Microsatellite Instability–High Endometrial Cancers with MLH1 Promoter Hypermethylation Have Distinct Molecular and Clinical Profiles

This article is featured in Highlights of This Issue, p. 4159

DNA mismatch repair deficiency (MMR-D) or high microsatellite instability (MSI-H; ref. 1) is present in 20% to 40% of endometrial carcinomas and defines distinct molecular subclasses of endometrial cancer (2). The molecular mechanisms of MMR-D/MSI-H endometrial cancers include epigenetic MLH1 promoter hypermethylation (MLH1ph; 70%–75%), somatic (15%–20%), and germline (5%–10%) mutations in MLH1, MSH2, MSH6, PMS2, and/or EPCAM (3, 4).

MMR-D/MSI-H endometrial cancers exhibit specific pathologic features, including a predominance of tumor-infiltrating lymphocytes (TILs; refs. 3, 5). Accordingly, immune checkpoint inhibitors (ICIs) have been shown to be effective for patients with MSI-H endometrial cancers, and pembrolizumab and dostarlimab have been FDA approved for the treatment of recurrent MMR-D/MSI-H endometrial cancer (6, 7).

Differences in immunogenicity of tumors and patient outcomes between various MMR-D/MSI-H mechanisms are poorly understood (8, 9). Previously published data suggested that MMR-D endometrial cancers may have increased recurrence rates (10), in particular MLH1ph tumors (11). Data are conflicting, however, and other studies reported better outcomes for MSI-H endometrial cancer patients compared to those with copy-number high endometrial cancers, suggesting heterogeneity among MMR-D/MSI-H tumors (2). The potential differences in outcomes of MSI-H endometrial cancers, if confirmed, could have implications on treatment selection, given that prior clinical trials and subsequent approvals for ICIs have included all MMR-D endometrial cancers regardless of the underlying mechanism of MMR-D (6, 7). Whether the genomic landscapes vary according to MMR-D mechanism is currently unknown.

We sought to determine whether the clinicopathologic characteristics, genomic features [e.g., tumor mutational burden (TMB) and mutational signatures], and levels of immune infiltration would vary in endometrial cancer patients according to the mechanism of MMR-D (i.e., germline, somatic or MLH1ph). In addition, we investigated the impact of the MMR-D mechanism on oncologic outcomes of MSI-H endometrial cancers, including response to ICI.

Of 1,157 consented patients with endometrial cancer who underwent clinical FDA-authorized tumor-normal sequencing using Memorial Sloan Kettering-Integrated Mutation Profiling of Actionable Cancer Targets (MSK-IMPACT) between March 2015 and July 2020, 202 had MSI-H tumors using MSIsensor or pathogenic germline mutations affecting MLH1, MSH2, MSH6, PMS2, or EPCAM. Endometrial cancers with pathogenic POLE mutations (n = 6; ref. 12), non-uterine primary sites (n = 2), or no MMR immunohistochemistry (IHC) data (n = 10) were excluded, yielding a final cohort of 184 patients (Fig. 1). Molecular and clinicopathologic analyses were conducted using the entire cohort. To minimize referral biases, clinical outcomes were assessed only in patients who had their initial treatment at Memorial Sloan Kettering Cancer Center (MSK) and MSK-IMPACT sequencing within 6 months of diagnosis (n = 119; Fig. 1). Women who presented to MSK at time of recurrence were excluded from the analyses of clinical outcomes, as previously described (13). This study has been approved by the institutional review board (IRB) of MSK, and written informed consent was obtained from all patients. This study was conducted in accordance with the Declaration of Helsinki.

Massively parallel sequencing was performed on primary (n = 153) or recurrent (n = 31) endometrial cancers and matched normal blood using MSK-IMPACT targeting 341 (n = 1), 410 (n = 13), or 468 (n = 170) cancer-related genes (14, 15). The median sequencing depth was 637× (range, 129×–1,368×). Relevant genomic data derived from MSK-IMPACT included, but was not limited to, somatic and germline mutations in MLH1, MSH2, MSH6, PMS2, and EPCAM, MSIsensor score (MSIsensor, RRID:SCR_006418; refs. 16, 17), and TMB (somatic mutations per Mb). MSIsensor scores of ≥10 were considered MSI-H, ≥3 to <10 MSI-indeterminate, and <3 microsatellite stable (MSS; refs. 16, 17). Mutational signatures (COSMIC, version 3.1) were defined using deconstructSigs at default parameters for endometrial cancers with ≥20 single-nucleotide variants (SNV; ref. 18), as previously described (19). Germline sequencing was available for 179 (97%) patients, and for those with pathogenic germline variants, loss of heterozygosity (LOH) of the wild-type allele was assessed in tumors using FACETS (20).

IHC for MMR proteins MLH1, MSH2, MSH6, and PMS2 was performed in all 184 endometrial cancers to confirm the underlying MMR alteration (21). An aberrant staining pattern was defined as loss of nuclear immunoreactivity in the tumor cells. Nuclei of adjacent nonneoplastic tissue (e.g., immune and/or stromal cells) served as a positive internal control. Endometrial cancers exhibiting absent nuclear MLH1 and/or PMS2 staining by IHC underwent MLH1ph testing. MLH1ph status was determined by the bisulfite mediated detection of methylated cytosines, as described previously (22).

Endometrial cancers were classified by mechanism of MMR (i.e., germline, somatic or MLH1ph) and by affected gene (i.e., MLH1, MSH2, MSH6, PMS2); no pathogenic EPCAM alterations were identified. Endometrial cancers from patients harboring pathogenic/likely pathogenic germline variants in MMR genes were classified as MMR germline-altered (Lynch) according to the American College of Medical Genetics (ACMG) guidelines (23). For endometrial cancers harboring pathogenic somatic MMR mutations, tumors were classified according to the gene harboring two pathogenic alterations or based on the presence of a pathogenic mutation associated with LOH of the respective wild-type allele.

Hematoxylin and eosin–stained whole slide images were used to assess TILs within corresponding tumor regions sequenced by MSK-IMPACT. A semiquantitative estimation of the density and distribution of TILs within the intratumoral, stromal, and peritumoral compartments was performed by two gynecologic pathologists (K. Devereaux and L.H. Ellenson) independently, blinded from the results of the molecular analysis and following a previously reported scoring framework (24, 25). Intratumoral TILs were defined as lymphocytes admixed within malignant epithelium. A density (absent, 0; mild, 1; moderate, 2; marked, 3) and distribution (absent, 0; focal, 1; multifocal, 2; diffuse, 3) score was assigned. Stromal TILs were defined as lymphocytes within the stroma interdigitating and within the tumor, and an overall density percentage [(area occupied by stromal TILs/total area of stromal tissue) × 100] and distribution score (as described above) were assigned. Peritumoral TILs were defined as lymphocytes adjacent to the border of the tumor (<1 mm), and density and distribution scores (as described above) were assigned. A “TIL score” was generated by multiplying density and distribution scores (intratumoral range 0–9; stromal range 0–100; peritumoral range 0–9). Zones exhibiting crush artifact, poor preservation, and/or necrosis were excluded, as were endometrial biopsies and lymph node metastases (n = 11).

Clinicopathologic data were abstracted from electronic medical records. Stage at diagnosis was assigned following the International Federation of Gynecology and Obstetrics (FIGO) 2014 staging system (26). Tumor grade, histology, lymphovascular space invasion (LVSI), and microcystic, elongated, and fragmented (MELF) pattern data were abstracted from pathology reports. Adjuvant therapy was defined as any additional therapy given following surgery and included observation, chemotherapy, and intravaginal radiotherapy (IVRT)/external beam radiation therapy (EBRT). Body mass index (BMI) was collected from diagnosis, and obesity was defined as BMI ≥30 kg/m² (27).

To investigate the association of mechanism of MMR-D with clinical and treatment outcomes, patients who received their initial care at MSK were selected for this analysis, as described above (n = 119; Fig. 1). Progression-free survival (PFS) was measured from the date of diagnosis to date of progression, as determined by biopsy, surgical resection, or imaging. Overall survival (OS) was measured from the date of diagnosis to date of death. Patients alive and disease-free or alive with disease were censored for PFS and OS, respectively, at date of last-follow-up. A subset of patients receiving pembrolizumab for recurrence per FDA label were assessed for radiographic response following RECIST 1.1 criteria by independent, blinded radiology review (Y. Lakhman and W. Ma; ref. 28). Those receiving ICI on clinical trials were excluded. Clinical benefit was defined as patients who achieved a complete response, partial response, or stable disease as best response on therapy, while overall response rate was defined as patients who achieved a complete response or partial response as best response on therapy. Percent change in tumor volume was quantified by dividing the summation of target lesion diameter at best response by the summation of target lesion diameter at baseline (28).

Associations between MMR-D mechanism and continuous and categorical clinicopathologic variables were compared using Kruskal–Wallis and Fisher exact tests, respectively. TMB, MSIsensor score, mutational signatures, and TIL scores were compared between groups graphically and using appropriate nonparametric tests. For analyses of primary treatment after surgery, landmark analysis was applied with a 6-week landmark time, with 4 patients excluded due to early progression in PFS analyses (29). Survival curves were created using the Kaplan–Meier method. P values were generated by applying log-rank test for categorical variables and Wald-test based on Cox proportional hazards (CoxPH) model for continuous variables. Hazard ratios (HR) with 95% confidence intervals (CI) were obtained based on CoxPH model. A P value <0.05 was considered statistically significant. P values with fewer than 3 events were considered unreliable but are reported in a hypothesis-generating fashion. All statistical analyses were performed using R version 4.1.1 (https://cran.r-project.org/; CRAN, RRID:SCR_003005). In those with recurrent endometrial cancer treated with ICI, RECIST-defined responses were tabulated and plotted using swimmer's and waterfall plots.

Targeted sequencing data that support the findings of this study are available at cBioPortal [www.cbioportal.org, “Endometrial Carcinoma MSI (MSK, Clin Cancer Res 2022)”].

Of 184 endometrial cancer patients, 25 (14%) had germline MMR gene mutations, 39 (21%) had somatic MMR gene mutations, and 120 (65%) had MLH1ph (Fig. 1 and Table 1). Compared with patients harboring germline or somatic mutations, those with MLH1ph endometrial cancers were older at diagnosis (median 64 years MLH1 vs. 54 years germline/somatic; P < 0.001), more likely to be obese (58% MLH1ph vs. 42% germline and 26% somatic; P = 0.001), diagnosed at higher stages (P = 0.025) and had tumors with higher rates of LVSI (P = 0.033). There were no significant differences in histology, MELF pattern, or primary treatment between groups (Table 1). Also, within the group of endometrial cancers harboring MLH1 alterations, patients with MLH1ph endometrial cancers compared with those with MLH1 mutations (germline or somatic) were older at diagnosis (median 64 vs. 50 years; P < 0.001) and more likely to be obese (58% vs. 9%; P < 0.001). No other significant differences were observed (Supplementary Table S1).

The landscape of somatic mutations varied across endometrial cancers by MMR-D mechanism. In the germline group, tumors were enriched in ERBB2 (29%), ERBB3 (25%), and FBXW7 (46%) hotspot mutations compared with the MLH1ph group (3%, 2%, and 10%, respectively, P < 0.05). Alterations in JAK1, predominantly frameshift-indels, were enriched in MLH1ph endometrial cancers compared with the germline group (45% vs. 4%, P < 0.05). The frequencies of ARID1A, PTEN, and PIK3CA mutations were similar among the different MMR-D endometrial cancer groups (Fig. 2).

Median TMB was significantly lower among endometrial cancers with MLH1ph (32 mt/Mb, range 13–302 mt/Mb) compared with germline (44 mt/Mb, range 1–74 mt/MB) and somatic MMR mutations (48 mt/Mb, range 25–89 mt/Mb; P < 0.001; Supplementary Fig. S1A). All endometrial cancers in the MLH1ph and somatic groups were MSI-H; however, 83% (10/12) of endometrial cancers from MSH6 germline mutation carriers were classified as MSS or MSI-indeterminate, and 31% (4/13) of endometrial cancers from MLH1, PMS2, or MSH2 germline mutated endometrial cancers were classified as MSS or MSI-indeterminate (P < 0.02). Upon further review, in all MSS or MSI-indeterminate endometrial cancers from germline MMR mutation carriers, biallelic loss was observed in the tumor in 5 of 14 (36%) of cases. Low tumor content was detected in 9 of 14 (64%) of samples (Supplementary Table S2) potentially negatively impacting the accuracy of MSISensor and LOH assessments. Of note, among the 25 germline patients, there were 2 MMR-proficient and 3 equivocal MSH6-associated endometrial cancers as well as 2 equivocal MSH2-associated endometrial cancers (Supplementary Table S3).

Mutational signature analysis revealed a dominant aging-related signature 1 in 62% (15/25) of germline and 54% (20/39) of somatic MMR-D endometrial cancers compared with 23% (27/120) of MLH1ph endometrial cancers (P < 0.001). Conversely, 77% (88/120) of MLH1ph endometrial cancers had dominant MMR-D mutational signatures 6, 15, 20, or 26 compared with 38% (9/25) and 46% (17/39) of germline and somatic MMR-D endometrial cancers, respectively (P < 0.001; Table 1; Supplementary Fig. S1B).

While TILs were present in all endometrial cancer samples analyzed, TIL scores were significantly lower among endometrial cancers with MLH1ph compared to those with germline/somatic MMR mutations across all compartments (P = 0.002 intratumoral, P = 0.002 stromal, P = 0.001 peritumoral; Fig. 3; Supplementary Fig. S2).

Among 119 patients included in the survival analyses, 29 patients experienced progression; no deaths occurred without progression. Median follow-up for PFS was 25 months (range 1.7–60.0 months). In univariate analyses of PFS (Supplementary Table S4), MLH1ph was associated with inferior PFS (P = 0.005; Fig. 4), as was older age at diagnosis (P = 0.019), higher stage (P < 0.001) and presence of LVSI (P = 0.025). In contrast, TMB had borderline association with improved PFS (P = 0.06). In multivariate models including MMR-D mechanism, age and stage at diagnosis, and presence of LVSI, only stage was negatively associated with PFS (P < 0.001; Supplementary Table S4). Of note, an exploratory analysis revealed that two-year PFS was worse among patients with early-stage endometrial cancer (stage I/II) and MLH1ph (70.3%) compared with germline (100%) and somatic mutations (100%, Fig. 4C).

In the same cohort, 11 deaths were observed, and median follow-up for OS was 25.7 months (range 1.7–63.1 months; Fig. 4B). In univariate analyses of OS (Supplementary Table S5), higher stage was associated with worse OS (P = 0.002), and there was a trend towards higher TMB being associated with improved OS (P = 0.051). Multivariate analysis was not conducted due to the rarity of OS events.

In a subset of patients with recurrent disease treated with on-label pembrolizumab (n = 18), clinical benefit defined by RECIST 1.1 criteria was evaluated in a hypothesis-generating exploratory analysis among patients with germline (n = 1), somatic (n = 1), and MLH1ph endometrial cancers (n = 16). This subgroup included patients with grade 1 endometrioid (n = 5), grade 2 endometrioid (n = 4), grade 3 endometrioid (n = 6), undifferentiated/dedifferentiated (n = 2), and mixed (n = 1) histologies. Overall, 7 of 18 (39%) and 11 of 18 (61%) patients had stage I/II and stage III/IV disease at diagnosis, respectively. The median number of prior lines of cytotoxic therapy was 1 (range 0–2). One patient who was not a candidate for platinum-based chemotherapy received pembrolizumab for recurrence.

Clinical activity of pembrolizumab was consistent with published literature, and although limited by sample size, disease progression on pembrolizumab was observed in 4 of 16 (25%) patients with MLH1ph endometrial cancers with no progression observed for either the germline or somatic endometrial cancer patients (Fig. 5A and B). For the 16 patients with MLH1ph endometrial cancers, 2 had stable disease, 8 experienced partial response, and 2 had a complete response with a clinical benefit rate of 75% and overall response rate of 63%. The median TMB of MLH1ph endometrial cancers in patients receiving ICI was 33.8 mt/Mb (range 22.8–66.7 mt/Mb).

We demonstrate here that MLH1ph endometrial cancers have distinct clinicopathologic and molecular features as well as TIL densities when compared with germline and somatic MMR-mutated endometrial cancers, potentially affecting outcomes. MLH1ph endometrial cancers were observed more frequently in older and obese patients and were diagnosed at more advanced stages with higher rates of LVSI compared with endometrial cancer patients with germline and somatic MMR gene mutations. MLH1ph tumors exhibited a distinct mutational profile compared with the germline and somatic groups. TMB and TIL scores were also lower in MLH1ph compared with germline and somatic mutated endometrial cancers. Although PFS and OS were not significantly different between the various mechanisms of MMR-D and stage was the main driver of survival in multivariate models, when examining early-stage endometrial cancer only, MLH1ph was associated with worse PFS compared with germline and somatic MMR gene-mutated endometrial cancers.

Previous studies have reported conflicting findings regarding outcomes of MMR-D/MSI-H endometrial cancer compared to endometrial cancers of copy number-low or POLE molecular subtypes (5, 30–32), potentially reflecting the heterogeneity present in this group. Our findings corroborate and expand the results of previous studies, demonstrating that MLH1ph endometrial cancers are associated with high-risk features including older age, higher stage, obesity, and LVSI (3, 5, 10, 33), which may portend worse outcomes (34). Historically, early-stage endometrial cancer (stage I/II) is associated with 5-year survival rates of 87% to 96% (35); however, our study demonstrated a 2-year PFS rate of 70% in patients with MLH1ph endometrial cancer compared with 100% in both MMR germline and somatically mutated endometrial cancer. Although these results are limited given the sample size, our findings warrant additional research to confirm these findings in larger cohorts and to define whether patients with early-stage endometrial cancers with MLH1ph may benefit from novel therapies, such as radiation combined with immune checkpoint inhibition in the upfront setting (as in NRG-GY020 or NCT04774419) to prevent recurrence.

Our study revealed that MLH1ph endometrial cancers may not only have distinct molecular and immune profiles but also an enrichment of JAK1 mutations and lower TMB and TIL scores. This is consistent with previous reports on decreased presence of TILs (36) and PD-L1 expression (37) in MLH1ph compared with other MMR-D endometrial cancers as well as variation in TMB by underlying cause of MMR-D (38), suggesting distinct immunological entities with implications for treatment (39). Our data further the notion of a direct relationship between magnitude of TMB and response to immunotherapy across multiple cancer types, including endometrial cancer (40, 41). These findings indicate that substratification by MMR-D mechanism may be warranted in future trials, such as those investigating the impact of upfront ICI and radiation combinations for MMR-D ECs (NRG-GY020 and NCT04774419), those investigating the efficacy of dual immune checkpoint blockade (NRG-GY025), or those evaluating the efficacy of immune checkpoint blockade in combination with chemotherapy in the upfront or recurrent setting (NRG-GY018; ref. 42). Previous studies have described ICI response and resistance mechanisms in JAK1/2 (MMR-D colon cancer and melanoma), PBRM1 (clear cell renal carcinoma), and PTEN (acquired ICI resistance in uterine leiomyosarcoma; refs. 43–46). A recent phase II study of pembrolizumab in recurrent, MSI-H endometrial cancer found worse responses in tumors with MLH1ph compared with somatic MMR mutations and hypothesized that alterations in the JAK pathway may be associated with ICI resistance (47). We have also observed that MLH1ph endometrial cancers display a significantly higher rate of JAK1 alterations compared with germline and somatic MMR-D tumors, and future studies and clinical trials should investigate the need for more aggressive treatments and combination immunotherapy to overcome potential resistance and augment response.

Of note, we found similar outcomes between germline and somatic MMR-D endometrial cancer with comparable levels of TIL infiltration. The prevalence of mutations in genes commonly altered in endometrial cancers, including PTEN, PIK3CA, and ARID1A were similar across different MMR groups. However, MMR germline-mutated endometrial cancers were enriched in ERBB2 hotspot mutations, which have emerged as a target through irreversible HER2 tyrosine kinase inhibition in non–small cell lung and breast cancers (48). Additional investigation of this potentially targetable alteration in the setting of Lynch syndrome–associated endometrial cancers is still required, including determining the frequency of ERBB2 mutations and whether they represent early clonal/truncal versus late subclonal events.

Our study has several limitations beyond those inherent to retrospective research. The patient selection is restricted to those with endometrial cancers who underwent tumor-normal sequencing and may exclude patients who underwent germline only testing through other avenues, potentially underestimating the germline component of MMR-D endometrial cancers. The mutational analyses were restricted to 341–468 cancer-related genes, and it is possible that unexplored differences in genetic alterations may underpin differences observed between MMR-D mechanism. Although we utilized a comprehensive TIL score to correlate the immune microenvironment with the genomic findings, future assessments incorporating multiplex IHC to allow differentiation of CD4/CD8⁺ T-cell populations as well as other immune markers such as PD-L1 may be informative. Although independent radiologic review and RECIST classification were conducted to establish ICI response, imaging was obtained outside of a clinical trial setting and was thus subject to less standardized time intervals. In addition, we noted discrepancies between IHC-based MMR and molecular MSI assessments among groups, particularly in the germline endometrial cancer group. Of the MMR germline-mutated endometrial cancers 28% were MMR-proficient/equivocal on IHC, compared with none of either the somatic or MLH1ph groups, and 56% of germline MMR-D endometrial cancers had stable or indeterminate MSIsensor score. Together with previous reports showing that molecular MSI analysis has lower sensitivity for MMR-D detection in endometrial cancer than in colorectal cancer, and that a subset of molecular MSI-H cancers have retained/distinct MMR IHC expression patterns (49, 50), our work underpins the importance of conducting both IHC and MSI assessment in the diagnostic workup of endometrial cancers. Finally, given the sample size restrictions of this study, the exploratory analyses investigating the associations between MLH1ph endometrial cancers and PFS events in early-stage disease as well as the associations between subclassification of MMR-D endometrial cancers and response to ICI should be interpreted with caution and requires additional study using a larger cohort.

In summary, our findings suggest that subclassification of MMR-D endometrial cancer is likely warranted, given the clinicopathologic, genetic, and immunologic differences based on the mechanism of MMR-D reported in this study. These differences may influence clinical outcomes and response to treatment. Future therapeutic studies should account for these differences to design targeted interventions toward this high-risk group of MLH1ph endometrial cancers and explore biomarkers in addition to MMR-D/MSI and TMB for immune-based therapies.

Y.L. Liu reports grants from AstraZeneca, GlaxoSmithKline, and Repare Therapeutics outside the submitted work. M.M. Rubinstein reports grants from Merck and AstraZeneca outside the submitted work. C.F. Friedman reports personal fees from Seagen and Bristol Meyers Squibb, other support from Merck, and other support from Genentech outside the submitted work; and Research funds to institution from Merck, Genentech/Roche, Bristol Myers Squibb, Daiichi, and AstraZeneca. C. Aghajanian reports personal fees from Eisai/Merck, Mersana Therapeutics, Roche/Genentech, AbbVie, AstraZeneca/Merck, and Repare, grants from AbbVie, Clovis, Genentech, and AstraZeneca, and other support from GOG Foundation, Board of Directors (unpaid) outside the submitted work. N.R. Abu-Rustum reports grants from GRAIL outside the submitted work. Z.K. Stadler reports other support from Genentech/Roche, Adverum, Gyroscope Therapeutics, RegenexBio, Neurogene, Outlook Therapeutics, Optos Plc, and Regeneron outside the submitted work. J.S. Reis-Filho reports personal fees from Personalis during the conduct of the study, personal fees from Paige, Repare Therapeutics, Goldman Sachs, Grupo Oncoclinicas, Roche Tissue Diagnostics, MSD, Daiichi, and Personalis outside the submitted work. D. Zamarin reports grants from National Cancer Institute during the conduct of the study. D. Zamarin also reports grants and personal fees from Genentech, AstraZeneca, and Synthekine; grants from Plexxikon; personal fees from Memgen, Xencor, Targovax, Tessa Therapeutics, Agenus, Celldex, Crown Biosciences, GSK, Astellas, and Takeda; other support from Immunos and Accurius; and personal fees and other support from Calidi Biotherapeutics outside the submitted work; and reports a patent for Newcastle Disase Virus for cancer therapy licensed and with royalties paid from Merck. Y. Lakhman reports grants from the National Institutes of Health/National Cancer Institute Cancer Center Support Grant (P30 CA008748), personal fees from Calyx Clinical Trial Solutions, and other support from Y-mAbs Therapeutics outside the submitted work. B. Weigelt reports personal fees from Repare Therapeutics outside the submitted work. No disclosures were reported by the other authors.

B.L. Manning-Geist: Conceptualization, data curation, validation, investigation, writing–original draft, writing–review and editing. Y.L. Liu: Conceptualization, data curation, investigation, writing–original draft, writing–review and editing. K.A. Devereaux: Investigation, methodology, writing–review and editing, pathology review. A. Da Cruz Paula: Formal analysis, investigation, writing–review and editing. Q.C. Zhou: Formal analysis, writing–review and editing. W. Ma: Investigation, writing–review and editing. P. Selenica: Formal analysis, investigation, writing–review and editing. O. Ceyhan-Birsoy: Data curation, writing–review and editing. L.A. Moukarzel: Conceptualization, writing–review and editing. T. Hoang: Formal analysis, writing–review and editing. S. Gordhandas: Data curation, writing–review and editing. M.M. Rubinstein: Resources, writing–review and editing. C.F. Friedman: Resources, writing–review and editing. C. Aghajanian: Resources, writing–review and editing. N.R. Abu-Rustum: Resources, writing–review and editing. Z.K. Stadler: Conceptualization, writing–review and editing. J.S. Reis-Filho: Conceptualization, writing–review and editing. A. Iasonos: Formal analysis, writing–review and editing. D. Zamarin: Resources, writing–review and editing. L.H. Ellenson: Investigation, methodology, writing–review and editing. Y. Lakhman: Investigation, writing–review and editing. D.L. Mandelker: Conceptualization, data curation, supervision, investigation, writing–review and editing. B. Weigelt: Conceptualization, data curation, supervision, investigation, writing–original draft, writing–review and editing.

Research reported in this publication was funded in part by a Cancer Center Support Grant of the National Institutes of Health (NIH)/National Cancer Institute (NCI) (Grant No. P30CA008748). J.S. Reis-Filho and B. Weigelt are funded in part by NIH/NCI P50 CA247749 01 and Breast Cancer Research Foundation grants, and B. Weigelt by Cycle for Survival. The funders of this study had no role in the design of the study; the collection, analysis, and interpretation of the data; the writing of the manuscript; and the decision to submit the manuscript for publication.

The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked advertisement in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.