WO2024124132A1

WO2024124132A1 - Machine learning methods for predicting cell phenotype using holographic imaging

Info

Publication number: WO2024124132A1
Application number: PCT/US2023/083122
Authority: WO
Inventors: Sarah RAJANI; David Reiss; Maryam NEZAFAT
Original assignee: Juno Therapeutics Inc
Current assignee: Juno Therapeutics Inc
Priority date: 2022-12-09
Filing date: 2023-12-08
Publication date: 2024-06-13
Anticipated expiration: 2025-06-09
Also published as: KR20250121074A; EP4630782A1; CN121100267A; JP2026501122A

Abstract

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

Description

MACHINE LEARNING METHODS FOR PREDICTING CELL PHENOTYPE USING HOLOGRAPHIC IMAGING

Cross-Reference to Related Applications

[0001] This application claims priority to U.S. Provisional Application No. 63/431,635, filed December 9, 2022, entitled “MACHINE LEARNING METHODS FOR PREDICTING CELL PHENOTYPE USING HOLOGRAPHIC IMAGING,” and U.S. Provisional Application No. 63/537,467, filed September 8, 2023, entitled “MACHINE LEARNING METHODS FOR PREDICTING CELL PHENOTYPE USING HOLOGRAPHIC IMAGING,” the contents of which are incorporated by reference in their entirety for all purposes.

Incorporation By Reference of Sequence Listing

[0002] The present application is being filed along with a Sequence Listing in electronic format. The Sequence Listing is provided as a file entitled 735042024740SeqList.xml, created December 7, 2023, which is 38,941 bytes in size. The information in the electronic format of the Sequence Listing is incorporated by reference in its entirety.

Field

[0003] The present disclosure relates to methods for determining a cell phenotype, such as activation state, of a population of T cells. In some aspects, the provided methods are label- free methods. In some aspects, the provided methods can be used for monitoring a cell phenotype, such as activation state, of a culture of T cells. Also provided herein are computing devices for use in performing the provided methods.

Background

[0004] Existing methods for determining a cell phenotype, such as activation state, of T cells, for example during in vitro or ex vivo culture of the T cells, can be time-consuming or require specialized reagents, instrumentation, or trained operators. In some instances, existing methods involve directly manipulating or labelling T cells, for example in a manner involving incubation with immunoaffinity-based reagents that can interfere with the quality or function of the T cells. Such methods can pose risks of contamination of or damage to cells prior to any downstream processing. Improved methods for accurately determining a cell phenotype, such as activation state, of T cells are needed. Such methods are useful, for example, in the production of cell products, for instance cell therapy products that contain T cells. Provided herein are methods and computing devices that meet such needs.

Summary

[0005] Provided herein in some embodiments is a method for determining a cell phenotype of a population of T cells, comprising determining, for a population of cells comprising T cells, a population-level output measure of expression of a marker expressed by T cells with a cell phenotype of interest, wherein the population-level output measure is determined based on a plurality of population-level statistics, wherein: each population-level statistic is of one or more input measures for a cellular feature of a plurality of cellular features derived from holographic information obtained for the population of cells, the plurality of population-level statistics comprising one or more population-level statistics for each of the plurality of cellular features; and each input measure is from an individual cell of the population of cells.

[0006] In some of any embodiments, the method is for determining the activation state of the population of T cells, and the marker is expressed by activated T cells. In some of any embodiments, the marker is CD137 (4-1BB).

[0007] In some of any embodiments, the method is for determining the memory phenotype of the population of T cells, and the marker, e.g., CCR7, is expressed by T cells having a central memory phenotype or a stem cell memory phenotype.

[0008] Also provided herein in some embodiments is a method for determining the activation state of a population of T cells, comprising determining, for a population of cells comprising T cells, a population-level output measure of expression of a marker expressed by activated T cells, wherein the population-level output measure is determined based on a plurality of population-level statistics, wherein: each population-level statistic is of one or more input measures for a cellular feature of a plurality of cellular features derived from holographic information obtained for the population of cells, the plurality of population-level statistics comprising one or more population-level statistics for each of the plurality of cellular features; and each input measure is from an individual cell of the population of cells.

[0009] In some of any embodiments, the method is for determining recombinant receptor expression of the population of T cells, and the marker is a recombinant receptor introduced into T cells of the population of cells prior to when the holographic information is obtained. In some of any embodiments, the recombinant receptor is a chimeric antigen receptor (CAR) or an engineered T cell receptor (TCR).

[0010] Also provided herein in some embodiments is a method for determining recombinant receptor expression of a population of T cells, comprising determining, for a population of cells comprising T cells, a population-level output measure of expression of a recombinant receptor introduced into T cells of the population of T cells, wherein the population-level output measure is determined based on a plurality of population-level statistics, wherein: each population-level statistic is of one or more input measures for a cellular feature of a plurality of cellular features derived from holographic information obtained for the population of cells, the plurality of population-level statistics comprising one or more population-level statistics for each of the plurality of cellular features; and each input measure is from an individual cell of the population of cells.

[0011] In some of any embodiments, the method comprises determining the plurality of population-level statistics from the one or more input measures for each of the plurality of cellular features.

[0012] In some of any embodiments, the method comprises determining the one or more input measures for each of the plurality of cellular features from the holographic information.

[0013] In some of any embodiments, the method comprises obtaining the holographic information.

[0014] Also provided herein in some embodiments is a method for determining a cell phenotype of a population of T cells, comprising: (a) obtaining holographic information for a population of cells comprising T cells; (b) determining one or more input measures for each of a plurality of cellular features derived from the holographic information, wherein each input measure is from an individual cell in the population of cells; (c) determining a plurality of population-level statistics, wherein each population-level statistic is of the one or more input measures for a cellular feature of the plurality of cellular features, and the plurality of population-level statistics comprises one or more population-level statistics for each of the plurality of cellular features; and (d) determining, based on the plurality of population-level statistics, a population-level output measure of expression of a marker expressed by T cells with a cell phenotype of interest for the population of cells. [0015] In some of any embodiments, the method is for determining the activation state of the population of T cells, and the marker is expressed by activated T cells. In some of any embodiments, the marker is CD137 (4-1BB).

[0016] Also provided herein in some embodiments is a method for determining the activation state of a population of T cells, comprising: (a) obtaining holographic information for a population of cells comprising T cells; (b) determining one or more input measures for each of a plurality of cellular features derived from the holographic information, wherein each input measure is from an individual cell in the population of cells; (c) determining a plurality of population-level statistics, wherein each population-level statistic is of the one or more input measures for a cellular feature of the plurality of cellular features, and the plurality of population-level statistics comprises one or more population-level statistics for each of the plurality of cellular features; and (d) determining, based on the plurality of population-level statistics, a population-level output measure of expression of a marker expressed by activated T cells for the population of cells.

[0017] In some of any embodiments, the method is for determining recombinant receptor expression of the population of T cells, and the marker is a recombinant receptor introduced into T cells of the population of cells prior to when the holographic information is obtained. In some of any embodiments, the recombinant receptor is a chimeric antigen receptor (CAR) or an engineered T cell receptor (TCR).

[0018] Also provided herein in some embodiments is a method for determining recombinant receptor expression of a population of T cells, comprising: (a) obtaining holographic information for a population of cells comprising T cells; (b) determining one or more input measures for each of a plurality of cellular features derived from the holographic information, wherein each input measure is from an individual cell in the population of cells; (c) determining a plurality of population-level statistics, wherein each population-level statistic is of the one or more input measures for a cellular feature of the plurality of cellular features, and the plurality of population-level statistics comprises one or more populationlevel statistics for each of the plurality of cellular features; and (d) determining, based on the plurality of population-level statistics, a population-level output measure of expression for the population of cells of a recombinant receptor introduced into T cells of the population of cells. [0019] In some of any embodiments, the method further comprises engineering the population of T cells following the determining of the population-level output measure.

[0020] In some of any embodiments, the holographic information is obtained by differential digital holographic microscopy (DDHM). In some of any embodiments, the obtaining the holographic information comprises imaging the population of cells using DDHM.

[0021] In some of any embodiments, the one or more population-level statistics for at least one, optionally each, of the plurality of cellular features comprise one or more quantiles of the one or more input measures of the cellular feature. In some of any embodiments, the one or more population-level statistics for each of the plurality of cellular features comprise one or more quantiles of the one or more input measures of the cellular feature. In some of any embodiments, the one or more quantiles are selected from (compise one or more of) the 0.01, 0.1, 0.5, 0.9, and 0.99 quantiles of the one or more input measures of the cellular feature.

[0022] In some of any embodiments, the one or more population-level statistics for at least one, optionally each, of the plurality of cellular features are determined by applying a distribution-based pooling filter to the one or more input measures of the cellular feature. In some of any embodiments, the one or more population-level statistics for each of the plurality of cellular features are determined by applying a distribution-based pooling filter to the one or more input measures of the cellular feature.

[0023] In some of any embodiments, the population-level output measure is the number of cells, the percentage of cells, the proportion of cells, or the density of cells of the population of cells that express the marker.

[0024] In some of any embodiments, the population of cells is from a culture of cells being cultured in vitro or ex vivo.

[0025] In some of any embodiments, the holographic information is obtained during the in vitro or ex vivo culture of the culture of cells.

[0026] In some of any embodiments, the population-level output measure is of expression of the marker during the in vitro or ex vivo culture of the culture of cells.

[0027] In some of any embodiments, the in vitro or ex vivo culture is under conditions to expand T cells of the culture of cells.

[0028] In some of any embodiments, the in vitro or ex vivo culture is in a bioreactor. [0029] In some of any embodiments, the population of cells are incubated under T cell stimulating conditions prior to when the holographic information is obtained. In some of any embodiments, the method comprises incubating the population of cells under T cell stimulating conditions prior to when the holographic information is obtained.

[0030] In some of any embodiments, the incubation is prior to the in vitro or ex vivo culture.

[0031] In some of any embodiments, the T cell stimulating conditions comprise incubation in the presence of T cell stimulatory agents that induce a primary activation signal and a costimulatory signal in T cells. In some of any embodiments, the T cell stimulatory agents comprise an anti-CD3 antibody or antibody fragment. In some of any embodiments, the T cell stimulatory agents comprise an anti-CD28 antibody or antibody fragment. In some of any embodiments, the T cell stimulatory agents are immobilized on a bead.

[0032] In some of any embodiments, the T cell stimulatory agents are immobilized on an oligomeric streptavidin mutein reagent.

[0033] In some of any embodiments, a recombinant receptor is introduced into T cells of the population of cells prior to when the holographic information is obtained. In some of any embodiments, the method comprises introducing a recombinant receptor into T cells of the population of cells prior to when the holographic information is obtained.

[0034] In some of any embodiments, the introducing comprising contacting the population of cells with an agent comprising a polynucleotide encoding the recombinant receptor.

[0035] In some of any embodiments, the recombinant receptor is a chimeric antigen receptor (CAR) or an engineered T cell receptor (TCR).

[0036] In some of any embodiments, the population of cells is enriched for T cells. In some of any 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 population of cells are T cells.

[0037] In some of any embodiments, the population-level output measure is determined by providing the plurality of population-level statistics as input to a machine learning model trained to predict population-level output measures of expression of the marker based on population-level statistics of the plurality of cellular features.

[0038] In some of any embodiments, the machine learning model is trained using a dataset of reference population-level statistics, wherein: for each of a first plurality of reference populations of cells comprising T cells, the dataset of reference population-level statistics comprises one or more reference population-level statistics for each of the plurality of cellular features, wherein: each reference population-level statistic is of one or more reference input measures for a cellular feature of the plurality of cellular features derived from holographic information obtained for the reference population of cells; and each reference input measure is from an individual cell of the reference population of cells.

[0039] In some of any embodiments, the machine learning model is trained using a dataset of reference population-level output measures, wherein for each of a second plurality of reference populations of cells, the dataset of reference population-level output measures comprises a reference population-level output measure of expression of the marker for the reference population of cells, wherein: the first and second pluralities of reference populations of cells are from reference cultures of cells being cultured in vitro or ex vivo', and reference populations of cells from the first plurality of reference populations of cells are from the same reference culture of cells as reference populations of cells from the second plurality of reference populations of cells.

[0040] Also provided herein in some embodiments is a method for training a machine learning model that predicts a cell phenotype of a population of T cells, comprising training a machine learning model using: (i) a dataset of reference population-level statistics, wherein for each of a first plurality of reference populations of cells comprising T cells, the dataset of reference population-level statistics comprises one or more reference population-level statistics for each of a plurality of cellular features, wherein: each reference population-level statistic is of one or more reference input measures for a cellular feature of the plurality of cellular features derived from holographic information obtained for the reference population of cells; and each reference input measure is from an individual cell of the reference population of cells; and (ii) a dataset of reference population-level output measures, wherein for each of a second plurality of reference populations of cells, the dataset of reference population-level output measures comprises a reference population-level output measure of expression of a marker expressed by T cells with a cell phenotype of interest for the reference population of cells, wherein: the first and second pluralities of reference populations of cells are from reference cultures of cells being cultured in vitro or ex vivo', and reference populations of cells from the first plurality of reference populations of cells are from the same reference culture of cells as reference populations of cells from the second plurality of reference populations of cells; whereby the machine learning model is trained to predict population-level output measures of expression of the marker based on population-level statistics of the plurality of cellular features.

[0041] In some of any embodiments, the method is for determining the activation state of the population of T cells, and the marker is expressed by activated T cells. In some of any embodiments, the marker is CD137 (4-1BB).

[0042] Also provided herein in some embodiments is a method for training a machine learning model that predicts the activation state of a population of T cells, comprising training a machine learning model using: (i) a dataset of reference population-level statistics, wherein for each of a first plurality of reference populations of cells comprising T cells, the dataset of reference population-level statistics comprises one or more reference populationlevel statistics for each of a plurality of cellular features, wherein: each reference populationlevel statistic is of one or more reference input measures for a cellular feature of the plurality of cellular features derived from holographic information obtained for the reference population of cells; and each reference input measure is from an individual cell of the reference population of cells; and (ii) a dataset of reference population-level output measures, wherein for each of a second plurality of reference populations of cells, the dataset of reference population-level output measures comprises a reference population-level output measure of expression of a marker expressed by activated T cells for the reference population of cells, wherein: the first and second pluralities of reference populations of cells are from reference cultures of cells being cultured in vitro or ex vivo', and reference populations of cells from the first plurality of reference populations of cells are from the same reference culture of cells as reference populations of cells from the second plurality of reference populations of cells; whereby the machine learning model is trained to predict populationlevel output measures of expression of the marker based on population-level statistics of the plurality of cellular features.

[0043] In some of any embodiments, the method is for determining recombinant receptor expression of the population of T cells, and the marker is a recombinant receptor introduced into T cells of the population of cells prior to when the holographic information is obtained. In some of any embodiments, the recombinant receptor is a chimeric antigen receptor (CAR) or an engineered T cell receptor (TCR).

[0044] Also provided herein in some embodiments is a method for training a machine learning model that predicts recombinant receptor expression of a population of T cells, comprising training a machine learning model using: (i) a dataset of reference populationlevel statistics, wherein for each of a first plurality of reference populations of cells comprising T cells, the dataset of reference population-level statistics comprises one or more reference population-level statistics for each of a plurality of cellular features, wherein: each reference population-level statistic is of one or more reference input measures for a cellular feature of the plurality of cellular features derived from holographic information obtained for the reference population of cells; and each reference input measure is from an individual cell of the reference population of cells; and (ii) a dataset of reference population-level output measures, wherein for each of a second plurality of reference populations of cells, the dataset of reference population-level output measures comprises a reference population-level output measure of expression for the reference population of cells of a recombinant receptor introduced into T cells of the reference population of cells, wherein: the first and second pluralities of reference populations of cells are from reference cultures of cells being cultured in vitro or ex vivo', and reference populations of cells from the first plurality of reference populations of cells are from the same reference culture of cells as reference populations of cells from the second plurality of reference populations of cells; whereby the machine learning model is trained to predict population-level output measures of expression of the recombinant receptor based on population-level statistics of the plurality of cellular features.

[0045] In some of any embodiments, the plurality of cellular features comprise one or of intensity skewness, intensity correlation, intensity homogeneity, intensity maximum, intensity minimum, intensity entropy, intensity contrast, phase entropy, cell area, and radius mean.

[0046] In some of any embodiments, the plurality of cellular features are selected from (comprises one or more of) intensity maximum, intensity minimum, intensity entropy, intensity contrast, phase entropy, cell area, and radius mean. Also provided herein in some embodiments is a method for determining the activation state of a population of T cells, comprising determining, for a population of cells comprising T cells, expression of a marker expressed by activated T cells, wherein the marker is CD 137, and the determining is based on one or more input measures for each of a plurality of cellular features derived from holographic information obtained for the population of cells, wherein: the plurality of cellular features are selected from (comprises one or more of) intensity maximum, intensity minimum, intensity entropy, intensity contrast, phase entropy, cell area, and radius mean; and each input measure is from an individual cell of the population of cells.

[0047] In some of any embodiments, the plurality of cellular features comprises intensity maximum, intensity minimum, intensity entropy, intensity contrast, phase entropy, cell area, and radius mean.

[0048] Provided herein in some embodiments is a method for determining the activation state of a population of T cells, comprising determining, for a population of cells comprising T cells, expression of a marker expressed by activated T cells, wherein the marker is CD137 (4- IBB), and the determining is based on one or more input measures for each of a plurality of cellular features derived from holographic information obtained for the population of cells, wherein: the plurality of cellular features comprises intensity maximum, intensity minimum, intensity entropy, intensity contrast, phase entropy, cell area, and radius mean; and each input measure is from an individual cell of the population of cells.

[0049] Also provided herein, in some embodiments, is a method for determining the activation state of a population of T cells, comprising determining, for a population of cells comprising T cells, a population-level output measure of expression of a marker expressed by activated T cells, wherein the marker is CD137 (4-1BB), and the population-level output measure is determined based on one or more input measures for each of a plurality of cellular features derived from holographic information obtained for the population of cells, wherein: the plurality of cellular features comprises intensity maximum, intensity minimum, intensity entropy, intensity contrast, phase entropy, cell area, and radius mean; and each input measure is from an individual cell of the population of cells.

[0050] In some of any embodiments, the plurality of cellular features comprises one or more of intensity skewness, intensity correlation, and intensity homogeneity. In some of any embodiments, the plurality of cellular features comprises intensity skewness, intensity correlation, and intensity homogeneity.

[0051] Also provided herein in some embodiments is a method for determining the activation state of a population of T cells, comprising determining, for a population of cells comprising T cells, expression of a marker expressed by activated T cells, wherein the marker is CD137 (4-1BB), and the determining is based on one or more input measures for each of a plurality of cellular features derived from holographic information obtained for the population of cells, wherein: the plurality of cellular features comprises intensity skewness, intensity correlation, and intensity homogeneity; and each input measure is from an individual cell of the population of cells.

[0052] Also provided herein, in some embodiments, is a method for determining the activation state of a population of T cells, comprising determining, for a population of cells comprising T cells, a population-level output measure of expression of a marker expressed by activated T cells, wherein the marker is CD137 (4-1BB), and the population-level output measure is determined based on one or more input measures for each of a plurality of cellular features derived from holographic information obtained for the population of cells, wherein: the plurality of cellular features comprises intensity skewness, intensity correlation, and intensity homogeneity; and each input measure is from an individual cell of the population of cells.

[0053] In some of any embodiments, the plurality of cellular features comprises one or more of intensity maximum, intensity minimum, intensity entropy, intensity contrast, phase entropy, cell area, and radius mean. In some of any embodiments, the plurality of cellular features comprises intensity maximum, intensity minimum, intensity entropy, intensity contrast, phase entropy, cell area, and radius mean.

[0054] In some of any embodiments, the plurality of cellular features comprises one or more of peak area, phase average uniformity, intensity geometric mean, minimum optical height, and normalized optical height. In some of any embodiments, the plurality of cellular features comprises peak area, phase average uniformity, intensity geometric mean, minimum optical height, and normalized optical height.

[0055] In some of any embodiments, the plurality of cellular features comprises one or more of cell area, perimeter, mean intensity, normalized peak area, and equivalent peak diameter. In some of any embodiments, the plurality of cellular features comprises cell area, perimeter, mean intensity, normalized peak area, and equivalent peak diameter.

[0056] Also provided herein, in some embodiments, is a method for determining the memory phenotype of a population of T cells, comprising determining, for a population of cells 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 the determining is based on one or more input measures for each of a plurality of cellular features derived from holographic information obtained for the population of cells, wherein: the plurality of cellular features comprises one or more of cell area, perimeter, mean intensity, normalized peak area, and equivalent peak diameter; and each input measure is from an individual cell of the population of cells.

[0057] In some of any embodiments, the plurality of cellular features comprises intensity maximum, intensity minimum, intensity entropy, intensity contrast, phase entropy, cell area, and radius mean. In some of any embodiments, the plurality of cellular features comprises one or more of peak area, phase average uniformity, intensity geometric mean, minimum optical height, and normalized optical height.

[0058] In some of any embodiments, the plurality of cellular features comprises peak area, phase average uniformity, intensity geometric mean, minimum optical height, and normalized optical height.

[0059] In some of any embodiments, the plurality of cellular features comprises one or more of cell area, perimeter, mean intensity, normalized peak area, and equivalent peak diameter. In some of any embodiments, the plurality of cellular features comprises cell area, perimeter, mean intensity, normalized peak area, and equivalent peak diameter.

[0060] Also provided herein, in some embodiments, is a method for determining the memory phenotype of a population of T cells, comprising determining, for a population of cells comprising T cells, a population-level output measure of expression of a marker expressed by central memory T cells or stem cell memory T cells, wherein the marker is CCR7, and the population-level output measure is determined based on one or more input measures for each of a plurality of cellular features derived from holographic information obtained for the population of cells, wherein: the plurality of cellular features comprises one or more of cell area, perimeter, mean intensity, normalized peak area, and equivalent peak diameter; and each input measure is from an individual cell of the population of cells.

[0061] In some embodiments, the plurality of features comprises cell area, perimeter, mean intensity, normalized peak area, and equivalent peak diameter. [0062] Also provided herein in some embodiments is a method for determining recombinant receptor expression of a population of T cells, comprising determining, for a population of cells comprising T cells, expression of a recombinant receptor introduced into T cells of the population of cells, wherein the determining is based on one or more input measures for each of a plurality of cellular features derived from holographic information obtained for the population of cells, wherein: the plurality of cellular features comprises peak area, phase average uniformity, intensity geometric mean, minimum optical height, and normalized optical height; and each input measure is from an individual cell of the population of cells.

[0063] In some of any embodiments, the one or more input measures for at least one, optionally each, of the plurality of cellular features are determined by providing the holographic information for individual cells of the population of cells to a convolutional neural network, wherein: the plurality of cellular features are cellular features extracted by the convolutional neural network; and the one or more input measures are determined from the convolutional neural network. In some of any embodiments, the one or more input measures for each of the plurality of cellular features are determined by providing the holographic information for individual cells of the population of cells to a convolutional neural network, wherein: the plurality of cellular features are cellular features extracted by the convolutional neural network; and the one or more input measures are determined from the convolutional neural network.

[0064] In some of any embodiments, the one or more reference input measures for at least one, optionally each, of the plurality of cellular features are determined by providing the holographic information for individual cells of the reference population of cells to a convolutional neural network, wherein: the plurality of cellular features are cellular features extracted by the convolutional neural network; and the one or more reference input measures are determined from the convolutional neural network. In some of any embodiments, the one or more reference input measures for each of the plurality of cellular features are determined by providing the holographic information for individual cells of the reference population of cells to a convolutional neural network, wherein: the plurality of cellular features are cellular features extracted by the convolutional neural network; and the one or more reference input measures are determined from the convolutional neural network. [0065] In some of any embodiments, the convolutional neural network is trained using a dataset of reference holographic information, wherein for each of a first plurality of reference populations of cells comprising T cells, the dataset of reference holographic information comprises holographic information obtained for individual cells of the reference population of cells.

[0066] In some of any embodiments, the one or more reference population-level statistics for at least one, optionally each, of the plurality of cellular features comprise one or more quantiles of the one or more reference input measures for the cellular feature. In some of any embodiments, the one or more reference population-level statistics for each of the plurality of cellular features comprise one or more quantiles of the one or more reference input measures for the cellular feature. In some of any embodiments, the one or more quantiles are selected from the 0.01, 0.1, 0.5, 0.9, and 0.99 quantiles of the one or more reference input measures for the cellular feature.

[0067] In some of any embodiments, the one or more reference population-level statistics for at least one, optionally each, of the plurality of cellular features are determined by applying a distribution-based pooling filter to the one or more reference input measures for the cellular feature. In some of any embodiments, the one or more reference population-level statistics for each of the plurality of cellular features are determined by applying a distribution-based pooling filter to the one or more reference input measures for the cellular feature.

[0068] Also provided herein in some embodiments is a method for training a machine learning model that predicts a cell phenotype of a population of T cells, comprising: (a) training a convolutional neural network using a dataset of reference holographic information, wherein for each of a first plurality of reference populations of cells comprising T cells, the dataset of reference holographic information comprises holographic information obtained for individual cells of the reference population of cells; (b) determining, from the convolutional neural network, one or more reference input measures for each cellular feature of a plurality of cellular features derived from the holographic information, wherein the plurality of cellular features are cellular features extracted by the convolutional neural network, and each reference input measure is from an individual cell of a reference population of cells; (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 the plurality of cellular features, each reference population-level statistic determined by applying a distribution-based pooling filter to the one or more reference input measures for a cellular feature of the plurality of cellular features; and (d) training a machine learning model using the dataset of reference population-level statistics and a dataset of reference populationlevel output measures, wherein for each of a second plurality of reference populations of cells, the dataset of reference population-level output measures comprises a reference population-level output measure of expression of a marker expressed by T cells with a cell phenotype of interest for the reference population of cells, wherein: the first and second pluralities of reference populations of cells are from reference cultures of cells being cultured in vitro or ex vivo', and reference populations of cells from the first plurality of reference populations of cells are from the same reference culture of cells as reference populations of cells from the second plurality of reference populations of cells; whereby the machine learning model is trained to predict population-level output measures of expression of the marker based on population-level statistics of the plurality of cellular features.

[0069] In some of any embodiments, the method is for determining the activation state of the population of T cells, and the marker is expressed by activated T cells. In some of any embodiments, the marker is CD137 (4-1BB).

[0070] Also provided herein in some embodiments is a method for training a machine learning model that predicts the activation state of a population of T cells, comprising: (a) training a convolutional neural network using a dataset of reference holographic information, wherein for each of a first plurality of reference populations of cells comprising T cells, the dataset of reference holographic information comprises holographic information obtained for individual cells of the reference population of cells; (b) determining, from the convolutional neural network, one or more reference input measures for each cellular feature of a plurality of cellular features derived from the holographic information, wherein the plurality of cellular features are cellular features extracted by the convolutional neural network, and each reference input measure is from an individual cell of a reference population of cells; (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 the plurality of cellular features, each reference population-level statistic determined by applying a distribution-based pooling filter to the one or more reference input measures for a cellular feature of the plurality of cellular features; and (d) training a machine learning model using the dataset of reference population-level statistics and a dataset of reference populationlevel output measures, wherein for each of a second plurality of reference populations of cells, the dataset of reference population-level output measures comprises a reference population-level output measure of expression of a marker expressed by activated T cells for the reference population of cells, wherein: the first and second pluralities of reference populations of cells are from reference cultures of cells being cultured in vitro or ex vivo', and reference populations of cells from the first plurality of reference populations of cells are from the same reference culture of cells as reference populations of cells from the second plurality of reference populations of cells; whereby the machine learning model is trained to predict population-level output measures of expression of the marker based on populationlevel statistics of the plurality of cellular features.

[0071] In some of any embodiments, the method is for determining recombinant receptor expression of the population of T cells, and the marker is a recombinant receptor introduced into T cells of the population of cells prior to when the holographic information is obtained. In some of any embodiments, the recombinant receptor is a chimeric antigen receptor (CAR) or an engineered T cell receptor (TCR).

[0072] Also provided herein in some embodiments is a method for training a machine learning model that predicts recombinant receptor of a population of T cells, comprising: (a) training a convolutional neural network using a dataset of reference holographic information, wherein for each of a first plurality of reference populations of cells comprising T cells, the dataset of reference holographic information comprises holographic information obtained for individual cells of the reference population of cells; (b) determining, from the convolutional neural network, one or more reference input measures for each cellular feature of a plurality of cellular features derived from the holographic information, wherein the plurality of cellular features are cellular features extracted by the convolutional neural network, and each reference input measure is from an individual cell of a reference population of cells; (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 the plurality of cellular features, each reference population-level statistic determined by applying a distribution-based pooling filter to the one or more reference input measures for a cellular feature of the plurality of cellular features; and (d) training a machine learning model using the dataset of reference population-level statistics and a dataset of reference populationlevel output measures, wherein for each of a second plurality of reference populations of cells, the dataset of reference population-level output measures comprises a reference population-level output measure for the reference population of cells of expression of a recombinant receptor introduced into T cells of the reference population of cells, wherein: the first and second pluralities of reference populations of cells are from reference cultures of cells being cultured in vitro or ex vivo', and reference populations of cells from the first plurality of reference populations of cells are from the same reference culture of cells as reference populations of cells from the second plurality of reference populations of cells; whereby the machine learning model is trained to predict population-level output measures of expression of the recombinant receptor based on population-level statistics of the plurality of cellular features.

[0073] In some of any embodiments, the one or more input measures for at least one, optionally each, of the plurality of cellular features are obtained from a fully connected layer of the convolutional neural network. In some of any embodiments, the one or more input measures for each of the plurality of cellular features are obtained from a fully connected layer of the convolutional neural network.

[0074] In some of any embodiments, the holographic information for the first plurality of reference populations of cells is obtained by DDHM.

[0075] In some of any embodiments, the holographic information comprises phase information and intensity information.

[0076] In some of any embodiments, the reference population-level output measure is the number of cells, the percentage of cells, the proportion of cells, or the density of cells of the reference population of cells that express the marker.

[0077] In some of any embodiments, the dataset of reference population-level statistics and the dataset of reference population-level output measures are time-matched.

[0078] In some of any embodiments, the in vitro or ex vivo culture of the reference cultures of cells is performed under the same or similar conditions as the in vitro or ex vivo culture of the culture of cells. [0079] In some of any embodiments, the holographic information for the first plurality of reference populations of cells is obtained during the in vitro or ex vivo culture of the reference cultures of cells.

[0080] In some of any embodiments, the dataset of reference population-level output measures are of expression of the marker during or after the in vitro or ex vivo culture of the reference cultures of cells.

[0081] In some of any embodiments, the dataset of reference population-level output measures is determined using fluorescence imaging of the second plurality of reference populations of cells. In some of any embodiments, the fluorescence imaging is by flow cytometry.

[0082] In some of any embodiments, the method is for determining the activation state of the population of T cells, and the marker is expressed by activated T cells.

[0083] In some of any embodiments, the marker is CD137 (4-1BB).

[0084] In some of any embodiments, the method is for determining the memory phenotype of the population of T cells, and the marker is expressed by T cells having a central memory phenotype or a stem cell memory phenotype. In some of any embodiments, the marker is expressed by T cells having a central memory phenotype. In some of any embodiments, the marker is expressed by T cells having a stem cell memory phenotype. In some of any embodiments, the marker is CCR7.

[0085] In some of any embodiments, the method is for determining recombinant receptor expression of the population of T cells, and the marker is a recombinant receptor introduced into T cells of the population of cells prior to when the holographic information is obtained. In some of any embodiments, the recombinant receptor is a chimeric antigen receptor (CAR) or an engineered T cell receptor (TCR).

[0086] In some of any embodiments, the first plurality of reference populations of cells are incubated under T cell stimulating conditions prior to when the holographic information for the first plurality of reference populations of cells is obtained.

[0087] In some of any embodiments, the incubation of the first plurality of reference populations of cells is performed under the same or similar conditions as the incubation of the population of cells. [0088] In some of any embodiments, the incubation of the first plurality of reference populations of cells is prior to the in vitro or ex vivo culture of the reference cultures of cells.

[0089] In some of any embodiments, the first and/or second plurality of reference populations of cells are enriched for T cells. In some of any embodiments, the first and second plurality of reference populations of cells are each enriched for T cells.

[0090] In some of any 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 populations of cells are T cells. In some of any 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 populations of cells are T cells.

[0091] Also provided herein in some embodiments is a method for monitoring a cell phenotype of a culture of T cells, comprising determining, for a population of cells from a culture of cells comprising T cells that is being cultured in vitro or ex vivo, a population-level output measure of expression of a marker expressed by T cells with a cell phenotype of interest, wherein the population-level output measure is determined according to some of any of the provided methods.

[0092] In some of any embodiments, the method is for monitoring the activation state of the population of T cells, and the marker is expressed by activated T cells. In some of any embodiments, the marker is CD137 (4-1BB).

[0093] In some of any embodiments,

[0094] Also provided herein in some embodiments is a method for monitoring the activation state of a culture of T cells, comprising determining, for a population of cells from a culture of cells comprising T cells that is being cultured in vitro or ex vivo, a populationlevel output measure of expression of a marker expressed by activated T cells, wherein the population-level output measure is determined according to some of any of the provided methods.

[0095] In some of any embodiments, the method is for monitoring recombinant receptor expression of the population of T cells, and the marker is a recombinant receptor introduced into T cells of the population of cells prior to when the holographic information is obtained. In some of any embodiments, the recombinant receptor is a chimeric antigen receptor (CAR) or an engineered T cell receptor (TCR). [0096] Also provided herein in some embodiments is a method for monitoring recombinant receptor expression of a culture of T cells, comprising determining, for a population of cells from a culture of cells comprising T cells that is being cultured in vitro or ex vivo, a population-level output measure of expression of a recombinant receptor introduced into T cells of the culture of cells, wherein the population-level output measure is determined according to some of any of the provided methods.

[0097] Also provided herein in some embodiments is a computing device comprising instructions in memory for performing some of any of the provided methods, the instructions comprising instructions for: (a) receiving the holographic information for individual cells of a population of cells comprising T cells, the one or more input measures for each of the plurality of cellular features for individual cells of a population of cells comprising T cells, or the plurality of population-level statistics for a population of cells comprising T cells; and (b) determining, according to the method, the population-level output measure for the population of cells from the holographic information, the one or more input measures for each of the plurality of cellular features, or the plurality of population-level statistics.

[0098] In some of any embodiments, the computing device further comprises in memory a machine learning model trained according to some of any of the provided methods, wherein the population-level output measure for the population of cells is determined using the machine learning model.

Brief Description of the Drawings

[0099] FIG. 1 and FIG. 2 show validation results for a machine learning method for determining the overall activation state (e.g., CD137 positivity) of T cell populations using holographic imaging. The T cell populations for FIG. 1 and FIG. 2 were subjected to different stimulation processes.

[0100] FIG. 3 shows validation results for a machine learning method for determining recombinant receptor expression of T cell populations using holographic imaging.

[0101] FIG. 4 shows validation results for a machine learning method for determining the memory phenotype of T cell populations using holographic imaging. Detailed Description

[0102] Provided herein are methods for determining a cell phenotype, such as activation state, of a population of cells containing T cells. In some aspects, the provided methods do not require labeling the cells. In some aspects, the provided methods do not comprise labeling the T cells. In some aspects, the provided methods are label-free methods. In some aspects, the provided methods can be used for monitoring a cell phenotype, such as activation state, of a culture of cells containing T cells, for instance during in vitro or ex vivo culture and prior to downstream processing steps for the T cells. In some embodiments, the determination of a cell phenotype, such as activation state, is performed using machine learning. Also provided herein are methods for training a machine learning model to predict a cell phenotype, such as activation state, of a population of cells containing T cells, as well as computing devices for use in performing any of the provided methods.

[0103] Existing methods for determining a cell phenotype, such as activation state, of T cells, for example during in vitro or ex vivo culture of the T cells, can be time-consuming or require specialized reagents, instrumentation, or trained operators. In some instances, existing methods involve directly manipulating or labelling T cells, for example in a manner involving incubation with immunoaffinity-based reagents, that may interfere with the quality or function of the T cells. Such methods may risk contamination of or damage to cells prior to any downstream processing.

[0104] The provided methods allow for the label-free determination and monitoring of a cell phenotype, such as activation state, of T cells. In some aspects, the provided methods involve the determination of activation state based on expression in T cells of CD 137 (4- 1BB). The results demonstrated herein show that CD137 is a particularly useful marker for predicting overall activation state, as the results are consistent with a finding that CD 137 expression is associated with changes in characteristics that can be monitored with holographic imaging, such as the size of the activated T cells. In some aspects, and without wishing to be bound by theory, the degree to which CD 137 expression dynamically varies over time during activation contributes to its usefulness as a marker for predicting activation state during the monitoring of cells by holographic imaging. In some such embodiments, variation in CD137 expression over time is associated with changes in cell size. In some such embodiments, variation in CD137 expression is associated one or more of cell size, intensity, texture (or smoothness), and circularity. In some embodiments, an increase in CD137 expression is associated with one or more of (e.g., at least one of, at least two of, at least three of, 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, an increase in CD 137 expression is associated with one or more of (e.g., at least one of, at least two of, at least three of, or all of ) decreased intensity, decreased size (e.g., cell area and/or cell radius), greater cell smoothess, and greater circularity.

[0105] The provided methods also allow for the label-free determination and monitoring of another cell phenotype, such as the memory phenotype, of T cells. In some aspects, the provided methods involve the determination of the memory phenotype of the population of T cells. In some embodiments, the determination is based on expression in T cells of CCR7, where CCR7-positive cells are typically considered to be of an earlier 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 a stem cell memory or a central memory phenotype, are reported to result in sustained in vivo response of CAR T cell therapies given their proliferative capacity and effector capabilities in both hematological and solid tumor environments. 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 show that CCR7 expression is associated with changes in characteristics that can be monitored with holographic imaging. In some such embodiments, variation in CCR7 expression is associated one or more of cell size, intensity, texture (or smoothness), and circularity. In some embodiments, an increase in CCR7 expression is associated with one or more of (e.g., at least two of, at least three of, at least four of, or all of) the features of cell area, intensity mean, perimeter, equivalent peak diameter, and peak area normalized. In some embodiments, an increase in CCR7 expression is associated with one or more of (e.g., at least two of, at least three of, at least four of, or all of) the features of (i) greater cell size, (ii) smaller mean intensity, (iii) larger phase correlation, and (iv) smaller normalized radius variance. In some embodiments, greater cell size is indicated by greater area, longer perimeter, and/or larger diameter. [0106] In some aspects, the provided methods further comprise engineering the T cells for which the phenotype is determined and/or monitored to produce a cell therapy product.

[0107] In some aspects, the provided methods involve determining, for a population of cells containing T cells, a population-level output measure of expression of a marker expressed by T cells, such as a marker expressed by activated T cells. For instance, in some embodiments, the population-level output measure is the number of cells, the percentage of cells, the proportion of cells, or the density of cells of the population of cells that express the marker.

[0108] In some aspects, the population-level output measure is determined based on holographic information obtained for individual cells of the population of cells. In some aspects, the population-level output measure is determined based on cellular features derived from the holographic information for the individual cells. In some aspects, the populationlevel output measure is determined based on population-level statistics of the cellular features. In some aspects, the population-level statistics summarize input measures of the cellular features across individual cells of the population of cells. For example, in some embodiments, the population-level statistics include quantile values of the input measures across individual cells of the population of cells for the cellular features.

[0109] In some aspects, population-level statistics of cellular features derived from holographic information obtained for individual cells of reference populations of T cells are used to train a machine learning model to predict population-level output measures of expression of the marker. In some embodiments, the machine learning model is trained using reference population-level output measures of expression of the marker for reference populations of cells. In some embodiments, the reference population-level output measures are determined using fluorescence imaging, such as by immunoaffinity-based fluorescent labelling for the marker.

[0110] As demonstrated herein, marker expression of individual cells, such as activation marker expression, was not required or used for developing 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. In addition, because the cells that were holographically imaged did not have to be fluorescently labelled as part of the provided methods, the impact on cellular feature measurements and model predictive accuracy of any random or systematic morphometric changes that can occur in labelled cells, compared to unlabelled cells, was avoided.

[0111] In addition, the same holographic imaging system used in accordance with the provided methods for obtaining holographic information for training the machine learning model can also be used for obtaining holographic information for prediction using the trained machine learning model. This is in contrast to other methods in which different imaging systems may be used for obtaining holographic information for model training and use, for instance a dual fluorescence-holographic imaging system for model training and a holographic imaging-only system for model use. Thus, in some aspects, the machine learning models of the provided methods are highly transferrable for use in subsequent applications of the models following model training, for instance compared to other methods involving use of different imaging systems before and after model training.

[0112] In some embodiments, holographic information is provided as input to a convolutional neural network. In some embodiments, the selection or design of cellular features derived from holographic information is not required prior to model training. Instead, in some embodiments, cellular features can be automatically extracted from holographic information as part of model training, for instance extracted by a convolutional neural network. In some embodiments, cellular features that are automatically extracted include those that may not be detectable by humans or may not be known to be predictive of expression of the marker. In some embodiments, cellular features that are automatically extracted are not detectable by humans based on visual inspection of holographic images. In some embodiments, one or more of the cellular features that are automatically extracted are not known to be predictive of expression of the marker (e.g., the activation marker). Thus, in some aspects, the provided methods are accurate, efficient, objective, and unbiased methods for predicting cell phenotype, such as activation state.

[0113] In some aspects, the methods described herein can be used to monitor and characterize T cell phenotype, such as activation state or memory status, using a nondestructive, non-damaging approach in which T cells (e.g., unlabelled T cells) can be imaged without damaging them (e.g., cells can be drawn from a bioreactor and returned to the bioreactor in an undamaged state). The provided methods can be used to monitor the dynamics of T cell phenotype, such as activation state or memory status, during cultivation without frequent cell sampling or arduous analytical techniques. In some aspects, the provided methods can be used to identify relationships between a cell phenotype (such as activation state or memory status) and outcomes, for instance how the cells evolve over time.

[0114] In some embodiments, the provided methods can be used to predict the quality of T cells subjected to a manufacturing process, e.g., the quality of T cells during or after the manufacturing process. In some aspects, a cell phenotype (such as activation state or memory status) as predicted by the provided methods can be used as a readout during manufacturing, e.g., of manufacturing success or of the success of a manufacturing step (e.g., of cell stimulation). In some embodiments, the provided methods can be used to monitor whether T cells are sufficiently activated, such as for expansion of the T cells to a desired threshold number, for instance to numbers needed for clinical doses of the T cells for a T cell therapy.

[0115] In some aspects, the activation of T cells can lead to the differentiation of T cells. Higher proportions of early memory T cells, such as naive-like T cells, in T cell therapies 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 the T cells, either directly or by monitoring the activation state of the T cells. In some embodiments, the activation state of the T cells is monitored to predict the memory status of the T cells.

[0116] In some aspects, the cell phenotype information obtained by the provided methods can be used during process development to optimize the duration or other conditions of the manufacturing process or steps thereof in order to improve the quality of processed T cells. In some instances, this information can be used to develop process control strategies in which, for example, when a predicted cell phenotype, such as activation state, falls outside a determined range, conditions of one or more (e.g., the current or a subsequent) manufacturing steps can be altered, e.g., the duration of the current or subsequent manufacturing step can be altered, to improve the final quality of the T cells being manufactured. For example, when a predicted cell phenotype, such as activation state, falls outside a determined range during cultivation, subsequent cultivation can, in some instances, be performed under perfusion conditions and/or in the presence of small molecules for, e.g., modulating T cell phenotype towards desired profiles. [0117] In some aspects, the cell phenotype information obtained by the provided methods can be used to assess or reduce batch-to-batch variability of T cells subjected to the manufacturing process. In some aspects, the cell phenotype information can be used to assess or reduce batch-to-batch variability of a drug product produced using the manufacturing process. For instance, by ensuring that T cells across different cell therapy manufacturing runs are at comparable activation states, the differentiation and memory status of the T cells can be kept consistent. This can reduce variability (e.g., patient-to-patient variability) in the resulting T cell therapies (see, e.g., Jiang et al., Journal of Pharmaceutical Sciences (2021) 110:1871-1876).

[0118] In some embodiments, the provided methods can be used to monitor the activation state of T cells prior to or following the engineering of the T cells. In some aspects, transgene expression can be higher in activated vs. non-activated T cells, such as following the viral transduction of the T cells (see, e.g., Ghassemi et al., Nature Biomedical Engineering (2022) 6:118-128). In some aspects, electroporation efficiency for engineering can be higher in activated vs. non-activated T cells (see, e.g., Zhang et al., BMC Biotechnology (2018) 18:4). In some embodiments, the T cells are monitored in accordance with the provided methods prior to engineering, for instance so that engineering can be initiated once the provided methods predict that the T cells are sufficiently activated for improved transgene expression. In some embodiments, the T cells are monitored in accordance with the provided methods following engineering, for instance to determine whether the T cells are or remain sufficiently activated following engineering to improve transgene expression.

[0119] All publications, including patent documents, scientific articles, 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 individually incorporated by reference. If a definition set forth herein is contrary to or otherwise inconsistent with a definition set forth in the patents, applications, published applications, and other publications that are herein incorporated by reference, the definition set forth herein prevails over the definition that is incorporated herein by reference.

[0120] The section headings used herein are for organizational purposes only and are not to be construed as limiting the subject matter described. I. METHODS FOR DETERMINING T CELL PHENOTYPE

[0121] In some embodiments, the provided methods involve determining a cell phenotype, such as activation state, of a population of cells that contains T cells. In some embodiments, the provided methods are for determining a cell phenotype, such as activation state, of a population of cells that contains T cells. Exemplary populations of cells are described in Section II. In some embodiments, the provided methods involve performing any of the cell processing steps described in Section II for the population of cells.

[0122] Provided herein, in some embodiments, is a method for determining a cell phenotype of a population of T cells, comprising determining, for a population of cells comprising T cells, a population-level output measure of expression of a marker expressed by T cells with a cell phenotype of interest, wherein the population-level output measure is determined based on a plurality of population-level statistics, wherein: each population-level statistic is of one or more input measures for a cellular feature of a plurality of cellular features derived from holographic information obtained for the population of cells, the plurality of population-level statistics comprising one or more population-level statistics for each of the plurality of cellular features; and each input measure is from an individual cell of the population of cells.

[0123] Also provided herein, in some embodiments, is a method for determining a cell phenotype of a population of T cells, comprising: (a) obtaining holographic information for a population of cells comprising T cells; (b) determining one or more input measures for each of a plurality of cellular features derived from the holographic information, wherein each input measure is from an individual cell in the population of cells; (c) determining a plurality of population-level statistics, wherein each population-level statistic is of the one or more input measures for a cellular feature of the plurality of cellular features, and the plurality of population-level statistics comprises one or more population-level statistics for each of the plurality of cellular features; and (d) determining, based on the plurality of population-level statistics, a population-level output measure of expression of a marker expressed by T cells with a cell phenotype of interest for the population of cells. In some embodiments, the term “determining” means predicting.

[0124] Accordingly, in some embodiments, provided herein is a method for predicting a cell phenotype of a population of T cells, comprising: (a) obtaining holographic information for a population of cells comprising T cells; (b) predicting one or more input measures for each of a plurality of cellular features derived from the holographic information, wherein each input measure is from an individual cell in the population of cells; (c) predicting a plurality of population-level statistics, wherein each population-level statistic is of the one or more input measures for a cellular feature of the plurality of cellular features, and the plurality of population-level statistics comprises one or more population-level statistics for each of the plurality of cellular features; and (d) predicting, based on the plurality of population-level statistics, a population-level output measure of expression of a marker expressed by T cells with a cell phenotype of interest for the population of cells.

[0125] Also provided herein, in some embodiments, is a method for training a machine learning model that predicts a cell phenotype of a population of T cells, comprising training a machine learning model using: (i) a dataset of reference population-level statistics, wherein for each of a first plurality of reference populations of cells comprising T cells, the dataset of reference population-level statistics comprises one or more reference population-level statistics for each of a plurality of cellular features, wherein: each reference population-level statistic is of one or more reference input measures for a cellular feature of the plurality of cellular features derived from holographic information obtained for the reference population of cells; and each reference input measure is from an individual cell of the reference population of cells; and (ii) a dataset of reference population-level output measures, wherein for each of a second plurality of reference populations of cells, the dataset of reference population-level output measures comprises a reference population-level output measure of expression of a marker expressed by T cells with a cell phenotype of interest for the reference population of cells, wherein: the first and second pluralities of reference populations of cells are from reference cultures of cells being cultured in vitro or ex vivo', and reference populations of cells from the first plurality of reference populations of cells are from the same reference culture of cells as reference populations of cells from the second plurality of reference populations of cells; whereby the machine learning model is trained to predict population-level output measures of expression of the marker based on population-level statistics of the plurality of cellular features.

[0126] Also provided herein, in some embodiments, is a method for determining the activation state of a population of T cells, comprising determining, for a population of cells comprising T cells, a population-level output measure of expression of a marker expressed by activated T cells, wherein the marker is CD137 (4-1BB), and the population-level output measure is determined based on one or more input measures for each of a plurality of cellular features derived from holographic information obtained for the population of cells, wherein: the plurality of cellular features comprises intensity maximum, intensity minimum, intensity entropy, intensity contrast, phase entropy, cell area, and radius mean; and each input measure is from an individual cell of the population of cells.

[0127] Also provided herein, in some embodiments, is a method for determining the activation state of a population of T cells, comprising determining, for a population of cells comprising T cells, a population-level output measure of expression of a marker expressed by activated T cells, wherein the marker is CD137 (4-1BB), and the population-level output measure is determined based on one or more input measures for each of a plurality of cellular features derived from holographic information obtained for the population of cells, wherein: the plurality of cellular features comprises intensity skewness, intensity correlation, and intensity homogeneity; and each input measure is from an individual cell of the population of cells.

[0128] In some embodiments, the term “determining” means predicting.

[0129] Accordingly, in some embodiments, provided herein is a method for predicting the activation state of a population of T cells, comprising predicting, for a population of cells comprising T cells, a population-level output measure of expression of a marker expressed by activated T cells, wherein the marker is CD137 (4-1BB), and the population-level output measure is predicted based on one or more input measures for each of a plurality of cellular features derived from holographic information obtained for the population of cells, wherein: the plurality of cellular features comprises intensity maximum, intensity minimum, intensity entropy, intensity contrast, phase entropy, cell area, and radius mean; and each input measure is from an individual cell of the population of cells.

[0130] Also provided herein, in some embodiments, is a method for predicting the activation state of a population of T cells, comprising predicting, for a population of cells comprising T cells, a population-level output measure of expression of a marker expressed by activated T cells, wherein the marker is CD137 (4-1BB), and the population-level output measure is predicted based on one or more input measures for each of a plurality of cellular features derived from holographic information obtained for the population of cells, wherein: the plurality of cellular features comprises intensity skewness, intensity correlation, and intensity homogeneity; and each input measure is from an individual cell of the population of cells.

[0131] Also provided herein, in some embodiments, is a method for determining the memory phenotype of a population of T cells, comprising determining, for a population of cells comprising T cells, a population-level output measure of expression of a marker expressed by central memory T cells or stem cell memory T cells, wherein the marker is CCR7, and the population-level output measure is determined based on one or more input measures for each of a plurality of cellular features derived from holographic information obtained for the population of cells, wherein: the plurality of cellular features comprises one or more of cell area, perimeter, mean intensity, normalized peak area, and equivalent peak diameter; and each input measure is from an individual cell of the population of cells. In some embodiments, the term “determining” means predicting.

[0132] Accordingly, provided herein, in some embodiments, is a method for predicting the memory phenotype of a population of T cells, comprising predicting, for a population of cells comprising T cells, a population-level output measure of expression of a marker expressed by central memory T cells or stem cell memory T cells, wherein the marker is CCR7, and the population-level output measure is predicted based on one or more input measures for each of a plurality of cellular features derived from holographic information obtained for the population of cells, wherein: the plurality of cellular features comprises one or more of cell area, perimeter, mean intensity, normalized peak area, and equivalent peak diameter; and each input measure is from an individual cell of the population of cells.

[0133] Also provided herein, in some embodiments, is a method for training a machine learning model that predicts a cell phenotype of a population of T cells, comprising: (a) training a convolutional neural network using a dataset of reference holographic information, wherein for each of a first plurality of reference populations of cells comprising T cells, the dataset of reference holographic information comprises holographic information obtained for individual cells of the reference population of cells; (b) determining, from the convolutional neural network, one or more reference input measures for each cellular feature of a plurality of cellular features derived from the holographic information, wherein the plurality of cellular features are cellular features extracted by the convolutional neural network, and each reference input measure is from an individual cell of a reference population of cells; (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 the plurality of cellular features, each reference population-level statistic determined by applying a distribution-based pooling filter to the one or more reference input measures for a cellular feature of the plurality of cellular features; and (d) training a machine learning model using the dataset of reference population-level statistics and a dataset of reference populationlevel output measures, wherein for each of a second plurality of reference populations of cells, the dataset of reference population-level output measures comprises a reference population-level output measure of expression of a marker expressed by T cells with a cell phenotype of interest for the reference population of cells, wherein: the first and second pluralities of reference populations of cells are from reference cultures of cells being cultured in vitro or ex vivo', and reference populations of cells from the first plurality of reference populations of cells are from the same reference culture of cells as reference populations of cells from the second plurality of reference populations of cells; whereby the machine learning model is trained to predict population-level output measures of expression of the marker based on population-level statistics of the plurality of cellular features.

[0134] Also provided herein, in some embodiments is a method for determining recombinant receptor expression of a population of T cells, comprising determining, for a population of cells comprising T cells, expression of a recombinant receptor introduced into T cells of the population of cells, wherein the determining is based on one or more input measures for each of a plurality of cellular features derived from holographic information obtained for the population of cells, wherein: the plurality of cellular features comprises peak area, phase average uniformity, intensity geometric mean, minimum optical height, and normalized optical height; and each input measure is from an individual cell of the population of cells. In some embodiments, the term “determining” means predicting.

[0135] Accordingly, provided herein, in some embodiments, is a method for predicting recombinant receptor expression of a population of T cells, comprising predicting, for a population of cells comprising T cells, expression of a recombinant receptor introduced into T cells of the population of cells, wherein the determining is based on one or more input measures for each of a plurality of cellular features derived from holographic information obtained for the population of cells, wherein: the plurality of cellular features comprises peak area, phase average uniformity, intensity geometric mean, minimum optical height, and normalized optical height; and each input measure is from an individual cell of the population of cells.

[0136] In some embodiments, the cell phenotype is expression of a marker by cells of the population of cells. 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 population of cells, the degree to which marker-expressing cells are present in the population of cells, or the degree to which marker-expressing cells are present in T cells of the population of cells. In some embodiments, the presence or absence of marker-expressing cells in the population of cells is determined. In some embodiments, the degree to which marker-expressing cells are present in the population of cells is determined. In some embodiments, the degree to which marker-expressing cells are present in T cells of the population of cells is determined.

[0137] In some embodiments, the cell phenotype is based on expression of one or more of a combination of markers by cells of the population of cells. In some embodiments, the cell phenotype is based on a population-leve output measure of expression of one or more of a combination of markers by cells of the population of cells. In some embodiments, the markers are expressed on the surface of T cells. In some embodiments, the cell phenotype is expression of some markers of the combination of markers and non-expression of the remaining markers of the combination of markers. In some embodiments, the cell phenotype is a population-level output measure of expression of some markers of the combination of markers and non-expression of the remaining markers of the combination of markers. In some embodiments, the cell phenotype is expression of each of the combination of markers. In some embodiments, the cell phenotype is a population-leve output measure of expression of each of the combination of markers.

[0138] In some embodiments, reference herein to “marker” also refers to a combination of markers, such as any combination of the markers described herein. In some embodiments, reference herein to “expression of the marker”, “marker-expressing cells”, “cells expressing the marker”, or similar also refers to combination marker expression and/or non-expression consistent with the cell phenotype of interest. For instance, for the cell phenotype CCR7+CD45RA-, “marker-expressing cells” can refer to CCR7+CD45RA- cells. Likewise, for the cell phenotype CCR7+CD27+, “marker-expressing cells” can refer to CCR7+CD27+ cells.

[0139] In some embodiments, the provided methods involve determining the activation state of the population of cells. In some embodiments, “activation state” refers to the presence or absence of activated T cells in the population of cells, the degree to which activated T cells are present in the population of cells, or the degree to which activated T cells are present in T cells of the population of cells. In some embodiments, the presence or absence of activated T cells in the population of cells is determined. In some embodiments, the degree to which activated T cells are present in the population of cells is determined. In some embodiments, the degree to which activated T cells are present in T cells of the population of cells is determined. In some embodiments, such determinations are predictions.

[0140] In some embodiments, the provided methods involve determining the memory status of the population of cells. In some embodiments, “memory status” refers to the presence or absence of T cells of a particular memory phenotype in the population of cells, the degree to which T cells of a particular memory phenotype are present in the population of cells, or the degree to which T cells of a particular memory phenotype are present in T cells of the population of cells. In some embodiments, the presence or absence of T cells of a particular memory phenotype in the population of cells is determined. In some embodiments, the degree to which T cells of a particular memory phenotype are present in the population of cells is determined. In some embodiments, the degree to which T cells of a particular memory phenotype are present in T cells of the population of cells is determined.

[0141] In some embodiments, the memory phenotype is that of naive-like T cells. In some embodiments, the presence or absence of naive-like T cells in the population of cells is determined. In some embodiments, the degree to which naive-like T cells are present in the population of cells is determined. In some embodiments, the degree to which naive-like T cells are present in T cells of the population of cells 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 the population of cells is determined. In some embodiments, the degree to which T cells having a central memory phenotype or a stem cell memory phenotype are present in the population of cells is determined. In some embodiments, the degree to which T cells having a central memory phenotype or a stem cell memory phenotype are present in T cells of the population of cells is determined. In some embodiments, the determination is a prediction.

[0142] In some embodiments, the memory phenotype is that of central memory T cells. In some embodiments, the presence or absence of central memory T cells in the population of cells is determined. In some embodiments, the degree to which central memory T cells are present in the population of cells is determined. In some embodiments, the degree to which central memory T cells are present in T cells of the population of cells is determined.

[0143] In some embodiments, the memory phenotype is that of effector memory T cells. In some embodiments, the presence or absence of effector memory T cells in the population of cells is determined. In some embodiments, the degree to which effector memory T cells are present in the population of cells is determined. In some embodiments, the degree to which effector memory T cells are present in T cells of the population of cells is determined.

[0144] In some embodiments, the memory phenotype is that of terminally differentiated T cells. In some embodiments, the presence or absence of terminally differentiated T cells in the population of cells is determined. In some embodiments, the degree to which terminally differentiated T cells are present in the population of cells is determined. In some embodiments, the degree to which terminally differentiated T cells are present in T cells of the population of cells is determined.In some embodiments, the provided methods involve determining, for the population of cells, a population-level output measure indicative of the cell phenotype, e.g., activation state, of the population of cells. In some embodiments, the population-level output measure indicates the presence or absence of marker-expressing cells in the population of cells. In some embodiments, the population-level output measure indicates the degree to which marker-expressing cells are present in the population of cells. In some embodiments, the population-level output measure indicates the degree to which marker-expressing cells are present in T cells of the population of cells.

[0145] In some embodiments, the population-level output measure is the number of marker-expressing cells, the percentage of marker-expressing cells, the proportion of markerexpressing cells, or the density of marker-expressing cells in the population of cells. In some embodiments, the population-level output measure 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 in T cells of the population of cells. In some embodiments, the population-level output measure is the percentage of marker-expressing cells in T cells of the population of cells.

[0146] In some embodiments, the marker is expressed on the surface of T cells.

[0147] In some embodiments, the marker is CD3. In some embodiments, the marker is CD4. In some embodiments, the marker is CD8.

[0148] 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.

[0149] In some embodiments, the recombinant receptor is a T cell receptor (TCR). Exemplary TCRs are described in Section II-C-3.

[0150] In some embodiments, the provided methods involve determining, for the population of cells, a population-level output measure indicative of the activation state of the population of cells. In some embodiments, the population-level output measure indicates the presence or absence of activated T cells in the population of cells. In some embodiments, the population-level output measure indicates the degree to which activated T cells are present in the population of cells. In some embodiments, the population-level output measure indicates the degree to which activated T cells are present in T cells of the population of cells.

[0151] In some embodiments, the population-level output measure 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 in the population of cells. In some embodiments, the population-level output measure 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 in T cells of the population of cells. In some embodiments, the population-level output measure is the percentage of activated T cells in T cells of the population of cells.

[0152] In some embodiments, the provided methods involve determining, for the population of cells, a population-level output measure of expression of a marker indicative of T cell activation. In some embodiments, “activation state” refers to the presence of T cells expressing the marker in the population of cells, the degree to which T cells expressing the marker are present in the population of cells, or the degree to which T cells expressing the marker are present in T cells of the population of cells.

[0153] In some embodiments, the marker is a marker that is expressed by activated T cells (an “activation marker”). In some embodiments, “activated T cell” refers to a T cell expressing 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 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 CD 137 and one or more further activation markers. In some embodiments, the one or more further activation markers comprise CD25 and/or CD69.

[0154] In other embodiments, the marker is a marker that is expressed by non-activated T cells. In some embodiments, “activated T cell” refers to a T cell not expressing such a marker.

[0155] In some embodiments, the provided methods involve determining, for the population of cells, a population-level output measure indicative of the memory status of the population of cells. In some embodiments, the population-level output measure indicates the presence or absence of T cells of a particular memory phenotype in the population of cells. In some embodiments, the population-level output measure indicates the degree to which T cells of a particular memory phenotype are present in the population of cells. In some embodiments, the population-level output measure indicates the degree to which T cells of a particular memory phenotype are present in T cells of the population of cells. Early memory phenotypes, such as a stem cell memory or a central memory phenotype, are reported to result in sustained in vivo response of CAR T cell therapies given their proliferative capacity and effector capabilities in both hematological and solid tumor environments. See Gargett et al., Cytotherapy, 21: 593-602 (2019). CCR7 positive cells are typically considered to be of an earlier memory phenotype.

[0156] In some embodiments, the population-level output measure 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 in the population of cells. In some embodiments, the population-level output measure 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 in T cells of the population of cells. In some embodiments, the population-level output measure is the percentage of T cells of a particular memory phenotype in T cells of the population of cells.

[0157] In some embodiments, the memory phenotype is that of naive-like T cells. In some embodiments, the population-level output measure 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 in the population of cells. In some embodiments, the population-level output measure 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 in T cells of the population of cells. In some embodiments, the population-level output measure is the percentage of naive-like T cells in T cells of the population of cells.

[0158] In some embodiments, the memory phenotype is that of central memory phenotype or stem cell memory phenotype.

[0159] In some embodiments, the memory phenotype is that of central memory T cells. In some embodiments, the population-level output measure 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 in the population of cells. In some embodiments, the population-level output measure 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 in T cells of the population of cells. In some embodiments, the populationlevel output measure is the percentage of central memory T cells in T cells of the population of cells.

[0160] In some embodiments, the memory phenotype is that of stem cell memory T cells. In some embodiments, the population-level output measure 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 in the population of cells. In some embodiments, the population-level output measure 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 in T cells of the population of cells. In some embodiments, the population-level output measure is the percentage of stem cell memory T cells in T cells of the population of cells.

[0161] In some embodiments, the memory phenotype is that of effector memory T cells. In some embodiments, the population-level output measure 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 in the population of cells. In some embodiments, the population-level output measure 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 in T cells of the population of cells. In some embodiments, the populationlevel output measure is the percentage of effector memory T cells in T cells of the population of cells.

[0162] In some embodiments, the memory phenotype is that of terminally differentiated T cells. In some embodiments, the population-level output measure 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 in the population of cells. In some embodiments, the population-level output measure 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 in T cells of the population of cells. In some embodiments, the population-level output measure is the percentage of terminally differentiated T cells in T cells of the population of cells.

[0163] In some embodiments, the provided methods involve determining, for the population of cells, a population-level output measure of expression of a marker indicative of T cell memory status. In some embodiments, “memory status” refers to the presence of T cells expressing the marker in the population of cells, the degree to which T cells expressing the marker are present in the population of cells, or the degree to which T cells expressing the marker are present in T cells of the population of cells.

[0164] In some embodiments, the marker is a marker that is expressed by naive-like T cells (a “naive-like marker”). In some embodiments, “naive-like T cell” refers to a T cell expressing such a marker. In some embodiments, the marker is CCR7. In some embedments, the marker is CD27. In some embodiments, the marker is CD45RA.

[0165] In other embodiments, the marker is a marker that is expressed by non-naive-like T cells. In some embodiments, “naive-like T cell” refers to a T cell not expressing such a marker.

[0166] In some embodiments, the combination of markers is any combination of the markers described herein.

[0167] In some embodiments, the combination of markers is a combination of markers indicative of the memory phenotype of T cells. In some embodiments, the combination of markers is a combination of markers that is expressed by T cells of a particular memory phenotype.

[0168] In some embodiments, the memory phenotype is of an earlier memory phenotype. In some embodiments, the earlier memory phenotype is a central memory phenotype or a stem cell memory phenotype. In some embodiments, the earlier memory phenotype is a central memory phenotype. In some embodiments, the earlier memory phenotype is a stem cell memory phenotype. In some embodiments, the marker that is expressed by the earlier memory phenotype, e.g., the central memory phenotype or the stem cell memory phenotype, is CCR7.

[0169] In some embodiments, the memory phenotype is that of 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+.

[0170] In some embodiments, the memory phenotype is that of central memory T cells. In some embodiments, the central memory T cells are CCR7+CD45RA-.

[0171] In some embodiments, the memory phenotype is that of effector memory T cells. In some embodiments, the effector memory T cells are CCR7-CD45RA-.

[0172] In some embodiments, the memory phenotype is that of terminally differentiated T cells. In some embodiments, the effector memory T cells are CCR7-CD45RA+.

[0173] In some embodiments, the population-level output measure indicates the presence of T cells expressing the marker in the population of cells. In some embodiments, the population-level output measure indicates the degree to which T cells expressing the marker are present in the population of cells. In some embodiments, the population-level output measure indicates the degree to which T cells expressing the marker are present in T cells of the population of cells.

[0174] In some embodiments, the population-level output measure is the number of cells, the percentage of cells, the proportion of cells, or the density of cells in the population of cells that express the marker. In some embodiments, the population-level output measure is the number of cells, the percentage of cells, the proportion of cells, or the density of cells in T cells of the population of cells that express the marker. In some embodiments, the populationlevel output measure is the percentage of cells in T cells of the population of cells that express the marker.

[0175] In some embodiments, the cell phenotype, e.g., activation state, is determined based on holographic information obtained for the population of cells, e.g., for individual cells of the population of cells. In some embodiments, the population-level output measure is determined based on holographic information obtained for the population of cells, e.g., for individual cells of the population of cells. In some embodiments, the provided methods involve obtaining holographic information for the population of cells, e.g., for individual cells of the population of cells. In some embodiments, the holographic information includes holographic information for one or more individual cells of the population of cells. In some embodiments, the holographic information includes holographic information for each of a plurality of individual cells of the population of cells. Exemplary holographic information and methods for obtaining same are described in Section I-A.

[0176] In some embodiments, the cell phenotype, e.g., activation state, is determined also based on non-holographic information obtained for the population of cells. In some embodiments, the population-level output measure is determined also based on nonholographic information obtained for the population of cells. In some embodiments, the nonholographic information includes non-holographic information for one or more individual cells of the population of cells. In some embodiments, the non-holographic information includes non-holographic information for each of a plurality of individual cells of the population of cells.

[0177] In some embodiments, the cell phenotype, e.g., activation state, is determined based on one or more input measures for each of one or more cellular features. In some embodiments, the population-level output measure is determined based on one or more input measures for each of one or more cellular features. In some embodiments, “cellular feature” refers to a characteristic of an individual cell. In some embodiments, the one or more cellular features are derived from holographic information. In some embodiments, each cellular feature is derived from holographic information obtained for an individual cell.

[0178] In some embodiments, “input measure” refers to a value that is determined, e.g., measured or calculated, for a feature, e.g., a cellular feature. In some embodiments, the provided methods involve determining one or more input measures for each of one or more cellular features. In some embodiments, each input measure is determined from holographic information obtained for the population of cells, e.g., for individual cells of the population of cells. In some embodiments, each input measure is from an individual cell of the population of cells. In some embodiments, each input measure is determined from holographic information obtained for an individual cell of the population of cells. In some embodiments, each input measure is measured from the holographic information. In some embodiments, each input measure is calculated using the holographic information. Exemplary cellular features and methods for determining input measures for same are described in Section I-B.

[0179] In some embodiments, the cell phenotype, e.g., activation state, is determined also based on one or more other input measures for each of one or more other features derived from non-holographic information. In some embodiments, the population-level output measure is determined also based on one or more other input measures 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.

[0180] In some embodiments, the cell phenotype, e.g., activation state, is determined based on at least one population-level statistic. In some embodiments, the population-level output measure is determined based on at least one population-level statistic. In some embodiments, each population-level statistic is of one or more input measures for a cellular feature of the one or more cellular features. In some embodiments, each population-level statistic describes the one or more input measures. In some embodiments, each population-level statistic describes the distribution of the one or more input measures. In some embodiments, the provided methods involve determining each population-level statistic from the one or more input measures. In some embodiments, each population-level statistic is calculated using the one or more input measures. Exemplary population-level statistics are described in Section I- C. Exemplary methods for determining the population-level output measure based on the at least one population-level statistic are described in Section I-D.

[0181] 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 measure is determined based on one or more other population-level statistics. In some embodiments, the one or more other population-level statistics are of one or more other input measures for each of the one or more other features derived from non-holographic information.

[0182] In some embodiments, any number of the steps of the provided methods are performed in a closed system. In some embodiments, any number of the steps of the provided method are automated.

A. Holographic Information

[0183] In some embodiments, the cell phenotype, e.g., activation state, of the population of cells is determined based on holographic information obtained for the population of cells, e.g., for individual cells of the population of cells. In some embodiments, the population-level output measure is determined based on holographic information obtained for the population of cells, e.g., for individual cells of the population of cells. In some embodiments, the holographic information obtained for the population of cells includes holographic information for each of a plurality of individual cells of the population of cells. In some embodiments, the holographic information for each of the plurality of individual cells is obtained simultaneously or near simultaneously.

[0184] In some embodiments, the holographic information is obtained using methods that do not involve recording the projected image of the population of cells. In some embodiments, the holographic information includes a hologram obtained of the population of cells. In some embodiments, the holographic information is obtained using a light wavefront of interest and a reference wavefront. In some embodiments, the holographic information includes phase information and intensity information obtained for the population of cells, for instance as described 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 the light wavefront of interest and the reference wavefront.

[0185] In some embodiments, an image of the population of cells can be reconstructed from the holographic information. In some embodiments, an image of the population of cells can be reconstructed from the hologram. In some embodiments, an image of the population of cells can be reconstructed from the phase information and the intensity information. In some embodiments, the reconstruction is performed using a computer and numerical algorithms.

[0186] In some embodiments, the provided methods involve obtaining the holographic information for the population of cells, e.g., for 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 for individual cells of the population of cells is obtained by segmenting holographic information obtained for the population of cells.

[0187] In some embodiments, the holographic information is obtained by iterative imaging of cells, e.g., individual cells or samples of cells drawn from a bioreactor that holds a population of cells (e.g., a population of cells in suspension).

[0188] In some embodiments, “holographic information” refers to a 2D image of a population of cells or of an individual cell that is obtained using holographic imaging. In some embodiments, the holographic information includes multiple 2D images of the population of cells or the individual cell that are obtained using holographic imaging. In some embodiments, the holographic information includes a 2D phase image of the population of cells or the individual cell. In some embodiments, the holographic information includes a 2D intensity image of the population of cells or the individual cell. In some embodiments, the holographic information includes a 2D phase image and a 2D intensity image of the population of cells or the individual cell.

[0189] In some embodiments, holographic information itself can be used to train machine learning models, e.g., convolutional neural networks, and holographic information can be provided as input to a machine learning model, e.g., convolutional neural network, in accordance with any of the provided methods. In other embodiments, cellular features derived from holographic information, such as any described in Section I-B, can be used to train machine learning models and provided as input to a machine learning model in accordance with any of the provided methods.

[0190] The imaging technique may use a digital device to acquire and, for example, record the output (e.g., results) of the 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). Thus, 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 results (e.g., output) of the imaging process.

[0191] To obtain holographic information associated with T cells in a non-damaging, nondestructive manner, T cells may be contained in a liquid, such as a culture media. In some embodiments, the T cells may be suspended in a liquid, e.g., a culture media, for imaging. Therefore, in some embodiments, the imaging technique is capable of imaging T cells contained and/or suspended in a liquid.

[0192] In some embodiments, the holographic information is obtained by interferometric microscopy, optical coherence tomography, diffraction phase microscopy, or digital holographic microscopy (DHM). In some embodiments, the holographic information is obtained by imaging the population of cells using interferometric microscopy, optical coherence tomography, diffraction phase microscopy, or DHM.

[0193] 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 the population of cells, e.g., individual cells of the population of cells, using DHM. In some embodiments, the DHM is traditional DHM, in-line DHM, or differential DHM (DDHM).

[0194] DHM is a technique which can allow for recording of a 3D sample or object without the need to scan the sample layer-by-layer. In this respect, DHM can be a superior technique to confocal microscopy. In DHM, holographic information can be recorded by a digital camera such as a CCD- or a CMOS-camera, which can subsequently be 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-Al.

[0195] To make a holographic representation, or hologram, traditionally a highly coherent light source such as laser-light can 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 can be sent via an optical system to the sample to interact with it, thereby altering the phase and amplitude of the light depending on the object’s optical properties and 3D shape. The object beam which has been reflected on or transmitted through the sample can then be made (e.g., by set of mirrors and/or beam splitters) to interfere with the reference beam to result in an interference pattern that can be digitally recorded. Since the hologram can be more accurate when object beam and reference beam have comparable amplitude, an absorptive element can be introduced in the reference beam, which can decrease its amplitude to the level of the object beam without altering the phase of the reference beam, or at most changing the phase globally. The recorded interference pattern can contain information on the phase and amplitude changes which depend on the object's optical properties and 3D shape.

[0196] An alternative way of making a hologram is by using an in-line holographic technique. In some aspects, in-line DHM is similar to traditional DHM, but does not split the beam, at least not by a beam splitter or other external optical element. In-line DHM can be used to look at a solution of particles, e.g. cells, in a fluid such that, for example, some part of the at least partially coherent light will pass through the sample without interacting with the particles (reference beam) and interfere with light that has interacted with the particles (object beam) to give rise to an interference pattern which can be recorded digitally and processed. In-line DHM can be used in transmission mode.

[0197] Another DHM technique is DDHM. In some aspects, DDHM is different to the other techniques in its use of reference and object beams. In a preferred set-up of DDHM, the sample can be illuminated by illumination means which include at least partially coherent light in reflection or in transmission mode. The reflected or transmitted sample beam can be sent through an objective lens and subsequently split in two by a beam splitter and sent along different paths in a differential interferometer, e.g., of the Michelson or Mach-Zehnder type. In one of the paths, a beam-bending element or tilting means can be inserted, e.g., a transparent wedge. The two beams can then be made to interfere with each other in the focal plane of a focusing lens, and the interference pattern in this focal plane can be recorded digitally and stored by, e.g., a CCD-camera connected to a computer. Due to the beambending element, the two beams can be slightly shifted in a controlled way, and the interference pattern can depend on the amount of shifting. Then, the beam-bending element can be turned, thereby altering the amount of shifting. The new interference pattern can also be recorded. This can be done a number of times (N), and from these N interference patterns, the gradient (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 of obtaining the phase gradient are also known, such as a Fourier transform data processing technique. The gradient of the phase can be integrated to give the phase as a function of position. The amplitude of the light as a function of position can be computed from the possibly but not necessarily weighted average of the amplitudes of the N recorded interference patterns. Since phase and amplitude are thus known, the same information can be obtained as in a direct holographic method (using a reference and an object beam), and a subsequent 3D reconstruction of the object can be performed.

[0198] In DDHM, an illumination means can include spatially and temporally partially coherent light. In some aspects, this is in contrast with other DHM methods that might use highly correlated laser light. Spatially and temporally partially coherent light can be produced by, e.g., a LED. A LED can be cheaper than a laser and produce light with a spectrum centered around a known wavelength, which can be spatially and temporally partially coherent, e.g., not as coherent as laser light, but still coherent enough to produce holographic images of sufficient quality for the applications at hand. In some aspects, LEDs can also have the advantage of being available for many different wavelengths and can be very small in size and easy to use or replace if necessary. Therefore, in some aspects, methods that use spatially and temporally partially coherent light for obtaining holographic images can lead to more cost-effective devices for implementing such methods. In some embodiments, the illumination means is a red LED.

[0199] The holographic images may undergo object segmentation and further analysis to obtain a plurality of features that quantitatively describe the imaged objects (e.g., T cells, cellular debris). As such, various features, such as cellular features described in Section I-B, may be directly assessed or calculated from DDHM using, for example, steps of image acquisition, image processing, image segmentation, and feature extraction. In some embodiments, a digital recording device is used to record holographic images. In some embodiments, a computer including algorithms for analyzing holographic images may be used. In some embodiments, a monitor and/or computer may be used for displaying the results of the holographic image analysis. In some embodiments, the analysis is automated (e.g., capable of being performed in the absence of user input)

[0200] Any type of DHM can be used in accordance with the provided methods. In some embodiments, the DHM is traditional DHM. In some embodiments, the DHM is in-line DHM. In some embodiments, the DHM is differential DHM.

[0201] Exemplary DHM systems for use in the provided methods, for instance the Ovizio iLine F (Ovizio Imaging Systems NV/SA, Brussels, Belgium), include those that can be used in conjunction with a bioreactor for the incubation of T cells. In some embodiments, the population of cells for which holographic information is obtained is from a culture of cells being cultured in vitro or ex vivo, for instance as described in Section II-D. In some embodiments, the in vitro or ex vivo culture is in a bioreactor. In some embodiments, the population of cells is moved from the bioreactor for imaging. In some embodiments, the imaging is performed using a digital holographic microscope that is connected to the bioreactor. The microscope can be any that is capable of obtaining phase information of a fluid sample and having illumination means. The microscope can be any that is used for DHM, including any for traditional DHM, in-line DHM, or DDHM.

[0202] In some embodiments, one or more fluidic systems capable of guiding fluid from the bioreactor to the microscope are connected to the bioreactor and microscope. In some embodiments, at least one fluidic system contains one or more tubes which may come in direct contact with fluid from the bioreactor. In some embodiments, at least one tube contains a part which is at least partially transparent for the illumination means of the microscope for obtaining holographic information of said fluid sample. In some embodiments, the tube contains a part which is at least partially transparent for the illumination means of the microscope and which contains a flow cell and/or a microfluidic system. In some embodiments, the flow cell and/or microfluidic system contains a cross section in which the height and/or width varies along the cross section. In some aspects, this allows for obtaining clear holographic images for a variety of concentrations of objects suspended in the fluid. A high concentration of suspended objects could lead to a large number of objects being stacked on top of one another and could lead to difficulties in obtaining a holographic image, especially if the microscope works in transmission mode. A low concentration could result in the microscope obtaining holographic images of the fluid medium only and not of an object suspended in that medium. If the concentration is high, a holographic image can be obtained at the position where the height or width is small, thereby ensuring that not too many objects are stacked in the illumination beam. If the concentration is small, a holographic image can be obtained at the position where the height or width is large, thereby ensuring that at least one suspended object is in the illumination beam. In some embodiments, the microfluidic system contains a branching of the tube into multiple tubes of different cross sections, diameters, heights, and/or widths. Such an arrangement can allow for obtaining clear holographic images for a variety of concentrations of objects suspended in the fluid. In some embodiments, the cross section, diameter, height, and/or width of the flow cell and/or microfluidic system is chosen as a function of the size of the suspended objects and/or the size of the illumination beam of the microscope. In some embodiments, the narrowest dimension in a cross section of the flow cell and/or microfluidic system is larger than 10 micrometer, more preferably larger than 30 micrometer, even more preferably larger than 50 micrometer, and/or the largest dimension in a cross section of the flow cell and/or microfluidic system is smaller than 5000 micrometer, more preferably smaller than 3000 micrometer, even more preferably smaller than 2500 micrometer. In some embodiments, the microfluidic system is attached on a substrate, for instance to ease manufacturing and/or provide stability to the microfluidic system.

[0203] Although it is not necessary, it may be desired that a fluid flow is present in at least one of the fluidic systems. This can allow for sampling of the contents of the bioreactor in time and monitoring of different samples to obtain a better knowledge of the state and/or reactions of the bioreactor. A fluid flow may be present due to natural phenomenon such as convection, conduction, or radiation, by density or pressure differences induced by, e.g., the reactions taking place in the bioreactor or heat gradients, by gravity, etc. If a fluid flow is desired, but is not occurring spontaneously, or if the flow needs to be controlled, one or more pumping systems may be connected to the fluidic systems in order to induce a flow in said systems. Therefore, in some embodiments, at least one pumping system is connected to one or more fluidic systems and is capable of inducing a fluid flow in said fluidic systems.

[0204] In some embodiments, at least one fluidic system contains a fluid-tight flexible part which, when compressed, pulled, and/or pushed, results in a fluid flow in the fluidic system. As such, a fluid flow can be induced in the fluidic system without a high risk of leaks and without contamination of the actuator of the flow. In some embodiments, a pumping system is connected to the fluidic system and is capable of pulling and/or pushing the fluid-tight flexible part to induce a fluid flow in the fluidic system.

[0205] In some aspects, the contents of the bioreactor can be non-destructively monitored. Thereby, it is possible to re-introduce the populations of cells which are observed in the microscope to the bioreactor. Therefore, in some embodiments, at least one fluidic system forms a closed circuit between the bioreactor and the microscope and back to the bioreactor, e.g., the fluidic system is capable of guiding fluid from the bioreactor to the microscope and back.

B. Cellular Features

[0206] In some embodiments, the cell phenotype, e.g., activation state, of the population of cells is determined based on one or more input measures for each of one or more cellular features. In some embodiments, the population-level output measure is determined based on one or more input measures for each of one or more cellular features. In some embodiments, the one or more cellular features are derived from holographic information. In some embodiments, each input measure is determined from holographic information obtained for an individual cell of the population of cells. In some embodiments, determining the one or more input measures for each of the one or more cellular features involves one or more steps selected from holographic information acquisition, holographic information processing, holographic information segmentation, and cellular feature extraction. In some embodiments, image segmentation is used to determine holographic information associated with an individual cell. The process of segmentation is known in the art, for instance, from watershed treatment, clustering-based image threshold and using neural networks. Various algorithms exist for watershed treatment, including Meyer's flooding algorithm and optimal spanning forest algorithms (watershed cuts). Cluster-based image thresholding may employ Otsu’s method. In some embodiments, the holographic information includes the segmented images.

[0207] In some embodiments, the one or more input measures are a plurality of input measures. In some embodiments, the activation state is determined based on a plurality of input measures for each of one or more cellular features. In some embodiments, the population-level output measure is determined based on a plurality of input measures for each of one or more cellular features.

[0208] In some embodiments, the one or more cellular features are a plurality of cellular features. In some embodiments, the activation state is determined based on one or more input measures for each of a plurality of cellular features. In some embodiments, the populationlevel output measure is determined based on one or more input measures for each of a plurality of cellular features.

[0209] In some embodiments, the activation state is determined based on a plurality of input measures for each of a plurality of cellular features. In some embodiments, the population-level output measure is determined based on a plurality of input measures for each of a plurality of cellular features.

[0210] In some embodiments, input measures for some or all of the one or more cellular features are obtained for an individual cell. In some embodiments, input measures for each of the one or more cellular features are obtained for an individual cell.

[0211] In some embodiments, the one or more input measures for a cellular feature of the one or more cellular features include input measures from one or more individual cells of the population of cells. In some embodiments, the one or more input measures for a cellular feature of the one or more cellular features include input measures from a plurality of individual cells of the population of cells. In some embodiments, input measures for a first cellular feature are for a first plurality of individual cells, and input measures for a second cellular feature are for a second, distinct plurality of individual cells that can contain some or none of the first plurality of individual cells. In some embodiments, input measures for a first cellular feature are for a first plurality of individual cells, and input measures for a second cellular feature are also for the first plurality of individual cells. In some embodiments, the input measures for all of the cellular features are from the same plurality of individual cells. [0212] In some embodiments, the one or more input measures for a cellular feature of the one or more cellular features include input measures from some or all individual cells of the population of cells. In some embodiments, the one or more input measures for a cellular feature of the one or more cellular features include input measures from live cells only. In some embodiments, the live cells are classified as live using an automated method. In some embodiments, the live cells are classified as live by analysis of holographic images of cells of the population of cells, for instance, as exemplified herein (e.g., using the OsOne software for the classification of live cells). In some embodiments, the one or more input measures for a cellular feature of the one or more cellular features include input measures only from cells of at least a certain size. In some embodiments, size is determined by the radius mean cellular feature. In some embodiments, size is determined by the cell area cellular feature. In some embodiments, the one or more input measures for a cellular feature of the one or more cellular features include input measures only from cells of sufficient circularity, for instance as determined using the circularity cellular feature. In some embodiments, the one or more input measures for a cellular feature of the one or more cellular features include input measures only from cells having input measures for the intensity smoothness cellular feature below a certain threshold. In some embodiments, the one or more input measures for a cellular feature of the one or more cellular features include input measures only from cells satisfying all of the above criteria.

[0213] In some embodiments, the one or more input measures for a cellular feature of the one or more cellular features include one or more input measures only from cells that meet one or more (e.g., one, two, three, four, or all) of the following criteria: (i) classified as live, (ii) having radius mean cell feature measurements > 5, (iii) having intensity smoothness cell feature measurements < 0.03, (iv) having cell area cell feature measurements > 60, and (v) having circularity cell feature measurements > 0.5. In some embodiments, the cells meet all of the foregoing criteria.

[0214] In some embodiments, the one or more input measures for a cellular feature of the one or more cellular features include input measures from each of the individual cells of the population of cells.

[0215] In some embodiments, a cellular feature of the one or more cellular features is derived from phase information of the holographic information. In some embodiments, a cellular feature of the one or more cellular features is derived from intensity information of the holographic information. In some embodiments, a cellular feature of the one or more cellular features is derived from the phase information and the intensity information. In some embodiments, a cellular feature of the one or more cellular features is derived from an image of an individual cell that is reconstructed from the holographic information. Exemplary methods for determining cellular 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-Al. Exemplary software for determining cellular features from holographic information includes Ovizio OsOne (Ovizio Imaging Systems NV/SA, Brussels, Belgium).

[0216] In some embodiments, the one or more cellular features includes one or more morphological features, one or more optical features, one or more intensity texture features, one or more phase texture features, or a combination of any of the foregoing. In some embodiments, the one or more cellular features includes one or more morphological features, one or more intensity texture features, and one or more phase texture features. Exemplary cellular features and descriptions thereof are described in Table El.

[0217] In some embodiments, the one or more cellular features include one or more morphological features. In some embodiments, the one or more morphological features describe one or more characteristics of an individual cell’s physical shape. In some embodiments, the one or more morphological features are selected from aspect ratio, cell area, circularity, compactness, elongatedness, elongation, 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 include aspect ratio. In some embodiments, the one or more morphological features include cell area. In some embodiments, the one or more morphological features include circularity. In some embodiments, the one or more morphological features include compactness. In some embodiments, the one or more morphological features include elongatedness. In some embodiments, the one or more morphological features include elongation. In some embodiments, the one or more morphological features include diameter. In some embodiments, the one or more morphological features include hu moment 1. In some embodiments, the one or more morphological features include hu moment 2. In some embodiments, the one or more morphological features include hu moment 3. In some embodiments, the one or more morphological features include hu moment 4. In some embodiments, the one or more morphological features include hu moment 5. In some embodiments, the one or more morphological features include hu moment 6. In some embodiments, the one or more morphological features include hu moment 7. In some embodiments, the one or more morphological features include perimeter. In some embodiments, the one or more morphological features include radius mean. In some embodiments, the one or more morphological features include radius variance. In some embodiments, the one or more morphological features include normalized radius variance.

[0218] In some embodiments, the one or more morphological features include cell area and radius mean.

[0219] In some embodiments, the one or more cellular features include one or more optical features. In some embodiments, the one or more optical features describe one or more optical properties of an image of an individual cell. In some embodiments, the one or more optical features are selected from refraction peak diameter, intensity maximum, mean intensity, intensity minimum, mass excentricity, optical height maximum (radians), optical height maximum (pm), mean optical height (radians), mean optical height (pm), normalized optical height, optical height minimum (radians), optical height minimum (pm), optical volume, refraction peak surface, normalized peak area, aggregate size, refraction peak intensity, and normalized peak height. In some embodiments, the one or more optical features include refraction peak diameter. In some embodiments, the one or more optical features include intensity maximum. In some embodiments, the one or more optical features include mean intensity. In some embodiments, the one or more optical features include intensity minimum. In some embodiments, the one or more optical features include mass excentricity. In some embodiments, the one or more optical features include optical height maximum (radians). In some embodiments, the one or more optical features include optical height maximum (pm). In some embodiments, the one or more optical features include mean optical height (radians). In some embodiments, the one or more optical features include mean optical height (pm). In some embodiments, the one or more optical features include normalized optical height. In some embodiments, the one or more optical features include optical height minimum (radians). In some embodiments, the one or more optical features include optical height minimum (pm). In some embodiments, the one or more optical features include optical volume. In some embodiments, the one or more optical features include refraction peak surface. In some embodiments, the one or more optical features include normalized peak area. In some embodiments, the one or more optical features include aggregate size. In some embodiments, the one or more optical features include refraction peak intensity. In some embodiments, the one or more optical features include normalized peak height.

[0220] In some embodiments, the one or more cellular features include one or more intensity texture features. In some embodiments, the one or more intensity texture features describe one or more properties of the intensity information of an individual cell. In some embodiments, the one or more intensity texture features are selected from intensity variance, intensity average contrast, intensity average entropy, intensity average, intensity average 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 include intensity variance. In some embodiments, the one or more intensity texture features include intensity average contrast. In some embodiments, the one or more intensity texture features include intensity average entropy. In some embodiments, the one or more intensity texture features include intensity average. In some embodiments, the one or more intensity texture features include intensity average uniformity. In some embodiments, the one or more intensity texture features include intensity contrast. In some embodiments, the one or more intensity texture features include intensity correlation. In some embodiments, the one or more intensity texture features include intensity entropy. In some embodiments, the one or more intensity texture features include intensity homogeneity. In some embodiments, the one or more intensity texture features include intensity uniformity. In some embodiments, the one or more intensity texture features include intensity skewness. In some embodiments, the one or more intensity texture features include intensity smoothness.

[0221] In some embodiments, the one or more intensity texture features include maximum intensity, minimum intensity, intensity entropy, and intensity contrast.

[0222] In some embodiments, the one or more cellular features include one or more phase texture features. In some embodiments, the one or more phase texture features describe one or more properties 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 (pm), phase average contrast, phase average entropy, phase average, phase average 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 include optical height variance (radians). In some embodiments, the one or more phase texture features include optical height variance (pm). In some embodiments, the one or more phase texture features include phase average contrast. In some embodiments, the one or more phase texture features include phase average entropy. In some embodiments, the one or more phase texture features include phase average. In some embodiments, the one or more phase texture features include phase average uniformity. In some embodiments, the one or more phase texture features include phase contrast. In some embodiments, the one or more phase texture features include phase correlation. In some embodiments, the one or more phase texture features include phase entropy. In some embodiments, the one or more phase texture features include phase homogeneity. In some embodiments, the one or more phase texture features include phase skewness. In some embodiments, the one or more phase texture features include phase smoothness. In some embodiments, the one or more phase texture features include phase uniformity.

[0223] In some embodiments, the cell phenotype is that of activated T cells.

[0224] In some embodiments, the plurality of cellular features includes one or more of (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 cellular features includes intensity skewness, intensity correlation, intensity homogeneity, intensity maximum, intensity minimum, intensity entropy, intensity contrast, phase entropy, cell area, and radius mean.

[0225] In some embodiments, the plurality of cellular features are selected from (comprises one or more of) maximum intensity, minimum intensity, intensity entropy, intensity contrast, phase entropy, cell area, and radius mean. In some embodiments, the plurality of cellular features includes maximum intensity, minimum intensity, intensity entropy, intensity contrast, phase entropy, cell area, and radius mean. [0226] In some embodiments, the one or more cellular features includes one or more of (e.g., any combination of two, three, four, five, six, or all of) intensity maximum, intensity minimum, intensity entropy, intensity contrast, phase entropy, cell area, and radius mean.

[0227] In some embodiments, the plurality of cellular features includes one or more of (e.g., any combination of two or all of) intensity skewness, intensity correlation, and intensity homogeneity. In some embodiments, the plurality of cellular features includes intensity skewness, intensity correlation, and intensity homogeneity.

[0228] In some embodiments, the cell phenotype is a memory phenotype. In some embodiments, the memory phenotype is an earler 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, e.g., a stem cell memory phenotype or a central memory phenotype, and the marker is CCR7.

[0229] Accordingly, in some embodiments, the disclosed method is for determining the memory phenotype of the population of T cells, and 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.

[0230] In some embodiments, the plurality of cellular features comprises one or more of cell area, perimeter, mean intensity, normalized peak area, and equivalent peak diameter. In some embodiments, the plurality of cellular features comprises cell area. In some embodiments, the plurality of cellular features comprises perimeter. In some embodiments, the plurality of cellular features comprises mean intensity. In some embodiments, the plurality of cellular features comprises normalized peak area. In some embodiments, the plurality of cellular features comprises equivalent peak diameter. In some embodiments, the plurality of cellular features comprises cell area, perimeter, mean intensity, normalized peak area, and equivalent peak diameter.

[0231] In some embodiments, the cell phenotype is recombinant receptor expression, e.g., CAR or TCR expression. [0232] In some embodiments, the plurality of cellular features includes one or more of (e.g., any combination of two, three, four or all of) peak area, phase average uniformity, intensity geometric mean, minimum optical height, and normalized optical height. In some embodiments, the plurality of cellular features includes peak area, phase average uniformity, intensity geometric mean, minimum optical height, and normalized optical height. In some embodiments, the plurality of cellular features comprises one or more of larger features reflective of cell size (area, perimeter, diameter), smaller mean intensity features, larger phase correlation features, and smaller radius variance normalized features. In some embodiments, the plurality of cellular features comprises one or more of larger cell area, larger cell perimeter, larger cell diameter, smaller intensity mean, larger phase correlation, and smaller radius variance normalized.

[0233] In some embodiments, the plurality of cellular features are cellular features extracted by a machine learning model. In some embodiments, the extraction is automatic, e.g., without prior design or engineering of the cellular features. In some embodiments, the one or more input measures for at least one of the plurality of cellular features are determined by providing the holographic information for individual cells of the population of cells to the machine learning model. In some embodiments, the one or more input measures for each of the plurality of cellular features are determined by providing the holographic information for individual cells of the population of cells to the machine learning model. In some embodiments, the one or more input measures are determined from the machine learning model.

[0234] In some embodiments, the machine learning model is a deep learning model. In some embodiments, the machine learning model is a convolutional neural network. Advantages of convolutional neural networks include their ability to automatically detect and extract features important for classification or regression without human supervision and to detect features that may not be detectable by humans.

[0235] In some embodiments, the one or more input measures for at least one of the plurality of cellular features are obtained from a fully connected layer of the convolutional neural network. In some embodiments, the one or more input measures for each of the plurality of cellular features are obtained from a fully connected layer of the convolutional neural network. [0236] Architectures of convolutional neural networks suitable for feature extraction can be identified and designed by one of ordinary skill in the art. In some embodiments, the convolutional neural network is a LeNet, AlexNet, ResNet, GoogleNet/InceptionNet, MobileNetVl, ZfNet, Depth-based, Highway Network, Wide ResNet, VGG, PolyNet, Inception v2, Inception v3, Inception v4, Inception-ResNet, DenseNet, Pyramidal Net, Xception, Channel-Boosted, Residual Attention Neural Network, Attention-based, Feature- Map Exploitation-based, Squeeze-and-Excitation, or Competitive-Squeeze-and-Excitation convolutional neural network.

[0237] In some embodiments, the machine learning model, e.g., convolutional neural network, is trained using a dataset of reference holographic information. In some embodiments, for each of a first plurality of reference populations of cells containing T cells, the dataset of reference holographic information contains holographic information obtained for the reference population of cells, e.g., for individual cells of the reference population of cells. Exemplary reference populations of cells are described in Section I-D-l.

[0238] In some embodiments, the machine learning model, e.g., convolutional neural network, is trained using non-holographic images. In some embodiments, the machine learning model, e.g., convolutional neural network, is trained using holographic and nonholographic images.

C. Population-Level Statistics

[0239] In some embodiments, the cell phenotype, e.g., activation state, is determined based on at least one population-level statistic. In some embodiments, the population-level output measure is determined based on at least one population-level statistic. In some embodiments, each population-level statistic is of one or more input measures for a cellular feature of the one or more cellular features. In some embodiments, each population-level statistic describes the one or more input measures for the cellular feature of the one or more cellular features. In some embodiments, each population-level statistic describes the distribution of the one or more input measures for the cellular feature of the one or more cellular features.

[0240] In some embodiments, the at least one population-level statistic includes one or more population-level statistics for each of the one or more cellular features. In some embodiments, the at least one population-level statistic is a plurality of population-level statistics. In some embodiments, the plurality of population-level statistics includes one or more population-level statistics for each of the one or more cellular features. In some embodiments, the plurality of population-level statistics includes multiple population-level statistics for each of the one or more cellular features.

[0241] In some embodiments, the one or more population-level statistics for a first cellular feature are different statistics than the one or more population-level statistics for a second cellular feature. For instance, in some embodiments, the one or more population-level statistics for a first cellular feature include the mean of the one or more input features for the first cellular feature, and the one or more population-level statistics for a second cellular feature do not include the mean of the one or more input features for the second cellular feature. In some embodiments, the one or more population-level statistics for a first cellular feature are the same statistics as the population-level statistics for a second cellular feature. In some embodiments, the one or more population-level statistics are the same statistics for all cellular features.

[0242] 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 cellular features can be independently selected from any such population-level statistics, e.g., from any of the described exemplary population-level statistics. In some embodiments, the one or more population-level statistics for each of the one or more cellular features are the same and are selected from any of the described exemplary population-level statistics.

[0243] In some embodiments, the one or more population-level statistics for a cellular feature of the one or more cellular features include a population-level statistic that summarizes the one or more input measures of the cellular feature. In some embodiments, the population-level statistic is a summary statistic.

[0244] In some embodiments, the one or more population-level statistics for a cellular feature of the one or more cellular features include a measure of the central tendency of the one or more input measures of the cellular feature. In some embodiments, the one or more population-level statistics include one or more of a mean, median, and mode of the one or more input measures of the cellular feature.

[0245] In some embodiments, the one or more population-level statistics for a cellular feature of the one or more cellular features include a measure of the statistical dispersion of the one or more input measures of the cellular feature. In some embodiments, the one or more population-level statistics include one or more of the standard deviation, interquartile range, range, mean absolute difference, median absolute deviation, average absolute deviation, and distance standard deviation of the one or more input measures of the cellular feature.

[0246] In some embodiments, the one or more population-level statistics for a cellular feature of the one or more cellular features include a measure of the shape of the distribution of the one or more input measures of the cellular feature. In some embodiments, the one or more population-level statistics include one or both of the skewness and kurtosis of the one or more input measures of the cellular feature.

[0247] In some embodiments, the one or more population-level statistics for a cellular feature of the one or more cellular features include one or more quantiles of the one or more input measures of the cellular feature. In some embodiments, the one or more quantiles are selected from (include one or more of) the 0.01, 0.1, 0.5, 0.9, and 0.99 quantiles of the one or more input measures of the cellular feature. In some embodiments, the one or more quantiles include the 0.01, 0.1, 0.5, 0.9, and 0.99 quantiles of the one or more input measures of the cellular feature. In some embodiments, the one or more population-level statistics for each cellular feature of the one or more cellular features include the 0.01, 0.1, 0.5, 0.9, and 0.99 quantiles of the one or more input measures of the cellular feature.

[0248] In some embodiments, the one or more population-level statistics for a cellular feature of the one or more cellular features are determined by applying a pooling filter to the one or more input measures of the cellular feature. In some embodiments, the one or more population-level statistics for each cellular feature of the one or more cellular features are determined by applying a pooling filter to the one or more input measures of the cellular feature.

[0249] In some embodiments, the pooling filter summarizes the one or more input measures of the cellular feature. In some embodiments, the pooling filter is a point estimatebased pooling filter. In some embodiments, the pooling filter is a mean pooling filter. In some embodiments, the pooling filter is a maximum pooling filter.

[0250] In some embodiments, the pooling filter is a distribution-based pooling filter. In some embodiments, the distribution-based pooling filter is based on the estimated marginal distribution of the one or more input measures of the cellular features, such as described in Oner et al., arXiv:2006.01561. In some embodiments, the estimated marginal distribution is calculated using kernel density estimation, such as with a Gaussian kernel.

D. Determining Population-Level Output Measures

[0251] In some embodiments, the cell phenotype, e.g., activation state, of the population of cells is determined based on the population-level output measure determined for the population of cells. In some embodiments, the population-level output measure is determined based on the one or more input measures for each of the one or more cellular features. In some embodiments, the population-level output measure is determined based on the at least one population-level statistic determined for the population of cells.

[0252] 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 above the threshold value can indicate the presence of markerexpressing cells. For threshold values for other population-level statistics, a population-level statistic below the threshold value can indicate the presence of marker-expressing cells in the population of cells. In some embodiments, the threshold value for a cellular feature of the one or more cellular features 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 populations of cells containing T cells. In some embodiments, each of the plurality of reference populations of cells contains marker-expressing cells. In some embodiments, each of the plurality of reference populations of cells 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 populations of cells 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 populations of cells 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 T cells of each of the plurality of reference populations of cells are marker-expressing T cells. Exemplary reference populations of cells are described in Section I-D- 1.

[0253] In some embodiments, the population-level output measure indicates whether at least one population-level statistic indicates the presence of marker-expressing cells in the population of cells. In some embodiments, the population-level output measure 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 population of cells. In some embodiments, the population-level output measure indicates whether the majority of population-level statistics indicate the presence of marker-expressing cells in the population of cells. In some embodiments, the population-level output measure indicates whether each population-level statistic indicates the presence of marker-expressing cells in the population of cells.

[0254] 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 above the threshold value can indicate the presence of activated T cells. For threshold values for other population-level statistics, a population-level statistic below the threshold value can indicate the presence of activated T cells in the population of cells. In some embodiments, the threshold value for a cellular feature of the one or more cellular features 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 populations of cells containing T cells. In some embodiments, each of the plurality of reference populations of cells contains activated T cells. In some embodiments, each of the plurality of reference populations of cells contains T cells expressing the 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 populations of cells 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 populations of cells are T cells expressing the marker. In some embodiments, at least 2%, 4%, 6%, 8%, 10%, 20%, 30%, 40%, 50%, 60%, 70%, 80%, or 90% of T cells of each of the plurality of reference populations of cells are T cells expressing the marker. Exemplary reference populations of cells are described in Section I-D- 1.

[0255] In some embodiments, the population-level output measure indicates whether at least one population-level statistic indicates the presence of activated T cells in the population of cells. In some embodiments, the population-level output measure 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 population of cells. In some embodiments, the population-level output measure indicates whether the majority of population-level statistics indicate the presence of activated T cells in the population of cells. In some embodiments, the population-level output measure indicates whether each population-level statistic indicates the presence of activated T cells in the population of cells.

[0256] In some embodiments, the population-level output measure is determined using a trained machine learning model. Exemplary machine learning models and methods of training same 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).

[0257] In some embodiments, the population-level output measure is or is derived from an output of the trained machine learning model. In some embodiments, the one or more input measures for each of the one or more cellular features are provided as input to the trained machine learning model or to a process that includes the trained machine learning model. In some embodiments, the at least one population-level statistic for each of the one or more cellular features is provided as input to the trained machine learning model or to a process that includes the trained machine learning model. In some embodiments, as part of being provided as input to the trained machine learning model, the input measures or populationlevel statistics undergo one or more preprocessing steps. In some embodiments, the preprocessing steps include normalization steps. In some embodiments, the preprocessing steps include dimensionality reduction steps.

[0258] 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.

[0259] In some embodiments, the machine learning model is trained to predict the presence or absence of marker-expressing cells in populations of cells containing T cells, the degree to which marker-expressing cells are present in populations of cells containing T cells, or the degree to which marker-expressing cells are present in T cells of populations of cells containing T cells. In some embodiments, the machine learning model is trained to predict the presence or absence of marker-expressing T cells in populations of cells containing T cells, the degree to which marker-expressing T cells are present in populations of cells containing T cells, or the degree to which marker-expressing T cells are present in T cells of populations of cells containing T cells.

[0260] In some embodiments, the machine learning model is trained to predict the presence or absence of activated T cells in populations of cells containing T cells, the degree to which activated T cells are present in populations of cells containing T cells, or the degree to which activated T cells are present in T cells of populations of cells containing T cells. In some embodiments, the machine learning model is trained to predict the presence or absence of T cells expressing the marker indicative of T cell activation in populations of cells containing T cells, the degree to which T cells expressing the marker indicative of T cell activation are present in populations of cells containing T cells, or the degree to which T cells expressing the marker indicative of T cell activation are present in T cells of populations of cells containing T cells.

[0261] In some embodiments, the machine learning model is trained to predict based on input measures for the one or more cellular features. In some embodiments, the machine learning model is trained to predict based on population-level statistics for the one or more cellular features. In some embodiments, the machine learning model is trained to predict based on features that include the one or more cellular features as well as other features, for instance features of populations of cells that are not based on the characteristics of individual cells.

[0262] In some embodiments, the population-level output measure is determined by providing the plurality of population-level statistics as input to the machine learning model. In some embodiments, the machine learning model is trained to predict population-level output measures of expression of the marker based on population-level statistics of the plurality of cellular features.

[0263] 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.

[0264] 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. [0265] In some embodiments, the machine learning model is a regularized model. In some embodiments, the machine learning model is a lasso-regularized model. In some embodiments, the machine learning model is a ridge-regularized model. In some embodiments, the machine learning model is an elastic-net-regularized model.

[0266] 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 includes decision trees. In some embodiments, the machine learning model includes boosted decision trees. In some embodiments, the machine learning model is a random forest model. In some embodiments, the machine learning model is an ensemble model that includes 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 measure for a first marker, such as any described herein, and a second machine learning model of the ensemble model is trained to predict a populationlevel output measure for a second marker, such as any described herein, that is different from the first marker. In some embodiments, a first machine learning model of the ensemble model is trained to predict a population-level output measure for a marker, such as any described herein, and a second machine learning model of the ensemble model is trained to predict a population-level output measure for the same marker.

[0267] In some embodiments, the machine learning model is a multioutput model. In some embodiments, the multioutput model is a deep learning model, such as any described herein, for instance any of the convolutional neural networks described in Section I-B. In some embodiments, the multioutput model is trained to predict a population-level output measure for each of a plurality of different markers, which can be independently selected from any of the markers described herein.

1. Reference Datasets

[0268] In some embodiments, the threshold values are determined based on a dataset of reference input measures. In some embodiments, the machine learning model is trained using a dataset of reference input measures. In some embodiments, for each of a first plurality of reference populations of cells containing T cells, the dataset of reference input measures includes one or more reference input measures for each of the one or more cellular features.

[0269] In some embodiments, the threshold values are determined based on a dataset of reference population-level statistics. In some embodiments, the machine learning model is trained using a dataset of reference population-level statistics. In some embodiments, for each of a first plurality of reference populations of cells containing T cells, the dataset of reference population-level statistics includes one or more reference population-level statistics for each of the one or more cellular features. In some embodiments, each reference population-level statistic is of one or more reference input measures for a cellular feature of the plurality of cellular features.

[0270] In some embodiments, the provided methods are for training a machine learning model that predicts the cell phenotype, e.g., activation state, of a population of T cells. In some embodiments, the provide methods involve training a machine learning model using a dataset of reference input measures. In some embodiments, for each of a first plurality of reference populations of cells containing T cells, the dataset of reference input measures includes one or more reference input measures for each of one or more cellular features. Exemplary machine learning models are described in Section I-D.

[0271] In some embodiments, the provided methods involve training a machine learning model using a dataset of reference population-level statistics. In some embodiments, for each of a first plurality of reference populations of cells containing T cells, the dataset of reference population-level statistics includes one or more reference population-level statistics for each of the one or more cellular features. In some embodiments, each reference population-level statistic is of one or more reference input measures for a cellular feature of one or more cellular features. Exemplary machine learning models are described in Section I-D.

[0272] In some embodiments, the threshold values are determined based on a dataset of reference population-level output measures. In some embodiments, the machine learning model is trained using a dataset of reference population-level output measures. In some embodiments, the provided methods involve training the machine learning model using a dataset of reference population-level output measures. In some embodiments, for each of a second plurality of reference populations of cells, the dataset of reference population-level output measures includes a reference population-level output measure for the reference population of cells.

[0273] In some embodiments, the machine learning model is a multiple instance learning model. In some embodiments, the machine learning model is a deep multiple instance learning model.

[0274] In some embodiments, the provided methods involve (a) training a convolutional neural network using a dataset of reference holographic information, wherein for each of a first plurality of reference populations of cells containing T cells, the dataset of reference holographic information contains holographic information obtained for individual cells of the reference population of cells; (b) determining, from the convolutional neural network, one or more reference input measures for each cellular feature of a plurality of cellular features derived from the holographic information, wherein the plurality of cellular features are cellular features extracted by the convolutional neural network, and each reference input measure is from an individual cell of a reference population of cells; (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 cellular features; and (d) training a machine learning model using the dataset of reference population-level statistics and a dataset of reference population-level output measures, wherein for each of a second plurality of reference populations of cells, the dataset of reference population-level output measures contains a reference population-level output measure of expression of a marker for the reference population of cells. In some embodiments, the marker is a marker expressed by activated T cells.

[0275] In some embodiments, the machine learning model is trained to predict populationlevel output measures of expression of the marker based on population-level statistics of the plurality of cellular features.

[0276] In some embodiments, the first and second pluralities of reference populations of cells are the same. In some embodiments, the first and second pluralities of reference populations of cells are different from one another. In some embodiments, the dataset of reference population-level statistics and the dataset of reference population-level output measures are time-matched, e.g., include population-level statistics and population-level output measures that were measured at the same times or within one hour, 30 minutes, 20 minutes, 10 minutes, or 5 minutes of one another, whether from the same or different reference populations of cells.

[0277] In some embodiments, for each of the first plurality of reference populations of cells, each reference input measure is derived from holographic information obtained for the reference population of cells, e.g., for individual cells of the population of cells. In some embodiments, each reference input measure is from an individual cell of the reference population of cells.

[0278] In some embodiments, the first and/or second plurality of reference populations of cells includes at least 5, 10, 15, 20, 25, 30, 35, 40, 45, 50, 55, 60, 65, 70, 75, 80, 85, 90, 95, or 100 reference populations of cells. In some embodiments, each of the first and/or second plurality of reference populations of cells is any of the populations of cells described in Section II. In some embodiments, the provided methods involve performing any of the cell processing steps described in Section II for each of the first and/or second plurality of reference populations of cells. In some embodiments, the first and/or second plurality of reference populations of cells are each 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 populations of cells are T cells.

[0279] In some embodiments, the first and/or second plurality of reference populations of cells are from reference cultures of cells being cultured in vitro or ex vivo. In some embodiments, the first and second pluralities of reference populations of cells include reference populations from the same reference culture of cells. In some embodiments, the provided methods involve culturing the reference cultures of cells. Exemplary methods and conditions for culture are described in Section II-D. In some embodiments, the in vitro or ex vivo culture of the reference cultures of cells is performed under the same or similar conditions as the in vitro or ex vivo culture of the culture of cells.

[0280] In some embodiments, the first and/or second plurality of reference populations of cells are incubated under T cell stimulating conditions. In some embodiments, the incubation is prior to when the holographic information for the first plurality of reference populations of cells is obtained. In some embodiments, the provided methods involve incubating the first and/or second plurality of reference populations of cells 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 populations of cells is performed under the same or similar conditions as the incubation of the population of cells.

[0281] In some embodiments, the first and/or second plurality of reference populations of cells 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 populations of cells to express a recombinant protein. Exemplary engineering methods are described in Section II-C. In some embodiments, the engineering of the first and/or second plurality of reference populations of cells is performed under the same or similar conditions as the engineering of the population of cells. In some embodiments, the first and/or second plurality of reference populations of cells are engineered to express the same recombinant protein that the population of cells is engineered to express.

[0282] In some embodiments, the holographic information for the first plurality of reference populations of cells is obtained according to any of the methods described in Section I-A. In some embodiments, the provided methods involve obtaining the holographic information for the first plurality of reference populations of cells. In some embodiments, the holographic information for the first plurality of reference populations of cells is obtained during the in vitro or ex vivo culture of the reference cultures of cells. In some embodiments, multiple reference populations of cells of the first plurality of reference populations of cells are from the same reference culture of cells. In some embodiments, the holographic information for the multiple reference populations of cells is obtained at multiple timepoints during the in vitro or ex vivo culture of the reference culture of cells.

[0283] In some embodiments, the holographic information for the first plurality of reference populations of cells is obtained according to the method used to obtain the holographic information for the population of cells. In some embodiments, the holographic information for the first plurality of reference populations of cells is obtained by DHM.

[0284] In some embodiments, the reference input measures are determined according to any of the methods described in Section I-B. In some embodiments, the provided methods involve determining the reference input measures. In some embodiments, the dataset of reference input measures or reference population-level statistics includes only the one or more cellular features for which input measures or population-level statistics are obtained for the population of cells. In some embodiments, the dataset of reference input measures or reference population-level statistics includes other features in addition to the one or more cellular features. The other features can be determined from holographic or non-holographic information obtained for the first plurality of reference populations of cells. The other features can be based or not based on the characteristics of individual cells of the first plurality of reference populations of cells.

[0285] In some embodiments, the one or more reference population-level statistics for each of the one or more cellular features are any of the population-level statistics described in Section I-C. In some embodiments, the provided methods involve determining the one or more reference population-level statistics for each of the one or more cellular features. In some embodiments, for a cellular feature of the one or more cellular features, the one or more reference population-level statistics is the same as the one or more population-level statistics for the population of cells. In some embodiments, the one or more reference population-level statistics for a cellular feature of the one or more cellular features include one or more quantiles of the one or more reference input measures of the cellular 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 the one or more reference input measures of the cellular feature. In some embodiments, the one or more quantiles include the 0.01, 0.1, 0.5, 0.9, and 0.99 quantiles of the one or more reference input measures of the cellular feature. In some embodiments, the one or more reference population-level statistics for each cellular feature of the one or more cellular features include the 0.01, 0.1, 0.5, 0.9, and 0.99 quantiles of the one or more reference input measures of the cellular feature.

[0286] In some embodiments, the one or more reference population-level statistics for a cellular feature of the one or more cellular features are determined by applying a pooling filter to the one or more reference input measures for the cellular feature. In some embodiments, each reference population-level statistic is determined by applying a distribution-based pooling filter to the one or more reference input measures for a cellular feature of the plurality of cellular features. 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 maximum pooling filter. In some embodiments, the pooling filter is a distribution-based pooling filter. [0287] In some embodiments, for each of the second plurality of reference populations of cells, the reference population-level output measure indicates the cell phenotype of the reference population of cells. In some embodiments, the reference population-level output measure indicates the presence or absence of marker-expressing cells in the reference population of cells. In some embodiments, the reference population-level output measure indicates the degree to which marker-expressing cells are present in the reference population of cells. In some embodiments, the reference population-level output measure indicates the degree to which marker-expressing cells are present in T cells of the reference population of cells.

[0288] In some embodiments, the reference population-level output measure 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 in the reference population of cells. In some embodiments, the reference population-level output measure 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 in T cells of the reference population of cells. In some embodiments, the reference population-level output measure is the percentage of marker-expressing in T cells of the reference population of cells.

[0289] In some embodiments, for each of the second plurality of reference populations of cells, the reference population-level output measure indicates the activation state of the reference population of cells. In some embodiments, the reference population-level output measure indicates the presence or absence of activated T cells in the reference population of cells. In some embodiments, the reference population-level output measure indicates the degree to which activated T cells are present in the reference population of cells. In some embodiments, the reference population-level output measure indicates the degree to which activated T cells are present in T cells of the reference population of cells.

[0290] In some embodiments, the reference population-level output measure 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 in the reference population of cells. In some embodiments, the reference population-level output measure 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 in T cells of the reference population of cells. In some embodiments, the reference population- level output measure is the percentage of activated T cells in T cells of the reference population of cells.

[0291] In some embodiments, the reference population-level output measure is of expression of the marker indicative of T cell activation by T cells of the reference population of cells. In some embodiments, the reference population-level output measure indicates the presence of T cells expressing the marker in the reference population of cells. In some embodiments, the reference population-level output measure indicates the degree to which T cells expressing the marker are present in the reference population of cells. In some embodiments, the reference population-level output measure indicates the degree to which T cells expressing the marker are present in T cells of the reference population of cells.

[0292] In some embodiments, the reference population-level output measure is the number of cells, the percentage of cells, the proportion of cells, or the density of cells of the reference population of cells that express the marker. In some embodiments, the reference populationlevel output measure is the number of cells, the percentage of cells, the proportion of cells, or the density of cells in T cells of the reference population of cells that express the marker. In some embodiments, the reference population-level output measure is the percentage of cells in T cells of the reference population of cells that express the marker.

[0293] In some embodiments, the dataset of reference population-level output measures are of expression of the marker during or after the in vitro or ex vivo culture of the reference cultures of cells. In some embodiments, multiple reference populations of cells of the second plurality of reference populations of cells are from the same reference culture of cells. In some embodiments, the reference population-level output measures for the multiple reference populations of cells is of expression of the marker at multiple timepoints during the in vitro or ex vivo culture of the reference culture of cells. In some embodiments, for a reference population of cells, the reference population-level output measure is of expression of the marker at the same timepoint that the holographic information for a reference population of cells of the first plurality of reference populations of cells is obtained.

[0294] In some embodiments, the dataset of reference population-level output measures is determined using fluorescence imaging of cells of the reference cultures of cells. In some embodiments, the provided methods involve determining the dataset of reference population- level output measures using fluorescence imaging of the second plurality of reference populations of cells. In some embodiments, the fluorescence imaging is by flow cytometry.

[0295] In some embodiments, one or more preparation and/or non-affinity-based cell separation steps are carried out prior to labelling 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 particular reagents. In some embodiments, the cells are separated based on one or more properties, such as density, adherent properties, size, sensitivity, and/or resistance to particular components. In some embodiments, the methods include density -based cell separation methods.

[0296] In some embodiments, the cells are labelled with one or more fluorescent markers (e.g., one or more fluorophores) that produce a fluorescent signal that can be measured by a flow cytometer. In some embodiments, the cells are labelled by incubating the cells with one or more staining reagents, in which each staining reagent contains a fluorescent signal or marker (e.g., fluorophore). The staining reagent can be any reagent for characterization, selection, or isolation of a particular cell type or subtype of cells. In some embodiments, the staining reagent contains an immunoaffinity -based reagent, such as an antibody.

[0297] In some embodiments, the staining reagent stains the cells based on the cells’ expression or expression level of the marker, e.g., the marker indicative of T cell activation. In some embodiments, the staining reagent contains one or more fluorescent markers that may be attached, such as by chemical conjugation, to a binding agent that is able to bind, such as specifically bind, to the marker. In some embodiments, the binding agent is a protein. In some embodiments, the binding agent is an antibody or an antigen-binding fragment. The fluorescent marker may be conjugated to the binding agent, e.g., antibody, by any method known in the art.

[0298] As is well known in the art, an “antibody” is an immunoglobulin (Ig) molecule capable of specific binding to a target, such as a carbohydrate, polynucleotide, lipid, or polypeptide, through at least one epitope recognition site, located in the variable region of the Ig molecule. As used herein, the term encompasses not only intact polyclonal or monoclonal antibodies, but also 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 with an antigen-binding fragment of the required specificity, chimeric antibodies, nanobodies, and any other modified configuration of the immunoglobulin molecule that comprises an antigen-binding site or fragment (epitope recognition site) of the required specificity. Minibodies comprising an scFv joined to a CH3 domain are also included herein.

[0299] 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 more frequently, more rapidly, with greater duration, and/or with greater affinity with a particular marker than it does with alternative markers. An antibody specifically binds or preferentially binds to a target if it binds with greater affinity, avidity, more readily, and/or with greater duration than it binds to other substances. It is also understood that specific binding or preferential binding does not necessarily require (although it can include) exclusive binding. Methods to determine such specific or preferential binding are also well known in the art, e.g., an immunoassay.

[0300] Antibodies for flow cytometry can be selected based on the marker being detected in the plurality of reference populations of cells, e.g., based on the marker that is indicative of T cell activation. In some embodiments, the marker is CD137 (4-1BB). Exemplary CD137- binding agents, such as anti-CD137 antibodies, for flow cytometry include clone 4B4-1 available from Miltenyi Biotec and clone 17B5 available from ThermoFisher Scientific.

[0301] 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.

[0302] In some embodiments, the binding agent, such as antibody, is conjugated to a fluorescent marker, such as a fluorophore. For instance, the cells may be incubated with one or more fluorescently labeled antibodies. In some embodiments, any fluorescent marker or fluorophore suitable for use with 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 colorants (e.g., 4 ', 6-diamidino-2-phenylindole (DAPI), SYT016, and propidium iodide (PI)), cell membrane stain (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-Diochlorofluorescien, ATTO 488, Chromeo 488, Dylight 488, HiEyte 488, Alexa Fluor 532, Alexa Fluor 555, ATTO 550, BODIPY TMR-X, CF 555, Chromeo 546, Cy3, TMR, TRITC, Dy547, Dy548, Dy549, HiEyte 555, Dylight 550, BODIPY 564, Alexa Fluor 568, Alexa Fluor 594, Rhodamine, Texas Red, Alexa Fluor 610, Alexa Fluor 633, Dylight 633, Alexa Fluor 647, APC, ATTO 655, CF633, CF640R, Chromeo642, Cy5, Dylight 650, Alexa Fluor 680, IRDye 680, Alexa Fluor 700 (AF700), Cy5.5, ICG, Alexa Fluor 750, Dylight 755, IRDye 750, Cy7, PE-Cy7, Cy7.5, Alexa Fluor 790, Dylight 800, IRDye 800, BV421, BV510, BV570, BV605, BV650, BV711, BV750, BV785, Qdot® 525, Qdot® 565, Qdot® 605, Qdot® 655, Qdot® 705, and Qdot® 800.

[0303] In some embodiments, the cell staining for flow cytometry involves incubation with the staining reagent, which in some embodiments is followed by washing steps and separation of cells bound to the staining reagent from those cells not bound to the staining reagent. In some embodiments, a volume of cells is mixed with an amount of a desired staining reagent and incubated under conditions for staining of the cells. In some embodiments, the staining is carried out at a temperature between 0°C and 25°C, such as at or about 4°C. In some embodiments, the staining is carried out for greater than 5 minutes, typically greater than 15 minutes. In some embodiments, the staining is carried out for between 15 minutes and 6 hours, such as between 30 minutes and 2 hours. In some embodiments, the staining is carried out, for example, at or about 15 minutes, 30 minutes, 1 hour, 1.5 hours, 2 hours, 2.5 hours, 3 hours, or any value between any of the foregoing. In some embodiments, one or more wash steps are carried out prior to introducing the cells into the flow cytometer for analysis. In some embodiments, the stained cells are introduced into a flow cytometer.

[0304] In some embodiments, the cells are prepared by suspending single cells at a density of 1 x 10⁶ to 1 x 10⁷ cells/mE in order to allow the cells to pass through the flow cytometer for reading. In some embodiments, this concentration of cells is called the fluid sheath. In some embodiments, the fluid sheath influences the rate of flow sorting, which typically progresses at around 2,000-20,000 cells per second. The cell sample's fluid sheath can be made of a phosphate buffered saline solution, but other solutions are available, as will be known and understood by those skilled in the art.

II. T CELL POPULATIONS

[0305] Exemplary populations of cells containing T cells and cell processing steps involving the populations of cells are described in this section. In some embodiments, each of the reference populations of cells referred to in Section I-D-l is individually selected from any of the described populations of cells 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 populations of cells. In some embodiments, the same or similar cell processing steps are performed for each of the plurality of reference populations of cells. In some embodiments, the same or similar cell processing steps are performed for each of the plurality of reference population of cells and for the population of cells.

[0306] In some embodiments, at least 2%, 4%, 6%, 8%, 10%, 20%, 30%, 40%, 50%, 60%, 70%, 80%, or 90% of the population of cells is T cells. In some embodiments, the population of cells 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 population of cells is T cells. Exemplary methods for the selection of T cells from, for example, mixed populations of cells are described in Section LA.

[0307] In some embodiments, the population of cells are incubated under T cell stimulating conditions. In some embodiments, the provided methods involve incubating the population of cells under T cell stimulating conditions. Exemplary T cell stimulating conditions are described in Section ILB.

[0308] In some embodiments, the population of cells are genetically engineered to express a recombinant protein. In some embodiments, the provided methods involve genetically engineering the population of cells to express a recombinant protein. Exemplary engineering methods are described in Section ILC.

[0309] In some embodiments, the population of cells is from a culture of cells being cultured in vitro or ex vivo. In some embodiments, the provided methods involve culturing the culture of cells. Exemplary conditions for in vitro or ex vivo culture are described in Section ILD. [0310] In some embodiments, the holographic information for the individual cells of the population of cells is obtained during the in vitro or ex vivo culture of the culture of cells. In some embodiments, the cell phenotype, e.g., activation state, of the population of cells during the in vitro or ex vivo culture is determined. In some embodiments, the population-level output measure indicates the cell phenotype, e.g., activation state, of the population of cells during the in vitro or ex vivo culture. In some embodiments, expression of the marker, e.g., the marker indicative of T cell activation, by the population of cells during the in vitro or ex vivo culture is determined. In some embodiments, the population-level output measure is of expression of the marker during the in vitro or ex vivo culture of the culture of cells.

[0311] In some embodiments, the culture of cells is incubated under T cell stimulating conditions. In some embodiments, the provided methods involve incubating the culture of cells under T cell stimulating conditions. Exemplary T cell stimulating conditions are described in Section II-B.

[0312] In some embodiments, the incubation under T cell stimulating conditions is prior to when the holographic information for the individual cells of the population of cells is obtained. In some embodiments, the incubation under T cell stimulating conditions is subsequent to when the holographic information for the individual cells of the population of cells is obtained. In some embodiments, the incubation under T cell stimulating conditions is prior to and subsequent to when the holographic information for the individual cells of the population of cells is obtained.

[0313] Sections II-A to II-C describe exemplary methods and reagents for selecting, stimulating, and culturing populations of cells containing T cells, respectively. 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 steps of stimulating T cells. In some embodiments, the provided methods involve steps of culturing T cells. In some embodiments, the provided methods involve steps of stimulating and culturing T cells.

[0314] In some embodiments, a step of selecting T cells is performed prior to steps of stimulating or culturing T cells. In some embodiments, a step of selecting T cells is performed subsequent to steps of stimulating or culturing T cells. [0315] In some embodiments, the incubation under T cell stimulating conditions is prior to the in vitro or ex vivo culture. In some embodiments, at least a portion of the in vitro or ex vivo culture is under T cell stimulating conditions. In some embodiments, all of the in vitro or ex vivo culture is under T cell stimulating conditions. In some embodiments, the incubation under T cell stimulating conditions is subsequent to the in vitro or ex vivo culture.

[0316] In some embodiments, some or all of the steps of selecting, stimulating, and culturing T cells are carried out at a temperature that is above room temperature, for instance at a physiological temperature. In some embodiments, the temperature is between about 35 °C and about 39°C, such as at or about 37°C.

[0317] In some embodiments, any number of the steps of the provided methods are performed in a closed system. In some embodiments, any number of the steps of the provided method are automated.

A. Selection

[0318] In some embodiments, the population of cells is from a biological sample. In some embodiments, the population of cells are primary cells. In some embodiments, the primary cells are primary cells from a human subject. The biological samples can include tissue, fluid, and other samples taken directly from the subject. The biological sample can be a sample obtained directly from a biological source or a sample that is processed. Exemplary biological samples include body fluids, such as blood, plasma, serum, cerebrospinal fluid, synovial fluid, urine, and sweat; tissue; and organ samples, including processed samples derived therefrom. Exemplary biological samples also include whole blood, peripheral blood mononuclear cells (PBMCs), leukocytes, bone marrow, thymus, tissue biopsy, tumor, leukemia, lymphoma, lymph node, gut associated lymphoid tissue, mucosa associated lymphoid tissue, spleen, other lymphoid tissues, liver, lung, stomach, intestine, colon, kidney, pancreas, breast, bone, prostate, cervix, testes, ovaries, tonsil, or other organ, or cells derived therefrom.

[0319] In some examples, cells from the circulating blood of the subject are obtained by, e.g., apheresis or leukapheresis. The biological samples can contain lymphocytes, including T cells, monocytes, granulocytes, B cells, other nucleated white blood cells, red blood cells, and/or platelets, and in some aspects contain cells other than red blood cells and platelets. [0320] 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 white blood cell 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 leukaphresis product.

[0321] In some embodiments, the cells obtained from the subject are washed to, e.g., remove the plasma fraction and to place the cells in an appropriate buffer or media for subsequent processing steps. In some embodiments, the cells are washed with phosphate buffered saline (PBS). In some embodiments, the wash solution lacks calcium, magnesium, and/or many or all divalent cations. In some aspects, a washing step is accomplished using a semi- automated “flow-through” centrifuge (for example, the Cobe 2991 cell processor, Baxter) according to the manufacturer's instructions. In some aspects, a washing step is accomplished by tangential flow filtration (TFF) according to the manufacturer's instructions. In some embodiments, the cells are resuspended in a variety of biocompatible buffers after washing, such as Ca²⁺/Mg²⁺ free PBS. In some embodiments, components of a blood cell biological sample are removed, and the cells are directly resuspended in culture media. In some embodiments, the biological sample (e.g., an apheresis product or a leukapheresis product) is washed in order to remove one or more anti-coagulants, such as heparin, added during apheresis or leukapheresis.

[0322] In some embodiments, the selection of cells includes one or more preparation and/or non-affinity based cell separation steps. In some examples, 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 particular reagents. In some examples, the cells are separated based on one or more properties, such as density, adherent properties, size, sensitivity, and/or resistance to particular components. In some embodiments, the methods involve density-based cell separation methods, such as the preparation of white blood cells from peripheral blood by lysing the red blood cells and centrifugation through a Percoll or Ficoll gradient.

[0323] In some embodiments, a cryopreserved and/or cryoprotected apheresis product or leukapheresis product is thawed. In some embodiments, the thawed cell composition is subjected to dilution (e.g., with a serum-free medium) and/or wash (e.g., with a serum-free medium), which in some cases can remove or reduce unwanted or undesired components. In some cases, the dilution and/or wash removes or reduces the presence of a cryoprotectant, e.g. DMSO, contained in the thawed sample, which otherwise may negatively impact cellular viability, yield, or recovery upon extended room temperature exposure. In some embodiments, the dilution and/or wash allows media exchange of a thawed cryopreserved product into a serum-free medium, e.g., one described in US-20210207080.

[0324] Exemplary methods and reagents for the selection of T cells for producing the populations of cells containing T cells, including by positive or negative selection, 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.

[0325] In some embodiments, at least a portion of the selection step includes incubation of cells with a selection reagent. In some embodiments, the incubation is with a selection reagent or reagents, e.g., as part of selection methods which may be performed using one or more selection reagents for selection of one or more different cell types based on the expression or presence in or on the cell of one or more specific molecules, such as surface markers, e.g., surface proteins, intracellular markers, or nucleic acid. 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 affinity- or immunoaffinity-based separation. For example, the selection in some aspects includes incubation with a reagent or reagents for separation of cells and cell populations based on the cells’ expression or expression level of one or more markers, typically cell surface markers, for example, by incubation with a binding partner, e.g., antibody, that specifically binds to such markers, followed generally by washing steps and separation of cells having bound the binding partner, from those cells having not bound to the binding partner.

[0326] In some aspects of such processes, a volume of cells is mixed with an amount of a desired affinity-based selection reagent. The immunoaffinity-based selection can be carried out using any system or method that results in a favorable energetic interaction between the cells being separated and the molecule specifically binding to the marker on the cell, e.g., the binding partner on a solid surface, e.g., particle. In some embodiments, methods are carried out using particles such as beads, e.g., magnetic beads, that are coated with a selection agent (e.g., antibody) specific to the marker of the cells. The particles (e.g., beads) can be incubated or mixed with cells in a container, such as a tube or bag, while shaking or mixing, with a constant cell density-to-particle (e.g., bead) ratio to aid in promoting energetically favored interactions.

[0327] In some embodiments, the total duration of the incubation with the selection reagent is from or from about 5 minutes to 6 hours, such as 30 minutes to 3 hours, for example, at least or about at least 30 minutes, 60 minutes, 120 minutes, or 180 minutes.

[0328] In some embodiments, after the incubation and/or mixing of the cells and selection reagent and/or reagents, the incubated cells are subjected to a separation to select for cells based on the presence or absence of the particular selection reagent or reagents. In some embodiments, after incubation with the selection reagents, incubated cells, including cells in which the selection reagent has bound are transferred into a system for immunoaffinity-based separation of the cells. In some embodiments, the system for immunoaffinity-based separation is or contains a magnetic separation column.

[0329] Such separation steps can be based on positive selection, in which the cells having bound the selection reagents, e.g., antibody, are retained for further use, and/or negative selection, in which the cells having not bound to the selection reagent, e.g., antibody, are retained. In some examples, both fractions are retained for further use. In some embodiments, the process steps further include negative and/or positive selection of the incubated and cells, such as using a system or apparatus that can perform an affinity-based selection. In some embodiments, isolation is carried out by enrichment for a particular cell population by positive selection, or depletion of a particular cell population, by negative selection. In some embodiments, positive or negative selection is accomplished by incubating cells with one or more antibodies or other binding agents that specifically bind to one or more surface markers expressed or expressed (marker+) at a relatively higher level (marker¹¹¹⁸¹¹) on the positively or negatively selected cells, respectively.

[0330] The separation need not result in 100 % enrichment or removal of a particular cell population or cells expressing a particular selection marker. For example, positive selection of or enrichment for cells of a particular type, such as those expressing a selection marker, refers to increasing the number or percentage of such cells, but need not result in a complete absence of cells not expressing the selection marker. Likewise, negative selection, removal, or depletion of cells of a particular type, such as those expressing a selection marker, refers to decreasing the number or percentage of such cells, but need not result in a complete removal of all such cells.

[0331] In some examples, multiple rounds of separation steps are carried out, where 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 examples, a single separation step can deplete cells expressing multiple markers simultaneously, such as by incubating cells with a plurality of antibodies or other binding partners, each specific for a marker targeted for negative selection. Likewise, multiple cell types can simultaneously be positively selected by incubating cells with a plurality of antibodies or other binding partners expressed on the various cell types. In certain embodiments, separation steps are repeated and or performed more than once, where the positively or negatively selected fraction from one step is subjected to the same separation step, such as a repeated positive or negative selection. In some examples, 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 the negatively selected cells from the negatively selected fraction. In certain embodiments, one or more separation steps are performed two times, three times, four times, five times, six times, seven times, eight times, nine times, ten times, or more than ten times. In certain embodiments, the one or more selection steps are performed and/or repeated between one and ten times, between one and five times, or between three and five times.

[0332] In some embodiments, T cells are separated from a PBMC, apheresis, or leukapheresis sample by negative selection of markers expressed on non-T cells, such as B cells, monocytes, or other white blood cells, such as CD 14. In some aspects, a CD3+ selection step is used to generate a population enriched in 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 previous, concurrent, or subsequent selection for another marker (e.g., CD4 and/or CD8) in order to generate a population enriched in CD3+ T cells. In some embodiments, the population of cells is enriched in CD3+ T cells, which population is not subjected to a further selection, e.g., CD4+ and/or CD8+ selection, before being subjected to a step of, for example, stimulating the population of cells. In certain embodiments, the population enriched in CD3+ T cells has a ratio of between 1:10 and 10:1, between 1:5 and 5:1, between 4:1 and 1:4, between 1:3 and 3:1, between 2:1 and 1:2, between 1.5:1 and 1:1.5, between 1.25:1 and 1:1.25, between 1.2:1 and 1:1.2, between 1.1:1 and 1:1.1, or about 1:1 or 1:1 CD4+ T cells to CD8+ T cells.

[0333] In some embodiments, T cells are separated from a PBMC, apheresis, or leukapheresis sample by negative selection of markers expressed on non-T cells, such as B cells, monocytes, or other white blood cells, such as CD 14. In some aspects, a CD4+ or CD8+ selection step is used to separate CD4+ helper and CD8+ cytotoxic T cells. Such CD4+ and CD8+ populations can be further sorted into sub-populations by positive or negative selection for markers expressed or expressed to a relatively higher degree on one or more naive-like, memory, and/or effector T cell subpopulations.

[0334] In particular embodiments, a biological sample, e.g., a PBMC, apheresis, or leukapheresis sample, is subjected to selection of CD4+ T cells, where both the 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, where both the negative and positive fractions are retained. In certain embodiments, CD4+ T cells are selected from the negative fraction.

[0335] In some aspects, a CD8-based positive selection step is used to generate a population enriched in CD8+ T cells from a starting sample, such as a PBMC, apheresis, or leukaphresis 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 in order to generate a population enriched in CD4+ T cells. In some embodiments, cells from the population enriched in CD8+ T cells and cells from the population enriched in CD4+ T cells are mixed, combined, and/or pooled to generate a population containing CD4+ T cells and CD8+ T cells.

[0336] In some aspects, a CD4-based positive selection step is used to generate a population enriched in CD4+ T cells from a starting sample, such as a PBMC, apheresis, or leukaphresis 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 the 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 population enriched in CD8+ T cells. In some embodiments, cells from the population enriched in CD4+ T cells and cells from the population enriched in CD8+ T cells are mixed, combined, and/or pooled to generate a population containing CD8+ T cells and CD4+ T cells.

[0337] In certain embodiments, the population enriched in CD4+ T cells and the population enriched in CD8+ T cells are pooled, mixed, and/or combined prior to stimulating cells, e.g., culturing the cells under stimulating conditions such as described in Section I-B. In certain embodiments, the pooled, mixed, and/or combined cells or populations have a ratio of between 1:10 and 10:1, between 1:5 and 5:1, between 4:1 and 1:4, between 1:3 and 3:1, between 2:1 and 1:2, between 1.5:1 and 1:1.5, between 1.25:1 and 1:1.25, between 1.2:1 and 1:1.2, between 1.1:1 and 1:1.1, or about 1:1, or 1:1 CD4+ T cells to CD8+ T cells. In certain embodiments, the cells or populations are pooled, mixed, and/or combined in order to have a ratio of or of about 1:1 CD4+ T cells to CD8+ T cells in the pooled, mixed, and/or combined cell composition.

[0338] 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 in CD4+ T cells and a population enriched in CD8+ T cells, respectively. In some embodiments, the capacities of the CD4-specific selection agent and the CD8-specific selection agent are the same or substantially the same, for example, a unit volume or unit weight of the selection agents (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 the CD8-specific selection agent with the same or substantially the same capacity. For example, the volumes or weights of the CD4-specific selection agent and the CD8-specific selection agent can be at a ratio of about 5:1, 4:1, 3:1, 2:1, or 1:5:1.

[0339] In some aspects, the incubated sample or population of cells to be separated is incubated with a selection reagent containing small, magnetizable, or magnetically responsive material, such as magnetically responsive particles or microparticles, such as paramagnetic beads (e.g., Dynabeads or MACS® beads). The magnetically responsive material, e.g., particle, generally is directly or indirectly attached to a binding partner, e.g., an antibody, that specifically binds to a molecule, e.g., surface marker, present on the population of cells that it is desired to separate, e.g., that it is desired to negatively or positively select.

[0340] In some embodiments, the magnetic particle, e.g., bead, contains 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, e.g., those described in US-4452773 and in EP-452342. Colloidal sized particles, such as those described in US-4795698 and US-5200084 may also be used.

[0341] The incubation can be carried out under conditions whereby the antibodies or other binding partners, such as secondary antibodies or other reagents, which specifically bind to such antibodies or other binding partners, which are attached to the magnetic particle, e.g., bead, specifically bind to cell surface molecules if present on cells within the sample.

[0342] In certain embodiments, the magnetically responsive particles are coated in primary antibodies or other binding partners, secondary antibodies, lectins, enzymes, or streptavidin. In certain embodiments, the magnetic particles are attached to cells via a coating of primary antibodies specific for one or more markers. In certain embodiments, the cells, rather than the beads, are labeled with a primary antibody or binding partner, and then cell-type specific secondary antibody- or other binding partner (e.g., streptavidin)-coated magnetic particles, are added. In certain embodiments, streptavidin-coated magnetic particles are used in conjunction with biotinylated primary or secondary antibodies.

[0343] In some aspects, separation is achieved in a procedure in which the sample is placed in a magnetic field, and those cells having magnetically responsive or magnetizable particles attached thereto will be attracted to the magnet and separated from the unlabeled cells. For positive selection, cells that are attracted to the magnet are retained; for negative selection, cells that are not attracted (unlabeled cells) are retained. In some aspects, a combination of positive and negative selection is performed during the same selection step, where the positive and negative fractions are retained and further processed or subject to further separation steps.

[0344] In some embodiments, the affinity-based selection is via magnetic-activated cell sorting (MACS) (Miltenyi Biotech, Auburn, CA). Magnetic Activated Cell Sorting (MACS), e.g., CliniMACS systems, are capable of high-purity selection of cells having magnetized particles attached thereto. In certain embodiments, MACS operates in a mode wherein the non-target and target species are sequentially eluted after the application of the external magnetic field. That is, the cells attached to magnetized particles are held in place while the unattached species are eluted. Then, after this first elution step is completed, the species that were trapped in the magnetic field and were prevented from being eluted are freed in some manner such that they can be eluted and recovered. In certain embodiments, the non-target cells are labelled and depleted from the heterogeneous population of cells.

[0345] In some embodiments, the separation and/or isolation steps are carried out using magnetic beads in which immunoaffinity reagents are reversibly bound, such as via a peptide ligand interaction with a streptavidin mutein as described in US-20170037369. Exemplary of such magnetic beads are Streptamers®. In some embodiments, the separation and/or steps is carried out using magnetic beads, such as those commercially available from Miltenyi Biotec.

[0346] In some embodiments, the T cells are isolated, selected, or enriched by chromatographic isolation, such as by column chromatography including affinity chromatography or gel permeation chromatography. Such methods may be described as (traceless) cell affinity chromatography technology (CATCH) and may include any of the methods or techniques described in US- 10228312 and US-20170037369. Generally, a chromatographic method is a fluid chromatography, typically a liquid chromatography. In some aspects, the chromatography can be carried out in a flow through mode in which a fluid sample containing the cells is applied, for example, by gravity flow or by a pump on one end of a column containing the chromatography matrix and in which the fluid sample exists the column at the other end of the column. In addition, the chromatography can be carried out in an “up and down” mode in which a fluid sample containing the cells to be isolated is applied, for example, by a pipette on one end of a column containing the chromatography matrix packed within a pipette tip and in which the fluid sample enters and exists the chromatography matrix/pipette tip at the other end of the column. Alternatively, the chromatography can also be carried out in a batch mode in which the chromatography material (e.g., stationary phase) is incubated with the sample that contains the cells, for example, under shaking, rotating, or repeated contacting and removal of the fluid sample, for example, by means of a pipette. [0347] In some embodiments, the selection agent is contained in a chromatography column, e.g., bound directly or indirectly to the chromatography matrix (e.g., stationary phase). In some embodiments, the selection agent is present on the chromatography matrix (e.g., stationary phase) at the time the sample is added to the column. In some embodiments, the selection agent is capable of being bound indirectly to the chromatography matrix (e.g., stationary phase) through a reagent, e.g., selection reagent. In some embodiments, the selection reagent is bound covalently or non-covalently 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 covalent bonds. In some aspects, the selection reagent is reversibly immobilized on the chromatography matrix (e.g., stationary phase) non-covalently.

[0348] In some embodiments, the selection agent may be present, for example bound directly to (e.g., covalently or non-covalently) or indirectly via a selection reagent, on the chromatography matrix (e.g., stationary phase) at the time the sample is added to the chromatography column (e.g., stationary phase). Thus, upon addition of the sample, T cells can be bound by the selection agent and immobilized on the chromatography matrix (e.g., stationary phase) of the column. Alternatively, in some embodiments, the selection agent can be added to the sample. In this way, the selection agent binds to the T cells in the sample, and the sample can then be added to a chromatography matrix (e.g., stationary phase) containing the selection reagent, where the selection agent, already bound to the T cells, binds to the selection reagent, thereby immobilizing the target cells on the chromatography matrix (e.g., stationary phase).

[0349] In some embodiments, one or both of a first and/or second selection can employ a plurality of affinity chromatography matrices and/or antibodies, whereby the plurality of matrices and/or antibodies are serially connected. In some embodiments, the affinity chromatography matrix or matrices employed in selection adsorb or are capable of selecting or enriching at least about 50 x 10⁶ cells/mL, 100 x 10⁶ cells/mL, 200 x 10⁶ cells/mL or 400 x 10⁶ cells/mL. In some embodiments, the adsorption capacity can be modulated based on the diameter and/or length of the matrix. In some embodiments, the culture-initiating ratio of the selected or enriched population is achieved by choosing a sufficient amount of matrix and/or at a sufficient relative amount to achieve the culture-initiating ratio assuming based on, for example, the adsorption capacity of the matrix or matrices for selecting cells.

[0350] In some aspects, the chromatography matrix/stationary phase is a non-magnetic material or non-magnetisable material. Such material may include derivatized silica or a crosslinked gel. A crosslinked gel (which is typically manufactured in a bead form) may be based on a natural polymer, such as a crosslinked polysaccharide. Suitable examples include agarose gels or a gel of crosslinked dextrans. A crosslinked gel may also be based on a synthetic polymer, e.g., on a polymer class that does not occur in nature. Usually, such a synthetic polymer on which a stationary phase for cell separation is based is a polymer that has polar monomer units and which is therefore in itself polar.

[0351] Illustrative examples of suitable synthetic polymers are polyacrylamides, a styrene- divinylbenzene gel, and a copolymer of an acrylate and a diol or of an acrylamide and a diol. An illustrative example is a polymethacrylate gel, commercially available as a Fractogel®. A further example is a copolymer of ethylene glycol and methacrylate, commercially available as a Toyopearl®. In some embodiments, a stationary phase may also include natural and synthetic polymer components, such as a composite matrix or a composite or a co-polymer of a polysaccharide and agarose, e.g., a polyacrylamide/agarose composite, or of a polysaccharide and N,N'-methylenebisacrylamide. An illustrative example of a copolymer of a dextran and N,N'-methylenebisacryl-iamide is the above-mentioned Sephacryl® series of material. A derivatized silica may include silica particles that are coupled to a synthetic or to a natural polymer. Examples of such embodiments include polysaccharide grafted silica, polyvinyl-ipyrrolidone grafted silica, polyethylene oxide grafted silica, poly(2- hydroxyethylaspartamide) silica, and poly(N-isopropylacrylamide) grafted silica.

[0352] In some embodiments, the isolation and/or selection results in one or more populations of enriched T cells, e.g., CD3+ T cells, CD4+ T cells, and/or CD8+ T cells. In some embodiments, two or more separate population of enriched T cells are isolated, selected, enriched, or obtained from a single biological sample. In some embodiments, separate populations are isolated, selected, enriched, and/or obtained from separate biological samples collected, taken, and/or obtained from the same subject.

[0353] In certain embodiments, the isolation and/or selection results in one or more populations of enriched T cells that includes 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 at or at about 100% CD3+ T cells. In particular embodiment, the population of enriched T cells consists essentially of CD3+ T cells.

[0354] In certain embodiments, the isolation and/or enrichment results in a populations of enriched CD4+ T cells that includes 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 at or at about 100% CD4+ T cells. In certain embodiments, the population of CD4+ T cells includes 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 population of enriched T cells consists essentially of CD4+ T cells.

[0355] In certain embodiments, the isolation and/or enrichment results in a populations of enriched CD8+ T cells that includes 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 at or at about 100% CD8+ T cells. In certain embodiments, the population of CD8+ T cells 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 of or substantially free of CD4+ T cells. In some embodiments, the population of enriched T cells consists essentially of CD8+ T cells.

[0356] In some embodiments, the selection marker may be CD4 and the selection agent specifically binds CD4. In some aspects, the selection agent that specifically binds CD4 may be selected from the group consisting of an anti-CD4 antibody, a divalent antibody fragment of an anti-CD4 antibody, a monovalent antibody fragment of an anti-CD4 antibody, and a proteinaceous CD4 binding molecule with antibody-like binding properties. In some embodiments, an anti-CD4 antibody, divalent antibody fragment, or monovalent antibody fragment (e.g. anti-CD4 Fab fragment) can be derived from antibody 13B8.2 or a functionally active mutant of 13B8.2 that retains specific binding for CD4. For example, exemplary mutants of antibody 13B8.2 or ml3B8.2 are described in US-7482000, US-20140295458, US-10228312, and Bes et al., J Biol Chem (2003) 278:14265-14273. The mutant Fab fragment termed "m!3B8.2" carries the variable domain of the CD4 binding murine antibody 13B8.2 and a constant domain containing constant human CHI domain of type gamma for the heavy chain and the constant human light chain domain of type kappa, as described in US-7482000. In some embodiments, the anti-CD4 antibody, divalent antibody fragment, or monovalent antibody fragment, e.g., a mutant of antibody 13B8.2, contains the amino acid replacement H91A in the variable light chain, the amino acid replacement Y92A in the variable light chain, the amino acid replacement H35A in the variable heavy chain, and/or the amino acid replacement R53A in the variable heavy chain, each by Kabat numbering. In some aspects, compared to variable domains of the 13B8.2 Fab fragment in ml3B8.2, the His residue at position 91 of the light chain (position 93 in SEQ ID NO: 1) is mutated to Ala, and the Arg residue at position 53 of the heavy chain (position 55 in SEQ ID NO: 2) is mutated to Ala. In some embodiments, the reagent that is reversibly bound to the anti-CD4 antibody or fragment thereof is commercially available or derived from a reagent that is commercially available (e.g., catalog No. 6-8000-206, 6-8000-205, or 6-8002-100; IBA GmbH, Gottingen, Germany). In some embodiments, the selection agent comprises an anti-CD4 Fab fragment. In some embodiments, the anti-CD4 Fab fragment contains 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 contains the CDRs of the variable heavy chain having the sequence set forth in SEQ ID NO: 2 and the CDRs of the variable light chain having the sequence set forth in SEQ ID NO: 1.

[0357] In some embodiments, the selection marker may be CD 8 and the selection agent specifically binds CD8. In some aspects, the selection agent that specifically binds CD8 may be selected from the group consisting of an anti-CD8 antibody, a divalent antibody fragment of an anti-CD8 antibody, a monovalent antibody fragment of an anti-CD8 antibody, and a proteinaceous CD8 binding molecule with antibody-like binding properties. In some embodiments, the anti-CD8 antibody, divalent antibody fragment, or monovalent antibody fragment (e.g., anti-CD8 Fab fragment) can be derived from antibody OKT8 (e.g., ATCC CRL-8014) or a functionally active mutant thereof that retains specific binding for CD8. In some embodiments, the reagent that is reversibly bound to anti-CD8 or a fragment thereof is commercially available or derived from a reagent that is commercially available (e.g., catalog No. 6-8003 or 6-8000-201; IBA GmbH, Gottingen, Germany). In some embodiments, the selection agent contains an anti-CD8 Fab fragment. In some embodiments, the anti-CD8 Fab fragment contains 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 contains the CDRs of the variable heavy chain having the sequence set forth in SEQ ID NO: 3 and the CDRs of the variable light chain having the sequence set forth by SEQ ID NO: 4.

[0358] In some embodiments, the selection marker may be CD3 and the selection agent specifically binds CD3. In some aspects, the selection agent that specifically binds CD3 may be selected from the group consisting of an anti-CD3 antibody, a divalent antibody fragment of an anti-CD3 antibody, a monovalent antibody fragment of an anti-CD3 antibody, and a proteinaceous CD3 binding molecule with antibody-like binding properties. In some embodiments, the anti-CD3 antibody, divalent antibody fragment, or monovalent antibody fragment (e.g., anti-CD3 Fab fragment) can be derived from 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 for CD3. In some embodiments, the reagent that is reversibly bound to the anti-CD3 antibody or a fragment thereof is commercially available or derived from a reagent that is commercially available (e.g., catalog No. 6-8000-201 or 6-8001-100; IBA GmbH, Gottingen, 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 the CDRs of the variable heavy chain having the sequence set forth in SEQ ID NO: 5 and the CDRs of the variable light chain having the sequence set forth in SEQ ID NO: 6.

[0359] In any of the above examples, the divalent antibody fragment may be an F(ab’)2 fragment or a divalent single-chain Fv fragment. In some embodiments, the monovalent antibody fragment may 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 with antibody-like binding properties may be an aptamer, a mutein based on a polypeptide of the lipocalin family, a glubody, a protein based on the ankyrin scaffold, a protein based on the crystalline scaffold, an adnectin, or an avimer. B. Stimulation

[0360] In some embodiments, the population of cells are incubated under T cell stimulating conditions. In some embodiments, the provided methods involve incubating the population of cells under T cell stimulating conditions.

[0361] Exemplary methods and stimulatory reagents for the 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.

[0362] In some embodiments, the conditions for T cell stimulation can include one or more of particular media, 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 agents, designed to stimulate the T cells.

[0363] In some embodiments, the incubation is in basal media. In some embodiments, the basal media is serum-free. In some embodiments, the basal media is free of serum derived from human. In some embodiments, the basal media 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 media can also help maintain pH and osmolality. A wide variety of commercially available basal media are well known to those skilled in the art and include Dulbeccos' Modified Eagles Medium (DMEM), Roswell Park Memorial Institute Medium (RPMI), Iscove modified Dulbeccos' medium, and Hams medium. In some embodiments, the basal media is Iscove's Modified Dulbecco's Medium, RPMI- 1640, or a-MEM.

[0364] In some embodiments, the basal media is a balanced salt solution (e.g., PBS, DPBS, HBSS, or EBSS). In some embodiments, the basal media is selected from Dulbecco's Modified Eagle's Medium (DMEM), Minimal Essential Medium (MEM), Basal Medium Eagle (BME), F-10, F-12, RPMI 1640, Glasgow's Minimal Essential Medium (GMEM), alpha Minimal Essential Medium (alpha MEM), Iscove's Modified Dulbecco's Medium, and M199. In some embodiments, the base media is a complex medium (e.g., RPMI-1640 or IMDM). In some embodiments, the base media is OpTmizer™ CTS™ T-Cell Expansion Basal Medium (ThermoFisher). [0365] In some embodiments, the basal media is supplemented with additional additives. In some embodiments, the basal media is not supplemented with any additional additives. Additives to cell culture media include nutrients, sugars, e.g., glucose, amino acids, vitamins, and additives such as ATP and NADH.

[0366] In some embodiments, the incubation is in a serum- free medium, such as one described in US-20210207080.

[0367] In some embodiments, the incubation is 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 to and/or are capable of binding to receptors that are expressed by and/or are endogenous to T cells. In particular embodiments, the one or more recombinant cytokines include 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 in the presence of IL-2, IL- 15, and IL-7.

[0368] In certain embodiments, the amount or concentration of the one or more recombinant cytokines are measured and/or quantified with International Units (IU). International units may be used to quantify vitamins, hormones, cytokines, vaccines, blood products, and similar biologically active substances. In some embodiments, IU are or include units of measure of the potency of biological preparations by comparison to an international reference standard of a specific weight and strength, e.g., WHO 1st International Standard for Human IL-2, 86/504. International Units are the only recognized and standardized method to report biological activity units that are published and are derived from an international collaborative research effort. In particular embodiments, the IU for population, sample, or source of a cytokine may be obtained through product comparison testing with an analogous WHO standard product. For example, in some embodiments, the lU/mg of a population, sample, or source of human recombinant IL-2, IL-7, or IL-15 is compared to the WHO standard IL-2 product (NIBSC code: 86/500), the WHO standard IL- 17 product (NIBSC code: 90/530), and the WHO standard IL- 15 product (NIBSC code: 95/554), respectively. [0369] In some embodiments, the biological activity in lU/mg is equivalent to (ED50 in ng/mL)-l x 10⁶. In particular embodiments, the ED50 of recombinant human IL-2 or IL-15 is equivalent to the concentration required for the half-maximal stimulation of cell proliferation (XTT cleavage) with CTLL-2 cells. In certain embodiments, the ED50 of recombinant human IL-7 is equivalent to the concentration required for the half-maximal stimulation for proliferation of PHA-activated human peripheral blood lymphocytes. Details relating to assays and calculations of IU for IL-2 are discussed in Wadhwa et al., Journal of Immunological Methods (2013) 379 (1-2): 1-7 and Gearing and Thorpe, Journal of Immunological Methods (1988) 114 (l-2):3-9. Details relating to assays and calculations of IU for IL-15 are discussed in Soman et al., Journal of Immunological Methods (2009) 348 (1- 2):83-94.

[0370] 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 between 1 lU/mL and 1,000 lU/mL, between 10 lU/mL and 50 lU/mL, between 50 lU/mL and 100 lU/mL, between 100 lU/mL and 200 lU/mL, between 100 lU/mL and 500 lU/mL, between 250 lU/mL and 500 lU/mL, or between 500 lU/mL and 1,000 lU/mL.

[0371] 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 between 1 lU/mL and 500 lU/mL, between 10 lU/mL and 250 lU/mL, between 50 lU/mL and 200 lU/mL, between 50 lU/mL and 150 lU/mL, between 75 lU/mL and 125 lU/mL, between 100 lU/mL and 200 lU/mL, or between 10 lU/mL and 100 lU/mL. In particular embodiments, cells are stimulated or subjected to stimulation in the presence of recombinant IL-2 at a concentration at or at about 50 lU/mL, 60 lU/mL, 70 lU/mL, 80 lU/mL, 90 lU/mL, 100 lU/mL, 110 lU/mL, 120 lU/mL, 130 lU/mL, 140 lU/mL, 150 lU/mL, 160 lU/mL, 170 lU/mL, 180 lU/mL, 190 lU/mL, or 100 lU/mL. In some embodiments, the cells are stimulated or subjected to stimulation in the presence of or of about 100 lU/mL of recombinant IL-2, e.g., human recombinant IL-2.

[0372] 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 between 100 lU/mL and 2,000 lU/mL, between 500 lU/mL and 1,000 lU/mL, between 100 lU/mL and 500 lU/mL, between 500 lU/mL and 750 lU/mL, between 750 lU/mL and 1,000 lU/mL, or between 550 lU/mL and 650 lU/mL. In particular embodiments, the cells are stimulated or subjected to stimulation in the presence of IL-7 at a concentration at or at about 50 IU/mL,100 lU/mL, 150 lU/mL, 200 lU/mL, 250 lU/mL, 300 lU/mL, 350 lU/mL, 400 lU/mL, 450 lU/mL, 500 lU/mL, 550 lU/mL, 600 lU/mL, 650 lU/mL, 700 lU/mL, 750 lU/mL, 800 lU/mL, 750 lU/mL, 750 lU/mL, 750 lU/mL, or 1,000 lU/mL. In particular embodiments, the cells are stimulated or subjected to stimulation in the presence of or of about 600 lU/mL of recombinant IL-7, e.g., human recombinant IL-7.

[0373] 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 between 1 lU/mL and 500 lU/mL, between 10 lU/mL and 250 lU/mL, between 50 lU/mL and 200 lU/mL, between 50 lU/mL and 150 lU/mL, between 75 lU/mL and 125 lU/mL, between 100 lU/mL and 200 lU/mL, or between 10 lU/mL and 100 lU/mL. In particular embodiments, cells are stimulated or subjected to stimulation in the presence of recombinant IL- 15 at a concentration at or at about 50 lU/mL, 60 lU/mL, 70 lU/mL, 80 lU/mL, 90 lU/mL, 100 lU/mL, 110 lU/mL, 120 lU/mL, 130 lU/mL, 140 lU/mL, 150 lU/mL, 160 lU/mL, 170 lU/mL, 180 lU/mL, 190 lU/mL, or 200 lU/mL. In some embodiments, the cells are stimulated or subjected to stimulation in the presence of or of about 100 lU/mL of recombinant IL- 15, e.g., human recombinant IL- 15.

[0374] In some embodiments, the incubation is in the absence of recombinant cytokines.

[0375] In certain embodiments, the stimulation is performed under static conditions, such as conditions that do not involve centrifugation, shaking, rotating, rocking, or perfusion, e.g., continuous or semi-continuous perfusion of the media. In some embodiments, either prior to or shortly after, e.g., within 5, 15, or 30 minutes, of the initiation of the stimulation, 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 particular embodiments, the incubator is set at, at about, or at least 16°C, 24°C, or 35°C. In some embodiments, the incubator is set at 37°C, at about 37°C, or at 37°C ±2°C, ±1°C, ±0.5°C, or ±0.1 °C. In particular embodiments, the stimulation under static condition is performed in a cell culture bag placed in an incubator. In some embodiments, the culture bag is composed of a single-web polyolefin gas permeable film which enables monocytes, if present, to adhere to the bag surface. [0376] In some embodiments, the T cell stimulating conditions involve incubation in the presence of T cell stimulatory agents. In some embodiments, the T cell stimulatory agents bind to molecules expressed on the surface of T cells. In some embodiments, a T cell stimulatory agent of the T cell stimulatory agents induces a primary activation signal in T cells. In some embodiments, a T cell stimulatory agent of the T cell stimulatory agents induces a costimulatory signal in T cells. In some embodiments, the T cell stimulatory agents induce a primary activation signal and a costimulatory signal in T cells.

[0377] In some embodiments, the T cell stimulatory agent that induces a primary activation signal binds to a member of a TCR/CD3 complex in T cells. In some embodiments, the T cell stimulatory agent binds to CD3. In some embodiments, the T cell stimulatory agent is an anti- CD3 antibody, a divalent antibody fragment of an anti-CD3 antibody, a monovalent antibody fragment of an anti-CD3 antibody, or a proteinaceous CD3 binding molecule with antibodylike binding properties. In some embodiments, the T cell stimulatory agent is an anti-CD3 antibody or antibody fragment. In some embodiments, the anti-CD3 antibody, divalent antibody fragment of an anti-CD3 antibody, or monovalent antibody fragment of an anti-CD3 antibody (e.g., anti-CD3 Fab fragment) is derived from 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 for CD 3. In some embodiments, the T cell stimulatory agent is an anti-CD3 Fab. In some embodiments, the anti-CD3 Fab 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 contains the CDRs of the variable heavy chain having the sequence set forth in SEQ ID NO: 5 and the CDRs of the variable light chain having the sequence set forth in SEQ ID NO: 6.

[0378] In some embodiments, the T cell stimulatory agent that induces a costimulatory signal binds to a costimulatory molecule in T cells. In some embodiments, the costimulatory molecule is CD28, CD90 (Thy-1), CD95 (Apo-/Fas), CD137 (4-1BB), CD154 (CD40L), ICOS, LAT, CD27, 0X40, or HVEM.

[0379] In some embodiments, the costimulatory molecule is CD28. In some embodiments, the T cell stimulatory agent is an anti-CD28 antibody, a divalent 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 stimulatory agent is an anti-CD28 antibody or antibody fragment. In some embodiments, the anti-CD28 antibody, divalent antibody fragment of an anti-CD28 antibody, or monovalent antibody fragment of an anti-CD28 antibody (e.g., 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 NO: 7 and 8, respectively. In some embodiments, the T cell stimulatory agent 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 the CDRs of the variable heavy chain having the sequence set forth in SEQ ID NO: 7 and the CDRs of the variable light chain having the sequence set forth in SEQ ID NO: 8.

[0380] In some embodiments, the costimulatory molecule is CD90. In some embodiments, the T cell stimulatory agent is an anti-CD90 antibody, a divalent antibody fragment of an anti-CD90 antibody, a monovalent antibody fragment of an anti-CD90 antibody, or a proteinaceous CD90 binding molecule with antibody-like binding properties. In some embodiments, the T cell stimulatory agent is an anti-CD90 antibody or antibody fragment. In some embodiments, the T cell stimulatory agent is an anti-CD90 Fab. In some embodiments, the anti-CD90 antibody, divalent antibody fragment of an anti-CD90 antibody, or monovalent antibody fragment of an anti-CD90 antibody (e.g., anti-CD90 Fab fragment) is derived from the anti-CD90 antibody G7 (Biolegend, cat. no. 105201).

[0381] In some embodiments, the costimulatory molecule is CD95. In some embodiments, the T cell stimulatory agent is an anti-CD95 antibody, a divalent 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 stimulatory agent is an anti-CD28 antibody or antibody fragment. In some embodiments, the T cell stimulatory agent is an anti-CD95 Fab. In some embodiments, the anti-CD95 antibody, divalent antibody fragment of an anti-CD95 antibody, or monovalent antibody fragment of an anti-CD95 antibody (e.g., anti-CD95 Fab fragment) is derived from monoclonal mouse anti-human CD95 CH11 (Upstate Biotechnology, Lake Placid, NY), anti- CD95 mAh 7C11, or anti-APO-1, such as described in Paulsen et al., Cell Death & Differentiation (2011) 18(4):619-631.

[0382] In some embodiments, the costimulatory molecule is CD137. In some embodiments, the T cell stimulatory agent is an anti-CD137 antibody, a divalent antibody fragment of an anti-CD137 antibody, a monovalent antibody fragment of an anti-CD137 antibody, or a proteinaceous CD 137 binding molecule with antibody-like binding properties. In some embodiments, the T cell stimulatory agent is an anti-CD137 antibody or antibody fragment. In some embodiments, the T cell stimulatory agent is an anti-CD137 Fab. In some embodiments, the anti-CD137 antibody, divalent antibody fragment of an anti-CD137 antibody, or monovalent antibody fragment of an anti-CD137 antibody (e.g., anti-CD137 Fab fragment) is derived from LOB 12, IgG2a, or LOB 12.3, IgGl as described in Taraban et al., Eur J Immunol. (2002) 32(12):3617-27. See also, e.g., US-6569997, US-6303121, and Mittler et al., Immunol Res. (2004) 29(1-3): 197-208.

[0383] In some embodiments, the costimulatory molecule is CD40. In some embodiments, the T cell stimulatory agent is an anti-CD40 antibody, a divalent 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 stimulatory agent is an anti-CD40 antibody or antibody fragment. In some embodiments, the T cell stimulatory agent is an anti-CD40 Fab.

[0384] In some embodiments, the costimulatory molecule is CD40L. In some embodiments, the T cell stimulatory agent is an anti-CD40L antibody, a divalent antibody fragment of an anti-CD40L antibody, a monovalent antibody fragment of an anti-CD40L antibody, or a proteinaceous CD40L binding molecule with antibody-like binding properties. In some embodiments, the T cell stimulatory agent is an anti-CD40L antibody or antibody fragment. In some embodiments, the T cell stimulatory agent is an anti-CD40L Fab. In some embodiments, the anti-CD40L antibody, divalent antibody fragment of an anti-CD40L antibody, or monovalent antibody fragment of an anti-CD40L antibody (e.g., anti-CD40L Fab fragment) is derived from Hu5C8, as described in Blair et al., JEM (2000) 19(4):651- 660. See also, e.g., US-7563445, US20010026932, US7547438, and US-7172759. [0385] In some embodiments, the costimulatory molecule is ICOS. In some embodiments, the T cell stimulatory agent is an anti-ICOS antibody, a divalent antibody fragment of an anti- ICOS antibody, a monovalent antibody fragment of an anti-ICOS antibody, or a proteinaceous ICOS binding molecule with antibody-like binding properties. In some embodiments, the T cell stimulatory agent is an anti-ICOS antibody or antibody fragment. In some embodiments, the T cell stimulatory agent is an anti-ICO Fab. In some embodiments, the anti-ICOS antibody, divalent antibody fragment of an anti-ICOS antibody, or monovalent antibody fragment of an anti-ICOS antibody (e.g., anti-ICOS Fab fragment) is derived from any of the antibodies described in US-20080279851 and Deng et al., Hybrid Hybridomics (2004) 23(3): 176-82.

[0386] In some embodiments, the costimulatory molecule is Linker for Activation of T cells (LAT). In some embodiments, the T cell stimulatory agent is an anti-LAT antibody, a divalent 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 stimulatory agent is an anti-LAT antibody or antibody fragment. In some embodiments, the T cell stimulatory agent is an anti-LAT Fab.

[0387] In some embodiments, the costimulatory molecule is CD27. In some embodiments, the T cell stimulatory agent is an anti-CD27 antibody, a divalent antibody fragment of an anti-CD27 antibody, a monovalent antibody fragment of an anti-CD27 antibody, or a proteinaceous CD27 binding molecule with antibody-like binding properties. In some embodiments, the T cell stimulatory agent is an anti-CD27 antibody or antibody fragment. In some embodiments, the T cell stimulatory agent is an anti-CD27 Fab. In some embodiments, the anti-CD27 antibody, divalent antibody fragment of an anti-CD27 antibody, or monovalent antibody fragment of an anti-CD27 antibody (e.g., anti-CD27 Fab fragment) is derived from any of the antibodies described in US-8481029.

[0388] In some embodiments, the costimulatory molecule is 0X40. In some embodiments, the T cell stimulatory agent is an anti-OX40 antibody, a divalent antibody fragment of an anti-OX40 antibody, a monovalent antibody fragment of an anti-OX40 antibody, or a proteinaceous 0X40 binding molecule with 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, divalent antibody fragment of an anti-OX40 antibody, or monovalent antibody fragment of an anti-OX40 antibody (e.g., anti-OX40 Fab fragment) is derived from any of the antibodies described in US-9475880 and Melero et al., Clin Cancer Res. (2013) 19(5): 1044-53.

[0389] In some embodiments, the costimulatory molecule is HVEM. In some embodiments, the T cell stimulatory agent is an anti-HVEM antibody, a divalent antibody fragment of an anti-HVEM antibody, a monovalent antibody fragment of an anti-HVEM antibody, or a proteinaceous HVEM binding molecule with antibody-like binding properties. In some embodiments, the T cell stimulatory agent is an anti-HVEM antibody or antibody fragment. In some embodiments, the T cell stimulatory agent is an anti-HVEM Fab. In some embodiments, the anti-HVEM antibody, divalent antibody fragment of an anti-HVEM antibody, or monovalent antibody fragment of an anti-HVEM antibody (e.g., anti-HVEM Fab fragment) is derived from any of the antibodies described in WO-2006054961, US-8188232, and Park et al., Cancer Immunol Immunother. (2012) 61(2):203-14.

[0390] In some embodiments, the T cell stimulatory agents are immobilized on a solid support. In some embodiments, the solid support is a bead. In some embodiments, the bead is biocompatible, e.g., composed of a material that is suitable for biological use. In some embodiments, the beads are non-toxic to T cells.

[0391] In some embodiments, the bead has a diameter of greater than about 0.001 pm, greater than about 0.01 pm, greater than about 0.1 pm, greater than about 1.0 pm, greater than about 10 pm, greater than about 50 pm, greater than about 100 pm, or greater than about 1000 pm and no more than about 1500 pm. In some embodiments, the bead has a diameter of about 1.0 pm to about 500 pm, about 1.0 pm to about 150 pm, about 1.0 pm to about 30 pm, about 1.0 pm to about 10 pm, about 1.0 pm to about 5.0 pm, about 2.0 pm to about 5.0 pm, or about 3.0 pm to about 5.0 pm. In some embodiments, the bead has a diameter of about 3 pm to about 5 pm. In some embodiments, the bead has a diameter of at least or at least about or about 0.001 pm, 0.01 pm, 0.1pm, 0.5pm, 1.0 pm, 1.5 pm, 2.0 pm, 2.5 pm, 3.0 pm, 3.5 pm, 4.0 pm, 4.5 pm, 5.0 pm, 5.5 pm, 6.0 pm, 6.5 pm, 7.0 pm, 7.5 pm, 8.0 pm, 8.5 pm, 9.0 pm, 9.5 pm, 10 pm, 12 pm, 14 pm, 16 pm, 18 pm, or 20 pm. In certain embodiments, the bead has a diameter of or about 4.5 pm. In certain embodiments, the bead has a diameter of or about 2.8 pm. [0392] In some embodiments, the bead has a density of 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 5g/cm³. In some embodiments, the bead has a density of between about 0.001 g/cm³ and about 100 g/cm³, about 0.01 g/cm³ and about 50 g/cm³, about 0.1 g/cm³ and about 10 g/cm³, about 0.1 g/cm³ and about .5 g/cm³, about 0.5 g/cm³ and about 1 g/cm³, about 0.5 g/cm³ and about 1.5 g/cm³, about 1 g/cm³ and about 1.5 g/cm³, about 1 g/cm³ and about 2 g/cm³, or about 1 g/cm³ and about 5 g/cm³. In some embodiments, the bead has 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 bead has a density of about 1.6 g/cm³. In particular embodiments, the bead has a density of about 1.5 g/cm³. In certain embodiments, the bead has a density of about 1.3 g/cm³.

[0393] In some embodiments, a plurality of the beads has a uniform density. In some embodiments, a uniform density has a density standard deviation of less than 10%, less than 5%, or less than 1% of the mean bead density.

[0394] In some embodiments, the bead reacts in a magnetic field. In some embodiments, the bead is a magnetic bead. In some embodiments, the bead is paramagnetic. In particular embodiments, the bead is superparamagnetic. In certain embodiments, the bead does not display any magnetic properties unless it is exposed to a magnetic field.

[0395] In particular embodiments, the bead contains a magnetic core. In particular embodiments, the bead contains a paramagnetic core. In particular 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 combinations thereof. In certain embodiments, the core contains metal oxides (e.g., iron oxides), ferrites (e.g., manganese ferrites, cobalt ferrites, and nickel ferrites), hematite, and/or metal alloys (e.g., CoTaZn). In some embodiments, the core contains one or more of a ferrite, a metal, a metal alloy, an 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 ( FC O4), maghemite (yFc2O3), and greigite (Fe3S4). In some embodiments, the core contains an iron oxide (e.g., FCTCE).

[0396] In some embodiments, the bead contains at least one material at or near the bead surface that can be coupled, linked, or conjugated to an agent. In some embodiments, the T cell stimulatory agents are immobilized on the bead via this material. In some embodiments, the bead is surface functionalized, e.g., has functional groups that are capable of forming a covalent bond with a binding molecule, e.g., a polynucleotide or a polypeptide. In particular embodiments, the bead has surface-exposed carboxyl, amino, hydroxyl, tosyl, epoxy, and/or chloromethyl groups. In particular embodiments, the bead has surface-exposed agarose and/or sepharose. In some embodiments, the bead has surface-exposed protein A, protein G, or biotin.

[0397] In certain embodiments, the bead contains a magnetic, paramagnetic, and/or superparamagnetic core that is covered by a surface functionalized coat or coating. In some embodiments, the coat can contain a material that can include a polymer, a polysaccharide, a silica, a fatty acid, a protein, a carbon, agarose, sepharose, or a combination thereof. In some embodiments, the polymer can be a polyethylene glycol, poly (lactic-co-glycolic acid), polyglutaraldehyde, polyurethane, polystyrene, or a polyvinyl alcohol. In certain embodiments, the outer coat or coating comprises polystyrene. In particular embodiments, the outer coating is surface functionalized.

[0398] In some embodiments, the population of cells containing T cells is incubated with a stimulatory reagent at a ratio of beads to cells at or at about 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. In particular embodiments, the ratio of beads to cells 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 ratio of beads to cells is about 1:1 or is 1:1.

[0399] In some embodiments, the T cell stimulatory agents are not immobilized on a solid support, e.g., not immobilized on a bead. In some embodiments, the T cell stimulatory agents are part of a stimulatory reagent that is in soluble form. Exemplary T cell stimulatory reagents in soluble form are described in US2021/0032297 (see also Poltorak et al., Scientific Reports (2020)). [0400] In some embodiments, the T cell stimulatory agents are immobilized on an oligomeric streptavidin mutein reagent. In some embodiments, the T cell stimulatory agents are reversibly immobilized on the oligomeric streptavidin mutein reagent. In some embodiments, the oligomeric streptavidin mutein reagent is an oligomer or polymer of a streptavidin mutein.

[0401] In some embodiments, the T cell stimulatory agents include binding partners that are reversibly bound to a streptavidin mutein molecule of the oligomeric streptavidin mutein reagent. In some embodiments, the binding partners are fused to the C-terminus of a heavy chain of the T cell stimulatory agents.

[0402] In some embodiments, the binding partners reversibly bind to a biotin-binding site of the streptavidin mutein. In some embodiments, the binding of the binding partners to the streptavidin mutein is disrupted by the presence of biotin. In some embodiments, the biotin is D-biotin.

[0403] 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 the sequence set forth in any of SEQ ID NO: 9-15. In some embodiments, the sequence of the streptavidin-binding peptide is set forth in any of SEQ ID NO: 9-15. In some embodiments, the streptavidin-binding peptide comprises the 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.

[0404] In some embodiments, the oligomeric streptavidin mutein reagent comprises between or between about 2,000 and 3,000 tetramers of the streptavidin mutein. In some embodiments, the oligomeric streptavidin mutein reagent comprises about 2,400 tetramers of the streptavidin mutein.

[0405] In some embodiments, the oligomeric streptavidin mutein reagent has a radius of between or between about 90 and 110 nm. In some embodiments, the oligomeric streptavidin mutein reagent has a radius of about 100 nm. In some embodiments, the radius is a hydrodynamic radius.

[0406] In some embodiments, individual molecules of the oligomeric streptavidin mutein reagent are crosslinked 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.

[0407] 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 the sequence set forth in any of SEQ ID NO: 9-15. In some embodiments, the sequence of the streptavidin-binding peptide is set forth in any of SEQ ID NO: 9-15.

[0408] 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 sequence of amino acids set forth in SEQ ID NO: 16.

[0409] In some embodiments, the streptavidin mutein comprises one or more mutations compared to a minimal streptavidin. In some embodiments, the minimal streptavidin begins N-terminally in the region of the amino acid positions 10 to 16 and terminates C-terminally in the region of the amino acid positions 133 to 142 compared to the sequence set forth in SEQ ID NO: 16. In some embodiments, the streptavidin mutein begins N-terminally in the region of the amino acid positions 10 to 16 and terminates C-terminally in the region of the amino acid positions 133 to 142 compared to the sequence set forth in SEQ ID NO: 16.

[0410] In some embodiments, the sequence of the minimal streptavidin is from position Alal3 to Serl39 of the sequence of amino acids set forth in SEQ ID NO: 16. In some embodiments, the minimal streptavidin has an N-terminal methionine residue instead of Alal3.

[0411] 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 sequence of amino acids set forth in SEQ ID NO: 17.

[0412] 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 sequence of amino acids set forth in SEQ ID NO: 18.

[0413] 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 sequence of amino acids 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 sequence of amino acids set forth in SEQ ID NO: 16.

[0414] In some embodiments, the streptavidin mutein further comprises the amino acid replacements Glut 17, Glyl20, and Try 121 at sequence positions corresponding to positions of the sequence of amino acids set forth in SEQ ID NO: 16.

[0415] In some embodiments, the streptavidin mutein comprises the sequence of amino acids set forth in any of SEQ ID NO: 21-28. In some embodiments, the streptavidin mutein comprises the sequence of amino acids set forth in SEQ ID NO: 21. In some embodiments, the sequence of the streptavidin mutein is set forth in SEQ ID NO: 21.

[0416] 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, for about, or for less than, 48 hours, 42 hours, 36 hours, 30 hours, 24 hours, 22 hours, 20 hours, 18 hours, 16 hours, or 12 hours. In some embodiments, the incubation is performed for between or between about 16 hours and 24 hours. In particular embodiments, the incubation is for between or between about 12 hours and 36 hours, 18 hours and 30 hours, or for or for about 24 hours. In some embodiments, the incubation is performed for, for about, or for less than 2 days. In some embodiments, the incubation is performed for, for about, or for less than one day.

C. Genetic Engineering

[0417] In some embodiments, the population of cells are genetically engineered to express a recombinant protein. In some embodiments, the provided methods involve genetically engineering the population of cells to express a recombinant protein. In some embodiments, a heterologous or recombinant polynucleotide encoding the recombinant protein is introduced into cells of the population of cells. Any method of introducing a heterologous or recombinant polynucleotide that would result in integration of the polynucleotide encoding the recombinant protein into the genome of a cell such as a T cell may be used, including viral and non-viral methods of genetic engineering. Introduction of the polynucleotides, e.g., heterologous or recombinant polynucleotides, encoding the recombinant protein into the cell may be carried out 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 those for transfer of heterologous polynucleotides encoding the recombinant proteins, including via viral, e.g., retroviral or lentiviral, transduction.

[0418] Exemplary methods for the genetic engineering of cells are described in US- 11400115, US-20190112576, US-20190136186, US-11274278, US-20210032297, US- 20200354677, US-20200384025, US-20210163893, and US-20220002669.

[0419] In some embodiments, the heterologous or recombinant polynucleotide encoding the recombinant protein is introduced using a non-viral method, 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 described in, e.g., US-10654928 and US-7446190. Transposon-based systems also are contemplated.

[0420] In particular 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., in Section II-B. In particular embodiments, the cells are genetically engineered, transformed, or transduced at, at about, or within 72 hours, 60 hours, 48 hours, 36 hours, 24 hours, or 12 hours, inclusive, from the initiation of the stimulation. In particular embodiments, the cells are genetically engineered, transformed, or transduced at, at about, or within 3 days, two days, or one day, inclusive, from the initiation of the stimulation. In certain embodiments, the cells are genetically engineered, transformed, or transduced between or between about 12 hours and 48 hours, 16 hours and 36 hours, or 18 hours and 30 hours after the initiation of the stimulation. In particular embodiments, the cells are genetically engineered, transformed, or transduced between or between about 18 hours and 30 hours after the initiation of the stimulation. In particular embodiments, the cells are genetically engineered, transformed, or transduced at or at about 16 hours, 18 hours, 20 hours, 22 hours, or 24 hours after the initiation of the stimulation.

[0421] In some embodiments, the engineering, e.g., transduction, is performed for between 24 and 48 hours, between 36 and 12 hours, between 18 and 30 hours, or for or for about 24 hours. In some embodiments, the engineering, e.g., transduction, is performed for or for about 24 hours, 48 hours, or 72 hours, or for or for about 1 day, 2 days, or 3 days, respectively. In particular embodiments, the engineering, e.g., transduction, is performed for or for about 24 hours ± 6 hours, 48 hours ± 6 hours, or 72 hours ± 6 hours. In particular embodiments, the engineering, e.g., transduction, is performed for or for about 72 hours, 72 ± 4 hours, or for or for about 3 days.

[0422] In certain embodiments, methods for genetic engineering are carried out by contacting or introducing one or more cells of a population with a nucleic acid molecule or polynucleotide encoding the recombinant protein. In certain embodiments, the nucleic acid molecule or polynucleotide is heterologous to the cells. In particular embodiments, the heterologous nucleic acid molecule or heterologous polynucleotide is not native to the cells. In certain embodiments, the heterologous nucleic acid molecule or heterologous polynucleotide encodes a protein, e.g., a recombinant protein, that is not natively expressed by the cell. In particular 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.

[0423] In some embodiments, the cells are engineered, 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 engineered in the presence of polycations, fibronectin or fibronectin-derived fragments or variants, and/or RetroNectin. In particular embodiments, the cells are engineered in the presence of a polycation that is polybrene, DEAE-dextran, protamine sulfate, poly-L-lysine, or a cationic liposome. In particular embodiments, the cells are engineered in the presence of protamine sulfate.

[0424] In some embodiments, the genetic engineering, e.g., transduction, is carried out in any of the media described in Section II-B. In some embodiments, the genetic engineering, e.g., transduction, is carried out in serum free media, e.g, any as described in US- 20210207080.

[0425] In some embodiments, the genetic engineering, e.g., transduction, is carried out in the presence of one or more recombinant cytokines. In some embodiments, the genetic engineering, e.g., transduction, is carried out in the presence of any of the recombinant cytokines described in Section II-B and at any of the concentrations described in Section II- B. [0426] In some embodiments, the cells are genetically engineered, transformed, or transduced in the presence of the same or similar media as was present during the stimulation. In some embodiments, the cells are genetically engineered, transformed, or transduced in media having the same cytokines as the media present during stimulation. In certain embodiments, the cells are genetically engineered, transformed, or transduced, in media having the same cytokines at the same concentrations as the media present during stimulation.

[0427] In some embodiments, genetically engineering the cells is or includes introducing the polynucleotide, e.g., the heterologous or recombinant polynucleotide, into the cells by transduction. In some embodiments, the cells are transduced or subjected to transduction with a viral vector. In particular embodiments, the cells are transduced or subjected to transduction with a viral vector. In some embodiments, the virus is a retroviral vector, such as a gammaretroviral vector or a lentiviral vector. Methods of lentiviral transduction are known. Exemplary methods are described in, e.g., 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.

[0428] In some embodiments, the transduction is carried out by contacting one or more cells of a population with a nucleic acid molecule encoding the recombinant protein. In some embodiments, the contacting can be effected with centrifugation, such as spinoculation (e.g., centrifugal inoculation). Such methods include any of those as described in US- 10428351. Exemplary centrifugal chambers include those produced and sold by Biosafe SA, including those for use with the Sepax® and Sepax® 2 system, including an A-200/F and A-200 centrifugal chambers and various kits for use with such systems. Exemplary chambers, systems, and processing instrumentation and cabinets are described, for example, in US- 6123655, US-6733433, US-20080171951, and US-US6733433. Exemplary kits for use with such systems include single-use kits sold by BioSafe SA under product names CS-430.1, CS- 490.1, CS-600.1 and CS-900.2.

[0429] In particular embodiments, genetic engineering, such as by transforming (e.g., transducing) the cells with a viral vector, further includes one or more steps of incubating the cells after the introducing or contacting of the cells with the viral vector. In some embodiments, cells, e.g., cells of the transformed cell population (also called “transformed cells”), are incubated subsequent to processes for genetically engineering, transforming, transducing, or transfecting the cells to introduce the viral vector into the cells.

[0430] In some embodiments, the cells, e.g. transformed cells, are incubated after the introducing of the heterologous or recombinant polynucleotide, e.g., viral vector particles, is carried out without further processing of the cells. In particular embodiments, prior to the incubating, the cells are washed, such as to remove or substantially remove exogenous or remaining polynucleotides encoding the heterologous or recombinant polynucleotide, e.g. viral vector particles, such as those remaining in the media after the genetic engineering process following the spinoculation.

[0431] In some embodiments, the further incubation is effected under conditions to result in integration of the viral vector into a host genome of one or more of the cells. For example, the further incubation provides time for the viral vector that may be bound to the T cells following transduction, e.g., via spinoculation, to integrate within the genome of the cell to delivery the gene of interest. In some aspects, the further incubation is carried out under conditions to allow the cells, e.g. transformed cells, to rest or recover in which the culture of the cells during the incubation supports or maintains the health of the cells. In particular embodiments, the cells are incubated under static conditions, such as conditions that do not involve centrifugation, shaking, rotating, rocking, or perfusion, e.g., continuous or semi- continuous perfusion of the media.

[0432] It is within the level of a skilled artisan to assess or determine if the incubation has resulted in integration of viral vector particles into a host genome, and hence to empirically determine the conditions for a further incubation. In some embodiments, integration of a viral vector into a host genome can be assessed by measuring the level of expression of a recombinant protein, such as a heterologous protein, encoded by a nucleic acid contained in the genome of the viral vector particle following incubation. A number of well-known methods for assessing expression level of recombinant molecules may be used, such as detection by affinity-based methods, e.g., immunoaffinity-based methods, e.g., in the context of cell surface proteins, such as by flow cytometry. In some examples, the expression is measured by detection of a transduction marker and/or reporter construct. In some embodiments, nucleic acid encoding a truncated surface protein is included within the vector and used as a marker of expression and/or enhancement thereof. [0433] In certain embodiments, the incubation is performed under static conditions, such as conditions that do not involve centrifugation, shaking, rotating, rocking, or perfusion, e.g., continuous or semi-continuous perfusion of the media. In some embodiments, either prior to or shortly after, e.g., within 5, 15, or 30 minutes, the initiation of the incubation, 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 into the container under closed or sterile conditions. In some embodiments, the container, e.g., the vial or bag, is then placed into an incubator for all or a portion of the incubation. In particular embodiments, incubator is set at, at about, or at least 16°C, 24°C, or 35°C. In some embodiments, the incubator is set at 37°C, at about 37°C, or at 37°C ±2°C, ±1°C, ±0.5°C, or ±0.1 °C.

[0434] In some embodiments, the incubation is performed in serum free media. In some embodiments, the serum free media is a defined and/or well-defined cell culture media. In certain embodiments, the serum free media is a controlled culture media that has been processed, e.g., filtered to remove inhibitors and/or growth factors. In some embodiments, the serum free media contains proteins. In certain embodiments, the serum-free media may contain serum albumin, hydrolysates, growth factors, hormones, carrier proteins, and/or attachment factors.

[0435] In some embodiments, the further incubation is carried out in any of the media described in Section II-B. In some embodiments, the further incubation is carried out in serum free media, e.g, any as described in US-20210207080.

[0436] In some embodiments, the further incubation is carried out in the presence of one or more recombinant cytokines. In some embodiments, the further incubation is carried out in the presence of any of the recombinant cytokines described in Section II-B and at any of the concentrations described in Section II-B.

[0437] In some embodiments, the further incubation is in the presence of the same or similar media as was present during the engineering. In some embodiments, the further incubation is in media having the same cytokines as the media present during engineering. In certain embodiments, the further incubation is in media having the same cytokines at the same concentrations as the media present during engineering. [0438] In particular embodiments, the cells are further incubated in the absence of cytokines. In particular embodiments, the cells are further incubated in the absence of any recombinant cytokine. In particular embodiments, the cells are further incubated in the absence of recombinant IL-2, IL-7, and IL- 15.

[0439] In certain embodiments, the cells are further incubated after the introducing of the polynucleotide encoding the heterologous or recombinant protein, e.g., viral vector, for, for about, or for at least 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. In certain embodiments, the cells are further incubated after the introducing of the polynucleotide encoding the heterologous or recombinant protein, e.g., viral vector, for, for about, or for at least one day, 2 days, 3 days, 4 days, or more than 4 days. In some embodiments, the further incubating is performed for an amount of time between 30 minutes and 2 hours, between 1 hour and 8 hours, between 6 hours and 12 hours, between 12 hours and 18 hours, between 16 hours and 24 hours, between 18 hours and 30 hours, between 24 hours and 48 hours, between 24 hours and 72 hours, between 42 hours and 54 hours, between 60 hours and 120 hours between 96 hours and 120 hours, between 90 hours and between 1 days and 7 days, between 3 days and 8 days, between 1 day and 3 days, between 4 days and 6 days, or between 4 days and 5 days prior to the genetic engineering. In some embodiments, the further incubating is for or for about between 18 hours and 30 hours. In particular embodiments, the further incubating is for or for about 24 hours or for for for about one day.

[0440] In certain embodiments, the total duration of the further incubation is, is about, or is at least 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 further incubation is, is about, or is at least one day, 2 days, 3 days, 4 days, or 5 days. In particular embodiments, the further incubation is completed at, at about, 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 particular embodiments, the further incubation is completed at, at about, or within one day, 2 days, 3 days, 4 days, or 5 days. In some embodiments, the total duration of the further incubation is between or between about 12 hour and 120 hours, 18 hour and 96 hours, 24 hours and 72 hours, or 24 hours and 48 hours, inclusive. In some embodiments, the total duration of the further

Ill incubation is between or about between 1 hour and 48 hours, 4 hours and 36 hours, 8 hours and 30 hours or 12 hours and 24 hours, inclusive. In particular embodiments, the further incubation is performed for or for about 24 hours, 48 hours, or 72 hours, or for or for about 1 day, 2 days, or 3 days, respectively. In particular embodiments, the further incubation is performed for 24 hours ± 6 hours, 48 hours ± 6 hours, or 72 hours ± 6 hours. In particular embodiments, the further incubation is performed for or for about 72 hours or for or for about 3 days.

[0441] In some embodiments, the further incubation is completed between or between about 24 hour and 120 hours, 36 hour and 108 hours, 48 hours and 96 hours, or 48 hours and 72 hours, inclusive, after the initiation of the stimulation. In some embodiments, the further incubation is completed at, about, or within 120 hours, 108 hours, 96 hours, 72 hours, 48 hours, or 36 hours from the initiation of the stimulation. In some embodiments, the further incubation is completed at, about, or within 5 days, 4.5 days, 4 days, 3 days, 2 dayrs, or 1.5 days from the initiation of the stimulation. In particular embodiments, the further incubation is completed after hours 24 hours ± 6 hours, 48 hours ± 6 hours, or 72 hours ± 6 hours after the initiation of the stimulation. In some embodiments, the further incubation is completed after or after about 72 hours or after or after about 3 days.

1. Viral Particles

[0442] 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 virus particle. In some embodiments, the viral particle is a viral vector, such as a vector derived from simian virus 40 (SV40), adenoviruses, 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.

[0443] In some embodiments, the retroviral vector has a long terminal repeat sequence (LTR), e.g., a retroviral vector derived from the 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 retroviruses include those derived from any avian or mammalian cell source. The retroviruses can be amphotropic, meaning that they are capable of infecting host cells of several species, including humans. In one embodiment, the gene to be expressed replaces the retroviral gag, pol and/or env sequences. A number of illustrative 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; Bums et al. (1993) Proc. Natl. Acad. Sci. USA 90:8033-8037; and Boris-Lawrie and Temin (1993) Cur. Opin. Genet. Develop. 3:102- 109.

[0444] The viral vector genome can be constructed in a plasmid form that can be transfected into a packaging or producer cell line. In some embodiments, the nucleic acid encoding a recombinant protein, such as a recombinant receptor, is inserted or located in a region of the viral vector, such as in a non-essential region of the viral genome. In some embodiments, the nucleic acid is inserted into the viral genome in the place of certain viral sequences to produce a vims that is replication defective.

[0445] Any of a variety of known methods can be used to produce retroviral particles whose genome contains an RNA copy of the viral vector genome. In some embodiments, at least two components are involved in making a virus-based gene delivery system: first, packaging plasmids, encompassing the structural proteins as well as the enzymes necessary to generate a viral vector particle, and second, the viral vector itself, e.g., the genetic material to be transferred. Biosafety safeguards can be introduced in the design of one or both of these components.

[0446] In some embodiments, the packaging plasmid can contain all retroviral, such as HIV-1, proteins other than envelope proteins (Naldini et al., 1998). In other embodiments, viral vectors can lack additional viral genes, such as those that are associated with vimlence, e.g., vpr, vif, vpu and nef, and/or Tat, a primary transactivator of HIV. In some embodiments, lentiviral vectors, such as HIV-based lentiviral vectors, contain only three genes of the parental vims: gag, pol and rev, which reduces or eliminates the possibility of reconstitution of a wild-type virus through recombination. [0447] In some embodiments, the viral vector genome is introduced into a packaging cell line that contains all the components necessary to package viral genomic RNA, transcribed from the viral vector genome, into viral particles. Alternatively, the viral vector genome may contain one or more genes encoding viral components in addition to the one or more sequences, e.g., recombinant nucleic acids, of interest. In some aspects, in order to prevent replication of the genome in the target cell, however, endogenous viral genes required for replication are removed and provided separately in the packaging cell line.

[0448] In some embodiments, a packaging cell line is transfected with one or more plasmid vectors containing the components necessary to generate the particles. In some embodiments, a packaging cell line is transfected with a plasmid containing the viral vector genome, including the LTRs, the cis-acting packaging sequence and the sequence of interest, i.e. a nucleic acid encoding an antigen receptor, such as a CAR; and one or more helper plasmids encoding the virus enzymatic and/or structural components, such as Gag, pol and/or rev. In some embodiments, multiple vectors are utilized to separate the various genetic components that generate the retroviral vector particles. In some such embodiments, providing separate vectors to the packaging cell reduces the chance of recombination events that might otherwise generate replication competent viruses. In some embodiments, a single plasmid vector having all of the retroviral components can be used.

[0449] In some embodiments, the retroviral vector particle, such as lentiviral vector particle, is pseudotyped to increase the transduction efficiency of host cells. For example, a retroviral vector particle, such as a lentiviral vector particle, in some embodiments is pseudotyped with a VSV-G glycoprotein, which provides a broad cell host range extending the cell types that can be transduced. In some embodiments, a packaging cell line is transfected with a plasmid or polynucleotide encoding a non-native envelope glycoprotein, such as to include xenotropic, polytropic or amphotropic envelopes, such as Sindbis virus envelope, GALV or VSV-G.

[0450] In some embodiments, the packaging cell line provides the components, including viral regulatory and structural proteins, that are required in trans for the packaging of the viral genomic RNA into lentiviral vector particles. In some embodiments, the packaging cell line may be any cell line that is 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.

[0451] In some embodiments, the packaging cell line stably expresses the viral protein(s). For example, in some aspects, a packaging cell line containing the gag, pol, rev and/or other structural genes but without the LTR and packaging components can be constructed. In some embodiments, a packaging cell line can be transiently transfected with nucleic acid molecules encoding one or more viral proteins along with the viral vector genome containing a nucleic acid molecule encoding a heterologous protein, and/or a nucleic acid encoding an envelope glycoprotein.

[0452] In some embodiments, the viral vectors and the packaging and/or helper plasmids are introduced via transfection or infection into the packaging cell line. The packaging cell line can produce viral vector particles that contain the viral vector genome. Methods for transfection or infection are well known. Examples include calcium phosphate, DEAE- dextran and lipofection methods, electroporation and microinjection.

[0453] When a recombinant plasmid and the retroviral LTR and packaging sequences are introduced into a special cell line (e.g., by calcium phosphate precipitation for example), the packaging sequences may permit the RNA transcript of the recombinant plasmid to be packaged into viral particles, which then may be secreted into the culture media. The media containing the recombinant retroviruses in some embodiments is then collected, optionally concentrated, and used for gene transfer. For example, in some aspects, after cotransfection of the packaging plasmids and the transfer vector to the packaging cell line, the viral vector particles are recovered from the culture media and titered by standard methods used by those of skill in the art.

[0454] In some embodiments, a retroviral vector, such as a lentiviral vector, can be produced in a packaging cell line, such as an exemplary HEK 293T cell line, by introduction of plasmids to allow generation of lentiviral particles. In some embodiments, a packaging cell is transfected and/or contains a polynucleotide encoding gag and pol, and a polynucleotide encoding a recombinant receptor, such as an antigen receptor, for example, 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 two days after transfection of cells, e.g. HEK 293T cells, the cell supernatant contains recombinant lentiviral vectors, which can be recovered and titered.

[0455] Recovered and/or produced retroviral vector particles can be used to transduce target cells using the methods as described. Once in the target cells, the viral RNA is reverse- transcribed, imported into the nucleus and stably integrated into the host genome. One or two days after the integration of the viral RNA, the expression of the recombinant protein, e.g. antigen receptor, such as CAR, can be detected.

[0456] 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) are introduced into the cell via retroviral or lentiviral vectors, or via transposons (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).

[0457] In some embodiments, the one or more polynucleotide(s) or vector(s) encoding a recombinant receptor and/or additional polypeptide(s) are introduced into cells, e.g., T cells, prior to elution, cultivating, and/or expansion. This introduction of the polynucleotide(s) or vector(s) can be carried out with any suitable retroviral vector. In some embodiments, following engineering, resulting genetically engineered cells can be liberated from the initial stimulus (e.g., anti-CD3/anti-CD28 stimulus) and subsequently be stimulated in the presence of a second type of stimulus (e.g., via a de novo introduced recombinant receptor). This second type of stimulus may include an antigenic stimulus in form of a peptide/MHC molecule, the cognate (cross-linking) ligand of the genetically introduced receptor (e.g. natural antigen and/or ligand of a CAR) or any ligand (such as an antibody) that directly binds within the framework of the new receptor (e.g. by recognizing constant regions 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). [0458] In some cases, a vector may be used that does not require that the cells, e.g., T cells, are activated. In some such instances, the cells may be selected and/or transduced prior to activation.

[0459] In some embodiments, the polynucleotide encoding the recombinant receptor contains at least one promoter that is operatively linked to control expression of the recombinant receptor. In some examples, the polynucleotide contains two, three, or more promoters operatively linked to control expression of the recombinant receptor. In some embodiments, polynucleotide can contain regulatory sequences, such as transcription and translation initiation and termination codons, which are specific to the type of host (e.g., bacterium, fungus, plant, or animal) into which the polynucleotide is to be introduced, as appropriate and taking into consideration whether the polynucleotide is DNA- or RNA-based. In some embodiments, the polynucleotide can contain regulatory/control elements, such as a promoter, an enhancer, an intron, a polyadenylation signal, a Kozak consensus sequence, internal ribosome entry sites (IRES), a 2A sequence, and splice acceptor or donor. In some embodiments, the polynucleotide can contain a nonnative promoter operably linked to the nucleotide sequence encoding the recombinant receptor and/or one or more additional polypeptide(s). In some embodiments, the promoter is selected from among an RNA pol I, pol II or pol III promoter. In some embodiments, the promoter is recognized by RNA polymerase II (e.g., a CMV, SV40 early region or adenovirus major late promoter). In another embodiment, the promoter is recognized by RNA polymerase III (e.g., a U6 or Hl promoter). In some embodiments, the promoter can be a non- viral promoter or a viral promoter, such as a cytomegalovirus (CMV) promoter, an SV40 promoter, an RSV promoter, and a promoter found in the long-terminal repeat of the murine stem cell virus. Other known promoters also are contemplated.

[0460] In some embodiments, the promoter is or comprises a constitutive promoter. Exemplary constitutive promoters include simian virus 40 early promoter (SV40), cytomegalovirus immediate-early promoter (CMV), human Ubiquitin C promoter (UBC), human elongation factor la promoter (EFla), mouse phosphoglycerate kinase 1 promoter (PGK), and chicken P- Actin promoter coupled with CMV early enhancer (CAGG). In some embodiments, the constitutive promoter is a synthetic or modified promoter. In some embodiments, the promoter is or comprises an MND promoter, a synthetic promoter that contains the U3 region of a modified MoMuLV LTR with 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 human elongation factor 1 alpha (EFla) promoter or a modified form thereof or the MND promoter.

[0461] In another embodiment, the promoter is a regulated promoter (e.g., inducible promoter). In some embodiments, the promoter is an inducible promoter or a repressible promoter. In some embodiments, the promoter comprises a Lac operator sequence, a tetracycline operator sequence, a galactose operator sequence or a doxycycline operator sequence, or is an analog thereof or is capable of being bound by or recognized by a Lac repressor or a tetracycline repressor, or an analog thereof. In some embodiments, the polynucleotide does not include a regulatory element, e.g. promoter.

[0462] In some cases, the nucleic acid sequence encoding the recombinant receptor contains a signal sequence that encodes a signal peptide. In some aspects, the signal sequence may encode a signal peptide derived from a native polypeptide. In other aspects, the signal sequence may encode a heterologous or non-native signal peptide. In some cases, the nucleic acid sequence encoding the recombinant receptor, e.g., chimeric antigen receptor (CAR), contains a signal sequence that encodes a signal peptide. Exemplary signal peptides include, for example, a GMCSER alpha chain signal peptide or a CD8 alpha signal peptide.

[0463] 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) includes a transduction tag, a surrogate tag and/or a resistance tag or selection tag. Among additional nucleic acid sequences introduced, e.g., encoding for one or more additional polypeptide(s), include nucleic acid sequences that can improve the efficacy of therapy, such as by promoting viability and/or function of transferred cells; nucleic acid sequences to provide a genetic tag for selection and/or evaluation of the cells, such as to assess in vivo survival or localization; nucleic acid sequences to improve safety, for example, by making the cell susceptible to negative selection in vivo as described by 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 describing the use of bifunctional selectable fusion genes derived from fusing a dominant positive selectable tag with a negative selectable tag, and US Patent No. 6,040,177.

[0464] In some embodiments, the tag is a transduction tag or a surrogate tag. A transduction tag or a surrogate tag can be used to detect cells that have been introduced with the polynucleotide, e.g., a polynucleotide encoding a recombinant receptor. In some embodiments, the transduction tag can indicate or confirm modification of a cell. In some embodiments, the surrogate tag is a protein that is made to be co-expressed on the cell surface with the recombinant receptor, e.g. CAR. In particular 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 that encodes 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 causes ribosome skipping, such as a 2A sequence. Extrinsic tag genes may in some cases be utilized in connection with engineered cell to permit detection or selection of cells and, in some cases, also to promote cell elimination and/or cell suicide.

[0465] Exemplary surrogate tags can include truncated forms of cell surface polypeptides, such as truncated forms that are non-functional and to not transduce or are not capable of transducing a signal or a signal ordinarily transduced by the full-length form of the cell surface polypeptide, and/or do not or are not capable of internalizing. Exemplary truncated cell surface polypeptides including truncated forms of growth factors or other receptors such as a truncated human epidermal growth factor receptor 2 (tHER2), a truncated epidermal growth factor receptor (tEGFR) or a prostate-specific membrane antigen (PSMA) or modified form thereof, such as a truncated PSMA (tPSMA). In some aspects, tEGFR may contain an epitope recognized by the antibody cetuximab (Erbitux®) or other therapeutic anti-EGFR antibody or binding molecule, which can be used to identify or select cells that have been engineered with the tEGFR construct and an encoded exogenous protein, and/or to eliminate or separate 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. surrogate tag, includes all or part e.g., truncated form) of CD34, a NGFR, a CD 19 or a truncated CD19, e.g., a truncated non-human CD19.

[0466] 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 the fluorescent proteins. In some embodiments, the tag is or comprises an enzyme, such as a luciferase, the lacZ gene from E. coli, alkaline phosphatase, secreted embryonic alkaline phosphatase (SEAP), chloramphenicol acetyl transferase (CAT). Exemplary light-emitting reporter genes include luciferase (luc), P-galactosidase, chloramphenicol acetyltransferase (CAT), P-glucuronidase (GUS) or variants thereof. In some aspects, expression of the enzyme can be detected by addition of a substrate that can be detected upon the expression and functional activity of the enzyme.

[0467] In some embodiments, the tag is a resistance maker or selection tag. In some embodiments, the resistance tag or selection tag is or comprises a polypeptide that confers resistance to exogenous agents or drugs. 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 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.

[0468] Any of the recombinant receptors and/or the additional polypeptide(s) described herein can be encoded by one or more polynucleotides containing one or more nucleic acid sequences encoding recombinant receptors, in any combinations, orientation or arrangements. For example, one, two, three or more polynucleotides can encode one, two, three or more different polypeptides, e.g., recombinant receptors or portions or components thereof, and/or one or more additional polypeptide(s), e.g., a tag and/or an effector molecule. In some embodiments, one polynucleotide contains a nucleic acid sequence encoding a recombinant receptor, e.g., CAR, or portion 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., CAR, or portion 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 the one or more additional polypeptide(s) are operably linked to two different promoters. In some embodiments, the nucleic acid encoding the recombinant receptor is present upstream of the nucleic acid encoding the one or more additional polypeptide(s). In some embodiments, the nucleic acid encoding the recombinant receptor is present downstream of the nucleic acid encoding one or more additional polypeptide(s).

[0469] In certain cases, one polynucleotide contains nucleic acid sequences encode 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, and each can be individually transferred or introduced into the cell for expression in the cell. In some embodiments, the nucleic acid sequences encoding the tag and the nucleic acid sequences encoding the recombinant receptor are present or inserted at different locations within the genome of the cell. In some embodiments, the nucleic acid sequences encoding the tag and the nucleic acid sequences encoding the recombinant receptor are operably linked to two different promoters.

[0470] In some embodiments, such as those where the polynucleotide contains a first and second nucleic acid sequence, the coding sequences encoding each of the different polypeptide chains can be operatively linked to a promoter, which can be the same or different. In some embodiments, the nucleic acid molecule can contain a promoter that drives the expression of two or more different polypeptide chains. In some embodiments, such nucleic acid molecules can be multicistronic (bicistronic or tricistronic, see e.g., U.S. Patent No. 6,060,273). In some embodiments, the nucleic acid sequences encoding the recombinant receptor and the nucleic acid sequences encoding the one or more additional polypeptide(s) are operably linked to the same promoter and are optionally separated by an internal ribosome entry site (IRES), or a nucleic acid encoding a self-cleaving peptide or a peptide that causes ribosome skipping, such as a 2A element. For example, an exemplary tag, and optionally a ribosome skipping sequence sequence, can be any as disclosed in PCT Pub. No. WO2014031687.

[0471] In some embodiments, transcription units can be engineered as a bicistronic unit containing an IRES, which allows coexpression of gene products (e.g. encoding the recombinant receptor and the additional polypeptide) by a message from a single promoter. Alternatively, in some cases, a single promoter may direct expression of an RNA that contains, in a single open reading frame (ORF), two or three genes (e.g. encoding the tag and encoding the recombinant receptor) separated from one another by sequences encoding a selfcleavage peptide (e.g., 2A sequences) or a protease recognition site (e.g., furin). The ORF thus encodes a single polypeptide, which, either during (in the case of 2A) or after translation, is processed into the individual proteins. In some cases, the peptide, such as a T2A, can cause the ribosome to skip (ribosome skipping) synthesis of a peptide bond at the C-terminus of a 2A element, leading to separation 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 system disclosed herein include 2A sequences from the foot- and-mouth disease virus (F2A), equine rhinitis A virus (E2A), Thosea asigna virus (T2A), and porcine teschovirus- 1 (P2A), for instance as described in U.S. Patent Pub. No.

20070116690.

[0472] In some embodiments, the polynucleotide encoding the recombinant receptor and/or additional polypeptide is contained in a vector or can be cloned into one or more vector(s). In some embodiments, the one or more vector(s) can be used to transform or transfect a host cell, e.g., a cell for engineering. 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 vectors can be prepared using standard recombinant DNA techniques. [0473] In some embodiments, the vector can be a vector of the pUC series (Fermentas Life Sciences), the pBluescript series (Stratagene, LaJolla, Calif.), the pET series (Novagen, Madison, Wis.), the pGEX series (Pharmacia Biotech, Uppsala, Sweden), or the pEX series (Clontech, Palo Alto, Calif.). In some cases, bacteriophage vectors, such as XG10, XGT11, XZapII (Stratagene), XEMBL4, and XNM1149, also can be used. In some embodiments, plant expression vectors can be used and include pBIOl, pBI101.2, pBI101.3, pBI121 and pBIN19 (Clontech). In some embodiments, animal expression vectors include pEUK-Cl, pMAM and pMAMneo (Clontech).

2. Chimeric Antigen Receptors

[0474] 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, chimeric receptors, such as a chimeric antigen receptors, contain one or more domains that combine a ligand-binding domain (e.g., antibody or antibody fragment) that provides specificity for a desired antigen (e.g., tumor antigen) with intracellular signaling domains. In some embodiments, the intracellular signaling domain is an activating intracellular domain portion, such as a T cell activating domain, providing a primary activation signal. In some embodiments, the intracellular signaling domain contains or additionally contains a costimulatory signaling domain to facilitate effector functions. In some embodiments, chimeric receptors when genetically engineered into immune cells can modulate T cell activity, and, in some cases, can modulate T cell differentiation or homeostasis, thereby resulting in genetically engineered cells with improved longevity, survival and/or persistence in vivo, such as for use in adoptive cell therapy methods.

[0475] Exemplary antigen receptors, including CARs, and methods for engineering and introducing such receptors into cells, include those 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, WG-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., 2012 October; 24(5): 633-39; and Wu et al., Cancer, 2012 March 18(2): 160-75. In some aspects, the antigen receptors include a CAR as described in U.S. Patent No.: 7,446,190, and those described in International Patent Application Publication No.: WO/2014055668 Al.

Examples of the CARs include CARs as disclosed in any of the aforementioned publications, such as WO2014031687, US 8,339,645, US 7,446,179, US 2013/0149337, U.S. Patent No.: 7,446,190, US Patent No.: 8,389,282, Kochenderfer et al., 2013, Nature Reviews Clinical Oncology, 10, 267-276 (2013); Wang et al. (2012) J. hnmunother. 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, U.S. Patent No.: 7,446,190, and US Patent No.: 8,389,282.

[0476] Exemplary antigen receptors, e.g., CARs, also include any described in 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_l): 1932 (2019); or can contain any of the antibodies or antigen-binding fragments described in U.S. Patent No. 8,153,765; 8,603477, 8,008,450; U.S. Pub. No. US20120189622 or US20100260748; and International PCT Publication Nos. W02006099875, W02009080829, WO2012092612, W02014210064.

[0477] Further exemplary antigen receptors, e.g., CARs, such as anti-BCMA CARs, include the CARs of idecabtagene vicleucel, ABECMA®, BCMA02, JCARH125, JNJ- 68284528 (LCAR-B38M; ciltacabtagene autoleucel; CARVYKTI™) (Janssen/Legend), P- BCMA-101 (Poseida), PBCAR269A (Poseida), P-BCMA-Allol (Poseida), Allo-715 (Pfizer/Allogene), CT053 (Carsgen), Descartes-08 (Cartesian), PHE885 (Novartis), ARI-002 (Hospital Clinic Barcelona, ID IB APS), and CTX120 (CRISPR Therapeutics). In a particular embodiment, the CAR is the CAR of idecabtagene vicleucel cells. In a particular embodiment, the CAR is the CAR of ABECMA® cells (cells used in ABECMA® immunotherapy). In a particular embodiment, the CAR is the CAR of ciltacabtagene autoleucel cells. In a particular embodiment, the CAR is the CAR of CARVYKTI™ cells (cells used in CARVYKTI™ immunotherapy ). [0478] Exemplary antigen receptors, e.g., CARs, also include the CARs of FDA-approved products BREYANZI® (lisocabtagene maraleucel), TECARTUS™ (brexucabtagene autoleucel), KYMRIAH™ (tisagenlecleucel), and YESCARTA™ (axicabtagene ciloleucel), ABECMA® (idecabtagene vicleucel), and CARVYKTI™ (ciltacabtagene autoleucel). In some of any of the provided embodiments, the CAR is the CAR of BREYANZI® (lisocabtagene maraleucel), TECARTUS™ (brexucabtagene autoleucel), KYMRIAH™ (tisagenlecleucel), YESCARTA™ (axicabtagene ciloleucel), ABECMA® (idecabtagene vicleucel), or CARVYKTI™ (ciltacabtagene autoleucel). In some of any of the provided embodiments, the CAR is the CAR of BREYANZI® (lisocabtagene maraleucel, see Sehgal et al., 2020, Journal of Clinical Oncology 38:15_suppl, 8040; Teoh et al., 2019, Blood 134(Supplement_l):593; and Abramson et al., 2020, The Lancet 396(10254): 839-852). In some of any of the provided embodiments, the CAR is the CAR of TECARTUS™ (brexucabtagene 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 the CAR of YESCARTA™ (axicabtagene 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 the CAR of ABECMA® (idecabtagene vicleucel, 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 any of the provided embodiments, the CAR is the CAR of CARVYKTI™ (ciltacabtagene autoleucel, see Berdeja et al., Lancet. 2021 Jul 24;398(10297):314-324; and Martin, Abstract #549 [Oral], presented at 2021 American Society of Hematology (ASH) Annual Meeting & Exposition)).

[0479] 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 the disease or condition, e.g., the tumor or pathogenic cells, as compared to normal or non-targeted cells or tissues. In other embodiments, the antigen is expressed on normal cells and/or is expressed on the engineered cells.

[0480] In some embodiments, the antigen is or includes avP6 integrin (avb6 integrin), B cell maturation antigen (BCMA), B7-H3, B7-H6, carbonic anhydrase 9 (CA9, also known as CAIX or G250), a cancer-testis antigen, cancer/testis antigen IB (CTAG, also known as NY- ESO-1 and LAGE-2), carcinoembryonic antigen (CEA), a 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 mutation (EGFR vIII), epithelial glycoprotein 2 (EPG-2), epithelial glycoprotein 40 (EPG-40), ephrinB2, 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), a folate binding protein (FBP), folate receptor alpha, ganglioside GD2, O-acetylated GD2 (OGD2), ganglioside GD3, glycoprotein 100 (gplOO), glypican-3 (GPC3), G Protein Coupled Receptor 5D (GPRC5D), Her2/neu (receptor tyrosine kinase erb-B2), Her3 (erb-B3), Her4 (erb-B4), erbB dimers, Human high molecular weight-melanoma- associated antigen (HMW-MAA), hepatitis B surface antigen, Human leukocyte antigen Al (HLA-A1), Human leukocyte antigen A2 (HLA-A2), IL- 22 receptor alpha(IL-22Ra), IL- 13 receptor alpha 2 (IL-13Ra2), kinase insert domain receptor (kdr), kappa light chain, LI cell adhesion molecule (LI -CAM), CE7 epitope of LI -CAM, Leucine Rich Repeat Containing 8 Family Member A (LRRC8A), Lewis Y, Melanoma-associated antigen (MAGE)-Al, MAGE-A3, MAGE-A6, MAGE-A10, mesothelin (MSLN), c-Met, murine cytomegalovirus (CMV), mucin 1 (MUC1), MUC16, natural killer group 2 member D (NKG2D) ligands, melan A (MART-1), neural cell adhesion molecule (NCAM), oncofetal antigen, Preferentially expressed antigen of melanoma (PRAME), progesterone receptor, a 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 biotinylated molecules, and/or molecules expressed by HIV, HCV, HBV or other pathogens. Antigens targeted by the receptors in some embodiments include antigens associated with a B cell malignancy, such as any of a number of known B cell marker. In some embodiments, the antigen is or includes CD20, CD19, CD22, R0R1, CD45, CD21, CD5, CD33, Igkappa, Iglambda, CD79a, CD79b or CD30. In some embodiments, the antigen is or includes a pathogen-specific or pathogen-expressed antigen. In some embodiments, the antigen is a viral antigen (such as a viral antigen from HIV, HCV, HBV, etc.), bacterial antigens, and/or parasitic antigens.

[0481] The chimeric receptors, such as CARs, generally include an extracellular antigen binding domain, such as a portion of an antibody molecule, generally a variable heavy (VH) chain region and/or variable light (VL) chain region of the antibody, e.g., an scFv antibody fragment.

[0482] 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 fragment antigen binding (Fab) fragments, F(ab’)2 fragments, Fab’ fragments, Fv fragments, recombinant IgG (rlgG) fragments, heavy chain variable (VH) regions capable of specifically binding the antigen, single chain antibody fragments, including single chain variable fragments (scFv), and single domain antibodies e.g., sdAb, sdFv, nanobody) fragments. 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 encompass 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 sub-class, including IgG and sub-classes thereof, IgM, IgE, IgA, and IgD.

[0483] The terms “complementarity determining region,” and “CDR,” synonymous with “hypervariable region” or “HVR,” are known in the art to refer to non-contiguous sequences of amino acids within antibody variable regions, which confer antigen specificity and/or binding affinity. In general, there are three CDRs in each heavy chain variable region (CDR- Hl, CDR-H2, CDR-H3) and three CDRs in each light chain variable region (CDR-L1, CDR- L2, CDR-L3). “Framework regions” and “FR” are known in the art to refer to the non-CDR portions of the variable regions of the heavy and light chains. In general, there are four FRs in each full-length heavy chain variable region (FR-H1, FR-H2, FR-H3, and FR-H4), and four FRs in each full-length light chain variable region (FR-L1, FR-L2, FR-L3, and FR-L4).

[0484] The precise amino acid sequence boundaries of a given CDR or FR can be readily determined using any of a number of well-known schemes, including those described by Kabat et al. (1991), “Sequences of Proteins of Immunological Interest,” 5th Ed. Public Health Service, National Institutes of Health, Bethesda, MD (“Kabat” numbering scheme); Al- Lazikani et al., (1997) JMB 273,927-948 (“Chothia” numbering scheme); 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(l):55-77 (“IMGT” numbering scheme); Honegger A and Pliickthun 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).

[0485] The boundaries of a given CDR or FR may vary depending on the scheme used for identification. For example, the Kabat scheme is based on structural alignments, while the Chothia scheme is based on structural information. Numbering for both the Kabat and Chothia schemes is based upon the most common antibody region sequence lengths, with insertions accommodated by insertion letters, for example, “30a,” and deletions appearing in some antibodies. The two schemes place certain insertions and deletions (“indels”) at different positions, resulting in differential numbering. The Contact scheme is based on analysis of complex crystal structures and is similar in many respects to the Chothia numbering scheme. The AbM scheme is a compromise between Kabat and Chothia definitions based on that used by Oxford Molecular’s AbM antibody modeling software. [0486] Table 1, below, lists exemplary position boundaries of CDR-L1, CDR-L2, CDR-L3 and CDR-H1, CDR-H2, CDR-H3 as identified by Kabat, Chothia, AbM, and Contact schemes, respectively. For CDR-H1, residue numbering is listed using both the Kabat and Chothia numbering schemes. FRs are located between CDRs, for example, with FR-L1 located before CDR-L1, FR-L2 located between CDR-L1 and CDR-L2, FR-L3 located between CDR-L2 and CDR-L3 and so forth. It is noted that because the shown Kabat numbering scheme places insertions at H35A and H35B, the end of the Chothia CDR-H1 loop when numbered using the shown Kabat numbering convention varies between H32 and H34, depending on the length of the loop.

Table 1. Boundaries of CDRs according to various numbering schemes.

1 - Kabat et al. (1991), “Sequences of Proteins of Immunological Interest,” 5th Ed. Public Health Service, National Institutes of Health, Bethesda, MD

2 - Al-Lazikani et al., (1997) JMB 273,927-948

[0487] Thus, unless otherwise specified, a “CDR” or “complementary determining region,” or individual specified CDRs (e.g., CDR-H1, CDR-H2, CDR-H3), of a given antibody or region thereof, such as a variable region thereof, should be understood to encompass a (or the specific) complementary determining region as defined by any of the aforementioned schemes, or other known schemes. For example, where it is stated that a particular CDR (e.g., a CDR-H3) contains the amino acid sequence of a corresponding CDR in a given VH or VL region amino acid sequence, it is understood that such a CDR has a sequence of the corresponding CDR (e.g. , CDR-H3) within the variable region, as defined by any of the aforementioned schemes, or other known schemes. In some embodiments, specific CDR sequences are specified. Exemplary CDR sequences of antibodies are described using various numbering schemes, although it is understood that an antibody can include CDRs as described according to any of the other aforementioned numbering schemes or other numbering schemes known to a skilled artisan.

[0488] Likewise, unless otherwise specified, a FR or individual specified FR(s) (e.g., FR- Hl, FR-H2, FR-H3, FR-H4), of a given antibody or region thereof, such as a variable region thereof, should be understood to encompass a (or the specific) framework region as defined by any of the known schemes. In some instances, the scheme for identification of a particular CDR, FR, or FRs or CDRs is specified, such as the CDR as defined by the Kabat, Chothia, AbM or Contact method, or other known schemes. In other cases, the particular amino acid sequence of a CDR or FR is given.

[0489] 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 antigen. The variable regions of the heavy chain and light chain (VH and VL, respectively) of a native antibody generally have similar structures, with each domain comprising four conserved framework regions (FRs) and three CDRs. (See, e.g., Kindt et al. Kuby Immunology, 6th ed., W.H. Freeman and Co., page 91 (2007). A single VH or VL domain may be sufficient to confer antigenbinding specificity. Furthermore, antibodies that bind a particular antigen may be isolated using a VH or VL domain from an antibody that binds the antigen to screen a library of complementary VL or VH domains, respectively. See, e.g., Portolano et al., J. Immunol. 150:880-887 (1993); Clarkson et al., Nature 352:624-628 (1991).

[0490] Among the antibodies included in the CARs 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 the antigen to which the intact antibody binds. Examples of antibody fragments include Fv, Fab, Fab’, Fab’-SH, F(ab’)2; diabodies; linear antibodies; heavy chain variable (VH) regions, single-chain antibody molecules such as scFvs and single-domain antibodies comprising only the VH region; and multispecific antibodies formed from antibody fragments. In some embodiments, the antigenbinding domain in the CARs is or comprises an antibody fragment comprising a variable heavy chain (VH) and a variable light chain (VL) region. In particular embodiments, the antibodies are single-chain antibody fragments comprising a heavy chain variable (VH) region and/or a light chain variable (VL) region, such as scFvs. [0491] In some embodiments, the chimeric antigen receptor includes an extracellular portion containing an antibody or antibody fragment. In some aspects, the chimeric antigen receptor includes an extracellular portion containing the antibody or fragment and an intracellular signaling domain. In some embodiments, the antibody or fragment includes an scFv.

[0492] In some embodiments, the antibody portion of the recombinant receptor, e.g., CAR, further includes 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 of a human IgG, such as IgG4 or IgGl. In some aspects, the portion of the constant region serves as a spacer region between the antigen-recognition component, e.g., scFv, and transmembrane domain. The spacer can be of a length that provides for increased responsiveness of the cell following antigen binding, as compared to in the absence of the spacer. Exemplary spacers include those described in Hudecek et al. (2013) Clin. Cancer Res., 19:3153, international patent application publication number W02014031687, U.S. Patent No. 8,822,647 or published app. No. US2014/0271635.

[0493] In some embodiments, the antigen receptor comprises an intracellular domain linked directly or indirectly to the extracellular domain. In some embodiments, the chimeric antigen receptor includes a transmembrane domain linking the extracellular domain and the intracellular signaling domain. In some embodiments, the intracellular signaling domain comprises an IT AM. For example, in some aspects, the antigen recognition domain (e.g. extracellular domain) generally is linked to one or more intracellular signaling components, such as signaling components that mimic activation through an antigen receptor complex, such as a TCR complex, in the case of a CAR, and/or signal via another cell surface receptor. In some embodiments, the chimeric receptor comprises a transmembrane domain linked or fused between the extracellular domain (e.g. scFv) and intracellular signaling domain. Thus, in some embodiments, the antigen-binding component (e.g., antibody) is linked to one or more transmembrane and intracellular signaling domains.

[0494] In one embodiment, a transmembrane domain that naturally is associated with one of the domains in the receptor, e.g., CAR, is used. In some instances, the transmembrane domain is selected or modified by amino acid substitution to avoid binding of such domains to the transmembrane domains of the same or different surface membrane proteins to minimize interactions with other members of the receptor complex.

[0495] The transmembrane domain in some embodiments is derived either from a natural or from a synthetic source. Where the source is natural, the domain in some aspects is derived from any membrane-bound or transmembrane protein. Transmembrane regions include those derived from (i.e. comprise at least the transmembrane region(s) of) 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 the transmembrane domain in some embodiments is synthetic. In some aspects, the synthetic transmembrane domain comprises predominantly hydrophobic residues such as leucine and valine. In some aspects, a triplet of phenylalanine, tryptophan and valine will be found at each end of a synthetic transmembrane domain. In some embodiments, the linkage is by linkers, spacers, and/or transmembrane domain(s). In some aspects, the transmembrane domain contains a transmembrane portion of CD28.

[0496] In some embodiments, the extracellular domain and transmembrane domain can be linked directly or indirectly. In some embodiments, the extracellular domain and transmembrane are linked by a spacer, such as any described herein. In some embodiments, the receptor contains extracellular portion of the molecule from which the transmembrane domain is derived, such as a CD28 extracellular portion.

[0497] Among the intracellular signaling domains are those that mimic or approximate a signal through a natural antigen receptor, a signal through such a receptor in combination with a costimulatory receptor, and/or a signal through a costimulatory receptor alone. In some embodiments, a short oligo- or polypeptide linker, for example, a linker of between 2 and 10 amino acids in length, such as one containing glycines and serines, e.g., glycine- serine doublet, is present and forms a linkage between the transmembrane domain and the cytoplasmic signaling domain of the CAR.

[0498] T cell activation is in some aspects described as being mediated by two classes of cytoplasmic signaling sequences: those that initiate antigen-dependent primary activation through the TCR (primary cytoplasmic signaling sequences), and those that act in an antigenindependent manner to provide a secondary or co- stimulatory signal (secondary cytoplasmic signaling sequences). In some aspects, the CAR includes one or both of such signaling components.

[0499] The receptor, e.g., the CAR, generally includes at least one intracellular signaling component or components. In some aspects, the CAR includes a primary cytoplasmic signaling sequence that regulates primary activation of the TCR complex. Primary cytoplasmic signaling sequences that act in a stimulatory manner may contain signaling motifs which are known as immunoreceptor tyrosine-based activation motifs or IT AMs. Examples of ITAM containing primary cytoplasmic signaling sequences include those derived from CD3 zeta chain, FcR gamma, CD3 gamma, CD3 delta and CD3 epsilon. In some embodiments, cytoplasmic signaling molecule(s) in the CAR contain(s) a cytoplasmic signaling domain, portion thereof, or sequence derived from CD3 zeta.

[0500] In some embodiments, the receptor includes an intracellular component of a TCR complex, such as a TCR CD3 chain that mediates T-cell activation and cytotoxicity, e.g., CD3 zeta chain. Thus, in some aspects, the antigen-binding portion is linked to one or more cell signaling modules. In some embodiments, cell signaling modules include CD3 transmembrane domain, CD3 intracellular signaling domains, and/or other CD3 transmembrane domains. In some embodiments, the receptor, e.g., CAR, further includes a portion of one or more additional molecules such as Fc receptor y, CD8, CD4, CD25, or CD 16. For example, in some aspects, the CAR or other chimeric receptor includes a chimeric molecule between CD3-zeta (CD3-Q or Fc receptor y and CD8, CD4, CD25 or CD16.

[0501] In some embodiments, upon ligation of the 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 the immune cell, e.g., T cell engineered to express the CAR. For example, in some contexts, the CAR induces a function of a T cell such as cytolytic activity or T-helper activity, such as secretion of cytokines or other factors. In some embodiments, a truncated portion of an intracellular signaling domain of an antigen receptor component or costimulatory molecule is used in place of an intact immuno stimulatory chain, for example, if it transduces the effector function signal. In some embodiments, the intracellular signaling domain or domains include the cytoplasmic sequences of the T cell receptor (TCR), and in some aspects also those of co-receptors that in the natural context act in concert with such receptors to initiate signal transduction following antigen receptor engagement.

[0502] In the context of a natural TCR, full activation generally requires not only signaling through the TCR, but also a costimulatory signal. Thus, in some embodiments, to promote full activation, a component for generating secondary or co- stimulatory signal is also included in the CAR. In other embodiments, the CAR does not include a component for generating a costimulatory signal. In some aspects, an additional CAR is expressed in the same cell and provides the component for generating the secondary or costimulatory signal.

[0503] In some embodiments, the chimeric antigen receptor contains an intracellular domain of a T cell costimulatory molecule. In some embodiments, the CAR includes a signaling domain and/or transmembrane portion of a costimulatory receptor, such as CD28, 4-1BB, 0X40, DAP10, and ICOS. In some aspects, the same CAR includes both the 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 intracellular signaling domain. In some aspects, the T cell costimulatory molecule is CD28 or 41BB.

[0504] In some embodiments, the activating domain is included within one CAR, whereas the costimulatory component is provided by another CAR recognizing another antigen. In some embodiments, the CARs include activating or stimulatory CARs, costimulatory CARs, both expressed on the same cell (see WO2014/055668). In some aspects, the cells include one or more stimulatory or activating CAR and/or a costimulatory CAR. In some embodiments, the cells further include inhibitory CARs (iCARs, see Fedorov et al., Sci. Transl. Medicine, 5(215) (December, 2013), such as a CAR recognizing an antigen other than the one associated with and/or specific for the disease or condition whereby an activating signal delivered through the disease-targeting CAR is diminished or inhibited by binding of the inhibitory CAR to its ligand, e.g., to reduce off-target effects.

[0505] In some embodiments, the two receptors induce, respectively, an activating and an inhibitory signal to the cell, such that ligation of one of the receptor to its antigen activates the cell or induces a response, but ligation of the second inhibitory receptor to its antigen induces a signal that suppresses or dampens that response. Examples are combinations of activating CARs and inhibitory CARs (iCARs). Such a strategy may be used, for example, to reduce the likelihood of off-target effects in the context in which the activating CAR binds an antigen expressed in a disease or condition but which is also expressed on normal cells, and the inhibitory receptor binds to a separate antigen which is expressed on the normal cells but not cells of the disease or condition.

[0506] In some aspects, the chimeric receptor is or includes an inhibitory CAR (e.g. iCAR) and includes intracellular components that dampen or suppress an immune response, such as an ITAM- and/or co stimulatory-promoted response in the cell. Exemplary 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 receptors, EP2/4 Adenosine receptors including A2AR. In some aspects, the engineered cell includes an inhibitory CAR including a signaling domain of or derived from such an inhibitory molecule, such that it serves to dampen the response of the cell, for example, that induced by an activating and/or costimulatory CAR.

[0507] 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 CD 137 (4- IBB, TNFRSF9) co- stimulatory domains, linked to a CD3 zeta intracellular domain.

[0508] In some embodiments, the CAR encompasses one or more, e.g., two or more, costimulatory domains and an activation domain, e.g., primary activation domain, in the cytoplasmic portion. Exemplary CARs include intracellular components of CD3-zeta, CD28, and 4- IBB.

[0509] In some embodiments, the antigen receptor further includes a tag and/or cells expressing the CAR or other antigen receptor further includes a surrogate tag, such as a cell surface tag, which may be used to confirm transduction or engineering of the cell to express the receptor. In some aspects, the tag includes all or part (e.g., truncated form) of CD34, a NGFR, or epidermal growth factor receptor, such as truncated version of such a cell surface receptor (e.g., tEGFR). In some embodiments, the nucleic acid encoding the tag is operably linked to a polynucleotide encoding for a linker sequence, such as a cleavable linker sequence, e.g., T2A. For example, a tag, and optionally a linker sequence, can be any as disclosed in published patent application No. WO2014031687. For example, the tag can be a truncated EGFR (tEGFR) that is, optionally, linked to a linker sequence, such as a T2A cleavable linker sequence.

[0510] In some embodiments, the tag is a molecule, e.g., cell surface protein, not naturally found on T cells or not naturally found on the surface of T cells, or a portion thereof. In some embodiments, the molecule is a non-self molecule, e.g., non-self protein, i.e., one that is not recognized as “self’ by the immune system of the host into which the cells will be adoptively transferred.

[0511] In some embodiments, the tag serves no therapeutic function and/or produces no effect other than to be used as a tag for genetic engineering, e.g., for selecting cells successfully engineered. In other embodiments, the tag may be a therapeutic molecule or molecule otherwise exerting some desired effect, such as a ligand for a cell to be encountered in vivo, such as a costimulatory or immune checkpoint molecule to enhance and/or dampen responses of the cells upon adoptive transfer and encounter with ligand.

[0512] In some cases, CARs are referred to as first, second, and/or third generation CARs. In some aspects, a first generation CAR is one that solely provides a CD3-chain induced signal upon antigen binding; in some aspects, a second-generation CARs is one that provides such a signal and costimulatory signal, such as one including an intracellular signaling domain from a costimulatory receptor such as CD28 or CD 137; in some aspects, a third generation CAR is one that includes multiple costimulatory domains of different costimulatory receptors.

[0513] For example, in some embodiments, the CAR contains an antibody, e.g., an antibody fragment, such as an scFv, specific to an antigen including any as described, a transmembrane domain that is or contains a transmembrane portion of CD28 or a functional variant thereof, and an intracellular signaling domain containing a signaling portion of CD28 or functional variant thereof and a signaling portion of CD3 zeta or functional variant thereof. In some embodiments, the CAR contains an antibody, e.g., antibody fragment, such as an scFv, specific to an antigen including any as described, a transmembrane domain that is or contains a transmembrane portion of CD28 or a functional variant thereof, and an intracellular signaling domain containing a signaling portion of a 4- IBB or functional variant thereof and a signaling portion of CD3 zeta or functional variant thereof. In some such embodiments, the receptor further includes 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.

[0514] For example, in some embodiments, the CAR includes an antibody such as an antibody fragment, including scFvs, 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 includes an antibody or fragment, such as scFv, a spacer such as any of the Ig-hinge containing spacers, a CD28-derived transmembrane domain, a 4-lBB-derived intracellular signaling domain, and a CD3 zeta-derived signaling domain.

[0515] Exemplary surrogate tags can include truncated forms of cell surface polypeptides, such as truncated forms that are non-functional and to not transduce or are not capable of transducing a signal or a signal ordinarily transduced by the full-length form of the cell surface polypeptide, and/or do not or are not capable of internalizing. Exemplary truncated cell surface polypeptides including truncated forms of growth factors or other receptors such as a truncated human epidermal growth factor receptor 2 (tHER2), a truncated epidermal growth factor receptor (tEGFR) or a prostate-specific membrane antigen (PSMA) or modified form thereof. tEGFR may contain an epitope recognized by the antibody cetuximab (Erbitux®) or other therapeutic anti-EGFR antibody or binding molecule, which can be used to identify or select cells that have been engineered to express the tEGFR construct and an encoded exogenous protein, and/or to eliminate or separate 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. surrogate tag, includes all or part (e.g., truncated form) of CD34, a NGFR, a CD19 or a truncated CD19, e.g., a truncated non-human CD19, or epidermal growth factor receptor (e.g., 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 a luciferase, the lacZ gene from E. coli, alkaline phosphatase, secreted embryonic alkaline phosphatase (SEAP), chloramphenicol acetyl transferase (CAT). Exemplary light-emitting reporter genes include luciferase (luc), P-galactosidase, chloramphenicol acetyltransferase (CAT), P-glucuronidase (GUS) or variants thereof.

[0516] In some embodiments, the tag is a resistance tag or selection tag. In some embodiments, the resistance tag or selection tag is or comprises a polypeptide that confers resistance to exogenous agents or drugs. 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 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.

[0517] In some embodiments, the nucleic acid encoding the tag is operably linked to a polynucleotide encoding for a linker sequence, such as a cleavable linker sequence, e.g., a T2A. For example, a tag, and optionally a linker sequence, can be any as disclosed in PCT Pub. No. WO2014031687.

3. T Cell Receptors

[0518] In some embodiments, the recombinant protein is a recombinant receptor. In some embodiments, the recombinant protein is a T cell receptor (TCR) or antigen-binding portion thereof that recognizes an peptide epitope or T cell epitope of a target polypeptide, such as an antigen of a tumor, viral or autoimmune protein. In some aspects, the TCR is or includes a recombinant TCR.

[0519] In some embodiments, a “T cell receptor” or “TCR” is a molecule that contains a variable a and P chains (also known as TCRa and TCRp, respectively) or a variable y and 5 chains (also known as TCRa and TCRp, respectively), or antigen-binding portions thereof, and which is capable of specifically binding to a peptide bound to an MHC molecule. In some embodiments, the TCR is in the aP form. Typically, TCRs that exist in aP and y5 forms are generally structurally similar, but T cells expressing them may have distinct anatomical locations or functions. A TCR can be found on the surface of a cell or in soluble form. Generally, a TCR is found on the surface of T cells (or T lymphocytes) where it is generally responsible for recognizing antigens bound to major histocompatibility complex (MHC) molecules.

[0520] Unless otherwise stated, the term “TCR” should be understood to encompass full TCRs as well as antigen-binding portions or antigen-binding fragments thereof. In some embodiments, the TCR is an intact or full-length TCR, including TCRs in the aP form or y5 form. In some embodiments, the TCR is an antigen-binding portion that is less than a full- length TCR but that binds to a specific peptide bound in an MHC molecule, such as binds to an MHC-peptide complex. In some cases, an antigen-binding portion or fragment of a TCR can contain only a portion of the structural domains of a full-length or intact TCR, but yet is able to bind the peptide epitope, such as MHC-peptide complex, to which the full TCR binds. In some cases, an antigen-binding portion contains the variable domains of a TCR, such as variable a chain and variable P chain of a TCR, sufficient to form a binding site for binding to a specific MHC-peptide complex. Generally, the variable chains of a TCR contain complementarity determining regions involved in recognition of the peptide, MHC and/or MHC-peptide complex.

[0521] In some embodiments, the variable domains of the TCR contain hypervariable loops, or complementarity determining regions (CDRs), which generally are the primary contributors to antigen recognition and binding capabilities and specificity. In some embodiments, a CDR of a TCR or combination thereof forms all or substantially all of the antigen-binding site of a given TCR molecule. The various CDRs within a variable region of a TCR chain generally are separated by framework regions (FRs), which generally display less variability among TCR molecules as compared to the CDRs (see, e.g., lores 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 main CDR responsible for antigen binding or specificity, or is the most important among the three CDRs on a given TCR variable region for antigen recognition, and/or for interaction with the processed peptide portion of the peptide-MHC complex. In some contexts, the CDR1 of the alpha chain can interact with the N-terminal part of certain antigenic peptides. In some contexts, CDR1 of the beta chain can interact with the C-terminal part of the peptide. In some contexts, CDR2 contributes most strongly to or is the primary CDR responsible for the interaction with or recognition of the MHC portion of the MHC -peptide complex. In some embodiments, the variable region of the P-chain can contain a further hypervariable region (CDR4 or HVR4), which generally is involved in superantigen binding and not antigen recognition (Kotb (1995) Clinical Microbiology Reviews, 8:411-426).

[0522] In some embodiments, a TCR also can 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 can possess one N-terminal immunoglobulin variable domain, one immunoglobulin constant domain, a transmembrane region, and a short cytoplasmic tail at the C-terminal end. In some embodiments, a TCR is associated with invariant proteins of the CD3 complex involved in mediating signal transduction.

[0523] In some embodiments, a TCR chain contains one or more constant domain. For example, the extracellular portion of a given TCR chain (e.g., a-chain or P-chain) can contain two immunoglobulin-like domains, such as a variable domain (e.g., Va or VP; typically amino acids 1 to 116 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., a-chain constant domain or Ca, typically positions 117 to 259 of the chain based on Kabat numbering or P chain constant domain or Cp, typically positions 117 to 295 of the chain based on Kabat) adjacent to the cell membrane. For example, in some cases, the extracellular portion of the TCR formed by the two chains contains two membrane-proximal constant domains, and two membrane-distal variable domains, which variable domains each contain CDRs. The constant domain of the TCR may contain short connecting sequences in which a cysteine residue forms a disulfide bond, thereby linking the two chains of the TCR. In some embodiments, a TCR may have an additional cysteine residue in each of the a and P chains, such that the TCR contains two disulfide bonds in the constant domains.

[0524] In some embodiments, the TCR chains contain 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 like CD3 and subunits thereof. For example, a TCR containing constant domains with a transmembrane region may anchor the protein in the cell membrane and associate with invariant subunits of the CD3 signaling apparatus or complex. The intracellular tails of CD3 signaling subunits (e.g. CD3y, CD35, CD3s and CD3(^ chains) contain one or more immunoreceptor tyrosine-based activation motif or IT AM that are involved in the signaling capacity of the TCR complex.

[0525] In some embodiments, the TCR may be a heterodimer of two chains a and P (or optionally y and 5) or it may be a single chain TCR construct. In some embodiments, the TCR is a heterodimer containing two separate chains (a and P chains or y and 5 chains) that are linked, such as by a disulfide bond or disulfide bonds.

[0526] In some embodiments, the TCR can be generated from a known TCR sequence(s), such as sequences of Va,P chains, for which a substantially full-length coding sequence is readily available. Methods for obtaining full-length TCR sequences, including V chain sequences, from cell sources are well known. In some embodiments, nucleic acids encoding the TCR can be obtained from a variety of sources, such as by polymerase chain reaction (PCR) amplification of TCR-encoding nucleic acids within or isolated from a given cell or cells, or synthesis of publicly available TCR DNA sequences.

[0527] In some embodiments, the recombinant receptors include recombinant TCRs and/or TCRs cloned from naturally occurring T cells. 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 the cells. In some embodiments, the TCR clone for a target antigen has been generated in transgenic mice 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.

[0528] In some embodiments, the TCR is obtained from a biological source, such as from cells such as from a T cell (e.g. cytotoxic T cell), T-cell hybridomas or other publicly available source. In some embodiments, the T-cells can be obtained from in vivo isolated cells. In some embodiments, the TCR is a thymically selected TCR. In some embodiments, the TCR is a neoepitope-restricted TCR. In some embodiments, the T- cells 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 sequence of the TCR.

[0529] In some embodiments, the TCR is generated from a TCR identified or selected from screening a library of candidate TCRs against a target polypeptide antigen, or target T cell epitope thereof. TCR libraries can be generated by amplification of the repertoire of Va and VP from T cells isolated from a subject, including cells present in PBMCs, spleen or other lymphoid organ. In some cases, T cells can be amplified from tumor-infiltrating lymphocytes (TILs). In some embodiments, TCR libraries can be generated from CD4+ or CD8+ cells. In some embodiments, the TCRs can be amplified from a T cell source of a normal of healthy subject, i.e. normal TCR libraries. In some embodiments, the TCRs can be amplified from a T cell source of a diseased subject, i.e. diseased TCR libraries. In some embodiments, degenerate primers are used to amplify the gene repertoire of Va and VP, such as by RT-PCR in samples, such as T cells, obtained from humans. In some embodiments, scTv libraries can be assembled from naive Va and VP libraries in which the amplified products are cloned or assembled to be separated by a linker. Depending on the source of the subject and cells, the libraries can be HLA allele- specific. Alternatively, in some embodiments, TCR libraries can be generated by mutagenesis or diversification of a parent or scaffold TCR molecule. In some aspects, the TCRs are subjected to directed evolution, such as by mutagenesis, e.g., of the a or P chain. In some aspects, particular residues within CDRs of the TCR are altered. In some embodiments, selected TCRs can be modified by affinity maturation. In some embodiments, antigen- specific T cells may be selected, such as by screening to assess CTL activity against the peptide. In some aspects, TCRs, e.g. present on the antigen- specific T cells, may be selected, such as by binding activity, e.g., particular affinity or avidity for the antigen.

[0530] In some embodiments, the TCR or antigen-binding portion thereof is one that has been modified or engineered. In some embodiments, directed evolution methods are used to generate TCRs with altered properties, such as with higher affinity for a specific MHC- peptide complex. 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, display approaches involve engineering, or modifying, a known, parent or reference TCR. For example, in some cases, a wild-type TCR can be used as a template for producing mutagenized TCRs in which in one or more residues of the CDRs are mutated, and mutants with an desired altered property, such as higher affinity for a desired target antigen, are selected.

[0531] Exemplary TCR T cell therapies for use in accordance with the methods provided herein are known in the art. TCR T cell therapies suitable for use in accordance with the methods provided herein include any 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), the contents of each of which are incorporated by reference herein in their entirety.

[0532] Exemplary TCR T cell therapies that target PRAME include those investigated or being investigated in clinical trials NCT03503968 and NCT02743611. Exemplary TCR T cell therapies that target MAGE-A3/A6 include those investigated or being investigated in clinical trial NCT03139370. Exemplary TCR T cell therapies that target CEA include those investigated or being investigated in clinical trial NCT00923806. Exemplary TCR T cell therapies that target MAGE-A3/12 include those investigated or being investigated in clinical trial NCT01273181. Exemplary TCR T cell therapies that target MAGE-A10 include those investigated or being investigated in clinical trials NCT02592577, NCT03391791, and NCT02989064. Exemplary TCR T cell therapies that target NY-ESO-1 include those investigated or being investigated in clinical trials NCT01343043, NCT03029273, NCT03462316, NCT01892293, NCT01352286, NCT01567891, NCT01350401, NCT02588612, NCT03691376, and NCT03168438. Exemplary TCR T cell therapies that target AFP include those investigated or being investigated in clinical trial NCT03132792. Exemplary TCR T cell therapies that target HA-1 include those investigated or being investigated in clinical trial NCT03326921. Exemplary TCR T cell therapies that target WT1 include those investigated or being investigated in clinical trials NCT02550535 and NCT02770820. Exemplary TCR T cell therapies that target GplOO include those investigated or being investigated in clinical trials NCT00923195 and NCT02889861. Exemplary TCR T cell therapies that target CMV include those investigated or being investigated in clinical trial NCT02988258. Exemplary TCR T cell therapies that target MART-1 include those investigated or being investigated in clinical trial NCT00091104. Exemplary TCR T cell therapies that target HBV include those investigated or being investigated in clinical trial NCT02719782. Exemplary TCR T cell therapies that target P53 include those investigated or being investigated in clinical trial NCT00393029. Exemplary TCR T cell therapies that target HPV-16 E6 include those investigated or being investigated in clinical trials NCT03578406 and NCT02280811. Exemplary TCR T cell therapies that target HPV-16 E7 include those investigated or being investigated in clinical trial NCT02858310. Exemplary TCR T cell therapies that target SL9 include those investigated or being investigated in clinical trial NCT00991224. Exemplary TCR T cell therapies that target TGFpiI include those investigated or being investigated in clinical trial NCT03431311. Exemplary TCR T cell therapies that target MCPy V include those investigated or being investigated in clinical trial NCT03747484. Exemplary TCR T cell therapies that target TRAIL include those investigated or being investigated in clinical trial NCT00923390. Exemplary TCR T cell therapies that target EBV include those investigated or being investigated in clinical trial NCT03648697. Exemplary TCR T cell therapies that target KRAS include those investigated or being investigated in clinical trials NCT03190941 and NCT03745326.

[0533] In some embodiments, peptides of a target polypeptide for use in producing or generating a TCR of interest are known or can be readily identified as a matter of routine. In some embodiments, peptides suitable for use in generating TCRs or antigen-binding portions can be determined based on the presence of an HLA-restricted motif in a target polypeptide of interest, such as a target polypeptide described below. In some embodiments, peptides are identified using computer prediction models as a matter of routine. In some embodiments, for predicting MHC class I binding sites, such models include, but are not limited to, ProPredl (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 choice of MHC antigen for use preparing a TCR or other MHC -peptide binding molecule.

[0534] HLA-A0201 -binding motifs and the cleavage sites for proteasomes and immune- proteasomes using computer prediction models are known. For predicting MHC class I binding sites, such models include, but are not limited to, ProPredl (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)

[0535] In some embodiments, the TCR or antigen binding portion thereof may be a recombinantly produced natural protein or mutated form thereof in which one or more property, such as binding characteristic, has been altered. In some embodiments, a TCR may be derived from one of various animal species, such as human, mouse, rat, or other mammal. A TCR may be cell-bound or in soluble form. In some embodiments, for purposes of the provided methods, the TCR is in cell-bound form expressed on the surface of a cell.

[0536] 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, a dTCR or scTCR have the structures as described in WO 03/020763, WO 04/033685, WO2011/044186.

[0537] In some embodiments, the TCR contains a sequence corresponding to the transmembrane sequence. In some embodiments, the TCR does contain a sequence corresponding to cytoplasmic sequences. In some embodiments, the TCR is capable of forming a TCR complex with CD3. In some embodiments, any of the TCRs, including a dTCR or scTCR, can be linked to signaling domains that yield an active TCR on the surface of a T cell. In some embodiments, the TCR is expressed on the surface of cells.

[0538] In some embodiments a dTCR contains a first polypeptide wherein a sequence corresponding to a TCR a chain variable region sequence is fused to the N terminus of a sequence corresponding to a TCR a chain constant region extracellular sequence, and a second polypeptide wherein a sequence corresponding to a TCR P chain variable region sequence is fused to the N terminus a sequence corresponding to a TCR P chain constant region extracellular sequence, the first and second polypeptides being linked by a disulfide bond. In some embodiments, the bond can correspond to the native inter-chain disulfide bond present in native dimeric aP TCRs. In some embodiments, the interchain disulfide bonds are not present in a native TCR. For example, in some embodiments, one or more cysteines can be incorporated into the constant region extracellular sequences of dTCR polypeptide pair. In some cases, both a native and a non-native disulfide bond may be desirable. In some embodiments, the TCR contains a transmembrane sequence to anchor to the membrane.

[0539] In some embodiments, a dTCR contains a TCR a chain containing a variable a domain, a constant a domain and a first dimerization motif attached to the C-terminus of the constant a domain, and a TCR P chain comprising a variable P domain, a constant P domain and a first dimerization motif attached to the C-terminus of the constant P domain, wherein the first and second dimerization motifs easily interact to form a covalent bond between an amino acid in the first dimerization motif and an amino acid in the second dimerization motif linking the TCR a chain and TCR P chain together.

[0540] In some embodiments, the TCR is a scTCR. Typically, a scTCR can be generated using suitable known methods, See e.g., Soo Hoo, W. F. et al., PNAS (USA) 89, 4759 (1992); Wiilfing, C. and Pliickthun, A., J. Mol. Biol. 242, 655 (1994); Kurucz, I. et al., PNAS (USA) 90 3830 (1993); International published PCT Nos. WO 96/13593, WO 96/18105, W099/60120, WO99/18129, WO 03/020763, WO2011/044186; and Schlueter, C. J. et al., J. Mol. Biol. 256, 859 (1996). In some embodiments, a scTCR contains an introduced nonnative disulfide interchain bond to facilitate the association of the TCR chains (see e.g. International published PCT No. WO 03/020763). In some embodiments, a scTCR is a nondisulfide linked truncated TCR in which heterologous leucine zippers fused to the C-termini thereof facilitate chain association (see e.g. International published PCT No. W099/60120). In some embodiments, a scTCR contain a TCRa variable domain covalently linked to a TCRP variable domain via a peptide linker (see e.g., International published PCT No. WO99/18129).

[0541] In some embodiments, a scTCR contains a first segment constituted by an amino acid sequence corresponding to a TCR a chain variable region, a second segment constituted by an amino acid sequence corresponding to a TCR P chain variable region sequence fused to the N terminus of an amino acid sequence corresponding to a TCR P chain constant domain extracellular sequence, and a linker sequence linking the C terminus of the first segment to the N terminus of the second segment.

[0542] In some embodiments, a scTCR contains a first segment constituted by an a chain variable region sequence fused to the N terminus of an a chain extracellular constant domain sequence, and a second segment constituted by a P chain variable region sequence fused to the N terminus of a sequence P chain extracellular constant and transmembrane sequence, and, optionally, a linker sequence linking the C terminus of the first segment to the N terminus of the second segment.

[0543] In some embodiments, a scTCR contains a first segment constituted by a TCR P chain variable region sequence fused to the N terminus of a P chain extracellular constant domain sequence, and a second segment constituted by an a chain variable region sequence fused to the N terminus of a sequence a chain extracellular constant and transmembrane sequence, and, optionally, a linker sequence linking the C terminus of the first segment to the N terminus of the second segment.

[0544] In some embodiments, the linker of a scTCRs that links the first and second TCR segments can be any linker capable of forming a single polypeptide strand, while retaining TCR binding specificity. In some embodiments, the linker sequence may, for example, have 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 so that the variable region sequences thereof are orientated for such binding. Hence, in some cases, the linker has a sufficient length to span the distance between the C terminus of the first segment and the N terminus of the second segment, or vice versa, but is not too long to block or reduces bonding of the scTCR to the target ligand. In some embodiments, the linker can contain from or from about 10 to 45 amino acids, such as 10 to 30 amino acids or 26 to 41 amino acids residues, for example 29, 30, 31 or 32 amino acids.

[0545] In some embodiments, the scTCR contains a covalent disulfide bond linking a residue of the immunoglobulin region of the constant domain of the a chain to a residue of the immunoglobulin region of the constant domain of the P chain. In some embodiments, the interchain disulfide bond in a native TCR is not present. For example, in some embodiments, one or more cysteines can be incorporated into the constant region extracellular sequences of the first and second segments of the scTCR polypeptide. In some cases, both a native and a non-native disulfide bond may be desirable.

[0546] In some embodiments of a dTCR or scTCR containing introduced interchain disulfide bonds, the native disulfide bonds are not present. In some embodiments, the one or more of the native cysteines forming a native interchain disulfide bonds are substituted to another residue, such as to a serine or alanine. In some embodiments, an introduced disulfide bond can be formed by mutating non-cysteine residues on the first and second segments to cysteine. Exemplary non-native disulfide bonds of a TCR are described in published International PCT No. W02006/000830.

[0547] In some embodiments, the TCR or antigen-binding fragment thereof exhibits an affinity with an equilibrium binding constant for a target antigen of between or between about 10-5 and 10-12 M and all individual values and ranges therein. In some embodiments, the target antigen is an MHC-peptide complex or ligand.

D. Culture

[0548] In some embodiments, the population of cells is from a culture of cells being cultured in vitro or ex vivo. In some embodiments, the provided methods involve culturing in vitro or ex vivo the culture of cells. In some embodiments, the population of cells is removed from the culture of cells in order to obtain the holographic information. In some embodiments, the holographic information is obtained in and the in vitro or ex vivo culture takes place in a closed system.

[0549] Exemplary methods and systems for the 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.

[0550] In some embodiments, the in vitro or ex vivo culture is in basal media. In some embodiments, the basal media is serum-free. In some embodiments, the basal media is free of serum derived from human. In some embodiments, the basal media 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 media can also help maintain pH and osmolality. A wide variety of commercially available basal media are well known to those skilled in the art and include Dulbeccos' Modified Eagles Medium (DMEM), Roswell Park Memorial Institute Medium (RPMI), Iscove modified Dulbeccos' medium, and Hams medium. In some embodiments, the basal media is Iscove's Modified Dulbecco's Medium, RPMI- 1640, or a-MEM.

[0551] In some embodiments, the basal media is a balanced salt solution (e.g., PBS, DPBS, HBSS, or EBSS). In some embodiments, the basal media is selected from Dulbecco's Modified Eagle's Medium (DMEM), Minimal Essential Medium (MEM), Basal Medium Eagle (BME), F-10, F-12, RPMI 1640, Glasgow's Minimal Essential Medium (GMEM), alpha Minimal Essential Medium (alpha MEM), Iscove's Modified Dulbecco's Medium, and M199. In some embodiments, the base media is a complex medium (e.g., RPMI-1640 or IMDM). In some embodiments, the base media is OpTmizer™ CTS™ T-Cell Expansion Basal Medium (ThermoFisher).

[0552] In some embodiments, the basal media is supplemented with additional additives. In some embodiments, the basal media is not supplemented with any additional additives. Additives to cell culture media include nutrients, sugars, e.g., glucose, amino acids, vitamins, and additives such as ATP and NADH.

[0553] In some embodiments, the in vitro or ex vivo culture is in the presence of one or more recombinant cytokines. In some embodiments, the culture is carried out in the presence of any of the recombinant cytokines described in Section II-B and at any of the concentrations 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 in the presence of IL-2, IL- 15, and IL-7.

[0554] In some embodiments, the in vitro or ex vivo culture is in the absence of recombinant cytokines.

[0555] In some embodiments, the in vitro or ex vivo culture is under conditions to expand T cells of the culture of cells. In some embodiments, the expansion conditions include one or more of particular media, 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 agents designed to promote growth, division, and/or expansion of the T cells. In some embodiments, the in vitro or ex vivo culture is carried out for a time period until a desired or threshold density, concentration, or number of T cells is achieved.

[0556] In some embodiments, the culture is performed under conditions that generally include a temperature suitable for the growth of primary immune cells, such as human T lymphocytes, for example, at least about 25 degrees Celsius, generally at least about 30 degrees, and generally at or about 37 degrees Celsius. In some embodiments, the culture of cells 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 carried out for a time period until the culture results in a desired or threshold density, concentration, or number of cells. In some embodiments, the culture is greater than or greater than about or is for about 24 hours, 48 hours, 72 hours, 96 hours, 5 days, 6 days, 7 days, 8 days, 9 days or more.

[0557] In some embodiments, the in vitro or ex vivo culture is in a bioreactor. Examples of bioreactors suitable for the in vitro or ex vivo culture include GE Xuri W25, GE Xuri W5, Sartorius BioSTAT RM 20 | 50, Finesse SmartRocker Bioreactor Systems, and Pall XRS Bioreactor Systems. In some embodiments, the bioreactor is used to perfuse and/or mix the culture of cells during at least a portion of the in vitro or ex vivo culture. In some embodiments, the population of cells is removed from the culture of cells in the bioreactor in order to obtain the holographic information. In some embodiments, the bioreactor is connected to a microscope for obtaining the holographic information, for instance as described in Section I-A.

[0558] In some embodiments, cells cultured while enclosed, connected, and/or under control of a bioreactor undergo expansion during the culture more rapidly than cells that are cultured without a bioreactor, e.g., cells that are cultured under static conditions such as without mixing, rocking, motion, and/or perfusion. In some embodiments, cells cultured while enclosed, connected, and/or under control of a bioreactor reach or achieve a 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 while enclosed, connected, and/or under control of a bioreactor reach or achieve a threshold expansion, cell count, and/or density 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 than cells cultured in an exemplary and/or alternative process where cells are not cultured while enclosed, connected, and/or under control of a bioreactor.

[0559] In some embodiments, the mixing is or includes rocking and/or motioning. In some cases, the bioreactor can be subject to motioning or rocking, which, in some aspects, can increase oxygen transfer. Motioning the bioreactor may include rotating along a horizontal axis, rotating along a vertical axis, a rocking motion along a tilted or inclined horizontal axis of the bioreactor or any combination thereof. In some embodiments, at least a portion of the culture is carried out with rocking. The rocking speed and rocking angle may be adjusted to achieve a desired agitation. In some embodiments the rock 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 rock angle is between 6-16°. In other embodiments, the rock angle is between 7-16°. In other embodiments, the rock angle is between 8-12°. In some embodiments, the rock rate 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 rock rate is between 4 and 12 rpm, such as between 4 and 6 rpm, inclusive.

[0560] In some embodiments, the bioreactor maintains the temperature at or near 37 °C and CO2 levels at or near 5% with a steady air flow at, at about, or at least 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 2.0 L/min. In certain embodiments, at least a portion of the culture is performed with perfusion, such as with a rate of 290 ml/day, 580 ml/day, and/or 1160 ml/day, e.g., depending on the timing in relation to the start of the culture and/or density of the cultured cells. In some embodiments, at least a portion of the cell culture expansion is performed with a rocking motion, such as at an angle of between 5° and 10°, such as 6°, at a constant rocking speed, such as a speed of between 5 and 15 RPM, such as 6 RMP or 10 RPM.

[0561] In some embodiments, at least a portion of the culture step is performed under constant perfusion, e.g., a perfusion at a slow steady rate. In some embodiments, the perfusion is or include an outflow of liquid e.g., used media, and an inflow of fresh media. In certain embodiments, the perfusion replaces used media with fresh media. In some embodiments, at least a portion of the culture is performed under perfusion at a steady rate of or of about or of at least 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.

[0562] In some embodiments, the in vitro or ex vivo culture is carried out under conditions in which there is minimal or no further expansion of T cells of the culture of cells. In some embodiments, the culture of cells is not incubated under conditions that increase the amount of T cells during the in vitro or ex vivo culture. In some embodiments, the culture of cells is incubated under conditions that may result in expansion, but are not carried out for purposes of expanding T cells of the culture of cells.

[0563] In some embodiments, the in vitro or ex vivo culture occurs in an incubator. In some embodiments, the culture of cells is transferred into a container for the in vitro or ex vivo culture. In some embodiments, the container is a vial. In particular embodiments, the container is a bag. In some embodiments, the culture of cells is transferred into the container under closed or sterile conditions. In some embodiments, the container, e.g., the vial or bag, is then placed into an incubator for all or a portion of the in vitro or ex vivo culture. In particular embodiments, the incubator is set at, at about, or at least 16°C, 24°C, or 35°C. In some embodiments, the incubator is set at 37°C, at about at 37°C, or at 37°C ±2°C, ±1°C, ±0.5°C, or ±0.1 °C.

[0564] In certain embodiments, the in vitro or ex vivo culture is performed under static conditions, such as conditions that do not involve centrifugation, shaking, rotating, rocking, or perfusion of media. In some embodiments, the in vitro or ex vivo culture is performed under gentle mixing conditions, e.g., involving rocking.

[0565] In some embodiments, the in vitro or ex vivo culture is for at least 12 hours. In some embodiments, the culture is performed for or for about, or for less than, 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.

[0566] In some embodiments, the in vitro or ex vivo culture is for no more than 14 days. In some embodiments, the in vitro or ex vivo culture is for no more than 7 days. In some embodiments, the in vitro or ex vivo culture is for between or between about 0.5 and 12 days, 0.5 and 10 days, 0.5 and 8 days, 0.5 and 6 days, 0.5 and 4 days, 0.5 and 2 days, 1 and 12 days, 1 and 10 days, 1 and 8 days, 1 and 6 days, 1 and 4 days, 1 and 2 days, 2 and 12 days, 2 and 10 days, 2 and 8 days, 2 and 6 days, 2 and 4 days, 4 and 12 days, 4 and 10 days, 4 and 8 days, 4 and 6 days, 6 and 12 days, 6 and 10 days, 6 and 8 days, 8 and 12 days, 8 and 10 days, or 10 and 12 days, each inclusive.

[0567] In some embodiments, the in vitro or ex vivo culture is for no more than 7 days. In some embodiments, the in vitro or ex vivo culture is for between or between about 0.5 and 7 days, 0.5 and 6 days, 0.5 and 5 days, 0.5 and 4 days, 0.5 and 3 days, 0.5 and 2 days, 0.5 and 1 day, 1 and 7 days, 1 and 6 days, 1 and 5 days, 1 and 4 days, 1 and 3 days, 1 and 2 days, 2 and 7 days, 2 and 6 days, 2 and 5 days, 2 and 4 days, 2 and 3 days, 3 and 7 days, 3 and 6 days, 3 and 5 days, 3 and 4 days, 4 and 7 days, 4 and 6 days, 4 and 5 days, 5 and 7 days, 5 and 6 days, or 6 and 7 days, each inclusive.

III. COMPUTING DEVICES

[0568] Also provided herein are computing devices that can be used in performing any of the provided methods. In some embodiments, the computing device has instructions in memory for performing any of the provided methods.

[0569] In some embodiments, the instructions include those for receiving the holographic information for individual cells of a population of cells containing T cells. In some embodiments, the instructions include those for receiving the one or more input measures for each of the plurality of cellular features for individual cells of a population of cells containing T cells. In some embodiments, the instructions include those for determining the one or more input measures for each of the plurality of cellular features for individual cells of a population of cells containing T cells. In some embodiments, the instructions include those for receiving the plurality of population-level statistics for a population of cells comprising T cells. In some embodiments, the instructions include those for determining the plurality of populationlevel statistics for a population of cells comprising T cells.

[0570] In some embodiments, the computing device has in memory a machine learning model trained according to any of the provided methods.

[0571] In some embodiments, the computing device has instructions in memory for training a machine learning model according to any of the provided methods.

[0572] In some embodiments, the instructions include those for receiving the holographic information for individual cells of a plurality of reference populations of cells containing T cells. In some embodiments, the instructions include those for receiving the dataset of reference input measures for a plurality of reference populations of cells containing T cells. In some embodiments, the instructions include those for determining the dataset of reference input measures for a plurality of reference populations of cells containing T cells. In some embodiments, the instructions include those for receiving the dataset of reference populationlevel statistics for a plurality of reference populations of cells comprising T cells. In some embodiments, the instructions include those for determining the dataset of reference population-level statistics for a plurality of reference populations of cells comprising T cells. In some embodiments, the instructions include those for receiving the dataset of reference population-level output measures for a plurality of reference populations of cells comprising T cells.

[0573] In some embodiments, the instructions include those for training a machine learning model, according to any of the provided methods, using the datasets of reference input measures, reference population-level statistics, and/or reference population-level output measures.

[0574] In some embodiments, the instructions include those for determining, according to any of the provided methods, the population-level output measure for the population of cells from the holographic information, the one or more input measures for each of the plurality of cellular features, or the plurality of population-level statistics. In some embodiments, the population-level output measure for the population of cells is determined using the machine learning model. In some embodiments, the instructions include those for determining the population-level output measure for the population of cells using the machine learning model.

[0575] A typical system for the provided computing devices 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 users and/or software applications. The system may include one or more output devices for presenting output to users and/or software applications. In some embodiments, the output devices may include a monitor capable of displaying to a user graphical representation of the described analytic functionality.

[0576] The described functionality may be supported using a computer including a suitable system processor 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 SEMPRON, PHENOM, OPTERON, ATHLON X2 or ATHLON 64 X2 (AMD Corp., Sunnyvale, Calif.), although other general purpose processors could be used. In some embodiments, the functionality, as further described below, may be distributed across multiple processing elements. The term processing element may refer to (1) a process running on a particular piece, or across particular pieces, of hardware, (2) a particular piece of hardware, or either (1) or (2) as the context allows. Some implementations can include one or more limited special purpose processors such as a digital signal processor (DSP), application specific integrated circuits (ASIC) or a field programmable gate arrays (FPGA). Further, some implementations can use combinations of general purpose and special purpose processors.

[0577] The environment can further include a SDS that could include a variety of primary and secondary storage elements. In one implementation, the SDS would include registers and RAM as part of the primary storage. The primary storage may in some implementations include other forms of memory such as cache memory, non-volatile memory (e.g., FLASH, ROM, EPROM, etc.), etc. The SDS may also include secondary storage including single, multiple and/or varied servers and storage elements. For example, the SDS may use internal storage devices connected to the system processor. In implementations where a single processing element supports all of the functionality, a local hard disk drive may serve as the secondary storage of the SDS, and a disk operating system executing on such a single processing element may act as a data server receiving and servicing data requests.

[0578] It will be understood by those skilled in the art that the different information used in the systems and methods provided herein may be logically or physically segregated within a single device serving as secondary storage for the SDS; multiple related data stores accessible through a unified management system, which together serve as the SDS; or multiple independent data stores individually accessible through disparate management systems, which may in some implementations be collectively viewed as the SDS. The various storage elements that comprise the physical architecture of the SDS may be centrally located or distributed across a variety of diverse locations.

[0579] In addition, or instead, the functionality and approaches discussed above can be stored in and/or on one or more computer readable storage media. Such media can include primary storage and/or secondary storage integrated with and/or within the computer such as RAM and/or a magnetic disk, and/or separable from the computer such as on a solid state device or removable magnetic or optical disk. The media can use any technology as would be known to those skilled in the art, including ROM, RAM, magnetic, optical, paper, and/or solid state media technology. IV. DEFINITIONS

[0580] Unless defined otherwise, all terms of art, notations, and other technical and scientific terms or terminology used herein are intended to have the same meaning as is commonly understood by one of ordinary skill in the art to which the claimed subject matter pertains. In some cases, terms with commonly understood meanings are defined herein for clarity and/or for ready reference, and the inclusion of such definitions herein should not necessarily be construed to represent a substantial difference over what is generally understood in the art.

[0581] As used herein, the singular forms “a,” “an,” and “the” include plural referents unless the context clearly dictates otherwise. For example, “a” or “an” means “at least one” or “one or more.” It is understood that aspects and variations described herein include “consisting” and/or “consisting essentially of’ aspects and variations.

[0582] 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 merely for convenience and brevity and should not be construed as an inflexible limitation on the scope of the claimed subject matter. Accordingly, the description of a range should be considered to have specifically disclosed all the possible sub-ranges as well as individual numerical values within that range. For example, where a range of values is stated, it is understood that smaller ranges between each intervening value between the upper and lower limit of the stated range and any other intervening value in the stated range is encompassed within the claimed subject matter. The upper and lower limits of these smaller ranges may independently be included in the smaller ranges and are also encompassed within the claimed subject matter, subject to any specifically excluded limit in the stated range. Where the stated range includes one or both of the limits, ranges excluding either or both of those included limits are also included in the claimed subject matter. This applies regardless of the breadth of the stated range.

[0583] The term “about” as used herein refers to the usual error range for the respective value readily known. Reference to “about” a value or parameter herein includes and describes embodiments that are directed to that value or parameter per se. For example, description referring to “about X” includes description of “X”. [0584] As used herein, a statement that a cell or population of cells is “positive” for a particular marker refers to the detectable presence on or in the cell of a particular marker, typically a surface marker. When referring to a surface marker, the term refers to the presence of surface expression as detected by flow cytometry, for example, by staining with an antibody that specifically binds to the marker and detecting said antibody, wherein the staining is detectable by flow cytometry at a level substantially above the staining detected carrying out the same procedure with an isotype-matched control under otherwise identical conditions and/or at a level substantially similar to that for a cell known to be positive for the marker, and/or at a level substantially higher than that for a cell known to be negative for the marker.

[0585] As used herein, a statement that a cell or population of cells is “negative” for a particular marker refers to the absence of substantial detectable presence on or in the cell of a particular marker, typically a surface marker. When referring to a surface marker, the term refers to the absence of surface expression as detected by flow cytometry, for example, by staining with an antibody that specifically binds to the marker and detecting said antibody, wherein the staining is not detected by flow cytometry at a level substantially above the staining detected carrying out the same procedure with an isotype-matched control under otherwise identical conditions, and/or at a level substantially lower than that for a cell known to be positive for the marker, and/or at a level substantially similar as compared to that for a cell known to be negative for the marker.

[0586] As used herein, a composition refers to any mixture of two or more products, substances, or compounds, including cells. It may be a solution, a suspension, liquid, powder, a paste, aqueous, non-aqueous, or any combination of the foregoing.

[0587] As used herein, a “subject” is a mammal, such as a human or other animal, and typically is human.

[0588] As used herein, the term “determining” means predicting. For instance, in some embodiments, determining a cell phenotype of a population of T cells, e.g., determining the activation state of a population of T cells or determining the memory phenotype of a population of T cells, means predicting the cell phenotype of a population of T cells, e.g., predicting the activation state of a population of T cells or predicting the memory phenotype of a population of T cells; determining one or more input measures means predicting one or more input measures; determining a plurality of population-level statistics means predicting a plurality of population-level statistics; and determining a population-level output measure of expression means predicting a population-level output measure of expression.

V. EXEMPLARY EMBODIMENTS

[0589] Among the provided embodiments are:

1. A method for determining the activation state of a population of T cells, comprising determining, for a population of cells comprising T cells, a population-level output measure of expression of a marker expressed by activated T cells, wherein the population-level output measure is determined based on a plurality of population-level statistics, wherein: each population-level statistic is of one or more input measures for a cellular feature of a plurality of cellular features derived from holographic information obtained for the population of cells, the plurality of population-level statistics comprising one or more population-level statistics for each of the plurality of cellular features; and each input measure is from an individual cell of the population of cells.

2. The method of embodiment 1, wherein the method comprises determining the plurality of population-level statistics from the one or more input measures for each of the plurality of cellular features.

3. The method of embodiment 1 or embodiment 2, wherein the method comprises determining the one or more input measures for each of the plurality of cellular features from the holographic information.

4. The method of any one of embodiments 1-3, wherein the method comprises obtaining the holographic information.

5. A method for determining the activation state of a population of T cells, comprising:

(a) obtaining holographic information for a population of cells comprising T cells;

(b) determining one or more input measures for each of a plurality of cellular features derived from the holographic information, wherein each input measure is from an individual cell in the population of cells;

(c) determining a plurality of population-level statistics, wherein each populationlevel statistic is of the one or more input measures for a cellular feature of the plurality of cellular features, and the plurality of population-level statistics comprises one or more population-level statistics for each of the plurality of cellular features; and

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

6. The method of any one of embodiments 1-5, the method further comprises engineering the population of T cells following the determining of the population-level output measure.

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

8. The method of any one of embodiments 4-7, wherein the obtaining the holographic information comprises imaging the population of cells using DDHM.

9. The method of any one of embodiments 1-8, wherein the one or more population-level statistics for at least one, optionally each, of the plurality of cellular features comprise one or more quantiles of the one or more input measures of the cellular feature.

10. The method of embodiment 9, wherein the one or more quantiles are selected from the 0.01, 0.1, 0.5, 0.9, and 0.99 quantiles of the one or more input measures of the cellular feature.

11. The method of any one of embodiments 1-8, wherein the one or more population-level statistics for at least one, optionally each, of the plurality of cellular features are determined by applying a distribution-based pooling filter to the one or more input measures of the cellular feature.

12. The method of any one of embodiments 1-11, wherein the population-level output measure is the number of cells, the percentage of cells, the proportion of cells, or the density of cells of the population of cells that express the marker.

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

14. The method of embodiment 13, wherein the holographic information is obtained during the in vitro or ex vivo culture of the culture of cells. 15. The method of embodiment 13 or embodiment 14, wherein the populationlevel output measure is of expression of the marker during the in vitro or ex vivo culture of the culture of cells.

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

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

18. The method of any one of embodiments 1-17, wherein the population of cells are incubated under T cell stimulating conditions prior to when the holographic information is obtained.

19. The method of any one of embodiments 1-18, wherein the method comprises incubating the population of cells under T cell stimulating conditions prior to when the holographic information is obtained.

20. The method of embodiment 18 or embodiment 19, wherein the incubation is prior to the in vitro or ex vivo culture.

21. The method of any one of embodiments 18-20, wherein the T cell stimulating conditions comprise incubation in the presence of T cell stimulatory agents that induce a primary activation signal and a costimulatory signal in T cells.

22. The method of embodiment 21, wherein the T cell stimulatory agents comprise an anti-CD3 antibody or antibody fragment.

23. The method of embodiment 21 or embodiment 22, wherein the T cell stimulatory agents comprise an anti-CD28 antibody or antibody fragment.

24. The method of any one of embodiments 21-23, wherein the T cell stimulatory agents are immobilized on a bead.

25. The method of any one of embodiments 1-24, wherein the population of cells is enriched for T cells.

26. The method of 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 population of cells are T cells.

27. The method of any one of embodiments 1-26, wherein the population-level output measure is determined by providing the plurality of population-level statistics as input to a machine learning model trained to predict population-level output measures of expression of the marker based on population-level statistics of the plurality of cellular features.

28. The method of embodiment 27, wherein the machine learning model is trained using a dataset of reference population-level statistics, wherein: for each of a first plurality of reference populations of cells comprising T cells, the dataset of reference population-level statistics comprises one or more reference populationlevel statistics for each of the plurality of cellular features, wherein: each reference population-level statistic is of one or more reference input measures for a cellular feature of the plurality of cellular features derived from holographic information obtained for the reference population of cells; and each reference input measure is from an individual cell of the reference population of cells.

29. The method of embodiment 28, wherein the machine learning model is trained using a dataset of reference population-level output measures, wherein for each of a second plurality of reference populations of cells, the dataset of reference population-level output measures comprises a reference population-level output measure of expression of the marker for the reference population of cells, wherein: the first and second pluralities of reference populations of cells are from reference cultures of cells being cultured in vitro or ex vivo', and reference populations of cells from the first plurality of reference populations of cells are from the same reference culture of cells as reference populations of cells from the second plurality of reference populations of cells.

30. A method for training a machine learning model that predicts the activation state of a population of T cells, comprising training a machine learning model using:

(i) a dataset of reference population-level statistics, wherein for each of a first plurality of reference populations of cells comprising T cells, the dataset of reference population-level statistics comprises one or more reference population-level statistics for each of a plurality of cellular features, wherein: each reference population-level statistic is of one or more reference input measures for a cellular feature of the plurality of cellular features derived from holographic information obtained for the reference population of cells; and each reference input measure is from an individual cell of the reference population of cells; and

(ii) a dataset of reference population-level output measures, wherein for each of a second plurality of reference populations of cells, the dataset of reference population-level output measures comprises a reference population-level output measure of expression of a marker expressed by activated T cells for the reference population of cells, wherein: the first and second pluralities of reference populations of cells are from reference cultures of cells being cultured in vitro or ex vivo', and reference populations of cells from the first plurality of reference populations of cells are from the same reference culture of cells as reference populations of cells from the second plurality of reference populations of cells; whereby the machine learning model is trained to predict population-level output measures of expression of the marker based on population-level statistics of the plurality of cellular features.

31. The method of any one of embodiments 1-30, wherein the plurality of cellular features are selected from intensity maximum, intensity minimum, intensity entropy, intensity contrast, phase entropy, cell area, and radius mean.

32. The method of any one of embodiments 1-31, wherein the plurality of cellular features comprises intensity maximum, intensity minimum, intensity entropy, intensity contrast, phase entropy, cell area, and radius mean.

33. A method for determining the activation state of a population of T cells, comprising determining, for a population of cells comprising T cells, expression of a marker expressed by activated T cells, wherein the marker is CD137, and the determining is based on one or more input measures for each of a plurality of cellular features derived from holographic information obtained for the population of cells, wherein: the plurality of cellular features comprises intensity maximum, intensity minimum, intensity entropy, intensity contrast, phase entropy, cell area, and radius mean; and each input measure is from an individual cell of the population of cells. 34. The method of any one of embodiments 1-30, wherein the one or more input measures for at least one, optionally each, of the plurality of cellular features are determined by providing the holographic information for individual cells of the population of cells to a convolutional neural network, wherein: the plurality of cellular features are cellular features extracted by the convolutional neural network; and the one or more input measures are determined from the convolutional neural network.

35. The method of any one of embodiments 28-30 and 34, wherein the one or more reference input measures for at least one, optionally each, of the plurality of cellular features are determined by providing the holographic information for individual cells of the reference population of cells to a convolutional neural network, wherein: the plurality of cellular features are cellular features extracted by the convolutional neural network; and the one or more reference input measures are determined from the convolutional neural network.

36. The method of embodiment 34 or embodiment 35, wherein the convolutional neural network is trained using a dataset of reference holographic information, wherein for each of a first plurality of reference populations of cells comprising T cells, the dataset of reference holographic information comprises holographic information obtained for individual cells of the reference population of cells.

37. The method of any one of embodiments 28-32 and 34-36, wherein the one or more reference population-level statistics for at least one, optionally each, of the plurality of cellular features comprise one or more quantiles of the one or more reference input measures for the cellular feature.

38. The method of embodiment 37, wherein the one or more quantiles are selected from the 0.01, 0.1, 0.5, 0.9, and 0.99 quantiles of the one or more reference input measures for the cellular feature.

39. The method of any one of embodiments 28-32 and 34-36, wherein the one or more reference population-level statistics for at least one, optionally each, of the plurality of cellular features are determined by applying a distribution-based pooling filter to the one or more reference input measures for the cellular feature.

40. A method for training a machine learning model that predicts the activation state of a population of T cells, comprising:

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

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

(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 the plurality of cellular features, each reference population-level statistic determined by applying a distribution-based pooling filter to the one or more reference input measures for a cellular feature of the plurality of cellular features; and

(d) training a machine learning model using the dataset of reference population-level statistics and a dataset of reference population-level output measures, wherein for each of a second plurality of reference populations of cells, the dataset of reference population-level output measures comprises a reference population-level output measure of expression of a marker expressed by activated T cells for the reference population of cells, wherein: the first and second pluralities of reference populations of cells are from reference cultures of cells being cultured in vitro or ex vivo', and reference populations of cells from the first plurality of reference populations of cells are from the same reference culture of cells as reference populations of cells from the second plurality of reference populations of cells; whereby the machine learning model is trained to predict population-level output measures of expression of the marker based on population-level statistics of the plurality of cellular features. 41. The method of any one of embodiments 34-40, wherein the one or more input measures for at least one, optionally each, of the plurality of cellular features are obtained from a fully connected layer of the convolutional neural network.

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

43. The method of any one of embodiments 1-42, wherein the holographic information comprises phase information and intensity information.

44. The method of any one of embodiments 29-32 and 34-43, wherein the reference population-level output measure is the number of cells, the percentage of cells, the proportion of cells, or the density of cells of the reference population of cells that express the marker.

45. The method of 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 measures are time-matched.

46. The method of any one of embodiments 29-32 and 34-45, wherein the in vitro or ex vivo culture of the reference cultures of cells is performed under the same or similar conditions as the in vitro or ex vivo culture of the culture of cells.

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

48. The method of any one of embodiments 29-32 and 34-47, wherein the dataset of reference population-level output measures are of expression of the marker during or after the in vitro or ex vivo culture of the reference cultures of cells.

49. The method of any one of embodiments 29-32 and 34-48, wherein the dataset of reference population-level output measures is determined using fluorescence imaging of the second plurality of reference populations of cells.

50. The method of embodiment 49, wherein the fluorescence imaging is by flow cytometry.

51. The method of any one of embodiments 1-32 and 34-50, wherein the marker is CD137 (4-1BB). 52. The method of any one of embodiments 28-32 and 34-51, wherein the first plurality of reference populations of cells are incubated under T cell stimulating conditions prior to when the holographic information for the first plurality of reference populations of cells is obtained.

53. The method of embodiment 52, wherein the incubation of the first plurality of reference populations of cells is performed under the same or similar conditions as the incubation of the population of cells.

54. The method of embodiment 52 or embodiment 53, wherein the incubation of the first plurality of reference populations of cells is prior to the in vitro or ex vivo culture of the reference cultures of cells.

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

56. The method of 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 populations of cells are T cells.

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

56.

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

(a) receiving the holographic information for individual cells of a population of cells comprising T cells, the one or more input measures for each of the plurality of cellular features for individual cells of a population of cells comprising T cells, or the plurality of population-level statistics for a population of cells comprising T cells; and

(b) determining, according to the method, the population-level output measure for the population of cells from the holographic information, the one or more input measures for each of the plurality of cellular features, or the plurality of population-level statistics. 59. The computing device of embodiment 58, wherein the computing device further comprises in memory a machine learning model trained according to the method of any one of embodiments 30-32 and 34-56, wherein the population-level output measure for the population of cells is determined using the machine learning model.

[0590] Also among the provided embodiments are:

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

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

2. A method for determining a cell phenotype of a population of T cells, comprising determining, for a population of cells comprising T cells, a population-level output measure of expression of a marker expressed by T cells with a cell phenotype of interest, wherein the population-level output measure is determined based on a plurality of population-level statistics, wherein: each population-level statistic is of one or more input measures for a cellular feature of a plurality of cellular features derived from holographic information obtained for the population of cells, the plurality of population-level statistics comprising one or more population-level statistics for each of the plurality of cellular features; and each input measure is from an individual cell of the population of cells.

3. The method of embodiment 2, wherein the method comprises determining the plurality of population-level statistics from the one or more input measures for each of the plurality of cellular features. 4. The method of embodiment 2 or embodiment 3, wherein the method comprises determining the one or more input measures for each of the plurality of cellular features from the holographic information.

5. The method of any one of embodiments 2-4, wherein the method comprises obtaining the holographic information.

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

7. The method of any one of embodiments 4-6, wherein the obtaining the holographic information comprises imaging the population of cells using DDHM.

8. The method of any one of embodiments 1-7, wherein the one or more population-level statistics for at least one, optionally each, of the plurality of cellular features comprise one or more quantiles of the one or more input measures of the cellular feature.

9. The method of embodiment 8, wherein the one or more quantiles comprise one or more of the 0.01, 0.1, 0.5, 0.9, and 0.99 quantiles of the one or more input measures of the cellular feature.

10. The method of any one of embodiments 1-7, wherein the one or more population-level statistics for at least one, optionally each, of the plurality of cellular features are determined by applying a distribution-based pooling filter to the one or more input measures of the cellular feature.

11. The method of any one of embodiments 1-10, wherein the population-level output measure is the number of cells, the percentage of cells, the proportion of cells, or the density of cells of the population of cells that express the marker.

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

13. The method of embodiment 12, wherein the holographic information is obtained during the in vitro or ex vivo culture of the culture of cells.

14. The method of embodiment 12 or embodiment 13, wherein the populationlevel output measure is of expression of the marker during the in vitro or ex vivo culture of the culture of cells. 15. The method of any one of embodiments 12-14, wherein the in vitro or ex vivo culture is under conditions to expand T cells of the culture of cells.

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

17. The method of any one of embodiments 1-16, wherein the population of cells are incubated under T cell stimulating conditions prior to when the holographic information is obtained.

18. The method of any one of embodiments 1-17, wherein the method comprises incubating the population of cells under T cell stimulating conditions prior to when the holographic information is obtained.

19. The method of embodiment 17 or embodiment 18, wherein the incubation is prior to the in vitro or ex vivo culture.

20. The method of any one of embodiments 17-19, wherein the T cell stimulating conditions comprise incubation in the presence of T cell stimulatory agents that induce a primary activation signal and a costimulatory signal in T cells.

21. The method of embodiment 20, wherein the T cell stimulatory agents comprise an anti-CD3 antibody or antibody fragment.

22. The method of embodiment 20 or embodiment 21, wherein the T cell stimulatory agents comprise an anti-CD28 antibody or antibody fragment.

23. The method of any one of embodiments 20-22, wherein the T cell stimulatory agents are immobilized on a bead.

24. The method of any one of embodiments 20-22, wherein the T cell stimulatory agents are immobilized on an oligomeric streptavidin mutein reagent.

25. The method of any one of embodiments 1-24, wherein a recombinant receptor is introduced into T cells of the population of cells prior to when the holographic information is obtained.

26. The method of any one of embodiments 1-25, wherein the method comprises introducing a recombinant receptor into T cells of the population of cells prior to when the holographic information is obtained. 27. The method of embodiment 25 or embodiment 26, wherein the introducing comprising contacting the population of cells with an agent comprising a polynucleotide encoding the recombinant receptor.

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

29. The method of 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 population of cells are T cells.

30. The method of any one of embodiments 1-29, wherein the population-level output measure is determined by providing the plurality of population-level statistics as input to a machine learning model trained to predict population-level output measures of expression of the marker based on population-level statistics of the plurality of cellular features.

31. The method of embodiment 30, wherein the machine learning model is trained using a dataset of reference population-level statistics, wherein: for each of a first plurality of reference populations of cells comprising T cells, the dataset of reference population-level statistics comprises one or more reference populationlevel statistics for each of the plurality of cellular features, wherein: each reference population-level statistic is of one or more reference input measures for a cellular feature of the plurality of cellular features derived from holographic information obtained for the reference population of cells; and each reference input measure is from an individual cell of the reference population of cells.

32. The method of embodiment 31, wherein the machine learning model is trained using a dataset of reference population-level output measures, wherein for each of a second plurality of reference populations of cells, the dataset of reference population-level output measures comprises a reference population-level output measure of expression of the marker for the reference population of cells, wherein: the first and second pluralities of reference populations of cells are from reference cultures of cells being cultured in vitro or ex vivo', and reference populations of cells from the first plurality of reference populations of cells are from the same reference culture of cells as reference populations of cells from the second plurality of reference populations of cells.

33. A method for training a machine learning model that predicts a cell phenotype of a population of T cells, comprising training a machine learning model using:

(ii) a dataset of reference population-level output measures, wherein for each of a second plurality of reference populations of cells, the dataset of reference population-level output measures comprises a reference population-level output measure of expression of a marker expressed by T cells with a cell phenotype of interest for the reference population of cells, wherein: the first and second pluralities of reference populations of cells are from reference cultures of cells being cultured in vitro or ex vivo', and reference populations of cells from the first plurality of reference populations of cells are from the same reference culture of cells as reference populations of cells from the second plurality of reference populations of cells; whereby the machine learning model is trained to predict population-level output measures of expression of the marker based on population-level statistics of the plurality of cellular features.

34. The method of any one of embodiments 1-33, wherein the plurality of cellular features comprise one or of intensity skewness, intensity correlation, intensity homogeneity, intensity maximum, intensity minimum, intensity entropy, intensity contrast, phase entropy, cell area, and radius mean. 35. The method of any one of embodiments 1-34, wherein the plurality of cellular features comprises intensity maximum, intensity minimum, intensity entropy, intensity contrast, phase entropy, cell area, and radius mean.

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

37. The method of any one of embodiments 1-36, wherein the plurality of cellular features comprises intensity skewness, intensity correlation, and intensity homogeneity.

38. A method for determining the activation state of a population of T cells, comprising determining, for a population of cells comprising T cells, a population-level output measure of expression of a marker expressed by activated T cells, wherein the marker is CD137 (4-1BB), and the population-level output measure is determined based on one or more input measures for each of a plurality of cellular features derived from holographic information obtained for the population of cells, wherein: the plurality of cellular features comprises intensity skewness, intensity correlation, and intensity homogeneity; and each input measure is from an individual cell of the population of cells.

39. The method of embodiment 38, wherein the plurality of cellular features comprises intensity maximum, intensity minimum, intensity entropy, intensity contrast, phase entropy, cell area, and radius mean.

40. The method of any one of embodiments 1-33, wherein the plurality of cellular features comprises one or more of peak area, phase average uniformity, intensity geometric mean, minimum optical height, and normalized optical height. 41. The method of any one of embodiments 1-33 and 40, wherein the plurality of cellular features comprises peak area, phase average uniformity, intensity geometric mean, minimum optical height, and normalized optical height.

42. The method of any one of embodiments 1-33, wherein the plurality of cellular features comprises one or more of cell area, perimeter, mean intensity, normalized peak area, and equivalent peak diameter.

43. The method of any one of embodiments 1-33 and 42, wherein the plurality of cellular features comprises cell area, perimeter, mean intensity, normalized peak area, and equivalent peak diameter.

44. A method for determining the memory phenotype of a population of T cells, comprising determining, for a population of cells comprising T cells, a population-level output measure of expression of a marker expressed by central memory T cells or stem cell memory T cells, wherein the marker is CCR7, and the population-level output measure is determined based on one or more input measures for each of a plurality of cellular features derived from holographic information obtained for the population of cells, wherein: the plurality of cellular features comprises one or more of cell area, perimeter, mean intensity, normalized peak area, and equivalent peak diameter; and each input measure is from an individual cell of the population of cells.

45. The method of embodiment 44, wherein the plurality of features comprises cell area, perimeter, mean intensity, normalized peak area, and equivalent peak diameter.

46. A method for determining recombinant receptor expression of a population of T cells, comprising determining, for a population of cells comprising T cells, expression of a recombinant receptor introduced into T cells of the population of cells, wherein the determining is based on one or more input measures for each of a plurality of cellular features derived from holographic information obtained for the population of cells, wherein: the plurality of cellular features comprises peak area, phase average uniformity, intensity geometric mean, minimum optical height, and normalized optical height; and each input measure is from an individual cell of the population of cells.

47. The method of any one of embodiments 1-33, wherein the one or more input measures for at least one, optionally each, of the plurality of cellular features are determined by providing the holographic information for individual cells of the population of cells to a convolutional neural network, wherein: the plurality of cellular features are cellular features extracted by the convolutional neural network; and the one or more input measures are determined from the convolutional neural network.

48. The method of any one of embodiments 31-33 and 47, wherein the one or more reference input measures for at least one, optionally each, of the plurality of cellular features are determined by providing the holographic information for individual cells of the reference population of cells to a convolutional neural network, wherein: the plurality of cellular features are cellular features extracted by the convolutional neural network; and the one or more reference input measures are determined from the convolutional neural network.

49. The method of embodiment 47 or embodiment 48, wherein the convolutional neural network is trained using a dataset of reference holographic information, wherein for each of a first plurality of reference populations of cells comprising T cells, the dataset of reference holographic information comprises holographic information obtained for individual cells of the reference population of cells.

50. The method of any one of embodiments 31-35, 37, 40-43, and 47-49, wherein the one or more reference population-level statistics for at least one, optionally each, of the plurality of cellular features comprise one or more quantiles of the one or more reference input measures for the cellular feature.

51. The method of embodiment 50, wherein the one or more quantiles are selected from the 0.01, 0.1, 0.5, 0.9, and 0.99 quantiles of the one or more reference input measures for the cellular feature.

52. The method of any one of embodiments 31-35, 37, 40-43, and 47-49, wherein the one or more reference population-level statistics for at least one, optionally each, of the plurality of cellular features are determined by applying a distribution-based pooling filter to the one or more reference input measures for the cellular feature. 53. A method for training a machine learning model that predicts a cell phenotype of a population of T cells, comprising:

(d) training a machine learning model using the dataset of reference population-level statistics and a dataset of reference population-level output measures, wherein for each of a second plurality of reference populations of cells, the dataset of reference population-level output measures comprises a reference population-level output measure of expression of a marker expressed by T cells with a cell phenotype of interest for the reference population of cells, wherein: the first and second pluralities of reference populations of cells are from reference cultures of cells being cultured in vitro or ex vivo', and reference populations of cells from the first plurality of reference populations of cells are from the same reference culture of cells as reference populations of cells from the second plurality of reference populations of cells; whereby the machine learning model is trained to predict population-level output measures of expression of the marker based on population-level statistics of the plurality of cellular features. 54. The method of any one of embodiments 47-53, wherein the one or more input measures for at least one, optionally each, of the plurality of cellular features are obtained from a fully connected layer of the convolutional neural network.

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

56. The method of any one of embodiments 1-55, wherein the holographic information comprises phase information and intensity information.

57. The method of any one of embodiments 32-35, 37, 40-43, and 47-56, wherein the reference population-level output measure is the number of cells, the percentage of cells, the proportion of cells, or the density of cells of the reference population of cells that express the marker.

58. The method of 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 populationlevel output measures are time-matched.

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

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

61. The method of any one of embodiments 32-35, 37, 40-43, and 47-60, wherein the dataset of reference population-level output measures are of expression of the marker during or after the in vitro or ex vivo culture of the reference cultures of cells.

62. The method of any one of embodiments 32-35, 37, 40-43, and 47-61, wherein the dataset of reference population-level output measures is determined using fluorescence imaging of the second plurality of reference populations of cells.

63. The method of embodiment 62, wherein the fluorescence imaging is by flow cytometry. 64. The method of any one of embodiments 1-35, 37, and 47-63, wherein the method is for determining the activation state of the population of T cells, and the marker is expressed by activated T cells.

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

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

67. The method of embodiment 66, wherein the marker is expressed by T cells having a central memory phenotype.

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

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

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

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

73. The method of embodiment 72, wherein the incubation of the first plurality of reference populations of cells is performed under the same or similar conditions as the incubation of the population of cells.

74. The method of embodiment 72 or embodiment 73, wherein the incubation of the first plurality of reference populations of cells is prior to the in vitro or ex vivo culture of the reference cultures of cells. 75. The method of any one of embodiments 31-35, 37, 40-43, and 47-74, wherein the first and/or second plurality of reference populations of cells are enriched for T cells.

76. The method of 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 populations of cells are T cells.

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

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

(b) determining, according to the method, the population-level output measure for the population of cells from the holographic information, the one or more input measures for each of the plurality of cellular features, or the plurality of population-level statistics.

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

VI. EXAMPLES

[0591] The following examples are included for illustrative purposes only and are not intended to limit the scope of the invention. Example 1: Machine Learning Method for Determining the Overall Activation State of T Cell Populations Using Holographic Imaging

[0592] A machine learning method for determining the overall activation state of T cell populations using holographic imaging was developed. To assess overall activation state, T cell populations were stimulated to induce activation and then were periodically assessed, during their cultivation in a bioreactor, for the percentage of cells exhibiting 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 the flow cytometry results of CD 137 expression as a surrogate for overall activation state, a machine learning method was trained using population-level statistics that summarized cell feature measurements across individual cells of the stimulated T cell populations. The cell feature measurements were derived from holographic information for the stimulated T cell populations that was captured during the cultivation.

[0593] In this approach, activation marker expression of individual cells was not required or used for model training, thereby obviating the need for any specialized equipment or reagents for capturing matched holographic and fluorescent images of individual cells. In addition, as fluorescently labelled cells were not holographically imaged, the impact on cell feature measurements and model predictive accuracy of any random or systematic morphometric changes in labelled cells, compared to un-labelled cells, was avoided.

[0594] In addition, with this approach, the same holographic imaging system used for obtaining images for training the machine learning model can be used for obtaining images for predicting overall T cell activation state using the trained machine learning model. This is in contrast to other methods in which different imaging systems may be used for obtaining images for model training and use, for instance a dual fluorescence-holographic imaging system for model training and a holographic imaging-only system for model use.

[0595] In this manner, an accurate, label-free method for monitoring overall T cell activation state was developed.

A. Stimulation Process

[0596] Primary T cells from healthy human subjects were cryofrozen, thawed, and subjected to stimulation to activate the T cells. Specifically, the T cells were stimulated for approximately 24 hours (between 18 and 30 hours) at about 37 °C and 5% CO2 in the presence of recombinant IL-2, recombinant IL-7, recombinant IL- 15, and super-paramagnetic polystyrene-coated beads that were conjugated to anti-CD3 and anti-CD28 monoclonal antibodies (Dynabeads, Invitrogen). The T cells were stimulated at a 1:1 bead-to-cell ratio in serum-free media. After stimulation, the T cells were then transduced by spinoculation for 60 minutes with a lentiviral vector encoding an exemplary recombinant protein, in this case one of two exemplary chimeric antigen receptors (CARs). Following spinoculation, the T cells were incubated for approximately 24 hours (between 18 and 30 hours) at about 37 °C and 5% CO2, after which the T cells were transferred to a rocking motion bioreactor and cultivated for expansion in serum- free media containing recombinant IL-2, IL-7, and IL- 15. The T cells were cultivated for expansion in the bioreactor for 3 days to 5 days before harvest. The anti- CD3/anti-CD28 beads were removed from the T cells following harvest.

[0597] This process was implemented for 15 primary T cell compositions from 15 human subjects. Cell feature measurements and activation marker expression estimates for each of the 15 primary T cell compositions were obtained during the cultivation step and used for model training and validation, as described in Sections B-D below.

B. Cell Feature Measurements

[0598] Holographic imaging of samples of the T cells was continuously performed using an exemplary digital holography microscopy (DHM) imaging system, in this case an Ovizio iLine F imaging system (Ovizio Imaging Systems NV/SA). The imaging system was connected to the bioreactor via a disposable tubing system. Every minute, a sample of the T cells was moved from the bioreactor, imaged using the DHM imaging system, and returned to the bioreactor.

[0599] Features of individual cells within each sample were derived from the captured holographic information (phase and intensity information). To do so, the captured information was segmented using OsOne software (Ovizio Imaging Systems NV/SA), which was then used to determine, per individual cell, the features listed in Table El.

Table El: Individual T Cell Features From Holographic Imaging

C. Activation Marker Expression

[0600] During the same time period in which T cells were being holographically imaged, samples of T cells from the bioreactor were collected every 24 hours, and the percentage of CD137-expressing activated T cells per sample was quantified using flow cytometry. To do so, CD137-expressing cells were fluorescently labelled using a fluorophore-conjugated anti- CD137 antibody (BioLegend, 4B4-1 clone conjugated to Brilliant Violet 421™, catalog number 309820). Flow cytometry data was acquired using the BD Biosciences Fortessa Dual Symphony instrument and BD Biosciences FACSDiva™ software, and flow analysis software (FlowJo vl0.6) was used to determine the percentage of CD137-expressing cells per sample. Time courses of the measured percentages were then interpolated in order to estimate the percentage of CD137-expressing cells at each hour of cultivation (at the same timepoints at which holographic information was captured).

D. Model Training and Validation

[0601] Supervised regression models for predicting the overall activation state of T cell populations were trained. The regression models were trained using time-matched (i) population-level statistics summarizing the cell feature measurements described in Section B and (ii) the estimated percentages of CD137-expressing cells described in Section C. For population-level statistics, the 0.01, 0.1, 0.5, 0.9, and 0.99 quantiles of cell feature measurements across cells were used for model training. Cell feature measurements were included during quantile calculation only for cells (i) classified as live, (ii) having radius mean cell feature measurements > 5, (iii) having intensity smoothness cell feature measurements < 0.03, (iv) having cell area cell feature measurements > 60, and (v) having circularity cell feature measurements > 0.5.

[0602] Several feature selection and regression modeling techniques were tested, including linear regression preceded by dimensionality reduction via principal components analysis; elastic net regression; boosted decision trees (software package “xgboost”); and random forests. For model validation, a leave-one-group-out cross-validation procedure was used in which per cycle of model training and testing, data from one of the 15 primary T cell compositions was held out and then used as validation data for model assessment.

[0603] The best predictive performance was achieved using boosted decision trees, results for which are shown in FIG. 1. FIG. 1 shows, for hold-out data from each of the subjects (data from each subject is designated by a unique symbol), the model-predicted percentages of CD137-expressing cells at each timepoint (y-axis) versus the measured percentages of CD137-expressing cells (x-axis), interpolated to those timepoints from daily flow cytometry measurements, as well as the best-fit regression line for all predicted percentages (y = 7.2 + 0.79 x, Radj² = 0.77). As shown in FIG. 1, model-predicted percentages were positively correlated with flow cytometry-based estimated percentages. The trained model achieved a root- mean-square deviation (RMSD) error of 8.2% on hold-out data.

[0604] The trained model was also assessed in order to determine the cell features that were most informative for predictive accuracy, primarily measured via RMSD and R² values. Gini index was used to assess cell feature informativeness. Based on this assessment, the cell features that were most informative for predictive accuracy included intensity maximum, intensity minimum, intensity entropy, intensity contrast, phase entropy, cellular area, and radius mean. Overall, cells predicted to be activated based on CD137 expression had higher intensity and were larger, more textured, and less circular, whereas cells predicted to be not activated based on CD 137 expression had lower intensity and were smaller, smoother, and more circular. Certain quantiles were informative as well. For example, the 0.90 quantile of the intensity contrast cell feature was ranked as most informative (> 20% relative importance).

[0605] These results validate that the overall activation state of T cell populations can be determined and monitored using holographic imaging during cultivation. In particular, the results show that CD 137 is a useful marker for predicting overall activation state. The results are also consistent with a finding that CD137 expression is associated with changes in the size of the activated T cells. Further, as demonstrated herein, this monitoring can be effected using a machine learning model trained using (i) population-level statistics summarizing individual cell feature measurements and (ii) percentages of cells expressing an exemplary activation marker, without having to label the individual cells.

[0606] While currently available methods to monitor and characterize the activation state of T cells can be time-consuming or require specialized reagents, instrumentation, or trained operators, the label-free method described herein can be used to monitor and characterize T cell activation using a non-destructive, non-damaging approach in which T cells from a bioreactor can be imaged and returned to the bioreactor in an undamaged state. The described method can be used to monitor the dynamics of T cell activation during cultivation without frequent cell sampling or arduous analytical techniques. This information can be used to assess the quality or batch-to-batch variability of T cells subjected to a manufacturing process, as well as to optimize the duration or other conditions of the manufacturing process or steps thereof in order to improve the quality or reduce the batch-to-batch variability of the processed T cells.

Example 2: Deep Learning Method for Determining The Overall Activation State of T Cell Populations Using Holographic Imaging

[0607] A deep learning method for determining the overall activation state of T cell populations directly from holographic images was developed. To do so, a deep multiple instance learning model was trained using the individual-cell holographic images and estimated percentages of CD137-expressing cells described in Example 1. Holographic images used for training included images only for cells (i) classified as live, (ii) having radius mean cell feature measurements > 5, (iii) having intensity smoothness cell feature measurements < 0.03, (iv) having cell area cell feature measurements > 60, and (v) having circularity cell feature measurements > 0.5.

[0608] The deep learning model was implemented based on the deep learning model described in Oner et al., 2022, Patterns 3: 100399 using an algorithm in which pooled “bags” of individual-cell holographic images associated with each imaging timepoint were used to train a deep learning regressor to predict the time-matched estimated percentages of CD 137- expressing cells from the bags of images. Specifically, the model included three components: a feature extractor module, a pooling filter, and a representation transformation module. The feature extractor module was implemented using a ResNetl8 convolutional neural network previously trained using 32x32-pixel images of segmented individual cells as input. After training, the ResNetl8 convolutional neural network received bags 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 ResNetl8 model. These extracted features were provided to the pooling filter, which calculated population-level statistics summarizing the extracted features across images in a bag. The calculated population-level statistics were the estimated marginal distributions of the extracted features across images in a bag. Finally, the population-level statistics for all bags were provided as input, along with the time-matched estimated percentages of CD 137 -expressing cells, to train the representation transformation module, implemented as a three-layer multi-layer perceptron. [0609] With this approach, using hold-out validation, the trained model achieved an RMSD error of 7.84%. Thus, an accurate, label-free approach for predicting the overall activation of T cell populations was developed. With this approach, the selection or design of individual cell features derived from the holographic images was not required prior to model training. Instead, individual cell features were automatically extracted from holographic images as part of model training.

[0610] In addition, the above method can be used for multiclass prediction using a representation transformation module with multiple output nodes for predicting expression of multiple cellular markers (e.g., CD137, CAR, and CD4).

Example 3: Determining The Overall Activation State of T Cell Populations Subjected to an Alternative Stimulation Process

[0611] The machine learning approach described in Example 1 was used to train and validate a model for predicting the overall activation state of T cell populations subjected to an alternative stimulation process.

[0612] Primary T cells from healthy human subjects were cryofrozen, thawed, and subjected to stimulation to activate the T cells. Specifically, the T cells were stimulated for approximately 24 hours (between 18 and 30 hours) at about 37 °C and 5% CO2 in the presence of recombinant IL-2, recombinant IL-7, recombinant IL- 15, and 4 pg/10⁶ cells of an anti-CD3/anti-CD28 stimulatory reagent prepared using an oligomeric streptavidin mutein reagent (WO 2018/197949; Poltorak et al., Scientific Reports (2020)). The oligomeric streptavidin mutein reagent had an average hydrodynamic radius of 90-120 nm and contained an average of 2000-2800 tetramers of a streptavidin mutein (Strep-Tactin® m2, SEQ ID NO: 21). The oligomeric streptavidin mutein reagent was mixed at room temperature with (i) an anti-CD3 Lab fragment individually fused at the carboxy-terminus of its heavy chain to a streptavidin-binding peptide sequence (Twin-Strep-tag®, SEQ ID NO: 15) and (ii) an anti- CD28 Lab fragment also individually fused at the carboxy-terminus of its heavy chain to a streptavidin-binding peptide sequence (Twin-Strep-tag®, SEQ ID NO: 15). The peptide- tagged Fab fragments were recombinantly produced (see International Patent App. Pub. Nos. WO 2013/011011 and WO 2013/124474). The anti-CD3 Fab fragment was derived from the 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 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 stimulatory reagent, 0.3 mg of oligomeric streptavidin mutein reagent, 0.5 pg of peptide-tagged anti-CD3 Fab fragments, and 0.5 pg of peptide-tagged anti-CD28 Fab fragment was used.

[0613] After stimulation, the T cells were engineered to express an exemplary recombinant protein, in this case a recombinant T cell receptor (TCR). Following engineering, the T cells were incubated for approximately 24 hours (between 18 and 30 hours) at about 37 °C and 5% CO2, after which the T cells were transferred to a rocking motion bioreactor and cultivated for expansion in serum- free media containing recombinant IL-2, IL-7, and IL- 15. The T cells were cultivated for expansion in the bioreactor for 5 days to 7 days before harvest.

[0614] This process was implemented for 16 primary T cell compositions from 16 human subjects. Cell feature measurements and activation marker expression estimates for each of the 16 primary T cell compositions were obtained during the cultivation step and used for model training and validation as described in Example 1. Holographic images used for training included images only for cells classified as live.

[0615] Several feature selection and regression modeling techniques were tested, including linear regression preceded by dimensionality reduction via principal components analysis; elastic net regression; boosted decision trees (software package “xgboost”); and random forests. For model validation, a leave-one-group-out cross-validation procedure was used in which per cycle of model training and testing, data from one of the 16 primary T cell compositions was held out and then used as validation data for model assessment.

[0616] The best predictive performance was achieved using boosted decision trees, results for which are shown in FIG. 2. FIG. 2 shows, for hold-out data from each of the subjects (data from each subject is designated by a unique symbol), the model-predicted percentages of CD137-expressing cells at each timepoint (y-axis) versus the measured percentages of CD137-expressing cells (x-axis), interpolated to those timepoints from daily flow cytometry measurements, as well as the best-fit regression line for all predicted percentages (y = 6.75 + 0.52 x, Radj² = 0.56). As shown in FIG. 2, model-predicted percentages were positively correlated with flow cytometry-based estimated percentages. The trained model achieved a root-mean-square deviation (RMSD) error of 7.9% on hold-out data.

[0617] The trained model was also assessed in order to determine the cell features that were most informative for predictive accuracy, primarily measured via RMSD and R² values. Shapley Additive Explanations (SHAP) values were used to assess cell feature informativeness. Based on this assessment, the cell features that were most informative 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.

[0618] These results further validate that the overall activation state of T cell populations can be determined and monitored using holographic imaging during cultivation.

Example 4: Machine Learning Method for Determining Recombinant Receptor Expression of T Cell Populations Using Holographic Imaging

[0619] The machine learning approach described in Example 1 was used to train and validate a model for predicting recombinant receptor expression in T cell populations.

[0620] For this model, data from 24 primary T cell compositions from 24 human subjects was used. Some of the primary T cell compositions underwent the stimulation process described in Example 1 and were engineered to express one of two exemplary chimeric antigen receptors (CARs) or an exemplary recombinant T cell receptor (TCR). The rest of the primary T cell compositions underwent the stimulation process described in Example 2 and were 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. Cell feature measurements and recombinant receptor expression estimates for each of the 24 primary T cell compositions were obtained during the cultivation step and used for model training and validation as described in Example 1. Holographic images used for training included images only for cells classified as live.

[0621] Several feature selection and regression modeling techniques were tested, including linear regression preceded by dimensionality reduction via principal components analysis; elastic net regression; boosted decision trees (software package “xgboost”); and random forests. For model validation, a leave-one-group-out cross-validation procedure was used in which per cycle of model training and testing, data from one of the 24 primary T cell compositions was held out and then used as validation data for model assessment.

[0622] The best predictive performance was achieved using boosted decision trees, results for which are shown in FIG. 3. FIG. 3 shows, for hold-out data from each of the subjects (data from each subject is designated by a unique symbol), the model-predicted percentages of recombinant receptor-expressing cells at each timepoint (y-axis) versus the measured percentages of recombinant receptor-expressing cells (x-axis), interpolated to those timepoints from daily flow cytometry measurements, as well as the best-fit regression line for all predicted percentages (y = 21.03 + 0.47 x, R_adj² = 0.2). As shown in FIG. 3, model- predicted percentages were positively correlated with flow cytometry-based estimated percentages. The trained model achieved a root-mean-square deviation (RMSD) error of 10.4% on hold-out data.

[0623] The trained model was also assessed in order to determine the cell features that were most informative for predictive accuracy, primarily measured via RMSD and R² values. SHAP values were used to assess cell feature informativeness. Based on this assessment, the cell features that were most informative for predictive accuracy included peak area (positively correlated with recombinant receptor expression), phase average uniformity (negatively correlated with recombinant receptor expression), intensity geometric mean (negatively correlated with recombinant receptor expression), minimum optical height (positively correlated with recombinant receptor expression), and normalized optical height (positively correlated with recombinant receptor expression).

[0624] These results validate that recombinant receptor expression of T cell populations can be determined and monitored using holographic imaging during cultivation. With the label-free method described herein, recombinant receptor expression can be monitored using a non-destructive, non-damaging approach in order to assess, for example, if T cells during cultivation have reached a threshold level of recombinant receptor expression prior to the T cells being harvested and, for instance, prepared for eventual patient administration. Example 5: Machine Learning Method for Determining the Earlier Memory Phenotypes of T Cell Populations Using Holographic Imaging

[0625] A machine learning method for determining the earlier memory phenotypes (e.g., stem cell memory and stem cell memory phenpotypes) of T cell populations using holographic imaging was developed. Earlier memory phenotypes have been reported to result in sustained in vivo response of CAR T cell therapies, given their proliferative capacity and effector capabilities in both hematologic and solid tumor environments (Gargett et al., 2019, Cytotherapy 21(6):593-602). Predicting CCR7 positivity can be used to understand the evolution of T cell phenotypes during manufacturing, without frequent sampling or arduous analytical techniques. The resulting information can then be further used to assess product quality and optimize the manufacturing process. The percentage of CCR7-positive cells during expansion can be used to understand the memory phenotype of the drug product and can inform the duration and/or the conditions of unit operations to ensure a high-quality consistent product. Additionally, this approach can enable multi-attribute T cell phenotype monitoring, which can be used to evaluate batch-to-batch variablitiy, enable the development of process control strategies that modify manufacturing conditions to minimize this variability in the final product.

[0626] To assess the earlier memory phenotype, T cell populations were stimulated to induce activation and then were periodically assessed, during their culviation in a bioreactor, at 24 hour intervals for the percentage of cells exhibiting surface expression of an exemplary T cell early memory phenotype marker, in this case CCR7. The percentage of cells expressing the CCR7 marker was quantified using flow cytometry.

[0627] To predict the flow cytometry results of CCR7 expression as a surrogate for the earlier memory phenotype, a machine learning method was trained using population-level statistics that summarized cell feature measurements across individual cells of the stimulated T cell populations. The cell feature measurements were derived from holographic information for the stimulated T cell populations that was captured during the cultivation.

[0628] In this approach, marker expression of individual cells was not required or used for model training, thereby obviating the need for any specialized equipment or reagents for capturing matched holographic and fluorescent images of individual cells. In addition, fluorescently labelled cells were not holographically imaged, the impact on cell feature measurements and model predictive accuracy of any random or systematic morphometric changes in labelled cells, compared to un-labelled cells, was avoided.

[0629] In addition, with this approach, the same holographic imaging system used for obtaining images for training the machine learning model can be used for obtaining images for predicting the earlier memory phenotype using the trained machine learning model. This is in contrast to other methods in which different imaging systems may be used for obtaining images for model training and use, for instance a dual fluorescence-holographic imaging system for model training and a holographic imaging-only system for model use.

[0630] In this manner, an accurate, label-free method for monitoring earlier T cell memory phenotype was developed.

A. Stimulation Process

[0631] Primary T cells from healthy human subjects were cryofrozen, thawed, and subjected to stimulation to activate the T cells. Specifically, the T cells were stimulated for approximately 24 hours (between 18 and 30 hours) at about 37 °C and 5% CO2 in the presence of recombinant IL-2, recombinant IL-7, recombinant IL- 15, and super-paramagnetic polystyrene-coated beads that were conjugated to anti-CD3 and anti-CD28 monoclonal antibodies (Dynabeads, Invitrogen). The T cells were stimulated at a 1:1 bead-to-cell ratio in serum-free media. After stimulation, the T cells were then transduced by spinoculation for 60 minutes with a lentiviral vector encoding an exemplary recombinant protein, in this case one of two exemplary chimeric antigen receptors (CARs). Following spinoculation, the T cells were incubated for approximately 24 hours (between 18 and 30 hours) at about 37 °C and 5% CO2, after which the T cells were transferred to a rocking motion bioreactor and cultivated for expansion in serum- free media containing recombinant IL-2, IL-7, and IL- 15. The T cells were cultivated for expansion in the bioreactor for 3 days to 5 days before harvest. The anti- CD3/anti-CD28 beads were removed from the T cells following harvest.

[0632] This process was implemented for 26 primary T cell compositions from 26 human subjects. Cell feature measurements and memory phenotype marker expression estimates for each of the 26 primary T cell compositions were obtained during the cultivation step and used for model training and validation, as described in Sections B-D below. B. Cell Feature Measurements

[0633] Holographic imaging of samples of the T cells was continuously performed using an exemplary digital holography microscopy (DHM) imaging system, in this case an Ovizio iLine F imaging system (Ovizio Imaging Systems NV/SA). The imaging system was connected to the bioreactor via a disposable tubing system. Every minute, a sample of the T cells was moved from the bioreactor, imaged using the DHM imaging system, and returned to the bioreactor.

[0634] Features of individual cells within each sample were derived from the captured holographic information. These features quantify individual cells imaged in terms of morphology, phase, and intensity texture, and are leveraged to predict the viability status of the cells within the culture, and providing an automated, label-free measurement of viable cell concentration during expansion. To derive these features from the captured holographic information, the captured information was segmented using OsOne software (Ovizio Imaging Systems NV/SA), which was then used to determine, per individual cell, the features listed above in Table El.

C. Memory Phenotype Marker Expression

[0635] During the same time period in which T cells were being holographically imaged, samples of T cells from the bioreactor were collected every 24 hours, and the percentage of CCR7-expressing activated T cells per sample was quantified using flow cytometry. The mixed population of T cells was exposed to purified IgG for saturation of the Fc receptors on the cells, preventing non-specific binding of the antibody of interest to other surface markers, such as CCR7.

[0636] Following this blocking step, antibody specific to CCR7 that is conjugated to a fluorophore, PE-Dazzle 594, catalog number 35326, was introduced to the samples. Cells that are positive for this surface marker will bind to the antibody and thus be fluorescently tagged. After an incubation period, the purified IgG and anti-CCR7 antibody was removed from the samples and the cells were fixed using BD Cytofix/Cytoperm solution following appropriate cell washing and fixing protocol steps. Flow cytometry data was acquired using the BD Fortessa Dual Symphony instrument and BD FACSDiva software. Cells from each sample are interrogated by lasers, promoting excitation of the fluorophore bound to the surface of the cell and emission of light signal. Laser voltages were set prior to data acquisition using rainbow calibration beads each time samples were run. Cell populations were then analyzed based on fluorescent signal to determine the fraction positive for CCR7 using BD Biosciences FACSDiva™ software and flow analysis software (FlowJo vl0.8). Time courses of the measured percentages were then interpolated, to estimate the percentage of CCR7-expressing cells at each hour of cultivation (at the same timepoints at which holographic information was captured). This was done for each sample collected daily during expansion and repeated across multiple healthy donors. The dataset generated for training the machine learning model included both donors transduced with a CAR construct and donors gene-edited through electroporation followed by transduction.

D. Model Training and Validation

[0637] Flow measurements involving the cellular percentage (%) positivity of CCR7, which were collected daily, were interpolated to obtain hourly measurements at the times which iLineF sampling occurred. At each iLineF sampling timepoint, using Ovizio OsOne software, standard cell features were extracted from the iLineF phase and intensity images, and filtered to remove dead cells. The average feature values were used as input to feature selection techniques and to then train a regression model to predict interpolated flow measurements. For model validation, leave-one-out sample cross validation was performed to evaluate the predictive accuracy of the model on data that was not included in training the model.

[0638] Supervised regression models for predicting the earlier memory phenotype of T cell populations were trained. The regression models were trained using time-matched (i) population-level statistics summarizing the cell feature measurements described in Section B and (ii) the estimated percentages of CCR7-expressing cells described in Section C. The training data comprised of samples transduced with a CAR construct, as well as samples that were gene-edited through electroporation, followed by transduction. For population-level statistics, the 0.01, 0.1, 0.5, 0.9, and 0.99 quantiles of cell feature measurements across cells were used for model training. Cell feature measurements were included during quantile calculation only for cells classified as live.

[0639] Several feature selection and regression modeling techniques were tested, including linear regression preceded by dimensionality reduction via principal components analysis; elastic net regression; boosted decision trees (software package “xgboost”); and random forests.

[0640] The best predictive performance was achieved using boosted decision trees, results for which are shown in FIG. 4. FIG. 4 shows, for hold-out data from each of the subjects (data from each subject is designated by a unique symbol), the model-predicted percentages of CCR7-expressing cells at each timepoint (y-axis) versus the measured percentages of CCR7-expressing cells (x-axis), interpolated to those timepoints from daily flow cytometry measurements, as well as the best-fit regression line for all predicted percentages (y = 30.2 + 0.62 x, R² = 0.61). As shown in FIG. 4, model-predicted percentages were positively correlated with flow cytometry-based estimated percentages for CCR7 positivity. The trained model was able to predict the cellular percentage (%) positivity of CCR7 at each iLineF sampling timepoint with a root- mean- square deviation (RMSD) error of 9.6% on hold-out data (FIG. 4).

[0641] The trained model was also assessed to determine the cell features that were most informative for predictive accuracy, primarily measured via RMSD and R² values. Gini index was used to assess cell feature informativeness. Based on this assessment, the cell features that were most informative for predictive accuracy included cell area, intensity mean, perimeter, equivalent peak diameter, and peak area normalized. Overall, cells predicted to be of an earlier memory phenotype based on CCR7 expression had larger features related to cell size (area, perimeter, diameter), smaller mean intensity features, larger phase correlation features, and smaller radius variance normalized features.

[0642] These results validate that the earlier memory phenotype of T cell populations can be determined and monitored using holographic imaging during cultivation. In particular, the results show that CCR7 is a useful marker for predicting earlier memory phenotype. The results are also consistent with a finding that CCR7 expression is associated with changes in the size of the activated T cells. Further, as demonstrated herein, this monitoring can be affected using a machine learning model trained using (i) population-level statistics summarizing individual cell feature measurements and (ii) percentages of cells expressing an exemplary memory phenotype marker, without having to label the individual cells.

[0643] Current methods to characterize CCR7 positivity are time-consuming, requiring specialized reagents, instrumentation and trained operators. The label-free classification method as described in this Example can be used to characterize cellular material using a non-destructive, non-damaging approach. Moreover, the cell population can be imaged and returned to the bioreactor in an undamaged state.

[0644] The present invention is not intended to be limited in scope to the particular disclosed embodiments, which are provided, for example, to illustrate various aspects of the invention. Various modifications to the compositions and methods described will become apparent from the description and teachings herein. Such variations may be practiced without departing from the true scope and spirit of the disclosure and are intended to fall within the scope of the present disclosure.

Sequences

Claims

Claims What is claimed is:

3. The method of claim 2, wherein the method comprises determining the plurality of population-level statistics from the one or more input measures for each of the plurality of cellular features.

4. The method of claim 2 or claim 3, wherein the method comprises determining the one or more input measures for each of the plurality of cellular features from the holographic information.

5. The method of any one of claims 2-4, wherein the method comprises obtaining the holographic information.

6. The method of any one of claims 1-5, wherein the holographic information is obtained by differential digital holographic microscopy (DDHM).

7. The method of any one of claims 4-6, wherein the obtaining the holographic information comprises imaging the population of cells using DDHM.

8. The method of any one of claims 1-7, wherein the one or more populationlevel statistics for at least one, optionally each, of the plurality of cellular features comprise one or more quantiles of the one or more input measures of the cellular feature.

9. The method of claim 8, wherein the one or more quantiles comprise one or more of the 0.01, 0.1, 0.5, 0.9, and 0.99 quantiles of the one or more input measures of the cellular feature.

10. The method of any one of claims 1-7, wherein the one or more populationlevel statistics for at least one, optionally each, of the plurality of cellular features are determined by applying a distribution-based pooling filter to the one or more input measures of the cellular feature.

11. The method of any one of claims 1-10, wherein the population-level output measure is the number of cells, the percentage of cells, the proportion of cells, or the density of cells of the population of cells that express the marker.

12. The method of any one of claims 1-11, wherein the population of cells is from a culture of cells being cultured in vitro or ex vivo.

13. The method of claim 12, wherein the holographic information is obtained during the in vitro or ex vivo culture of the culture of cells.

14. The method of claim 12 or claim 13, wherein the population-level output measure is of expression of the marker during the in vitro or ex vivo culture of the culture of cells.

15. The method of any one of claims 12-14, wherein the in vitro or ex vivo culture is under conditions to expand T cells of the culture of cells.

16. The method of any one of claims 12-15, wherein the in vitro or ex vivo culture is in a bioreactor.

17. The method of any one of claims 1-16, wherein the population of cells are incubated under T cell stimulating conditions prior to when the holographic information is obtained.

18. The method of any one of claims 1-17, wherein the method comprises incubating the population of cells under T cell stimulating conditions prior to when the holographic information is obtained.

19. The method of claim 17 or claim 18, wherein the incubation is prior to the in vitro or ex vivo culture.

20. The method of any one of claims 17-19, wherein the T cell stimulating conditions comprise incubation in the presence of T cell stimulatory agents that induce a primary activation signal and a costimulatory signal in T cells.

21. The method of claim 20, wherein the T cell stimulatory agents comprise an anti-CD3 antibody or antibody fragment.

22. The method of claim 20 or claim 21, wherein the T cell stimulatory agents comprise an anti-CD28 antibody or antibody fragment.

23. The method of any one of claims 20-22, wherein the T cell stimulatory agents are immobilized on a bead.

24. The method of any one of claims 20-22, wherein the T cell stimulatory agents are immobilized on an oligomeric streptavidin mutein reagent.

25. The method of any one of claims 1-24, wherein a recombinant receptor is introduced into T cells of the population of cells prior to when the holographic information is obtained.

26. The method of any one of claims 1-25, wherein the method comprises introducing a recombinant receptor into T cells of the population of cells prior to when the holographic information is obtained.

27. The method of claim 25 or claim 26, wherein the introducing comprising contacting the population of cells with an agent comprising a polynucleotide encoding the recombinant receptor.

28. The method of claim 26 or claim 27, wherein the recombinant receptor is a chimeric antigen receptor (CAR) or an engineered T cell receptor (TCR).

29. 'fhe method of any one of claims 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 population of cells are T cells.

30. The method of any one of claims 1-29, wherein the population-level output measure is determined by providing the plurality of population-level statistics as input to a machine learning model trained to predict population-level output measures of expression of the marker based on population-level statistics of the plurality of cellular features.

31. The method of claim 30, wherein the machine learning model is trained using a dataset of reference population-level statistics, wherein: for each of a first plurality of reference populations of cells comprising T cells, the dataset of reference population-level statistics comprises one or more reference populationlevel statistics for each of the plurality of cellular features, wherein: each reference population-level statistic is of one or more reference input measures for a cellular feature of the plurality of cellular features derived from holographic information obtained for the reference population of cells; and each reference input measure is from an individual cell of the reference population of cells.

32. The method of claim 31, wherein the machine learning model is trained using a dataset of reference population-level output measures, wherein for each of a second plurality of reference populations of cells, the dataset of reference population-level output measures comprises a reference population-level output measure of expression of the marker for the reference population of cells, wherein: the first and second pluralities of reference populations of cells are from reference cultures of cells being cultured in vitro or ex vivo', and reference populations of cells from the first plurality of reference populations of cells are from the same reference culture of cells as reference populations of cells from the second plurality of reference populations of cells.

34. The method of any one of claims 1-33, wherein the plurality of cellular features comprise one or of intensity skewness, intensity correlation, intensity homogeneity, intensity maximum, intensity minimum, intensity entropy, intensity contrast, phase entropy, cell area, and radius mean.

35. The method of any one of claims 1-34, wherein the plurality of cellular features comprises intensity maximum, intensity minimum, intensity entropy, intensity contrast, phase entropy, cell area, and radius mean.

37. The method of any one of claims 1-36, wherein the plurality of cellular features comprises intensity skewness, intensity correlation, and intensity homogeneity.

39. The method of claim 38, wherein the plurality of cellular features comprises intensity maximum, intensity minimum, intensity entropy, intensity contrast, phase entropy, cell area, and radius mean.

40. The method of any one of claims 1-33, wherein the plurality of cellular features comprises one or more of peak area, phase average uniformity, intensity geometric mean, minimum optical height, and normalized optical height.

41. The method of any one of claims 1-33 and 40, wherein the plurality of cellular features comprises peak area, phase average uniformity, intensity geometric mean, minimum optical height, and normalized optical height.

42. The method of any one of claims 1-33, wherein the plurality of cellular features comprises one or more of cell area, perimeter, mean intensity, normalized peak area, and equivalent peak diameter.

43. The method of any one of claims 1-33 and 42, wherein the plurality of cellular features comprises cell area, perimeter, mean intensity, normalized peak area, and equivalent peak diameter.

45. The method of claim 44, wherein the plurality of features comprises cell area, perimeter, mean intensity, normalized peak area, and equivalent peak diameter.

47. The method of any one of claims 1-33, wherein the one or more input measures for at least one, optionally each, of the plurality of cellular features are determined by providing the holographic information for individual cells of the population of cells to a convolutional neural network, wherein: the plurality of cellular features are cellular features extracted by the convolutional neural network; and the one or more input measures are determined from the convolutional neural network.

48. The method of any one of claims 31-33 and 47, wherein the one or more reference input measures for at least one, optionally each, of the plurality of cellular features are determined by providing the holographic information for individual cells of the reference population of cells to a convolutional neural network, wherein: the plurality of cellular features are cellular features extracted by the convolutional neural network; and the one or more reference input measures are determined from the convolutional neural network.

49. The method of claim 47 or claim 48, wherein the convolutional neural network is trained using a dataset of reference holographic information, wherein for each of a first plurality of reference populations of cells comprising T cells, the dataset of reference holographic information comprises holographic information obtained for individual cells of the reference population of cells.

50. The method of any one of claims 31-35, 37, 40-43, and 47-49, wherein the one or more reference population-level statistics for at least one, optionally each, of the plurality of cellular features comprise one or more quantiles of the one or more reference input measures for the cellular feature.

51. The method of claim 50, wherein the one or more quantiles are selected from the 0.01, 0.1, 0.5, 0.9, and 0.99 quantiles of the one or more reference input measures for the cellular feature.

52. The method of any one of claims 31-35, 37, 40-43, and 47-49, wherein the one or more reference population-level statistics for at least one, optionally each, of the plurality of cellular features are determined by applying a distribution-based pooling filter to the one or more reference input measures for the cellular feature.

53. A method for training a machine learning model that predicts a cell phenotype of a population of T cells, comprising:

(d) training a machine learning model using the dataset of reference population-level statistics and a dataset of reference population-level output measures, wherein for each of a second plurality of reference populations of cells, the dataset of reference population-level output measures comprises a reference population-level output measure of expression of a marker expressed by T cells with a cell phenotype of interest for the reference population of cells, wherein: the first and second pluralities of reference populations of cells are from reference cultures of cells being cultured in vitro or ex vivo', and reference populations of cells from the first plurality of reference populations of cells are from the same reference culture of cells as reference populations of cells from the second plurality of reference populations of cells; whereby the machine learning model is trained to predict population-level output measures of expression of the marker based on population-level statistics of the plurality of cellular features.

54. The method of any one of claims 47-53, wherein the one or more input measures for at least one, optionally each, of the plurality of cellular features are obtained from a fully connected layer of the convolutional neural network.

55. The method of any one of claims 31-35, 37, 40-43, and 47-54, wherein the holographic information for the first plurality of reference populations of cells is obtained by DDHM.

56. The method of any one of claims 1-55, wherein the holographic information comprises phase information and intensity information.

57. The method of any one of claims 32-35, 37, 40-43, and 47-56, wherein the reference population-level output measure is the number of cells, the percentage of cells, the proportion of cells, or the density of cells of the reference population of cells that express the marker.

58. The method of any one of claims 32-35, 37, 40-43, and 47-57, wherein the dataset of reference population-level statistics and the dataset of reference population-level output measures are time-matched.

59. The method of any one of claims 32-35, 37, 40-43, and 47-58, wherein the in vitro or ex vivo culture of the reference cultures of cells is performed under the same or similar conditions as the in vitro or ex vivo culture of the culture of cells.

60. The method of any one of claims 32-35, 37, 40-43, and 47-59, wherein the holographic information for the first plurality of reference populations of cells is obtained during the in vitro or ex vivo culture of the reference cultures of cells.

61. The method of any one of claims 32-35, 37, 40-43, and 47-60, wherein the dataset of reference population-level output measures are of expression of the marker during or after the in vitro or ex vivo culture of the reference cultures of cells.

62. The method of any one of claims 32-35, 37, 40-43, and 47-61, wherein the dataset of reference population-level output measures is determined using fluorescence imaging of the second plurality of reference populations of cells.

63. The method of claim 62, wherein the fluorescence imaging is by flow cytometry.

64. The method of any one of claims 1-35, 37, and 47-63, wherein the method is for determining the activation state of the population of T cells, and the marker is expressed by activated T cells.

65. The method of any one of claims 1-35, 37, and 47-64, wherein the marker is CD137 (4-1BB).

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

67. The method of claim 66, wherein the marker is expressed by T cells having a central memory phenotype.

68. The method of claim 66, wherein the marker is expressed by T cells having a stem cell memory phenotype.

69. The method of any one of claims 1-35, 42-63, and 66-68, wherein the marker is CCR7.

70. The method of any one of claims 1-33 and 40-63, wherein the method is for determining recombinant receptor expression of the population of T cells, and the marker is a recombinant receptor introduced into T cells of the population of cells prior to when the holographic information is obtained.

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

72. The method of any one of claims 31-35, 37, 40-43, and 47-71, wherein the first plurality of reference populations of cells are incubated under T cell stimulating conditions prior to when the holographic information for the first plurality of reference populations of cells is obtained.

73. The method of claim 72, wherein the incubation of the first plurality of reference populations of cells is performed under the same or similar conditions as the incubation of the population of cells.

74. The method of claim 72 or claim 73, wherein the incubation of the first plurality of reference populations of cells is prior to the in vitro or ex vivo culture of the reference cultures of cells.

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

76. The method of any one of claims 1 -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 populations of cells are T cells.

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

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

79. The computing device of claim 78, wherein the computing device further comprises in memory a machine learning model trained according to the method of any one of claims 33-35, 37, 40-43, and 47-76, wherein the population-level output measure for the population of cells is determined using the machine learning model.