using the trained machine learning model. This contrasts with other approaches that may use different imaging systems to obtain images for model training and model use, such as a dual fluorescence-holographic imaging system for model training and a dedicated holographic imaging system for model use.

In this way, we developed an accurate, label-free method to monitor the overall T cell activation status.

A. Stimulation Process

Primary T cells from healthy human subjects were cryopreserved, thawed, and subjected to stimulation to activate T cells. Specifically, T cells were stimulated in the presence of superparamagnetic polystyrene-coated beads (Dynabeads, Invitrogen) conjugated to recombinant IL-2, recombinant IL-7, recombinant IL-15, and anti-CD3 and anti-CD28 monoclonal antibodies for approximately 24 hours (18-30 hours) at about 37°C and 5% CO ₂ . T cells were stimulated in serum-free medium at a 1:1 bead-to-cell ratio. Following stimulation, T cells were then transduced with lentiviral vectors encoding exemplary recombinant proteins (in this case, one of two exemplary chimeric antigen receptors (CARs)) by spinoculation for 60 minutes. After spinoculation, T cells were incubated at approximately 37°C and 5% CO ₂ for approximately 24 hours (18–30 hours) before being transferred to a shaking motion bioreactor and cultured for expansion in serum-free medium containing recombinant IL-2, IL-7, and IL-15. T cells were cultured for expansion in the bioreactor for 3–5 days before harvest. Anti-CD3/anti-CD28 beads were removed from the T cells after harvest.

This process was implemented on 15 primary T cell compositions from 15 human subjects. Cellular signature measurements and activation marker expression estimates for each of the 15 primary T cell compositions were obtained during the cultivation phase and used for model training and validation as described in Sections B-D below.

B. Cellular trait measurements

Holographic imaging of T cell samples was performed continuously using an exemplary digital holographic microscopy (DHM) imaging system (in this case, the Obigio iLine F imaging system (Obigio Imaging Systems NV/SA)). The imaging system was connected to the bioreactor via a disposable tubing system. Every minute, the T cell sample was moved from the bioreactor, imaged using the DHM imaging system, and returned to the bioreactor.

The characteristics of individual cells within each sample were derived from the captured holographic information (phase and intensity information). To this end, OsOne software (Ovisio Imaging Systems NV/SA) was used to segment the captured information, which was then used to determine the characteristics listed in Table E1 for each individual cell.

Table E1: Individual T cell characteristics from holographic imaging

C. Expression of activation markers

During the same time period during which T cells were holographically imaged, T cell samples were collected from the bioreactor every 24 hours, and the percentage of CD137-expressing activated T cells per sample was quantified using flow cytometry. For this purpose, CD137-expressing cells were fluorescently labeled using a fluorophore-conjugated anti-CD137 antibody (BioLegend, clone 4B4-1 conjugated to Brilliant Violet 421™, catalog no. 309820). Flow cytometry data were acquired using a BD Biosciences Fortessa Dual Symphony instrument and BD Biosciences FACSDiva™ software, and the percentage of CD137-expressing cells per sample was determined using flow analysis software (FlowJo v10.6). The time course of the measured percentages was then interpolated to estimate the percentage of CD137-expressing cells at each cultivation time (the same time point at which the holographic information was captured).

D. Model Training and Validation

A supervised regression model was trained to predict the overall activation state of a T cell population. The regression model was trained using (i) time-matched population-level statistics summarizing the cell feature measurements described in Section B and (ii) the estimated percentage of CD137-expressing cells described in Section C. For the population-level statistics, the 0.01, 0.1, 0.5, 0.9, and 0.99 quantiles of the cell feature measurements across cells were used for model training. Cell feature measurements were included during quantile calculations only for (i) cells classified as alive, (ii) cells with a radius mean cell feature measurement > 5, (iii) cells with an intensity smoothness cell feature measurement < 0.03, (iv) cells with a cell area cell feature measurement > 60, and (v) cells with a circularity cell feature measurement > 0.5.

We tested several feature selection and regression modeling techniques, including linear regression after dimensionality reduction via principal component analysis; elastic net regression; boosted decision trees (using the software package "xgboost"); and random forests. For model validation, we used a leave-one-group-out cross-validation procedure, where data from one of the 15 primary T cell compositions was held out for each model training and testing cycle and then used as validation data for model evaluation.

Optimal prediction performance was achieved using a boosted decision tree, and the results are shown in Figure 1. Figure 1 shows the model-predicted percentage of CD137-expressing cells (y-axis) versus the measured percentage of CD137-expressing cells (x-axis) at each time point for the holdout data from each subject (data from each subject are designated by a unique symbol), interpolated from daily flow cytometry measurements to that time point, as well as the best-fit regression line for all predicted percentages (y = 7.2 + 0.79 x, Radj ² = 0.77). As shown in Figure 1, the model-predicted percentages were positively correlated with the flow cytometry-based estimated percentages. The trained model achieved a root mean square deviation (RMSD) error of 8.2% on the holdout data.

The trained model was also evaluated to determine the most informative cell features for predictive accuracy, primarily measured through RMSD and ^R2 values. The Gini index was used to assess cell feature informativeness. Based on this evaluation, the most informative cell features for predictive accuracy included intensity maxima, intensity minima, intensity entropy, intensity contrast, phase entropy, cell area, and mean radius. Overall, cells predicted to be activated based on CD137 expression had higher intensity, were larger, more textured, and less rounded, whereas cells predicted to be non-activated based on CD137 expression had lower intensity, were smaller, smoother, and more rounded. Certain quantiles were also informative. For example, the 0.90 quantile of the intensity-contrast cell feature was ranked as the most informative (>20% relative importance).

These results validate the feasibility of using holographic imaging to determine and monitor the overall activation status of T cell populations during cultivation. Specifically, the results demonstrate that CD137 is a useful marker for predicting overall activation status. Furthermore, the results are consistent with the finding that CD137 expression correlates with changes in the size of activated T cells. Furthermore, as demonstrated herein, such monitoring can be performed without labeling individual cells, using (i) population-level statistics summarizing individual cell characteristic measurements and (ii) a machine learning model trained on the percentage of cells expressing exemplary activation markers.

While currently available methods for monitoring and characterizing the activation status of T cells can be time-consuming or require specialized reagents, equipment, or trained personnel, the label-free method described herein can be used to monitor and characterize T cell activation using a non-destructive and non-damaging approach that allows T cells to be imaged from a bioreactor and returned to the bioreactor intact. The described method can be used to monitor the kinetics of T cell activation during cultivation without frequent cell sampling or demanding analytical techniques. This information can be used to assess the quality or batch-to-batch variability of T cells applied to the manufacturing process, as well as to optimize the duration or other conditions of the manufacturing process or its steps to improve the quality of processed T cells or reduce batch-to-batch variability.

Example 2: A deep learning method for determining the overall activation state of a T cell population using holographic imaging.

A deep learning method was developed to directly determine the overall activation state of a T cell population from holographic images. To this end, a deep multi-instance learning model was trained using the individual-cell holographic images and the estimated percentage of CD137-expressing cells described in Example 1. The holographic images used for training included only images of cells classified as (i) alive, (ii) having a radius average cell feature metric > 5, (iii) having an intensity smoothness cell feature metric < 0.03, (iv) having a cell area cell feature metric > 60, and (v) having a circularity cell feature metric > 0.5.

The deep learning model was implemented based on the deep learning model described in the literature [Oner et al., 2022, Patterns 3: 100399], using a pooled “bag” of individual-cell holographic images associated with each imaging time point to train a deep learning regressor to predict the time-matched estimated percentage of CD137-expressing cells from the bag of images. Specifically, the model comprised three components: a feature extractor module, a pooling filter, and a representation transformation module. The feature extractor module was implemented using a previously trained ResNet18 convolutional neural network with 32x32-pixel images of segmented individual cells as input. After training, the ResNet18 convolutional neural network received a bag of 64 segmented individual cell images as input. From these images, 128 individual cell features were automatically extracted from each image based on the 128 nodes in the last fully-connected layer of the trained ResNet18 model. These extracted features were fed into a pooling filter, which computed population-level statistics summarizing the extracted features across the images within the bag. The computed population-level statistics were the estimated marginal distributions of the extracted features across the images within the bag. Finally, the population-level statistics for each bag, along with the time-matched estimated percentage of CD137-expressing cells, were fed as input to train a representation transformation module implemented as a three-layer multilayer perceptron.

Using this approach, the trained model achieved a RMSD error of 7.84% using holdout validation. Therefore, we developed an accurate, label-free approach for predicting the overall activation of T cell populations. Using this approach, the selection or design of individual cell features derived from holographic images was not required prior to model training. Instead, individual cell features were automatically extracted from holographic images as part of model training.

Additionally, the method can be used for multi-class prediction using a representation transformation module with multiple output nodes to predict the expression of multiple cell markers (e.g., CD137, CAR and CD4).

Example 3: Determination of the overall activation state of a T cell population subjected to an alternative stimulation process.

Using the machine learning approach described in Example 1, a model was trained and validated to predict the overall activation state of a T cell population subjected to alternative stimulation processes.

Primary T cells from healthy human subjects were cryopreserved, thawed, and subjected to stimulation to activate T cells. Specifically, T cells were stimulated in the presence of 4 μg/106 cells of anti-CD3/anti-CD28 stimulation reagent prepared using recombinant IL-2, recombinant IL-7, recombinant IL- ¹⁵ , and oligomeric streptavidin mutein reagent (WO 2018/197949; Poltorak et al., Scientific Reports (2020)) at about 37°C and 5% _CO2 for approximately 24 h (18-30 h). The oligomeric streptavidin mutein reagent had an average hydrodynamic radius of 90-120 nm and contained an average of 2000-2800 streptavidin mutein tetramers (Strep-Tactin® m2, SEQ ID NO: 21). The oligomeric streptavidin mutein reagent was mixed at room temperature with (i) anti-CD3 Fab fragments individually fused to a streptavidin-binding peptide sequence (Twin-Strep-Tag®, SEQ ID NO: 15) at the carboxy-terminus of the heavy chain and (ii) anti-CD28 Fab fragments also individually fused to a streptavidin-binding peptide sequence (Twin-Strep-Tag®, SEQ ID NO: 15) at the carboxy-terminus of the heavy chain. Peptide-tagged Fab fragments were recombinantly produced (see International Patent Application Publication Nos. WO 2013/011011 and WO 2013/124474). The anti-CD3 Fab fragment was derived from a CD3-binding monoclonal antibody produced by the hybridoma cell line OKT3 (ATCC® CRL-8001™; see also U.S. Patent No. 4,361,549) and contained the heavy chain variable domain (SEQ ID NO: 5) and light chain variable domain (SEQ ID NO: 6) of the anti-CD3 antibody OKT3 described in Arakawa et al., J. Biochem. 120, 657-662 (1996). The anti-CD28 Fab fragment was derived from the antibody CD28.3 (deposited as a synthetic single-chain Fv construct under GenBank Accession No. AF451974.1; see also Vanhove et al., BLOOD, 15 July 2003, Vol. 102, No. 2, pages 564-570) and contained the heavy chain variable domain (SEQ ID NO: 7) and the light chain variable domain (SEQ ID NO: 8) of the anti-CD28 antibody CD28.3. To prepare the anti-CD3/anti-CD28 stimulation reagent, 0.3 mg of oligomeric streptavidin mutein reagent, 0.5 μg of peptide-tagged anti-CD3 Fab fragment, and 0.5 μg of peptide-tagged anti-CD28 Fab fragment were used.

After stimulation, T cells were engineered to express an exemplary recombinant protein (in this case, a recombinant T cell receptor (TCR)). After engineering, T cells were incubated at approximately 37°C and 5% CO ₂ for approximately 24 hours (18-30 hours) before being transferred to a shaking motion bioreactor and cultured for expansion in serum-free medium containing recombinant IL-2, IL-7, and IL-15. T cells were cultured for expansion in the bioreactor for 5-7 days prior to harvest.

This process was implemented on 16 primary T cell compositions from 16 human subjects. Cellular signature measurements and activation marker expression estimates for each of the 16 primary T cell compositions were obtained during the cultivation phase and used for model training and validation as described in Example 1. The holographic images used for training included only images of cells classified as live.

We tested several feature selection and regression modeling techniques, including linear regression after dimensionality reduction via principal component analysis; elastic net regression; boosted decision trees (using the software package "xgboost"); and random forests. For model validation, we used a leave-one-group-out cross-validation procedure, where data from one of the 16 primary T cell compositions was held out for each model training and testing cycle and then used as validation data for model evaluation.

Optimal prediction performance was achieved using a boosted decision tree, and the results are shown in Figure 2. Figure 2 shows the model-predicted percentage of CD137-expressing cells (y-axis) versus the measured percentage of CD137-expressing cells (x-axis) at each time point for the holdout data from each subject (data from each subject are designated by a unique symbol), interpolated from daily flow cytometry measurements to that time point, as well as the best-fit regression line for all predicted percentages (y = 6.75 + 0.52x, Radj ² = 0.56). As shown in Figure 2, the model-predicted percentages were positively correlated with the flow cytometry-based estimated percentages. The trained model achieved a root mean square deviation (RMSD) error of 7.9% on the holdout data.

The trained model was also evaluated to determine the most informative cell features for predictive accuracy, primarily measured through root mean square deviation (RMSD) and ^R2 values. The Shapley Additional Explanation (SHAP) value was used to assess cell feature informativeness. Based on this evaluation, the most informative cell features for predictive accuracy included intensity skewness, which was positively correlated with cell activation, as well as intensity correlation and intensity homogeneity, which were negatively correlated with cell activation.

These results further validate that holographic imaging can be used to determine and monitor the overall activation state of T cell populations during cultivation.

Example 4: Machine learning method for determining recombinant receptor expression in a T cell population using holographic imaging.

A model to predict recombinant receptor expression in a T cell population was trained and validated using the machine learning approach described in Example 1.

For this model, data from 24 primary T cell compositions from 24 human subjects were used. Some of the primary T cell compositions were subjected to the stimulation process described in Example 1 and engineered to express one of two exemplary chimeric antigen receptors (CARs) or an exemplary recombinant T cell receptor (TCR). The remaining primary T cell compositions were subjected to the stimulation process described in Example 2 and engineered to express an exemplary recombinant T cell receptor (TCR). Recombinant receptor (CAR or TCR) expression was assessed using fluorophore-conjugated anti-CAR or anti-TCR antibodies. Cellular signature measurements and recombinant receptor expression estimates for each of the 24 primary T cell compositions were obtained during the cultivation phase and used for model training and validation as described in Example 1. The holographic images used for training included only images of cells classified as live.

We tested several feature selection and regression modeling techniques, including linear regression after dimensionality reduction via principal component analysis; elastic net regression; boosted decision trees (using the software package "xgboost"); and random forests. For model validation, we used a leave-one-group-out cross-validation procedure, where data from one of the 24 primary T cell compositions was held out for each model training and testing cycle and then used as validation data for model evaluation.

Optimal prediction performance was achieved using a boosted decision tree, and the results are shown in Figure 3 . Figure 3 shows the model-predicted percentage of recombinant receptor-expressing cells (y-axis) at each time point versus the measured percentage of recombinant receptor-expressing cells (x-axis) for the holdout data from each subject (data from each subject are designated by a unique symbol), interpolated from daily flow cytometry measurements to that time point, as well as the best-fit regression line for all predicted percentages (y = 21.03 + 0.47x, Radj ² = 0.2). As shown in Figure 3 , the model-predicted percentages were positively correlated with the flow cytometry-based estimated percentages. The trained model achieved a root mean square deviation (RMSD) error of 10.4% on the holdout data.

The trained models were also evaluated to determine the most informative cell features for prediction accuracy, which were primarily measured through RMSD and ^R2 values. SHAP values were used to assess cell feature informativeness. Based on this evaluation, the most informative cell features for prediction accuracy included peak area (positively correlated with recombination receptor expression), phase average uniformity (negatively correlated with recombination receptor expression), intensity geometric mean (negatively correlated with recombination receptor expression), minimum optical height (positively correlated with recombination receptor expression), and normalized optical height (positively correlated with recombination receptor expression).

These results demonstrate that holographic imaging can be used to determine and monitor recombinant receptor expression in T cell populations during cultivation. The label-free method described herein allows for a non-destructive and non-damaging approach to monitor recombinant receptor expression, allowing for assessing whether T cells have reached a threshold level of recombinant receptor expression during cultivation, for example, before harvesting the cells and preparing them for ultimate patient administration.

Example 5: Machine learning method for determining the early memory phenotype of a T cell population using holographic imaging.

We developed a machine learning method using holographic imaging to determine the initial memory phenotype (e.g., stem cell memory and stem cell memory phenotype) of a T cell population. The initial memory phenotype has been reported to lead to sustained in vivo responses of CAR T cell therapy, given its proliferative and effector capabilities in both hematological and solid tumor settings (Gargett et al., 2019, Cytotherapy 21(6):593-602). Predicting CCR7 positivity can be used to understand the evolution of T cell phenotype during manufacturing without frequent sampling or demanding analytical techniques. The obtained information can then be further used to assess product quality and optimize manufacturing processes. The percentage of CCR7-positive cells during expansion can be used to understand the memory phenotype of a drug product and inform unit operation duration and/or conditions to ensure a high-quality, consistent product. Additionally, this approach may enable multi-attribute T cell phenotype monitoring, which can be used to assess batch-to-batch variability and develop process control strategies to modify manufacturing conditions to minimize this variability in the final product.

To assess the early memory phenotype, T cell populations were stimulated to induce activation, and the percentage of cells expressing surface expression of an exemplary T cell early memory phenotype marker (in this case, CCR7) was assessed periodically at 24-hour intervals during cultivation in a bioreactor. Flow cytometry was used to quantify the percentage of cells expressing the CCR7 marker.

To predict flow cytometry results of CCR7 expression as a surrogate for early memory phenotype, a machine learning method was trained using population-level statistics summarizing cell characteristic measurements across individual cells in a stimulated T cell population. Cell characteristic measurements were derived from holographic information about the stimulated T cell population captured during cultivation.

In this approach, marker expression of individual cells was not required or utilized in model training, thereby eliminating the need for any specialized equipment or reagents to capture matched holographic and fluorescent images of individual cells. Furthermore, fluorescently labeled cells were not imaged holographically, avoiding the impact of any random or systematic morphometric changes in labeled cells compared to unlabeled cells on cell feature measurements and model prediction accuracy.

Furthermore, using this approach, the same holographic imaging system used to obtain images for training a machine learning model can be used to obtain images for predicting early memory phenotypes using the trained machine learning model. This contrasts with other approaches that may use different imaging systems to obtain images for model training and use, such as a dual fluorescence-holographic imaging system for model training and a dedicated holographic imaging system for model use.

In this way, we developed an accurate, label-free method to monitor early T cell memory phenotype.

A. Stimulation Process

Primary T cells from healthy human subjects were cryopreserved, thawed, and subjected to stimulation to activate T cells. Specifically, T cells were stimulated for approximately 24 hours (18-30 hours) at approximately 37°C and 5% CO ₂ in the presence of superparamagnetic polystyrene-coated beads (Dynabeads, Invitrogen) conjugated to recombinant IL-2, recombinant IL-7, recombinant IL-15, and anti-CD3 and anti-CD28 monoclonal antibodies. T cells were stimulated in serum-free medium at a 1:1 bead-to-cell ratio. Following stimulation, T cells were then transduced with lentiviral vectors encoding exemplary recombinant proteins (in this case, one of two exemplary chimeric antigen receptors (CARs)) by spinoculation for 60 minutes. After spinoculation, T cells were incubated at approximately 37°C and 5% CO ₂ for approximately 24 hours (18–30 hours) before being transferred to a shaking motion bioreactor and cultured for expansion in serum-free medium containing recombinant IL-2, IL-7, and IL-15. T cells were cultured for expansion in the bioreactor for 3–5 days before harvest. Anti-CD3/anti-CD28 beads were removed from the T cells after harvest.

This process was implemented on 26 primary T cell compositions from 26 human subjects. Cell signature measurements and memory phenotype marker expression estimates for each of the 26 primary T cell compositions were obtained during the cultivation phase and used for model training and validation as described in Sections B-D below.

B. Cellular trait measurements

Holographic imaging of T cell samples was performed continuously using an exemplary digital holographic microscopy (DHM) imaging system (in this case, the Obigio iLine F imaging system (Obigio Imaging Systems NV/SA)). The imaging system was connected to the bioreactor via a disposable tubing system. Every minute, the T cell sample was moved from the bioreactor, imaged using the DHM imaging system, and returned to the bioreactor.

Characteristics of individual cells within each sample were derived from the captured holographic information. These characteristics were utilized to quantify individual imaged cells in terms of morphology, phase, and intensity texture, to predict the viability status of cells in culture, and to provide automated, label-free measurements of viable cell concentration during expansion. To derive these characteristics from the captured holographic information, OsOne software (Ovisio Imaging Systems NV/SA) was used to segment the captured information, which was then used to determine the characteristics listed in Table E1 for each individual cell.

C. Expression of memory phenotype markers

During the same time period during which T cells were holographically imaged, T cell samples were collected from the bioreactor every 24 hours, and the percentage of CCR7-expressing activated T cells per sample was quantified using flow cytometry. The mixed T cell population was exposed to purified IgG to saturate the Fc receptors on the cells, preventing nonspecific binding of the antibody of interest to other surface markers, such as CCR7.

After this blocking step, a CCR7-specific antibody conjugated to a fluorophore (PE-Dazzle 594, catalog no. 35326) was introduced into the samples. Cells positive for this surface marker would bind to the antibody and thus be fluorescently tagged. After the incubation period, the purified IgG and anti-CCR7 antibody were removed from the samples, and the cells were fixed using BD Cytofix/Cytoperm solution following the appropriate cell washing and fixation protocol steps. Flow cytometry data were acquired using a BD Fortessa Dual Symphony instrument and BD FACSDiva software. Cells from each sample were irradiated with a laser, promoting excitation of the fluorophore bound to the cell surface and emission of a light signal. The laser voltage was set for each sample run using rainbow calibration beads prior to data acquisition. The cell population was then analyzed based on the fluorescence signal, and the fraction positive for CCR7 was determined using BD Biosciences FACSDiva™ software and flow analysis software (FlowJo v10.8). The measured percentages were then interpolated over time to estimate the percentage of CCR7-expressing cells at each cultivation time (the same time point at which the holographic information was captured). This was performed for each sample collected daily during expansion and repeated for multiple healthy donors. The dataset generated to train the machine learning model included both donors transduced with the CAR construct and donors transduced after gene editing via electroporation.

D. Model Training and Validation

Flow measurements, which included the percentage (%) of CCR7-positive cells collected daily, were interpolated to obtain hourly measurements at the time of iLineF sampling. At each iLineF sampling time point, standard cell features were extracted from the iLineF phase and intensity images using Obigeo OsOne software, and dead cells were filtered out. The mean feature values were used as input for a feature selection technique, which was then used to train a regression model to predict the interpolated flow measurements. For model validation, leave-one-out sample cross-validation was performed to assess the model's predictive accuracy on data not included in model training.

A supervised regression model was trained to predict the early memory phenotype of T cell populations. The regression model was trained using (i) time-matched population-level statistics summarizing the cell signature measurements described in Section B and (ii) the estimated percentage of CCR7-expressing cells described in Section C. The training data consisted of samples transduced with CAR constructs as well as samples transduced after gene editing via electroporation. For population-level statistics, the 0.01, 0.1, 0.5, 0.9, and 0.99 quantiles of cell signature measurements across cells were used for model training. Cell signature measurements were included in quantile calculations only for cells classified as alive.

We tested several feature selection and regression modeling techniques, including linear regression after dimensionality reduction via principal component analysis; elastic net regression; boosted decision trees (software package "xgboost"); and random forests.

Optimal prediction performance was achieved using a boosted decision tree, and the results are shown in Figure 4 . Figure 4 shows the model-predicted percentage of CCR7-expressing cells (y-axis) versus the measured percentage of CCR7-expressing cells (x-axis) at each time point for the holdout data from each subject (data from each subject are designated by a unique symbol), interpolated from daily flow cytometry measurements to that time point, as well as the best-fit regression line for all predicted percentages (y = 30.2 + 0.62x, R ² = 0.61). As shown in Figure 4 , the model-predicted percentages were positively correlated with the flow cytometry-based estimated percentages for CCR7 positivity. The trained model was able to predict the percentage of CCR7-positive cells (%) at each iLineF sampling time point, with a root mean square deviation (RMSD) error of 9.6% on the holdout data (Figure 4 ).

The trained models were also evaluated to determine which cellular features were most informative for predictive accuracy, primarily measured by RMSD and ^R2 values. The Gini index was used to assess cellular feature informativeness. Based on this evaluation, the most informative cellular features for predictive accuracy included cell area, mean intensity, perimeter length, equivalent peak diameter, and normalized peak area. Overall, cells predicted to have an early memory phenotype based on CCR7 expression had larger features related to cell size (area, perimeter length, diameter), smaller mean intensity features, larger phase correlation features, and smaller radial variance normalized features.

These results validate the feasibility of using holographic imaging to determine and monitor the early memory phenotype of T cell populations during cultivation. Specifically, the results demonstrate that CCR7 is a useful marker for predicting the early memory phenotype. Furthermore, the results are consistent with the finding that CCR7 expression is associated with changes in the size of activated T cells. Furthermore, as demonstrated herein, such monitoring can be performed without labeling individual cells, using (i) population-level statistics summarizing individual cell characteristic measurements and (ii) a machine learning model trained on the percentage of cells expressing exemplary memory phenotype markers.

Current methods for characterizing CCR7 positivity are time-consuming and require specialized reagents, equipment, and trained personnel. The label-free sorting method described in this example can be used to characterize cell material using a non-destructive and non-damaging approach. Furthermore, cell populations can be imaged and returned to the bioreactor intact.

The present invention is not limited in scope to the specific disclosed embodiments provided to illustrate various aspects of the invention. Various modifications to the described compositions and methods will become apparent from the description and teachings herein. Such modifications may be made without departing from the true scope and spirit of the present disclosure and are intended to be included within the scope of the present disclosure.

Attach an electronic file of the sequence list

Claims (79)

A method for determining the cell phenotype of a T cell population, comprising:
(a) obtaining holographic information about a cell population including T cells;
(b) determining one or more input measurements for each of a plurality of cell characteristics derived from the holographic information, wherein each input measurement is from an individual cell within the cell population;
(c) a step of determining a plurality of population-level statistics, wherein each population-level statistic is for one or more input measurements of one of the plurality of cell characteristics, and the plurality of population-level statistics includes one or more population-level statistics for each of the plurality of cell characteristics; and
(d) determining a population-level output measure of the expression of a marker expressed by T cells having a cell phenotype of interest for the cell population based on multiple population-level statistics. A method of determining a cell phenotype of a T cell population, comprising: determining, for a cell population comprising T cells, a population-level output measurement of expression of a marker expressed by T cells having a cell phenotype of interest, wherein the population-level output measurement is determined based on a plurality of population-level statistics, wherein:
Each population-level statistic is for one or more input measurements of a cell characteristic among a plurality of cell characteristics derived from holographic information obtained for the cell population, wherein the plurality of population-level statistics includes one or more population-level statistics for each of the plurality of cell characteristics;
Each input measurement is from an individual cell in the cell population.
method. A method according to claim 2, wherein the method comprises determining a plurality of population-level statistics from one or more input measurements for each of the plurality of cell characteristics. A method according to claim 2 or 3, wherein the method comprises determining one or more input measurements for each of the plurality of cell characteristics from the holographic information. A method according to any one of claims 2 to 4, wherein the method comprises obtaining holographic information. A method according to any one of claims 1 to 5, wherein the holographic information is obtained by differential digital holographic microscopy (DDHM). A method according to any one of claims 4 to 6, wherein obtaining holographic information comprises imaging a cell population using DDHM. A method according to any one of claims 1 to 7, wherein at least one, and optionally one or more population-level statistics for each of the plurality of cell characteristics comprises one or more quantiles of one or more input measurements of the cell characteristic. A method according to claim 8, wherein the at least one quantile comprises at least one of the 0.01, 0.1, 0.5, 0.9 and 0.99 quantiles of at least one input measurement of a cell characteristic. A method according to any one of claims 1 to 7, wherein at least one, and optionally one or more population-level statistics for each of the plurality of cell characteristics, is determined by applying a distribution-based pooling filter to one or more input measurements of the cell characteristics. A method according to any one of claims 1 to 10, wherein the population-level output measurement is a number of cells, a percentage of cells, a proportion of cells, or a density of cells within a cell population expressing the marker. A method according to any one of claims 1 to 11, wherein the cell population is from a cell culture being cultured in vitro or ex vivo. A method according to claim 12, wherein holographic information is obtained during in vitro or ex vivo culture of a cell culture. A method according to claim 12 or 13, wherein the population-level output measurement is for the expression of a marker during in vitro or ex vivo cultivation of a cell culture. A method according to any one of claims 12 to 14, wherein the in vitro or ex vivo culture is performed under conditions that expand T cells in the cell culture. A method according to any one of claims 12 to 15, wherein the in vitro or ex vivo culture is performed in a bioreactor. A method according to any one of claims 1 to 16, wherein the cell population is incubated under T cell stimulating conditions before holographic information is obtained. A method according to any one of claims 1 to 17, wherein the method comprises incubating the cell population under T cell stimulating conditions before obtaining holographic information. A method according to claim 17 or 18, wherein the incubation is performed prior to in vitro or ex vivo culture. A method according to any one of claims 17 to 19, wherein the T cell stimulating condition comprises incubation in the presence of a T cell stimulator that induces a primary activation signal and a co-stimulatory signal in T cells. A method according to claim 20, wherein the T cell stimulator comprises an anti-CD3 antibody or antibody fragment. A method according to claim 20 or 21, wherein the T cell stimulator comprises an anti-CD28 antibody or antibody fragment. A method according to any one of claims 20 to 22, wherein the T cell stimulator is immobilized on a bead. A method according to any one of claims 20 to 22, wherein the T cell stimulator is immobilized on an oligomeric streptavidin mutein reagent. A method according to any one of claims 1 to 24, wherein the recombinant receptor is introduced into T cells of the cell population before holographic information is obtained. A method according to any one of claims 1 to 25, wherein the method comprises introducing a recombinant receptor into T cells of the cell population before holographic information is obtained. A method according to claim 25 or 26, wherein the introduction comprises contacting the cell population with an agent comprising a polynucleotide encoding a recombinant receptor. A method according to claim 26 or 27, wherein the recombinant receptor is a chimeric antigen receptor (CAR) or an engineered T cell receptor (TCR). A method according to any one of claims 1 to 28, wherein at least 50, 55, 60, 65, 70, 75, 80, 85, 90, 91, 92, 93, 94, 95, 96, 97, 98, 99 or 100% of the cell population are T cells. A method according to any one of claims 1 to 29, wherein the population-level output measurement is determined by providing a plurality of population-level statistics as input to a machine learning model trained to predict a population-level output measurement of marker expression based on the population-level statistics of the plurality of cell characteristics. In claim 30, the machine learning model is trained using a dataset of reference population-level statistics, wherein:
For each of the first plurality of reference cell populations comprising T cells, the dataset of reference population-level statistics comprises one or more reference population-level statistics for each of the plurality of cell characteristics, wherein:
Each reference population-level statistic is for one or more reference input measurements of one cell characteristic among a plurality of cell characteristics derived from holographic information obtained for the reference cell population;
Each reference input measurement is from an individual cell in the reference cell population.
method. In claim 31, the machine learning model is trained using a dataset of reference population-level output measurements, wherein for each of the second plurality of reference cell populations, the dataset of reference population-level output measurements includes reference population-level output measurements of expression of a marker for the reference cell population, wherein:
The first and second plurality of reference cell populations are from reference cell cultures cultured in vitro or ex vivo;
The reference cell population from the first plurality of reference cell populations is from the same reference cell culture as the reference cell population from the second plurality of reference cell populations.
method. A method of training a machine learning model to predict a cell phenotype of a T cell population, comprising training the machine learning model using the following dataset:
(i) a dataset of reference population-level statistics, wherein for each of a first plurality of reference cell populations comprising T cells, the dataset of reference population-level statistics comprises one or more reference population-level statistics for each of a plurality of cell characteristics, wherein:
Each reference population-level statistic is for one or more reference input measurements of one cell characteristic among a plurality of cell characteristics derived from holographic information obtained for the reference cell population;
Each reference input measurement is from an individual cell in the reference cell population; and
(ii) a dataset of reference population-level output measurements, wherein for each of the second plurality of reference cell populations, the dataset of reference population-level output measurements comprises a reference population-level output measurement of the expression of a marker expressed by T cells having a cell phenotype of interest for the reference cell population, wherein:
The first and second plurality of reference cell populations are from reference cell cultures cultured in vitro or ex vivo;
The reference cell population from the first plurality of reference cell populations is from the same reference cell culture as the reference cell population from the second plurality of reference cell populations;
Hereby, a machine learning model is trained to predict a population-level output measure of marker expression based on population-level statistics of multiple cell features.
method. A method according to any one of claims 1 to 33, wherein the plurality of cell characteristics include at least one of intensity skewness, intensity correlation, intensity homogeneity, intensity maximum, intensity minimum, intensity entropy, intensity contrast, phase entropy, cell area, and radius average. A method according to any one of claims 1 to 34, wherein the plurality of cell characteristics include intensity maxima, intensity minima, intensity entropy, intensity contrast, phase entropy, cell area, and radius average. A method of determining an activation state of a T cell population, comprising: determining, for a cell population comprising T cells, a population-level output measurement of expression of a marker expressed by activated T cells, wherein the marker is CD137 (4-1BB), and wherein the population-level output measurement is determined based on one or more input measurements for each of a plurality of cell characteristics derived from holographic information obtained for the cell population, wherein:
Multiple cell characteristics include intensity maximum, intensity minimum, intensity entropy, intensity contrast, phase entropy, cell area and radius mean;
Each input measurement is from an individual cell in the cell population.
method. A method according to any one of claims 1 to 36, wherein the plurality of cell characteristics include intensity skewness, intensity correlation, and intensity homogeneity. A method of determining an activation state of a T cell population, comprising: determining, for a cell population comprising T cells, a population-level output measurement of expression of a marker expressed by activated T cells, wherein the marker is CD137 (4-1BB), and wherein the population-level output measurement is determined based on one or more input measurements for each of a plurality of cell characteristics derived from holographic information obtained for the cell population, wherein:
Multiple cellular characteristics include intensity skewness, intensity correlation and intensity homogeneity;
Each input measurement is from an individual cell in the cell population.
method. A method according to claim 38, wherein the plurality of cell characteristics include intensity maxima, intensity minima, intensity entropy, intensity contrast, phase entropy, cell area, and radius average. A method according to any one of claims 1 to 33, wherein the plurality of cell characteristics comprise at least one of peak area, phase average uniformity, intensity geometric mean, minimum optical height, and normalized optical height. A method according to any one of claims 1 to 33 and 40, wherein the plurality of cell characteristics include peak area, phase average uniformity, intensity geometric mean, minimum optical height, and normalized optical height. A method according to any one of claims 1 to 33, wherein the plurality of cell characteristics include at least one of cell area, perimeter length, average intensity, normalized peak area, and equivalent peak diameter. A method according to any one of claims 1 to 33 and 42, wherein the plurality of cell characteristics include cell area, perimeter length, average intensity, normalized peak area, and equivalent peak diameter. A method of determining a memory phenotype of a T cell population, comprising: determining, for a cell population comprising T cells, a population-level output measurement of expression of a marker expressed by central memory T cells or stem cell memory T cells, wherein the marker is CCR7, and wherein the population-level output measurement is determined based on one or more input measurements for each of a plurality of cell characteristics derived from holographic information obtained for the cell population, wherein:
The multiple cell characteristics include one or more of cell area, perimeter length, mean intensity, normalized peak area, and equivalent peak diameter;
Each input measurement is from an individual cell in the cell population.
method. A method according to claim 44, wherein the plurality of characteristics include cell area, perimeter length, average intensity, normalized peak area, and equivalent peak diameter. A method of determining recombinant receptor expression in a T cell population, comprising: determining, for a cell population comprising T cells, expression of a recombinant receptor introduced into T cells of the cell population, wherein the determination is based on one or more input measurements for each of a plurality of cell characteristics derived from holographic information obtained for the cell population, wherein:
Multiple cell characteristics include peak area, phase average uniformity, intensity geometric mean, minimum optical height, and normalized optical height;
Each input measurement is from an individual cell in the cell population.
method. In any one of claims 1 to 33, at least one, optionally one or more input measurements of each of the plurality of cell characteristics is determined by providing holographic information about individual cells of the cell population to a convolutional neural network, wherein:
The multiple cell features are cell features extracted by a convolutional neural network;
One or more input measurements are determined from a convolutional neural network.
method. In any one of claims 31 to 33 and claim 47, at least one of the plurality of cell characteristics, optionally one or more reference input measurements for each, is determined by providing holographic information about individual cells of the reference cell population to a convolutional neural network, wherein:
The multiple cell features are cell features extracted by a convolutional neural network;
One or more reference input measurements are determined from a convolutional neural network.
method. A method according to claim 47 or 48, wherein the convolutional neural network is trained using a dataset of reference holographic information, wherein for each of the first plurality of reference cell populations including T cells, the dataset of reference holographic information includes holographic information obtained for individual cells of the reference cell population. A method according to any one of claims 31 to 35, 37, 40 to 43, and 47 to 49, wherein at least one, and optionally each, of the plurality of cell characteristics comprises one or more quantiles of one or more reference input measurements for the cell characteristic. A method according to claim 50, wherein one or more quantiles are selected from the 0.01, 0.1, 0.5, 0.9 and 0.99 quantiles of one or more reference input measurements for cell characteristics. A method according to any one of claims 31 to 35, 37, 40 to 43, and 47 to 49, wherein at least one, and optionally one or more, reference population-level statistics for each of the plurality of cell characteristics are determined by applying a distribution-based pooling filter to one or more reference input measurements for the cell characteristic. A method for training a machine learning model to predict a cell phenotype of a T cell population, comprising:
(a) a step of training a convolutional neural network using a dataset of reference holographic information, wherein for each of a first plurality of reference cell populations including T cells, the dataset of reference holographic information includes holographic information obtained for individual cells of the reference cell population;
(b) a step of determining, from a convolutional neural network, one or more reference input measurements for each cell feature among a plurality of cell features derived from holographic information, wherein the plurality of cell features are cell features extracted by the convolutional neural network, and each reference input measurement is from an individual cell of the reference cell population;
(c) determining a dataset of reference population-level statistics, wherein the dataset of reference population-level statistics comprises one or more reference population-level statistics for each of a plurality of cell characteristics, each reference population-level statistic being determined by applying a distribution-based pooling filter to one or more reference input measurements for one of the plurality of cell characteristics; and
(d) training a machine learning model using a dataset of reference population-level statistics and a dataset of reference population-level output measurements, wherein for each of the second plurality of reference cell populations, the dataset of reference population-level output measurements comprises reference population-level output measurements of expression of a marker expressed by T cells having a cell phenotype of interest for the reference cell population, wherein:
The first and second plurality of reference cell populations are from reference cell cultures cultured in vitro or ex vivo;
The reference cell population from the first plurality of reference cell populations is from the same reference cell culture as the reference cell population from the second plurality of reference cell populations;
Hereby, a machine learning model is trained to predict a population-level output measure of marker expression based on population-level statistics of multiple cell features.
method. A method according to any one of claims 47 to 53, wherein at least one, and optionally one or more, input measurements for each of the plurality of cell characteristics are obtained from a fully connected layer of a convolutional neural network. A method according to any one of claims 31 to 35, 37, 40 to 43, and 47 to 54, wherein holographic information for the first plurality of reference cell populations is obtained by DDHM. A method according to any one of claims 1 to 55, wherein the holographic information includes phase information and intensity information. A method according to any one of claims 32 to 35, 37, 40 to 43, and 47 to 56, wherein the reference population-level output measurement is the number of cells, the percentage of cells, the proportion of cells, or the density of cells in the reference cell population expressing the marker. A method according to any one of claims 32 to 35, 37, 40 to 43, and 47 to 57, wherein the dataset of reference population-level statistics and the dataset of reference population-level output measurements are time-matched. A method according to any one of claims 32 to 35, 37, 40 to 43, and 47 to 58, wherein the in vitro or ex vivo cultivation of the reference cell culture is performed under the same or similar conditions as the in vitro or ex vivo cultivation of the cell culture. A method according to any one of claims 32 to 35, 37, 40 to 43, and 47 to 59, wherein the holographic information for the first plurality of reference cell populations is obtained during in vitro or ex vivo culture of the reference cell culture. A method according to any one of claims 32 to 35, 37, 40 to 43, and 47 to 60, wherein the dataset of reference population-level output measurements relates to the expression of a marker during or after in vitro or ex vivo culturing of a reference cell culture. A method according to any one of claims 32 to 35, 37, 40 to 43, and 47 to 61, wherein the dataset of reference population-level output measurements is determined using fluorescence imaging of a second plurality of reference cell populations. A method according to claim 62, wherein the fluorescence imaging is by flow cytometry. A method according to any one of claims 1 to 35, 37, and 47 to 63, wherein the method is for determining the activation state of a T cell population, and the marker is expressed by activated T cells. A method according to any one of claims 1 to 35, 37, and 47 to 64, wherein the marker is CD137 (4-1BB). A method according to any one of claims 1 to 35 and 42 to 63, wherein the method is for determining the memory phenotype of a T cell population, and the marker is expressed by T cells having a central memory phenotype or a stem cell memory phenotype. A method according to claim 66, wherein the marker is expressed by T cells having a central memory phenotype. A method according to claim 66, wherein the marker is expressed by T cells having a stem cell memory phenotype. A method according to any one of claims 1 to 35, 42 to 63, and 66 to 68, wherein the marker is CCR7. A method according to any one of claims 1 to 33 and claims 40 to 63, wherein the method is for determining recombinant receptor expression in a T cell population, and the marker is a recombinant receptor introduced into T cells of the cell population before holographic information is obtained. A method according to any one of claims 46 to 63 and 70, wherein the recombinant receptor is a chimeric antigen receptor (CAR) or an engineered T cell receptor (TCR). A method according to any one of claims 31 to 35, 37, 40 to 43, and 47 to 71, wherein the first plurality of reference cell populations are incubated under T cell stimulating conditions before holographic information is obtained for the first plurality of reference cell populations. A method according to claim 72, wherein the incubation of the first plurality of reference cell populations is performed under the same or similar conditions as the incubation of the cell population. A method according to claim 72 or 73, wherein the incubation of the first plurality of reference cell populations is performed prior to in vitro or ex vivo culturing of the reference cell culture. A method according to any one of claims 31 to 35, 37, 40 to 43, and 47 to 74, wherein the first and/or second plurality of reference cell populations are enriched for T cells. A method according to any one of claims 31 to 35, 37, 40 to 43 and 47 to 75, wherein at least 50, 55, 60, 65, 70, 75, 80, 85, 90, 91, 92, 93, 94, 95, 96, 97, 98, 99 or 100% of the first and/or second plurality of reference cell populations are T cells. A method of monitoring a cell phenotype of a T cell culture, comprising: determining a population-level output measurement of the expression of a marker expressed by T cells having a cell phenotype of interest for a cell population from the cell culture comprising T cells cultured in vitro or ex vivo, wherein the population-level output measurement is determined according to the method of any one of claims 1 to 32, 34, 35, 37, 40 to 43, 47 to 52, and 54 to 76. A computing device comprising instructions in memory for performing the method of any one of claims 1 to 32, 34, 35, 37, 40 to 43, 47 to 52, and 54 to 76, wherein the instructions comprise the following instructions:
(a) instructions for receiving holographic information about an individual cell in a cell population comprising T cells, one or more input measurements for each of a plurality of cell characteristics for an individual cell in a cell population comprising T cells, or a plurality of population-level statistics for a cell population comprising T cells; and
(b) instructions for determining a population-level output measurement for a cell population, one or more input measurements for each of a plurality of cell characteristics, or a plurality of population-level statistics from the holographic information, according to the method. A computing device according to claim 78, wherein the computing device further comprises in a memory a machine learning model trained according to the method of any one of claims 33 to 35, 37, 40 to 43, and 47 to 76, wherein the population-level output measurement for the cell population is determined using the machine learning model.