WO2022074537A1

WO2022074537A1 - Biomarkers of tumor response to cdk4/6 inhibitors

Info

Publication number: WO2022074537A1
Application number: PCT/IB2021/059086
Authority: WO
Inventors: Jadwiga Renata BIENKOWSKA; Xinmeng MU; Zhou Zhu
Original assignee: Pfizer Inc.
Priority date: 2020-10-07
Filing date: 2021-10-04
Publication date: 2022-04-14

Abstract

This invention relates to biomarkers, gene signatures, and methods for the selection and treatment of patients having HR+, HER2- breast cancer who are likely to benefit from administration of a CDK4/6 inhibitor such as palbociclib.

Description

BIOMARKERS OF TUMOR RESPONSE TO CDK4/6 INHIBITORS

BACKGROUND OF THE INVENTION

Field of the Invention

The invention relates to biomarkers of tumor response, gene signatures, and methods for the selection and treatment of subjects having hormone receptor positive (HR+), human epidermal growth factor receptor 2 negative (HER2-) breast cancer who are likely to benefit from administration of a CDK4/6 inhibitor.

Description of Related Art

The cell cycle mechanism has been exploited by therapeutic agents that impair disparate aspects of cell division, including DNA synthesis, microtubule assembly, and DNA damage. More recently, targeted approaches, including CDK4/6 inhibitors such as palbociclib, have been developed. In clinical trials, palbociclib was found to significantly improve progression free survival (PFS) in hormone receptor positive (HR+), human epidermal growth factor receptor 2 negative (HER2-) advanced or metastatic breast cancer when combined with endocrine therapy. Palbociclib, ribociclib and abemaciclib have been approved for treatment of hormone receptor (HR)-positive, human epidermal growth factor receptor 2 negative (HER2-) advanced or metastatic breast cancer in combination with aromatase inhibitors, such as letrozole, in a first line setting and with fulvestrant in second or later lines of therapy in certain patients. (O’Leary et al. Treating cancer with selective CDK4/6 inhibitors. Nature Reviews (2016) 13:417-430).

While disease progression has been found to be significantly delayed by the addition of a CDK4/6 inhibitor to an anti-hormonal therapy, many patients ultimately progress after a period of stable disease/remission or do not respond to the combination treatment, representing acquired and inherent resistance, respectively. Resistance to anti-hormonal therapies have been partially attributed to non-synonymous mutations in the ligand binding domain of the ESR1 gene resulting in ligand independent activity of the mature transcription factor. See Robinson et al., Activating ESR1 mutations in hormone-resistant metastatic breast cancer. Nat Genet 45:1446-1451 , 2013; Toy et al., ESR1 ligand-binding domain mutations in hormone-resistant breast cancer. Nat Genet 45:1439-1445, 2013. Other resistance mechanisms observed for anti-hormonal therapy in breast cancer include upregulation of HER or FGFR family signaling and induction of the PI3K/AKT/MTOR pathway. See Giltnane et al., Genomic profiling of ER(+) breast cancers after short-term estrogen suppression reveals alterations associated with endocrine resistance. Sci Transl Med 9, 2017; Higgins and Baselga, Targeted therapies for breast cancer. J Clin Invest 121 :3797-3803, 2011.

A univariate analysis of 2,534 cancer-related genes was conducted in archival tumor samples from the PALOMA-3 trial (ClinicalTrials.gov identifier: NCT01942135), using the EdgeSeq Oncology Biomarker Panel (OBP) panel (HTG Molecular Diagnostics) for mRNA profiling. The association of gene expression with the effect of palbociclib on progression-free survival (PFS) was evaluated using Cox proportional hazards (Cox PH) regression analysis, with gene expression as a continuous variable or dichotomized by median. The analysis identified cyclin E1 (CCNE1) mRNA expression as strongly associated with palbociclib efficacy in previously treated HR+, HER2- metastatic breast cancer patients who failed on prior endocrine therapy. Patients with high baseline CCNE1 mRNA expression were less responsive to the combination of palbociclib plus fulvestrant relative to patients with low CCNE1 mRNA expression, implicating CDK2 activity in determining relative resistance to palbociclib in the second- line clinical setting. See Turner et al., Cyclin E1 Expression and Palbociclib Efficacy in Previously Treated Hormone Receptor-Positive Metastatic Breast Cancer. J. Clin. Oncol. 37:1169-1178, 2019.

However, CCNE1 mRNA expression level alone could not significantly stratify patient response from the PALOMA-2 trial (ClinicalTrials.gov identifier: NCT01740427), in patients with estrogen receptor positive (ER+), HER2- advanced breast cancer who had not received prior endocrine therapy for their advanced disease. These first-line patients were randomly assigned to receive palbociclib plus letrozole or placebo plus letrozole. See Finn et al., Biomarker Analyses of Response to Cyclin-Dependent Kinase 4/6 Inhibition and Endocrine Therapy in Women with Treatment-Naive Metastatic Breast Cancer. Clin Cancer Res 26: 110-121 , 2020.

The identification of multivariate gene expression signatures predictive of response to treatment with CDK4/6 inhibitors in HR+, HER2- breast cancer across multiple lines of therapy are needed.

BRIEF SUMMARY OF THE INVENTION

This invention relates to biomarkers of tumor response, gene signatures, and methods for the selection and treatment of subjects having hormone receptor positive (HR+), human epidermal growth factor receptor 2 negative (HER2-) breast cancer who are likely to benefit from administration of a CDK4/6 inhibitor. In one aspect, the invention provides a method of treating hormone receptorpositive (HR+), human epidermal growth factor receptor 2-negative (HER2-) breast cancer in a subject in need thereof, comprising:

(a) determining the expression level for each gene in a gene signature in a tumor sample from the subject;

(b) computing a composite score for the gene signature as a weighted sum of the expression levels for each gene in the gene signature;

(c) selecting the subject for treatment with a cyclin dependent kinase 4/6 (CDK 4/6) inhibitor and an endocrine therapeutic agent when the composite score is less than the median of a reference composite score computed for the gene signature in a reference breast cancer population; and

(d) administering to the selected subject an amount of a CDK4/6 inhibitor and an amount of an endocrine therapeutic agent, wherein the amounts together are effective to treat cancer; wherein the gene signature comprises five or more genes selected from the group consisting of: ABCB11 , BUB1 B, CCNE1 , CDH15, CDKN2D, COL11A1 , CPT1A, ERG, LRP5, MAP3K1 , MT1X, SLC38A2, STC2, TFDP2, TMPRSS2, and TRIB1.

In another aspect, the invention provides a method of selecting a subject having hormone receptor-positive (HR+), human epidermal growth factor receptor 2-negative (HER2-) breast cancer for treatment with a CDK4/6 inhibitor and an endocrine therapeutic agent, comprising:

(b) computing a composite score for the gene signature as a weighted sum of the expression levels for each gene in the gene signature; and

(c) selecting the subject for treatment when the composite score is less than the median of a reference composite score computed for the gene signature in a reference breast cancer population; wherein the gene signature comprises five or more genes selected from the group consisting of: ABCB11 , BUB1 B, CCNE1 , CDH15, CDKN2D, COL11A1 , CPT1A, ERG, LRP5, MAP3K1 , MT1X, SLC38A2, STC2, TFDP2, TMPRSS2, and TRIB1.

In another aspect, the invention provides a method of predicting whether a subject having hormone receptor-positive (HR+), human epidermal growth factor receptor 2-negative (HER2-) breast cancer will respond to treatment with a CDK4/6 inhibitor and an endocrine therapeutic agent, comprising: (a) determining the expression level for each gene in a gene signature in a tumor sample from the subject;

(c) comparing the composite score to a reference composite score computed for the gene signature in a reference breast cancer population; and

(d) predicting the subject will respond to treatment when the composite score is less than the median of the reference composite score; wherein the gene signature comprises five or more genes selected from the group consisting of: ABCB11 , BUB1 B, CCNE1 , CDH15, CDKN2D, COL11A1 , CPT1A, ERG, LRP5, MAP3K1 , MT1X, SLC38A2, STC2, TFDP2, TMPRSS2, and TRIB1.

In preferred embodiments of each of the aspects and embodiments described herein, the CDK4/6 inhibitor is palbociclib, or a pharmaceutically acceptable salt thereof.

In some embodiments of each of the aspects herein, the endocrine therapeutic agent is selected from the group consisting of an aromatase inhibitor, a selective estrogen receptor degrader (SERD), and a selective estrogen receptor modulator (SERM). In some embodiments, the endocrine therapeutic agent is selected from the group consisting of letrozole, anastrozole, exemestane and fulvestrant.

Each of the aspects and embodiments of the present invention described herein may be combined with one or more other embodiments of the present invention which is not inconsistent with the embodiment(s) with which it is combined.

BRIEF DESCRIPTION OF THE SEVERAL VIEWS OF THE DRAWINGS

FIG. 1 shows an 11 -gene exploratory signature from PALOMA-3 using the elastic net algorithm. Bar length (top x-axis) reflects the proportion of 100 bootstrap models containing the gene, while diamond points (bottom x-axis) reflect the average non-zero coefficient across bootstrap models.

FIG. 2 shows a hierarchical clustering heatmap of the 11 -gene exploratory signature and progression-free survival (PFS). Shown are patients treated with palbociclib + fulvestrant (A) or placebo + fulvestrant (B), in each case sorted by PFS excluding those with a follow-up time of <12 months and no event. Clustering was conducted using 1 -Pearson correlation as distance metric and average linkage.

FIG. 3 shows the association between the 11 -gene exploratory signature and clinical benefit (PFS) in metastatic samples from PALOMA-3 (A) and in the full data set from PALOMA-3 (B). Patients were dichotomized by median signature score into low (dashed line) and high (solid line) score groups. Also shown are hazard ratios for each gene in the 11 -gene exploratory signature using the full data set from PALOMA-3 (C).

FIG. 4 shows the association between the 11 -gene exploratory signature and clinical benefit (PFS) from the addition of palbociclib in an independent metastatic breast cancer cohort from PALOMA-2. Patients were dichotomized by median signature score into low (dashed line) and high (solid line) score groups.

FIG. 5 shows a 9-gene exploratory signature from PALOMA-3 using the elastic net algorithm. Bar length (top x-axis) reflects the proportion of 1 ,000 bootstrap models containing the gene, while circular points (bottom x-axis) reflect the average non-zero coefficient across bootstrap models.

FIG. 6 shows the association between the 9-gene exploratory signature and clinical benefit (PFS) in the full data set from PALOMA-3 (A) and in an independent metastatic breast cancer cohort from PALOMA-2 (B). Patients were dichotomized by median signature score into low (dashed line) and high (solid line) score groups.

FIG. 7 shows a 13-gene exploratory signature from PALOMA-3 using the elastic net algorithm. Bar length (top x-axis) reflects the proportion of 1 ,000 bootstrap models containing the gene, while circular points (bottom x-axis) reflect the average non-zero coefficient across bootstrap models.

FIG. 8 shows the association between the 13-gene exploratory signature and clinical benefit (PFS) in the full data set from PALOMA-3 after filtering out outliers from NMF liver factor (A), and in an independent metastatic breast cancer cohort from PALOMA-2 (B). Patients were dichotomized by median signature score into low (dashed line) and high (solid line) score groups.

FIG. 9 shows identification of samples with significant liver component in PALOMA-3 data set before (A) and after filtering out outliers from NMF liver factor (B).

DETAILED DESCRIPTION OF THE INVENTION

The present invention may be understood more readily by reference to the following detailed description of the embodiments of the invention and the Examples included herein. It is to be understood that the terminology used herein is for the purpose of describing specific embodiments only and is not intended to be limiting. It is further to be understood that unless specifically defined herein, the terminology used herein is to be given its traditional meaning as known in the relevant art.

The invention described herein may be suitably practiced in the absence of any element(s) not specifically disclosed herein. Thus, for example, in each instance herein any of the terms "comprising", "consisting essentially of", and "consisting of" may be replaced with either of the other two terms.

As used herein, the singular form "a", "an", and "the" include plural references unless indicated otherwise. For example, "a" substituent includes one or more substituents.

The term "about" means having a value falling within an accepted standard of error of the mean, when considered by one of ordinary skill in the art, typically such as ± 10%.

The term “administration” or “treatment” as it applies to an animal, human, experimental subject, cell, tissue, organ or biological fluid, refers to contact of an exogenous pharmaceutical, therapeutic or diagnostic agent, or composition, to the animal, human, experimental subject, cell, tissue, organ or biological fluid. Treatment of a cell encompasses contact of a reagent to the cell, as well as contact of a reagent to a fluid, where the fluid is in contact with the cell. “Administration” and “treatment” may also relate to in vitro and ex vivo treatment, e.g., of a cell, by a reagent, diagnostic, binding compound, or by another cell.

The term “cancer”, “cancerous”, “malignant” refer to or describe the physiological condition in mammals that is typically characterized by unregulated cell growth. As used herein “cancer” refers to any malignant and/or invasive growth or tumor caused by abnormal cell growth. As used herein “cancer” refers to solid tumors named for the type of cells that form them, cancer of blood, bone marrow, or the lymphatic system. Examples of solid tumors include but not limited to sarcomas and carcinomas. Examples of cancers of the blood include but not limited to leukemias, lymphomas and myeloma. The term “cancer” includes but is not limited to a primary cancer that originates at a specific site in the body, a metastatic cancer that has spread from the place in which it started to other parts of the body, a recurrence from the original primary cancer after remission, and a second primary cancer that is a new primary cancer in a person with a history of previous cancer of a different type from latter one. A subject may be identified as having de novo metastatic disease or after progression from an earlier-identified cancer.

The term “patient” or “subject” refer to any single subject for which therapy is desired or that is participating in a clinical trial, epidemiological study or used as a control, including humans and mammalian veterinary patients such as cattle, horses, dogs and cats. In preferred embodiments, the subject is a human. In some embodiments, the subject is a postmenopausal woman or a man. In other embodiments, the subject is a pre- or perimenopausal woman treated with a luteinizing hormone releasing hormone (LNRH) agonist, such as goserelin, so that their ovarian function is suppressed. Subjects may be treatment naive (i.e., the subject has not received prior treatment for advanced disease if metastatic, or for early disease if diagnosed with early breast cancer) or may have received one or more prior lines of treatment (i.e., in second or later line settings), such as one or more endocrine therapeutic agents or chemotherapeutic agents. In some embodiments, the subject is treated with an aromatase inhibitor as their initial endocrine therapeutic agent (i.e., as initial endocrine based therapy). In other embodiments, the subject has disease progression on or after treatment with an endocrine therapeutic agent in an adjuvant or metastatic setting.

The term “treat” or “treating” a cancer as used herein means to administer a combination therapy comprising a CDK4/6 inhibitor and an endocrine therapeutic agent, according to the present invention to a subject having cancer, or diagnosed with cancer, to achieve at least one positive therapeutic effect, such as, for example, reduced number of cancer cells, reduced tumor size, reduced rate of cancer cell infiltration into peripheral organs, or reduced rate of tumor metastases or tumor growth, reversing, alleviating, inhibiting the progress of, or preventing the disorder or condition to which such term applies, or one or more symptoms of such disorder or condition. The term "treatment", as used herein, unless otherwise indicated, refers to the act of treating as "treating" is defined immediately above. The term “treating” also includes adjuvant and neo-adjuvant treatment of a subject.

A “reference breast cancer population” refers to a population of individuals having HR+, HER2- breast cancer of which the subject is a member. For example, the reference breast cancer population could refer to a cohort of patients in a clinical trial enrolling the same or similar stage(s) of breast cancer as the subject. Alternatively, the reference breast cancer population could refer to a relevant population of individuals who have been successfully treated with a CDK4/6 inhibitor and endocrine therapy. For example, the reference breast cancer population could refer to the cohorts from the PALOMA-2 or PALOMA-3 studies for advanced or metastatic breast cancer, or to a clinical trial cohorts from a relevant early breast cancer trial.

A “control population” refers to a population of individuals who do not have cancer but are otherwise matched to the subject. The skilled person will be able to select an appropriate control population to provide the requisite reference value. For the purposes of this invention, beneficial or desired clinical results include, but are not limited to, one or more of the following: reducing the proliferation of (or destroying) neoplastic or cancerous cell; inhibiting metastasis or neoplastic cells; shrinking or decreasing the size of a tumor; remission of the cancer; decreasing symptoms resulting from the cancer; increasing the quality of life of those suffering from the cancer; decreasing the dose of other medications required to treat the cancer; delaying the progression of the cancer; curing the cancer; overcoming one or more resistance mechanisms of the cancer; and/or prolonging survival of patients the cancer. Positive therapeutic effects in cancer can be measured in a number of ways (see, for example, W. A. Weber, Assessing tumor response to therapy, J. Nucl. Med. 50 Suppl. 1 : 1 S-1 OS (2009). For example, with respect to tumor growth inhibition (T/C), according to the National Cancer Institute (NCI) standards, a T/C less than or equal to 42% is the minimum level of anti-tumor activity. A T/C <10% is considered a high anti-tumor activity level, with T/C (%) = median tumor volume of the treated I median tumor volume of the control x 100.

In some embodiments, the treatment achieved by a combination of the invention is defined by reference to any of the following: partial response (PR), complete response (CR), overall response (OR), progression free survival (PFS), disease free survival (DFS), invasive disease free survival (iDFS), and overall survival (OS). PFS, also referred to as “Time to Tumor Progression” indicates the length of time during and after treatment that the cancer does not grow and includes the amount of time patients have experienced a CR or PR, as well as the amount of time patients have experienced stable disease (SD). DFS refers to the length of time during and after treatment that the patient remains free of disease. iDFS is defined according to Hudis (J Clin Oncol 2007) as the time between randomization and first event (e.g., ipsi- or contralateral invasive inbreast or loco-regional recurrence, distant recurrence, death from breast cancer, death from non-breast cancer cause, death from unknown cause, invasive contralateral breast cancer, second primary invasive cancer (non-breast)). OS refers to a prolongation in life expectancy as compared to naive or untreated subjects or patients. In some embodiments, response to a combination of the invention is any of PR, CR, PFS, DFS, OR or OS that is assessed using Response Evaluation Criteria in Solid Tumors (RECIST) 1.1 response criteria.

The treatment regimen that is effective to treat a cancer patient may vary according to factors such as the disease state, age, and weight of the patient, and the ability of the therapy to elicit an anti-cancer response in the subject. While the aspects of the present invention may not be effective in achieving a positive therapeutic effect in every subject, it should do so in a statistically significant number of subjects as determined by any statistical test known in the art such as the Student’s t-test, the chi2- test the ll-test according to Mann and Whitney, the Kruskal-Wallis test (H-test), Jonckheere-Terpstrat-testy and the Wilcon on-test.

The terms “treatment regimen”, “dosing protocol” and “dosing regimen” may be used interchangeably to refer to the dose and timing of administration of each therapeutic agent administered according to the invention herein.

“Ameliorating” means a lessening or improvement of one or more symptoms upon treatment with a combination described herein, as compared to not administering the combination. “Ameliorating” also includes shortening or reduction in duration of a symptom.

As used herein, an “effective dosage” or “effective amount” of drug, compound or pharmaceutical composition is an amount sufficient to affect any one or more beneficial or desired, including biochemical, histological and I or behavioral symptoms, of the disease, its complications and intermediate pathological phenotypes presenting during development of the disease. For therapeutic use, an effective amount refers to that amount which will relieve to some extent one or more of the symptoms of the disorder being treated. In reference to the treatment of cancer, a therapeutically effective amount refers to that amount which has the effect of (1 ) reducing the size of the tumor, (2) inhibiting (that is, slowing to some extent, preferably stopping) tumor metastasis, (3) inhibiting to some extent (that is, slowing to some extent, preferably stopping) tumor growth or tumor invasiveness, (4) relieving to some extent (or, preferably, eliminating) one or more signs or symptoms associated with the cancer, (5) decreasing the dose of other medications required to treat the disease, and/or (6) enhancing the effect of another medication, and/or (7) delaying the progression of the disease in a patient.

An effective dosage can be administered in one or more administrations. For the purposes of this invention, an effective dosage of drug, compound, or pharmaceutical composition is an amount sufficient to accomplish prophylactic or therapeutic treatment either directly or indirectly. As is understood in the clinical context, an effective dosage of drug, compound or pharmaceutical composition may or may not be achieved in conjunction with another drug, compound or pharmaceutical composition.

“Tumor” as it applies to a subject diagnosed with, or suspected of having, a cancer refers to a malignant or potentially malignant neoplasm or tissue mass of any size and includes primary tumors and secondary neoplasms. A solid tumor is an abnormal growth or mass of tissue that usually does not contain cysts or liquid areas. Examples of solid tumors are sarcomas, carcinomas, and lymphomas. Leukaemia’s (cancers of the blood) generally do not form solid tumors (National Cancer Institute, Dictionary of Cancer Terms).

“Tumor burden” or “tumor load’, refers to the total amount of tumorous material distributed throughout the body. Tumor burden refers to the total number of cancer cells or the total size of tumor(s), throughout the body, including lymph nodes and bone marrow. Tumor burden can be determined by a variety of methods known in the art, such as, e.g., using calipers, or while in the body using imaging techniques, e.g., ultrasound, bone scan, computed tomography (CT), or magnetic resonance imaging (MRI) scans.

The term “tumor size” refers to the total size of the tumor which can be measured as the length and width of a tumor. Tumor size may be determined by a variety of methods known in the art, such as, e.g., by measuring the dimensions of tumor(s) upon removal from the subject, e.g., using calipers, or while in the body using imaging techniques, e.g., bone scan, ultrasound, CR or MRI scans.

The PALOMA-3 trial (ClinicalTrials.gov identifier: NCT01942135) was a Phase III double-blind study in women aged 18 years or older with HR+, HER2- metastatic breast cancer, who had progressed after previous endocrine therapy. 521 endocrine-pretreated patients were randomly assigned to receive palbociclib plus fulvestrant or placebo plus fulvestrant (palbociclib arm, 347 patients; placebo arm, 174 patients). The primary endpoint was PFS.

The PALOMA-2 trial (ClinicalTrials.gov identifier: NCT01740427) was a Phase III double-blind study in post-menopausal women with ER+, HER2- advanced breast cancer, who had not received prior treatment for advanced disease. 666 patients were randomly assigned to receive palbociclib plus letrozole or placebo plus letrozole (palbociclib arm, 444 patients; placebo arm, 222 patients). The primary endpoint was PFS.

The patient cohorts from the PALOMA-2 and PALOMA-3 trials, and their associated molecular profiles, were utilized to discover multivariate gene expression signatures capable of predicting strong versus weak response to palbociclib in combination with endocrine therapy, regardless of the line of therapy. An aggregated consideration of the features within such signatures revealed intrinsic resistance mechanisms to CDK4/6 inhibition that are common among HR+, HER2- breast cancer patients. Baseline tumor tissue samples from the PALOMA-3 trial were profiled using the EdgeSeq Oncology Biomarker Panel (OBP) panel (HTG Molecular Diagnostics), with 302 tumor samples evaluable for analysis. Anonymized gene expression datasets from the PALOMA-3 trial have been deposited in NCBI’s Gene Expression Omnibus (Turner et al., Cyclin E1 Expression and Palbociclib Efficacy in Previously Treated Hormone Receptor-Positive Metastatic Breast Cancer, Journal of Clinical Oncology 37:14, 1169- 1178, 2019) and are accessible through GEO Series accession number GSE128500 (https://www.ncbi. nlm.nih.gov/geo/query/acc.cgi?acc=GSE128500).

The number of evaluable samples from each arm of the PALOMA-3 trial, as well as the sites of tumor tissue collection, are summarized in Table 1 .

Table 1. PALOMA-3 Tumor Samples

Baseline tumor tissue samples from the PALOMA-2 trial were profiled using the EdgeSeq Oncology Biomarker Panel (OBP) panel (HTG Molecular Diagnostics), with 455 tumor samples evaluable for analysis. Anonymized gene expression datasets from the PALOMA-2 trial have been deposited in NCBI’s Gene Expression Omnibus (Finn et al., Biomarker Analyses of Response to Cyclin-Dependent Kinase 4/6 Inhibition and Endocrine Therapy in Women with Treatment-Naive Metastatic Breast Cancer, Clin. Cancer Res. 26:1 , 110-121 , 2020) and are accessible through GEO Series accession number GSE133394 (https://www.ncbi. nlm.nih.gov/geo/query/acc.cgi?acc=GSE133394).

The number of evaluable samples from each arm of the PALOMA-2 trial are summarized in Table 2 below. The tumor samples came from a mixture of primary or metastatic biopsies and the sites of tumor tissue collection were not recorded.

Table 2. PALOMA-2 Tumor Samples

Two distinct techniques, Elastic Net (EN) and Non-Negative Matrix Factorization (NMF), were employed to derive predictive multi-gene expression signatures associated with relative benefit from treatment with palbociclib in combination with endocrine therapy in a first-line or second-line setting.

The elastic net technique was employed to derive exploratory, predictive multigene expression features using patient samples from the PALOMA-3 study, which could then be tested for predictive power using the data set from the PALOMA-2 study. The machine-learning based elastic net regularization algorithm was applied. Clinical outcome association was performed by Cox PH regression using a number of bootstraps (e.g., 100 or 1 ,000 bootstraps) to provide a list of complementary gene expression features that comprise a model from each round. Genes were selected based on frequency observed in the bootstraps. Typically, at least 100 bootstraps are performed, and frequently 1 ,000 bootstraps are performed. In some methods, bootstraps are performed until the results have stabilized. See Goodhue et al., Does PLS have advantages for small sample size or non-normal data? MIS Quarterly, 36(3):981-1001 , 2012; Efron et al., (2004). Resampling methods of estimation. In N.J. Smelser, & P.B. Baltes (Eds.). International Encyclopedia of the Social & Behavioral Sciences (pp. 13216-13220). New York, NY: Elsevier.

The elastic net method linearly combines the L1 and L2 penalties of the lasso (least absolute shrinkage and selection operator) and ridge regression analysis methods. Lasso regression analysis performs variable selection and regularization using the penalty, which imposes sparsity among the coefficients to increase the accuracy and interpretability of the fitted model. The ridge regression method uses regularization to limit coefficient vector size. Elastic net regularization reduces overfitting by adding a complexity penalty that is part L1 and part L2 as a compromise between the ridge regression penalty (a = 0) and the Lasso penalty (a = 1 ):

It was previously demonstrated that the collection of tissue closer to treatment time facilitated the identification of predictive markers (Turner et al., 2019). Therefore, model derivation efforts were initially focused on metastatic tumor samples collected from patients in the palbociclib treatment arm of the PALOMA-3 cohort.

The machine-learning based elastic net regularization algorithm was applied, with clinical outcome association by Cox PH regression. The elastic net process involves:

(1 ) cross validation to select an optimal alpha (a) and lambda (X) pair with minimum error;

(2) with the best a and X selected, perform a number of bootstraps to obtain a list of complementary gene expression features that comprise a model from each round;

(3) select genes based on frequency observed in the bootstraps; and

(4) compute a composite signature score by weighting expression of signature genes by their coefficients.

Typically, at least 100 bootstraps are used, and frequently 1 ,000 bootstraps are used. In some methods, additional bootstraps are performed until the results have stabilized.

Gene Signatures

The elastic net method was applied to identify multivariate gene signatures capable of predicting palbociclib efficacy in HR+, HER2- breast cancer. Gene signatures were derived using expression data obtained from patient samples in the PALOMA-3 and PALOMA-2 trials using the EdgeSeq Oncology Biomarker Panel (OBP) panel (HTG Molecular Diagnostics).

Three multi-gene signatures were derived, encompassing different combinations of the following genes: ABCB11 , BUB1 B, CCNE1 , CDH15, CDKN2D, COL11A1 , CPT1A, ERG, LRP5, MAP3K1 , MT1X, SLC38A2, STC2, TFDP2, TMPRSS2, and TRIB1.

11-Gene Signature

As described in Example 1 , an 11 -gene signature was discovered that is predictive of palbociclib efficacy in both endocrine-refractory and advanced treatment naive HR+, HER2- advanced or metastatic breast cancer.

The 11 -gene signature was discovered using 92 metastatic samples from the palbociclib plus fulvestrant treatment arm of the PALOMA-3 trial in HR+, HER2- endocrine pretreated metastatic breast cancer patients.

Table 3

The 11 genes identified, CCNE1 , CDH15, CPT1A, ERG, LRP5, MAP3K1 , SLC38A2, STC2, TFDP2, TMPRSS2, and TRIB1 , were found to have non-zero coefficients in at least 40% of the models (FIG. 1 ). Six of the genes (CCNE1 , CPT1A, ERG, LRP5, TMPRSS2, and TRIB1 ) appear to be associated with relative resistance to palbociclib, while the remaining five genes (CDH15, MAP3K1 , SLC38A2, STC2, and TFDP2) appear to be associated with relative sensitivity to palbociclib (FIG. 1 ).

Hierarchical clustering was conducted using 1 -Pearson correlation as distance metric and average linkage. FIG. 2 shows a hierarchical clustering heatmap of the 11- gene exploratory signature and progression-free survival (PFS). Shown are patients treated with palbociclib + fulvestrant in FIG. 2(A) or placebo + fulvestrant FIG. 2(B), in each case sorted by PFS excluding those with an insufficient follow-up time of <12 months and no event.

The 11 -gene signature was applied to the 142 metastatic patient samples from both palbociclib treatment and placebo arms of the PALOMA-3 trial (Table 1 ). FIG. 3(A) shows the association between the 11 -gene exploratory signature and clinical benefit (PFS) as a Kaplan-Meier plot for the metastatic samples from PALOMA-3. The gene signature was further applied to the full cohort from the PALOMA-3 trial, including 302 primary and metastatic samples from the palbociclib treatment and placebo arms (Table 1 ). FIG. 3(B) shows the association between the 11 -gene exploratory signature and clinical benefit (PFS) as a Kaplan-Meier plot for the full PALOMA-3 data set, including both metastatic and primary tumor samples. Patients were dichotomized by median signature score into low (dashed line) and high (solid line) score groups. Also shown are hazard ratios for each gene in the 11 -gene exploratory signature using the full data set from PALOMA-3 in FIG. 3(C).

The ability of the gene signature to predict palbociclib efficacy across lines of therapy was demonstrated by application to an independent metastatic breast cancer cohort in the PALOMA-2 trial, which included ER+, HER2- advanced breast cancer patients who were treatment naive. Association between the 11 -gene exploratory signature and clinical benefit from the addition of palbociclib to letrozole in 454 samples from the palbociclib treatment and placebo plus letrozole control arms of the PALOMA-2 trial (Table 2) is shown in the Kaplan-Meier plot in FIG. 4. Patients were dichotomized by median signature score into low (dashed line) and high (solid line) score groups. The signature was also predictive based on continuous variable analysis (interaction P = 0.016 and P = 0.0031 before and after adjusting for baseline clinicopathologic characteristics respectively).

Some of the genes identified in the 11 -gene signature have a clear mechanistic link to the cell cycle, such as transcription factor DP-2 (TFDP2) which forms a heterodimer with the E2F transcription factors in regulating cell cycle progression from G1 to S phase (PMID: 8755520). Others could imply novel and previously unappreciated biology. For example, two of the genes, TMPRSS2 and ERG, are known fusion partners found in approximately half of prostate cancer patients, resulting in androgen receptor (AR) induced over-expression (PMID: 16254181 and 18283340). 9-Gene Signature

As described in Example 2, an 9-gene signature was discovered that is predictive of palbociclib efficacy in both endocrine-pretreated (second or later line) and treatment naive (first line) HR+, HER2- advanced or metastatic breast cancer.

The 9-gene signature was discovered using the 92 metastatic samples from the palbociclib plus fulvestrant treatment arm of the PALOMA-3 trial in HR+, HER2- endocrine pretreated metastatic breast cancer patients. Nine genes were selected using a C-lndex to build an optimal predictive model. Table 4

The 9 genes identified, CCNE1 , CPT1A, ERG, LRP5, MAP3K1 , SLC38A2, STC2, TFDP2, and TMPRSS2, were found to have non-zero coefficients in at least 10% of the bootstrap models (FIG. 5). Five of the genes (CCNE1 , CPT1A, ERG, LRP5, and TMPRSS2) appear to be associated with relative resistance to palbociclib. The remaining four genes (MAP3K1 , SLC38A2, STC2, and TFDP2) appear to be associated with relative sensitivity to palbociclib (FIG. 5).

FIG. 6(A) shows the association between the 9-gene exploratory signature and clinical benefit (PFS) as a Kaplan-Meier plot for the full PALOMA-3 data set, including 302 primary and metastatic samples from the palbociclib treatment and placebo arms (Table 1 ).

The ability of the gene signature to predict palbociclib efficacy across lines of therapy was demonstrated by application to and independent metastatic breast cancer cohort in the PALOMA-2 trial, which included ER+, HER2- advanced breast cancer patients who were treatment naive. Association between the 9-gene exploratory signature and clinical benefit from the combination of palbociclib plus letrozole in the PALOMA-2 trial is shown in the Kaplan-Meier plot in FIG. 6(B) ( interaction P=0.036).

Despite marked cross-cohort differences, including both the combination partner (fulvestrant versus letrozole) and the clinical setting (endocrine resistant vs. treatment naive), the 9-gene signature remained predictive of palbociclib efficacy in the PALOMA- 2 patient samples.

The 9-gene signature identified some of the same genes as in the 11 -gene signature, including transcription factor DP-2 (TFDP2) and the genes TMPRSS2 and ERG associated with AR induced over-expression.

13-Gene Signature

As described in Example 3, a 13-gene signature was discovered using the PALOMA-3 samples and then validated using the PALOMA-2 samples.

To further investigate multigene expression features that differentiate among the 302 primary and metastatic samples from the PALOMA-3 trial, a virtual microdissection approach, Non-Negative Matrix Factorization (NMF), was applied. NMF was conducted as described in Moffitt et al., Virtual microdissection identifies distinct tumor- and stroma-specific subtypes of pancreatic ductal adenocarcinoma. Nat Genet. 2015;47(10):1168-1178.). The NMF algorithm factorizes the gene expression matrix into two non-negative matrices: a gene factor matrix and a sample factor. Due to its constraints for all matrices to be non-negative, NMF is particularly suited to find biological processes that underlie expression data.

The virtual microdissection identified fourteen molecular factors that characterize biological processes differentiating baseline tumor samples of this cohort. One of the fourteen factors unambiguously represented the liver specific genes. Since a number of samples were from biopsies of liver metastases, a liver-specific factor was used to identify samples with significant liver component as described herein. Using this approach, 30 samples with significant liver content were identified and excluded from further analysis.

The Elastic Net methods described above were then reapplied to the remaining dataset after removing potential outliers for liver factors. This analysis provided a refined 13-gene signature (a = 1 , X = 0.204) with higher bootstrapping frequencies than the previous 9-gene signature. Thirteen genes were selected using a C-lndex to build an optimal predictive model.

Table 5

FIG. 7 shows the 13-gene exploratory signature, where bar length (top x-axis) reflects the proportion of 1000 bootstrap models containing the gene while purple points (bottom x-axis) reflect average non-zero coefficient across bootstrap models. The 13 genes identified, BUB1 B, CCNE1 , CDKN2D, ERG, LRP5, MT1X, TMPRSS2, TRIB1, ABCB11 , COL11A1 , MAP3K1 , STC2 and TFDP2, were found to have non-zero coefficients in the bootstrap models. Eight of the genes (BLIB1 B, CCNE1 , CDKN2D, ERG, LRP5, MT1X, TMPRSS2, and TRIB1 ) appear to be associated with relative resistance to palbociclib, while the remaining five genes (ABCB11 , COL11 A1 , MAP3K1 , STC2 and TFDP2) appear to be associated with relative sensitivity to palbociclib (FIG. 7).

The 13-gene signature was applied to the full cohort from the PALOMA-3 trial, including 302 primary and metastatic samples from the palbociclib treatment and placebo arms (Table 1). Data are provided in the Kaplan-Meier plots. FIG. 8(A) shows the association between the 13-gene exploratory signature and clinical benefit (PFS) in the full data set from PALOMA-3 after filtering out outliers from NMF liver factor (FIG. 9).

The 13-gene signature was further validated using independent samples from the PALOMA-2 trial. The ability of the gene signature to predict palbociclib efficacy across lines of therapy was demonstrated by application of the 13-gene signature to an independent metastatic breast cancer cohort in the PALOMA-2 trial. Association between the 13-gene exploratory signature and clinical benefit from the combination of palbociclib plus letrozole in the PALOMA-2 trial is shown in the Kaplan-Meier plot in FIG. 8(B).

Among the genes associated with palbociclib resistance, BUB1 B and CDKN2D are regulators of the cell cycle pathway, with the former inhibiting anaphase-promoting complex/cyclosome (Vleugel et al, 2015) and the latter an inhibitor to the activation of the CDK4/6 complex (Rolland et al, 2014). TRIB1 regulates the MAPK pathway (Jamieson et al, 2018).

Additional genes associated with relative palbociclib resistance in the 13-gene signature include CCNE1 , as well as targets of MYC and WNT signaling. See Birdsey et al., The endothelial transcription factor ERG promotes vascular stability and growth through Wnt/beta-catenin signaling. Dev Cell 32:82-96, 2015; Pacilli et al., Carnitineacyltransferase system inhibition, cancer cell death, and prevention of myc-induced lymphomagenesis. J Natl Cancer Inst 105:489-498, 2013; Sonntag et al., Cyclin E1 and cyclin-dependent kinase 2 are critical for initiation, but not for progression of hepatocellular carcinoma. Proc Natl Acad Sci II S A 115:9282-9287, 2018; Tomlins et al., Recurrent fusion of TMPRSS2 and ETS transcription factor genes in prostate cancer. Science 310:644-648, 2005; Vijayakumar et al., High-frequency canonical Wnt activation in multiple sarcoma subtypes drives proliferation through a TCF/beta-catenin target gene, CDC25A. Cancer Cell 19:601-612, 2011 ).

Genes associated with relative palbociclib sensitivity in the 13-gene signature have been associated with HR+ breast cancer etiology, either as modifiers of patient response to ESR1 antagonists, susceptibility to the disease or direct target genes. See Bouras et al., Stanniocalcin 2 is an estrogen-responsive gene coexpressed with the estrogen receptor in human breast cancer. Cancer Res 62:1289-1295, 2002; Griffith et al., The prognostic effects of somatic mutations in ER-positive breast cancer. Nat Commun 9: 3476, 2018; Morotti et al., Hypoxia-induced switch in SNAT2/SLC38A2 regulation generates endocrine resistance in breast cancer. Proc Natl Acad Sci II S A 116:12452-12461 , 2019; Subramanian et al., Gene set enrichment analysis: a knowledge-based approach for interpreting genome-wide expression profiles. Proc Natl Acad Sci II S A 102:15545-15550, 2005; and Toy et al., Activating ESR1 Mutations Differentially Affect the Efficacy of ER Antagonists. Cancer discovery 7:277-287, 2017.

While EdgeSeq profiling has been used to evaluate expression levels for the gene signatures provided herein, the skilled person would understand that other techniques (e.g., qPCR, RNAseq, or NanoString technology) could be employed to evaluate gene expression could be used to train the EN models.

Therapeutic Methods, Uses and Combinations

In one aspect, the invention provides a method of treating hormone receptorpositive (HR+), human epidermal growth factor receptor 2-negative (HER2-) breast cancer in a subject in need thereof, comprising:

(c) selecting the subject for treatment with a cyclin dependent kinase 4/6 (CDK 4/6) inhibitor and an endocrine therapeutic agent when the composite score is less than the median of a reference composite score computed for the gene signature in a reference breast cancer population; and (d) administering to the selected subject an amount of a CDK4/6 inhibitor and an amount of an endocrine therapeutic agent, wherein the amounts together are effective to treat cancer; wherein the gene signature comprises five or more genes selected from the group consisting of: ABCB11 , BUB1 B, CCNE1 , CDH15, CDKN2D, COL11A1 , CPT1A, ERG, LRP5, MAP3K1 , MT1X, SLC38A2, STC2, TFDP2, TMPRSS2, and TRIB1.

In some embodiments of this aspect, the subject is selected for treatment when the composite score is less than the first quartile of the reference composite score computed for the gene signature in the reference breast cancer population.

In another aspect, the invention provides a method of predicting whether a subject having hormone receptor-positive (HR+), human epidermal growth factor receptor 2-negative (HER2-) breast cancer will respond to treatment with a CDK4/6 inhibitor and an endocrine therapeutic agent, comprising:

(b) computing a composite score for the gene signature as a weighted sum of the expression levels for each gene in the gene signature; (c) comparing the composite score to a reference composite score computed for the gene signature in a reference breast cancer population; and

In some embodiments of this aspect, the subject is predicted to respond to treatment when the composite score is less than the first quartile of the reference composite score computed for the gene signature in the reference breast cancer population.

In a further aspect, the invention provides a method of predicting whether a subject having hormone receptor-positive (HR+), human epidermal growth factor receptor 2-negative (HER2-) breast cancer will be resistant to treatment with a CDK4/6 inhibitor and an endocrine therapeutic agent, comprising:

(d) predicting the subject will be resistant to treatment when the composite score is greater than the median of the reference composite score; wherein the gene signature comprises five or more genes selected from the group consisting of: ABCB11 , BUB1 B, CCNE1 , CDH15, CDKN2D, COL11A1 , CPT1A, ERG, LRP5, MAP3K1 , MT1X, SLC38A2, STC2, TFDP2, TMPRSS2, and TRIB1.

In some embodiments of this aspect, the subject is predicted to be resistant to treatment when the composite score is greater than the third quartile of the reference composite score computed for the gene signature in the reference breast cancer population.

In some embodiments of each of the methods, uses and combinations herein, the breast cancer is HR+, HER2- advanced or metastatic breast cancer. In some such embodiments, the HR+, HER2- advanced or metastatic breast cancer had progressed on prior endocrine therapy (i.e., the method relates to a second line or later line of therapy). In other such embodiments, the HR+, HER2- advanced or metastatic breast cancer is treatment naive (i.e. , the method relates to a first line of therapy).

In some embodiments of each of the methods herein, the breast cancer is HR+, HER2- early breast cancer. In some embodiments of each of the methods herein, the breast cancer is HR+, HER2- high risk early breast cancer. In some embodiments of each of the methods herein, the breast cancer is HR+, HER2- early breast cancer at high risk of relapse after showing less than pathological complete response to neoadjuvant chemotherapy. In some embodiments of each of the methods herein, the breast cancer is HR+, HER2- early breast cancer with residual invasive disease after neoadjuvant chemotherapy.

In some embodiments of each of the methods herein, the CDK4/6 inhibitor and the endocrine therapeutic agent are administered sequentially, simultaneously or concurrently. In some embodiments of each of the methods herein, the endocrine therapeutic agent is selected from the group consisting of an aromatase inhibitor, a selective estrogen receptor degrader (SERD), and a selective estrogen receptor modulator (SERM). In some such embodiments the endocrine therapeutic agent is selected from the group consisting of letrozole, anastrozole, exemestane and fulvestrant.

In some such embodiments, the aromatase inhibitor is selected from the group consisting of letrozole, anastrozole, and exemestane. In some such embodiments, the aromatase inhibitor is letrozole. In some embodiments, the endocrine therapeutic agent is a SERD. In some such embodiments, the SERD is selected from the group consisting of fulvestrant, elacestrant (RAD-1901 , Radius Health), SAR439859 (Sanofi), RG6171 (Roche), AZD9833 (AstraZeneca), AZD9496 (AstraZeneca), rintodestrant (G1 Therapeutics), ZN-c5 (Zentalis), LSZ102 (Novartis), D-0502 (Inventisbio), LY3484356 (Lilly), and SHR9549 (Jiansu Hengrui Medicine). In some such embodiments, the SERD is fulvestrant. In some embodiments, the endocrine therapeutic agent is a SERM. In some such embodiments, the SERM is selected from the group consisting of tamoxifen, raloxifene, toremifene, lasofoxifene, bazedoxifene and afimoxifene. In some such embodiments, the SERM is tamoxifen or raloxifene.

In preferred embodiments of each of the methods herein, the CDK4/6 inhibitor is palbociclib, or a pharmaceutically acceptable salt thereof.

In some embodiments of each of the methods herein, the gene signature comprises: (a) CCNE1 , CDH15, CPT1A, ERG, LRP5, MAP3K1 , SLC38A2, STC2, TFDP2, TMPRSS2, and TRIB1 ;

(b) CCNE1 , CPT1A, ERG, LRP5, MAP3K1 , SLC38A2, STC2, TFDP2, and TMPRSS2; or

(c) BUB1 B, CCNE1 , CDKN2D, ERG, LRP5, MT1X, TMPRSS2, TRIB1 , ABCB11 , COL1 1 A1 , MAP3K1 , STC2 and TFDP2.

In some embodiments of each of the methods herein, the gene signature consists essentially of:

(a) CCNE1 , CDH15, CPT1A, ERG, LRP5, MAP3K1 , SLC38A2, STC2, TFDP2, TMPRSS2, and TRIB1 ;

(b) CCNE1 , CPT1A, ERG, LRP5, MAP3K1 , SLC38A2, STC2, TFDP2, and TMPRSS2; or

In some embodiments of each of the methods herein, the gene signature consists of:

(b) CCNE1 , CPT1A, ERG, LRP5, MAP3K1 , SLC38A2, STC2, TFDP2, and TMPRSS2; or

In another aspect, the invention provides a method of treating hormone receptorpositive (HR+), human epidermal growth factor receptor 2-negative (HER2-) breast cancer in a subject in need thereof, comprising:

(a) determining the expression levels of CCNE1 , CDH15, CPT1A, ERG, LRP5, MAP3K1 , SLC38A2, STC2, TFDP2, TMPRSS2, and TRIB1 in a tumor sample from the subject;

(b) selecting the subject for treatment with a cyclin dependent kinase 4/6 (CDK 4/6) inhibitor and an endocrine therapeutic agent when:

(i) the expression levels of CCNE1 , CPT1A, ERG, LRP5, TMPRSS2, and TRIB1 in the tumor sample are downregulated relative to the expression levels in a non-tumor sample from the subject or relative to the median expression levels in a control population; and/or (ii) the expression levels of CDH15, MAP3K1 , SLC38A2, STC2, and TFDP2 in the tumor sample are upregulated relative to the expression levels in a non-tumor sample from the subject or relative to the median expression levels in a control population; and

(c) administering to the selected subject an amount of a CDK4/6 inhibitor and an amount of an endocrine therapeutic agent, wherein the amounts together are effective to treat cancer.

In some embodiments of this aspect, the subject is selected for treatment when:

(i) the expression levels of CCNE1 , CPT1A, ERG, LRP5, TMPRSS2, and TRIB1 in the tumor sample are downregulated relative to the expression levels in a non-tumor sample from the subject or relative to the first quartile expression levels in a control population; and/or

(ii) the expression levels of CDH15, MAP3K1 , SLC38A2, STC2, and TFDP2 in the tumor sample are upregulated relative to the expression levels in a non-tumor sample from the subject or relative to the first quartile expression levels in a control population.

(a) determining the expression levels of CCNE1 , CPT1A, ERG, LRP5, MAP3K1 , SLC38A2, STC2, TFDP2, and TMPRSS2 in a tumor sample from the subject;

(i) the expression levels of CCNE1 , CPT1A, ERG, LRP5, and TMPRSS2 in the tumor sample are downregulated relative to the expression levels in a non-tumor sample from the subject or the median expression levels in a control population; and/or

(ii) the expression levels of MAP3K1 , SLC38A2, STC2, and TFDP2 in the tumor sample are upregulated relative to the expression levels in a non- tumor sample from the subject or the median expression levels in a control population; and

In some embodiments of this aspect, the subject is selected for treatment when: (i) the expression levels of CCNE1 , CPT1A, ERG, LRP5, and TMPRSS2 in the tumor sample are downregulated relative to the expression levels in a non-tumor sample from the subject or the first quartile expression levels in a control population; and/or

(ii) the expression levels of MAP3K1 , SLC38A2, STC2, and TFDP2 in the tumor sample are upregulated relative to the expression levels in a nontumor sample from the subject or the first quartile expression levels in a control population.

(a) determining the expression levels of BLIB1 B, CCNE1 , CDKN2D, ERG, LRP5, MT1X, TMPRSS2, TRIB1 , ABCB11 , COL11A1 , MAP3K1 , STC2 and TFDP2 in a tumor sample from the subject;

(i) the expression levels of BUB1 B, CCNE1 , CDKN2D, ERG, LRP5, MT1X, TMPRSS2, and TRIB1 in the tumor sample are downregulated relative to the expression levels in a non-tumor sample from the subject or the median expression levels in a control population; and/or

(ii) the expression levels of ABCB11 , COL11 A1 , MAP3K1 , STC2 and TFDP2 in the tumor sample are upregulated relative to the expression levels in a non-tumor sample from the subject or the median expression levels in a control population; and

In some embodiments of this aspect, the subject is selected for treatment when:

(i) the expression levels of BUB1 B, CCNE1 , CDKN2D, ERG, LRP5, MT1X, TMPRSS2, and TRIB1 in the tumor sample are downregulated relative to the expression levels in a non-tumor sample from the subject or the first quartile expression levels in a control population; and/or

(ii) the expression levels of ABCB11 , COL11 A1 , MAP3K1 , STC2 and TFDP2 in the tumor sample are upregulated relative to the expression levels in a non-tumor sample from the subject or the first quartile expression levels in a control population.

In another aspect, the invention provides use of a combination comprising a CDK4/6 inhibitor and an endocrine therapeutic agent for treating hormone receptorpositive (HR+), human epidermal growth factor receptor 2-negative (HER2-) breast cancer in a subject in need thereof, comprising:

(c) selecting the subject for treatment with the combination when the composite score is less than the median of a reference composite score computed for the gene signature in a reference breast cancer population; wherein the gene signature comprises five or more genes selected from the group consisting of: ABCB11 , BUB1 B, CCNE1 , CDH15, CDKN2D, COL11A1 , CPT1A, ERG, LRP5, MAP3K1 , MT1X, SLC38A2, STC2, TFDP2, TMPRSS2, and TRIB1.

In another aspect, the invention provides a combination of a CDK4/6 inhibitor and an endocrine therapeutic agent for use in treating hormone receptor-positive (HR+), human epidermal growth factor receptor 2-negative (HER2-) breast cancer in a subject in need thereof, comprising:

(c) selecting the subject for treatment with the combination when the composite score is less than the median of a reference composite score computed for the gene signature in a reference breast cancer population; wherein the gene signature comprises five or more genes selected from the group consisting of: ABCB11 , BUB1 B, CCNE1 , CDH15, CDKN2D, COL11A1 , CPT1A, ERG, LRP5, MAP3K1 , MT1X, SLC38A2, STC2, TFDP2, TMPRSS2, and TRIB1. In some embodiments of this aspect, the subject is selected for treatment when the composite score is less than the first quartile of the reference composite score computed for the gene signature in the reference breast cancer population.

In some embodiments of each of the uses and combinations herein, the breast cancer is HR+, HER2- advanced or metastatic breast cancer. In some such embodiments, the HR+, HER2- advanced or metastatic breast cancer had progressed on prior endocrine therapy (i.e., the method relates to a second line or later line of therapy). In other such embodiments, the HR+, HER2- advanced or metastatic breast cancer is treatment naive (i.e., the method relates to a first line of therapy).

In some embodiments of each of the uses and combinations herein, the breast cancer is HR+, HER2- early breast cancer. In some embodiments of each of the uses and combinations herein, the breast cancer is HR+, HER2- high risk early breast cancer. In some embodiments of each of the uses and combinations herein, the breast cancer is HR+, HER2- early breast cancer at high risk of relapse after showing less than pathological complete response to neoadjuvant chemotherapy. In some embodiments of each of the uses and combinations herein, the breast cancer is HR+, HER2- early breast cancer with residual invasive disease after neoadjuvant chemotherapy.

In some embodiments of each of the uses and combinations herein, the CDK4/6 inhibitor and the endocrine therapeutic agent are administered sequentially, simultaneously or concurrently. In some embodiments of each of the methods herein, the endocrine therapeutic agent is selected from the group consisting of an aromatase inhibitor, a selective estrogen receptor degrader (SERD), and a selective estrogen receptor modulator (SERM). In some such embodiments the endocrine therapeutic agent is selected from the group consisting of letrozole, anastrozole, exemestane and fulvestrant.

In some embodiments of each of the uses and combinations herein, the CDK4/6 inhibitor is palbociclib, or a pharmaceutically acceptable salt thereof.

In some embodiments of each of the uses and combinations herein, the gene signature comprises:

(b) CCNE1 , CPT1A, ERG, LRP5, MAP3K1 , SLC38A2, STC2, TFDP2, and TMPRSS2; or

(c) BUB1 B, CCNE1 , CDKN2D, ERG, LRP5, MT1X, TMPRSS2, TRIB1 , ABCB11 ,

COL1 1 A1 , MAP3K1 , STC2 and TFDP2. In some embodiments of each of the uses and combinations herein, the gene signature consists essentially of:

(b) CCNE1 , CPT1A, ERG, LRP5, MAP3K1 , SLC38A2, STC2, TFDP2, and TMPRSS2; or

In some embodiments of each of the uses and combinations herein, the gene signature consists of:

(b) CCNE1 , CPT1A, ERG, LRP5, MAP3K1 , SLC38A2, STC2, TFDP2, and TMPRSS2; or

In a further aspect, the invention provides a kit for assaying a tumor sample to determine a composite score for a gene signature in the tumor sample, comprising a first set of probes for detecting the expression level of each gene in the gene signature, wherein the gene signature comprises:

(a) five or more genes selected from the group consisting of ABCB11 , BLIB1 B, CCNE1 , CDH15, CDKN2D, COL11A1 , CPT1A, ERG, LRP5, MAP3K1 , MT1X, SLC38A2, STC2, TFDP2, TMPRSS2, and TRIB1 ;

(b) CCNE1 , CDH15, CPT1A, ERG, LRP5, MAP3K1 , SLC38A2, STC2, TFDP2, TMPRSS2, and TRIB1 ;

(c) CCNE1 , CPT1A, ERG, LRP5, MAP3K1 , SLC38A2, STC2, TFDP2, and TMPRSS2; or

(d) BUB1 B, CCNE1 , CDKN2D, ERG, LRP5, MT1X, TMPRSS2, TRIB1 , ABCB11 , COL11 A1 , MAP3K1 , STC2 and TFDP2.

In some embodiments of this aspect, the invention provides a kit further comprising a second set of probes for detecting the expression level of a set of normalization genes in the tumor sample.

These and other aspects of the invention, including the exemplary specific embodiments listed below, will be apparent from the teachings contained herein. Example 1

Elastic Net Analysis: 11 -Gene Signature

An 11 -gene signature was derived from the PALOMA-3 cohort. Metastatic tissue samples from 92 patients in the palbociclib plus fulvestrant treatment arm were used to generate the signature. A machine-learning based elastic net regularization algorithm with clinical outcome association by Cox PH regression was applied. A 10-fold cross validation was run to select an a = 0.53 and X = 0.31 and then 100 bootstraps were performed to provide a list of complementary expression features comprising a model from each round.

Genes were selected based on frequency observed in the bootstraps. A composite signature score was computed by weighting expression of signature genes by their coefficients.

Eleven genes were found to have non-zero coefficients in at least 40% of the models. FIG. 1 shows the 11 -gene exploratory signature, where bar length (top x-axis) reflects the proportion of 100 bootstrap models containing the gene while purple points (bottom x-axis) reflect average non-zero coefficient across bootstrap models. Six of the genes (CCNE1 , CPT1A, ERG, LRP5, TMPRSS2, TRIB1 ) appear to be associated with relative resistance, while the remaining five genes (CDH15, MAP3K1 , SLC38A2, STC2, TFDP2) appear to be associated with relative sensitivity (FIG. 1 ). A composite signature score was computed by weighting each component gene by its average coefficient across bootstrap models.

A hierarchical clustering heatmap of the 11 -gene exploratory signature and progression-free survival (PFS) is provided in FIG. 2. Clustering was conducted using 1 -Pearson correlation as distance metric and average linkage. Shown in FIG. 2(A) are the metastatic samples across the palbociclib plus fulvestrant treatment arm of PALOMA-3 . Shown in FIG. 2(B) are placebo plus fulvestrant control arm. For each arm, patients are sorted by PFS, excluding those with an insufficient follow-up time of less than 12 months and no event.

The 11 -gene signature was applied to 142 metastatic patient samples from PALOMA-3, comprising both the palbociclib plus fulvestrant and the placebo plus fulvestrant arms (Table 1 ). Composite signature scores were computed by weighting each component gene by its average coefficient across bootstrap models. Composite scores for patients in the palbociclib arm and the placebo arm were dichotomized by median into high and low scoring groups. PFS in months was plotted in the Kaplan- Meier plot in FIG. 3(A) (interaction P = 3.53e-05): (1 ) palbociclib plus fulvestrant: high (n=47; mPFS = 7.3 mo.) - solid, blue line;

(2) palbociclib plus fulvestrant: low (n=45; mPFS = 16.1 mo.) - dashed, blue line;

(3) placebo plus fulvestrant: high (n=24; mPFS = 3.6 mo.) - solid, red line; and

(4) placebo plus fulvestrant: low (n=26; mPFS = 2.1 mo.) - dashed, red line.

All patient groups benefited from the palbociclib combination (HR<1 ) but to a varying extent (HR = 0.84 and 0.13 for higher and lower by median, respectively).

The 11 -gene signature was further applied to the full PALOMA-3 data set, comprising 302 patient samples, including both primary and metastatic samples (Table 1 ). Composite signature scores were computed by weighting each component gene by its average coefficient across bootstrap models. Composite scores for patients in the palbociclib arm and the placebo arm were dichotomized by median into high and low scoring groups. Median PFS (mPFS) in months was plotted in the Kaplan-Meier plot in FIG. 3(B) (interaction P = 3.31 e-05):

(1 ) palbociclib plus fulvestrant: high (n=102; mPFS = 9.3 mo.) - solid, blue line;

(2) palbociclib plus fulvestrant: low (n=92; mPFS = 16.6 mo.) - dashed, blue line;

(3) placebo plus fulvestrant: high (n=49; mPFS = 5.4 mo.) - solid, red line; and

(4) placebo plus fulvestrant: low (n=59; mPFS = 3.7 mo.) - dashed, red line.

All patient groups benefited from the palbociclib combination (HR<1 ) but to a varying extent (HR = 0.99 and 0.27 for higher and lower by median, respectively).

The association between gene expression level and efficacy for palbociclib treatment was assessed for the 11 component genes individually using the 302 samples from the full PALOMA-3 trial. The result was plotted as a Forest plot in FIG. 3(C). Gene expression levels were dichotomized by median expression, with hazard ratios (HRs) determined for PFS of palbociclib plus fulvestrant (PAL + Fulv) versus placebo plus fulvestrant (PCB + Fulv). HRs were derived from a Cox PH regression model. Interaction P value for statistical interaction between gene expression and treatment for the 11 genes in the signature (CCNE1 , CPT1A, ERG, LRP5, TMPRSS2, TRIB1 , CDH15, MAP3K1 , SLC38A2, STC2, and TFDP2).

The 11 -gene signature was further validated on an independent metastatic breast cancer cohort from first-line PALOMA-2 trial, where tumor samples came from a mixture of primary or metastatic biopsies and the origin of tissue samples were not recorded (Table 2). Composite signature scores were computed by weighting each component gene by its average coefficient across bootstrap models. Composite scores for patients in the palbociclib arm and the placebo arm were dichotomized by median into high and low scoring groups. Median PFS (mPFS) in months was plotted in the Kaplan-Meier plot in FIG. 4(A) (interaction P = 9.78e-03):

(1 ) palbociclib plus letrozole: high (n=156; mPFS = 19.2 mo.) - solid, blue line;

(2) palbociclib plus letrozole: low (n=146; mPFS = 33.1 mo.) - dashed, blue line;

(3) placebo plus letrozole: high (n=71 ; mPFS = 13.8 mo.) - solid, red line; and

(4) placebo plus letrozole: low (n=81 ; mPFS = 13.8 mo.) - dashed, red line.

Despite marked cross-cohort differences in combination partner (fulvestrant vs letrozole) and clinical setting (endocrine resistant vs treatment naive), the 11 -gene signature remained predictive of palbociclib efficacy (FIG. 4; interaction P=9.78E-3).

Signature-treatment statistical interaction was also apparent from continuous variable analysis (P = 0.016 and P = 0.0031 before and after adjusting for baseline clinicopathologic characteristics respectively). All patient groups benefited from the palbociclib combination (HR<1 ) but to a varying extent (HR = 0.79 and 0.43 for higher and lower by median respectively).

Example 2

Elastic Net Analysis: 9-Gene Signature

A 9-gene signature was derived from the PALOMA-3 cohort. Metastatic tissue samples from 92 patients in the palbociclib plus fulvestrant treatment arm were used to generate the signature. A machine-learning based elastic net regularization algorithm with clinical outcome association by Cox PH regression was applied. A 5-fold cross validation was run to select an a = 1 and X = 0.277 and then 1 ,000 bootstraps were performed to provide a list of complementary expression features comprising a model from each round.

Genes were selected based on frequency observed in the bootstraps and 9 genes were selected using a C-lndex to build optimal predictive model including CCNE1. A composite signature score was computed by weighting each component gene by its average coefficient across bootstrap models. The 9 selected genes included CCNE1 , CPT1A, ERG, LRP5, TMPRSS2, MAP3K1 , SLC38A2, STC2, and TFDP2 (FIG. 5). FIG. 5 shows the 9-gene exploratory signature, where bar length (top x-axis) reflects the proportion of 1000 bootstrap models containing the gene while purple points (bottom x-axis) reflect average non-zero coefficient across bootstrap models. Five of the genes (CCNE1 , CPT1A, ERG, LRP5, and TMPRSS2) appear to be associated with relative resistance to palbociclib, while the remaining four genes (MAP3K1 , SLC38A2, STC2, and TFDP2) appear to be associated with relative sensitivity to palbociclib.

The 9-gene signature was further applied to the full PALOMA-3 data set, comprising 302 patient samples, including both primary and metastatic samples (Table 1 ). Composite signature scores were computed by weighting each component gene by its average coefficient across bootstrap models.

Composite scores for patients in the palbociclib arm and the placebo arm were dichotomized by median into high and low scoring groups. Median PFS (mPFS) in months was plotted in the Kaplan-Meier plot in FIG. 5(B) (interaction P = 0.00011 ):

(1 ) palbociclib plus fulvestrant: high (n=99; mPFS = 7.62 mo.) - solid, blue line;

(2) palbociclib plus fulvestrant: low (n=95; mPFS = 16.62 mo.) - dashed, blue line;

(3) placebo plus fulvestrant: high (n=52; mPFS = 4.70 mo.) - solid, red line; and

(4) placebo plus fulvestrant: low (n=56; mPFS = 3.71 mo.) - dashed, red line.

The exploratory 9-gene signature was then tested on the independent metastatic breast cancer cohort from the PALOMA-2 clinical trial comprising 454 patient samples. Despite marked cross-cohort differences in combination partner (fulvestrant vs. letrozole) and clinical setting (endocrine resistant vs. treatment naive), the 9-gene signature remained predictive of palbociclib efficacy (FIG. 5(C)).

Composite scores for patients in the palbociclib arm and the placebo arm were dichotomized by median into high and low scoring groups. Median PFS (mPFS) in months was plotted in the Kaplan-Meier plot in FIG. 5(C) (interaction P = 0.036):

(1 ) palbociclib plus letrozole: high (n=154; mPFS = 19.2 mo.) - solid, blue line;

(2) palbociclib plus letrozole: low (n=149; mPFS = 30.3 mo.) - dashed, blue line;

(3) placebo plus letrozole: high (n=73; mPFS = 13.7 mo.) - solid, red line; and

(4) placebo plus letrozole: low (n=78; mPFS = 13.9 mo.) - dashed, red line.

Signature-treatment statistical interaction was also apparent from continuous variable analysis (P = 0.014). All patient groups benefited from the palbociclib combination (HR<1 ) but to a varying extent (HR = 0.77 and 0.46 for higher and lower by median respectively).

Example 3

Elastic Net Analysis: 13-Gene Signature

Non-Negative Matrix Factorization (NMF) was applied to the 302 primary and metastatic samples from the PALOMA-3 trial. The virtual microdissection identified fourteen molecular factors that characterize biological processes differentiating baseline tumor samples of this cohort. One of the fourteen factors unambiguously represented the liver specific genes. Since some samples were from biopsies of liver metastases, a liver-specific factor was used to identify samples with significant liver component as described in the methods below. Using this approach, 30 samples with significant liver content were identified and excluded from further analysis.

The Elastic Net method described in Example 2 was then reapplied to the remaining dataset after removing potential outliers for liver factors. A 5-fold cross validation was run to select an a = 1 and X = 0.204 and then 1 ,000 bootstraps were performed to provide a list of complementary expression features comprising a model from each round. This analysis provided a refined 13-gene signature with higher bootstrapping frequencies than the previous 9-gene signature. A composite signature score was computed by weighting each component gene by its average coefficient across bootstrap models. The 13 selected genes included LRP5, TRIB1 , TFDP2, MAP3K1 , ERG, TMPRSS2, BUB1 B, ABCB11 , MT1X, CCNE1 , CDKN2D, STC2 and COL11A1. FIG. 7 shows the 13-gene exploratory signature, where bar length (top x- axis) reflects the proportion of 1000 bootstrap models containing the gene while purple points (bottom x-axis) reflect average non-zero coefficient across bootstrap models. Eight of the genes (LRP5, TRIB1 , ERG, TMPRSS2, BUB1 B, MT1X, CCNE1 , and CDKN2D) appear to be associated with relative resistance to palbociclib, while the remaining five genes (TFDP2, MAP3K1 , ABCB11 , STC2 and COL11A1 ) appear to be associated with relative sensitivity to palbociclib (FIG. 7).

The 13-gene signature was further applied to the PALOMA-3 data set after removing samples with significant liver content, comprising 272 patient samples, including both primary and metastatic samples (Table 1 ). Composite signature scores were computed by weighting each component gene by its average coefficient across bootstrap models.

Composite scores for patients in the palbociclib arm and the placebo arm were dichotomized by median into high and low scoring groups. Median PFS (mPFS) in months in PALOMA-3 was plotted in the Kaplan-Meier plot in FIG. 8(A) (interaction P = 0.00078):

(1 ) palbociclib plus fulvestrant: high (n=91 ; mPFS = 9.23 mo.) - solid, blue line;

(2) palbociclib plus fulvestrant: low (n=84; mPFS = 16.13 mo.) - dashed, blue line;

(3) placebo plus fulvestrant: high (n=45; mPFS = 7.23 mo.) - solid, red line; and (4) placebo plus fulvestrant: low (n=52; mPFS = 3.71 mo.) - dashed, red line.

The transferability of the observed molecular factors’ association with palbociclib response was examined in connection with an independent cohort of ER+, HER2- advanced breast cancer patients from the PALOMA-2 study. This cohort represents a patient population at an earlier stage of HR+ metastatic breast cancer and was treated with palbociclib in combination with the aromatase inhibitor, letrozole, or letrozole plus placebo as a first line therapy. The 14 factors identified in PALOMA-3 cohort were projected onto the PALOMA-2 data using method described in Tamayo et. Al., Metagene projection for cross-platform, cross-species characterization of global transcriptional states. Proc. Natl. Acad. Sci. USA 104, 5959-5964, 2007. In this cohort, 37 samples were identified with significant liver content and excluded from further analyses as for PALOMA-3.

The exploratory 13-gene signature was then tested on the independent metastatic breast cancer cohort of 417 samples from PALOMA-2 after filtering out samples with significant liver content and remained predictive of palbociclib efficacy (FIG. 8(B)). Composite scores for patients in the palbociclib arm and the placebo arm were dichotomized by median into high and low scoring groups. Median PFS (mPFS) in months was plotted in the Kaplan-Meier plot in FIG. 8(B) (interaction P = 0.018):

(1 ) palbociclib plus letrozole: high (n=142; mPFS = 16.6 mo.) - solid, blue line;

(2) palbociclib plus letrozole: low (n=137; mPFS = 35.7 mo.) - dashed, blue line;

(3) placebo plus letrozole: high (n=66; mPFS = 13.7 mo.) - solid, red line; and placebo plus letrozole: low (n=72; mPFS = 16.5 mo.) - dashed, red line.

Signature-treatment statistical interaction was also apparent from continuous variable analysis (P = 0.062). All patient groups benefited from the palbociclib combination (HR<1 ) but to a varying extent (HR = 0.76 and 0.41 for higher and lower by median respectively). Methods Virtual microdissection analysis

NMF was applied to HTG gene expression of 302 patient samples from PALOMA-3. The NMF algorithm factorizes the gene expression matrix V of g genes and s samples into two non-negative matrices of k factors: gene factor matrix W of weights of n gene for k factors and sample factor matrix H of weights of m sample for k factors. W represents the expression pattern of the k parts and H represents the respective contribution of k parts in each sample or bulk tumor¹. Before applying NMF, the expression values for each patient were rank transformed. NMF was then performed on rank transformed gene expression matrix V_paloma-3with the R package “NMF” which used the “brunet” algorithm². 30 runs of NMF were performed and the factorization that achieved the lowest approximation error for subsequent analyses was chosen. To determine the optimal k, we computed the cophenetic coefficient and chose k=14 that maximized the coefficient score. See Brunet et al., Metagenes and molecular pattern discovery using matrix factorization. Proc Natl Acad Sci II S A 101 , 4164-4169 (2004).

To extract exemplar genes for each of the k factors, genes with their contribution to the corresponding basis components of gene factor greater than 0.2 were selected. Gene Set Enrichment Analysis (GSEA) (Subramanian, et al., Gene set enrichment analysis: a knowledge-based approach for interpreting genome-wide expression profiles. Proc Natl Acad Sci II S A 102(43): 15545-15550, 2005) was performed on the pre-ranked gene lists based on their gene factor weights and the MSigDB gene set collection to annotate the factors.

Factor 14 (F14) was found to be enriched in liver specific genes with top exemplar genes uniquely expressing in liver tissues according to GTEX. To further confirm this, t test was performed on all factors between samples labeled as metastatic samples and samples labeled as primary samples based on clinical annotation. F14 showed significantly higher weights in the metastatic samples (p=1.7e-5).

To reduce the noise contributed by high liver cell contamination in the tumor, samples with high liver content based on F14 were removed. The cutoff=0.478 was chosen as the upper quartile plus 1.5 fold IQR (inter-quartile range). Using this approach, 30 samples were removed from the PALOMA-3 cohort, of which 28 were labeled as metastatic samples in clinical annotation while 2 were labeled as primary samples.

Using the metagene projection methodology, the low-dimensional NMF representation derived from the PALOMA-3 HTG data was projected onto the PALOMA-2 HTG data. See Tamayo, P., Scanfeld, D., Ebert, B.L., Gillette, M.A., Roberts, C.W., and Mesirov, J.P. (2007). Metagene projection for cross-platform, crossspecies characterization of global transcriptional states. Proc. Natl. Acad. Sci. USA 104, 5959-5964.

This projection enabled direct comparison of the two cohorts. The equation: Hpaioma-2 = (Wpaioma-3 )"¹ x V_paloma-2 was obtained using the Moore-Penrose generalized pseudoinverse on the rank transformed HTG data from PALOMA-2. Ben- Israel & Greville (2003) Generalized Inverses: Theory and Applications (Springer, New York). Based on the projected liver factor F14 in PALOMA-2, the same cutoff=0.478 was applied and 37 samples with high liver content were removed.

Factor association with end-point PFS

The predictive and prognostic effect for each NMF factor was evaluated. A predictive biomarker refers to an NMF factor whose magnitude significantly correlates with palbociclib treatment outcome, manifest in the p-value associated with the factor x arm interaction term coefficient in the Cox PH model. A prognostic biomarker refers to an NMF factor whose magnitude significantly correlates with overall clinical outcome in a cohort, regardless of treatment arms (adjusting for treatment arms in a Cox PH model).

Predictive biomarkers were obtained from the interactive Cox PH model with PFS as the end-point eq. 1 : h(t|x) = h₀ (t) expCpiX-t + p₂x₂ + p₃XiX₂) eq. 1 where xi is a dichotomous variable with 1 representing palbociclib plus fulvestrant treatment and 0 representing control treatment and X2 is a dichotomous variable with values 1 or 0 indicating whether the biomarker (NMF factor) is above median value or below. Interaction p-value is the p value associated with p₃.

To determine prognostic biomarkers, the additive Cox PH regression model was used without an interaction term (eq. 2). The p-value associated with 2 was used to discover prognostic biomarkers. h(t|x) = h₀ (t) exp(p₂x₂) eq. 2

Gene Expression Analysis

The EdgeSeq Oncology Biomarker Panel (HTG Molecular Diagnostics, Inc) was used for mRNA profiling of 2,534 cancer-related genes. The EdgeSeq system uses targeted capture sequencing to quantitate mRNA expression levels of gene targets in FFPE tissues. The first section of breast cancer FFPE tissue was stained with hematoxylin and eosin (H&E). A board-certified pathologist assessed the tumor cell content and tissue necrosis. Tumor content was estimated based on the number of malignant cells as a percentage of all cells (i.e. , malignant plus normal cells in the tissue section). The acceptance criterion for analysis was set at >70% of tumor content. Necrosis was assessed based on the percentage of necrotic tissue within the total tissue area. The acceptance criterion for analysis was set at <20% necrosis. Macrodissection was performed on the tissue sections if the tumor content was <70% or if necrosis was >20%. Sample preparation was conducted following laboratory processes and manufactory protocols. Sequencing was performed on an Illumina NextSeq 500 Sequencer (Illumina, Inc.).

To normalize the sequencing data, probe counts were transformed into Iog2 counts per million. Expression values were quantile normalized. HTG Molecular Diagnostics, Inc., was blinded to patient information and clinical outcomes.

Alternative approaches to determine gene expression levels include may comprise performing RT-PCR, a hybridization, transcriptome analysis, RNAseq, a Northern blot, a Western blot, or an ELISA. For example, assaying the expression level of genes in a gene signature can comprise performing an array hybridization. In some embodiments, transcriptome analysis may comprise obtaining sequence information of expressed RNA molecules.

Claims

1. A method of treating hormone receptor-positive (HR+), human epidermal growth factor receptor 2-negative (HER2-) breast cancer in a subject in need thereof, comprising:

2. A method of selecting a subject having hormone receptor-positive (HR+), human epidermal growth factor receptor 2-negative (HER2-) breast cancer for treatment with a CDK4/6 inhibitor and an endocrine therapeutic agent, comprising:

38

3. A method of predicting whether a subject having hormone receptorpositive (HR+), human epidermal growth factor receptor 2-negative (HER2-) breast cancer will respond to treatment with a CDK4/6 inhibitor and an endocrine therapeutic agent, comprising:

4. A method of predicting whether a subject having hormone receptorpositive (HR+), human epidermal growth factor receptor 2-negative (HER2-) breast cancer will be resistant to treatment with a CDK4/6 inhibitor and an endocrine therapeutic agent, comprising:

5. The method of any one of claims 1 to 4, wherein the breast cancer is advanced or metastatic breast cancer.

39

6. The method of claim 5, wherein the advanced or metastatic breast cancer had progressed on prior endocrine therapy.

7. The method of claim 5, wherein the advanced or metastatic breast cancer is treatment naive.

8. The method of any one of claims 1 to 4, wherein the breast cancer is early breast cancer.

9. The method of any one of claims 1 to 8, wherein the CDK4/6 inhibitor and the endocrine therapeutic agent are administered sequentially, simultaneously or concurrently.

10. The method of any one of claims 1 to 9, wherein the endocrine therapeutic agent is selected from the group consisting of an aromatase inhibitor, a selective estrogen receptor degrader (SERD), and a selective estrogen receptor modulator (SERM).

11. The method of claim 10, wherein the endocrine therapeutic agent is selected from the group consisting of letrozole, anastrozole, exemestane and fulvestrant

12. The method of any one of claims 1 to 11 , wherein the CDK4/6 inhibitor is palbociclib, or a pharmaceutically acceptable salt thereof.

13. The method of any one of claims 1 to 12, wherein the gene signature comprises:

(a) CCNE1 , CDH15, CPT1A, ERG, LRP5, MAP3K1 , SLC38A2, STC2, TFDP2,

TMPRSS2, and TRIB1 ;

(b) CCNE1 , CPT1A, ERG, LRP5, MAP3K1 , SLC38A2, STC2, TFDP2, and TMPRSS2; or

14. The method of any one of claims 1 to 12, wherein the gene signature consists essentially of:

(a) CCNE1 , CDH15, CPT1A, ERG, LRP5, MAP3K1 , SLC38A2, STC2, TFDP2, TMPRSS2, and TRIB1 ; (b) CCNE1 , CPT1A, ERG, LRP5, MAP3K1 , SLC38A2, STC2, TFDP2, and TMPRSS2; or

15. The method of any one of claims 1 to 12, wherein the gene signature consists of:

(b) CCNE1 , CPT1A, ERG, LRP5, MAP3K1 , SLC38A2, STC2, TFDP2, and TMPRSS2; or

16. A method of treating hormone receptor-positive (HR+), human epidermal growth factor receptor 2-negative (HER2-) breast cancer in a subject in need thereof, comprising:

(i) the expression levels of CCNE1 , CPT1A, ERG, LRP5, TMPRSS2, and TRIB1 in the tumor sample are downregulated relative to the expression levels in a non-tumor sample from the subject or relative to the median expression levels in a control population; and/or

(ii) the expression levels of CDH15, MAP3K1 , SLC38A2, STC2, and TFDP2 in the tumor sample are upregulated relative to the expression levels in a non-tumor sample from the subject or relative to the median expression levels in a control population; and

17. A method of treating hormone receptor-positive (HR+), human epidermal growth factor receptor 2-negative (HER2-) breast cancer in a subject in need thereof, comprising:

(ii) the expression levels of MAP3K1 , SLC38A2, STC2, and TFDP2 in the tumor sample are upregulated relative to the expression levels in a nontumor sample from the subject or the median expression levels in a control population; and

18. A method of treating hormone receptor-positive (HR+), human epidermal growth factor receptor 2-negative (HER2-) breast cancer in a subject in need thereof, comprising:

(ii) the expression levels of ABCB11 , COL11 A1 , MAP3K1 , STC2 and TFDP2 in the tumor sample are upregulated relative to the expression levels in a non-tumor sample from the subject or the median expression levels in a control population; and (c) administering to the selected subject an amount of a CDK4/6 inhibitor and an amount of an endocrine therapeutic agent, wherein the amounts together are effective to treat cancer.

19. A kit for assaying a tumor sample to determine a composite score for a gene signature in the tumor sample, comprising a first set of probes for detecting the expression level of each gene in the gene signature, wherein the gene signature comprises:

(c) CCNE1 , CPT1A, ERG, LRP5, MAP3K1 , SLC38A2, STC2, TFDP2, and TMPRSS2; or

20. The kit for of claim 19, further comprising a second set of probes for detecting the expression level of a set of normalization genes in the tumor sample.

43