Morphologic and Genomic Characteristics of Breast Cancers Occurring in Individuals with Lynch Syndrome

Lynch syndrome is a cancer predisposition syndrome characterized by germline mutations in major DNA mismatch repair (MMR) genes (MLH1, MSH2, MSH6, or PMS2) and associated with a significantly increased risk of developing colorectal, endometrial, ovarian, small bowel, and ureteral cancers (1–3). The hallmark of Lynch syndrome spectrum cancers is defective DNA mismatch repair deficiency (MMRd) and high levels of microsatellite instability (MSI-H), which occur upon loss of the expression of MMR genes through a germline mutation and inactivation of the second allele through somatic mutation or loss of heterozygosity, and, more rarely, due to MLH1 gene promoter methylation.

Whether individuals with Lynch syndrome are at increased risk of developing breast cancer remains controversial (2, 4–12). Although previous studies have suggested a lack of association between pathogenic germline variants affecting MMR genes and breast cancer risk (7, 11, 12), whether breast cancers occurring in Lynch syndrome individuals are causally related to the genetic alteration of MMR genes is unclear (2). Loss of MMR protein expression has been reported in up to 50% of breast cancers from patients with Lynch syndrome (8, 9). Davies and colleagues reported 2 of 11 patients with MMRd breast cancers identified by whole-genome sequencing occurring in patients with Lynch syndrome (10). Other studies have found no evidence of MMRd in breast tumors derived from patients with Lynch syndrome or individuals with synchronous or metachronous breast plus colorectal cancer (11). In fact, a recent pan-cancer survey of over 15,000 tumors subjected to Memorial Sloan Kettering–Integrated Mutation Profiling of Actionable Cancer Targets (MSK-IMPACT) found no evidence of MMRd breast cancer occurring in patients with Lynch syndrome (12).

Defining whether breast cancers developing in the context of Lynch syndrome display MMRd is not a mere academic exercise, given that MSI-H has been approved as a pan-cancer biomarker of clinical benefit from immune checkpoint inhibitors (13). Hence, the identification of MMRd/MSI-H breast cancers in patients with Lynch syndrome has a direct therapeutic implication. Here, we sought to characterize the pathologic and genomic features of breast cancers occurring in patients with Lynch syndrome and to explore whether these breast cancers would be causally linked to the MMRd characteristic of Lynch syndrome–related cancers. By expanding on the analysis by Latham and colleagues (12), including ≥20,000 patients subjected to multigene germline genetic testing to identify primary breast cancers in patients with Lynch syndrome and focusing on the histologic, IHC, and genomic features of these cancers, we provide direct evidence supporting MMRd/MSI-H in a substantial subset of breast cancers in patients with Lynch syndrome.

The cohort consisted of 20,110 patients with cancer, who underwent multigene panel genetic testing for germline mutations at Memorial Sloan Kettering Cancer Center (MSKCC, New York, NY) from 2015 to present day, as described previously (14). The cohort includes 3,583 patients reported previously from January 2014 to June 2017 by Latham and colleagues (12), and additional 16,527 unique patients derived from the germline sequencing pool. The tumor somatic alterations by MSK-IMPACT with > 468 gene panels of the identified breast cancers were also analyzed.

Individuals with likely pathogenic/pathogenic germline variant in MLH1, MSH2, MSH6, or PMS2 who developed breast cancers were retrospectively identified. The genetic testing was performed in a Clinical Laboratory Improvement Amendments–certified laboratory using tumor/normal sequencing. All germline Lynch syndrome variants for patients in this cohort were reviewed by a board-certified molecular pathologist (D. Mandelker) and classified according to the American College of Medical Genetics and Genomics criteria (15) as likely pathogenic/pathogenic.

The study was conducted in accordance with the Declaration of Helsinki, and the protocol was approved by the Institutional Review Board (IRB) of MSKCC. Every patient included in this study signed a written informed consent form according to the protocol approved by the MSKCC IRB.

Histology of all breast cancers were reviewed by two pathologists (C.J. Schwartz and H. Zhang) and graded according to the Nottingham grading index (16). Clinicopathologic data were retrieved from the electronic medical records and curated by C.J. Schwartz.

Representative formalin-fixed, paraffin-embedded (FFPE) tissue blocks of invasive carcinomas from patients with Lynch syndrome were retrieved from the archive at MSKCC. Estrogen receptor (ER) and human epidermal growth factor receptor 2 (HER2) status were assessed by IHC and/or FISH as described previously (17), following the American Society of Clinical Oncology/College of American Pathologists guideline recommendations (18, 19). IHC staining for MMR proteins (MLH1, PMS2, MSH2, and MSH6) were performed according to methods described previously (20). Positive and negative controls, including internal positive controls, were assessed in every run and every case, respectively. Tumors with positive nuclear expression were classified as normal/retained MMR protein expression. All IHC reactions were reviewed by three pathologists (C.J. Schwartz., J. Shia, and H. Zhang). PD-L1 (SP142) IHC was performed using Ventana Benchmark Ultra System (Ventana Medical Systems) with antibody detection using the OptiView DAB IHC Detection Kit (Ventana Medical Systems), according to the manufacturer's manual (21). PD-L1 was assessed by two pathologists (C.J. Schwartz and H. Zhang) and scored as expression on stromal tumor-infiltrating immune cells occupying ≥1% of the tumor area as described previously (22, 23).

All Lynch syndrome breast cancers were evaluated for the presence of stromal tumor-infiltrating lymphocytes (TIL) following the recommendations by the International TILs Working Group 2014 (24). Three pathologists (C.J. Schwartz, H.Y. Wen, and H. Zhang) quantified the percentage of stromal area within the borders of the invasive tumor covered by TILs, and the average was reported. TILs outside of the tumor border, around ductal carcinoma in situ (DCIS) and normal lobules, and in tumor zones with crush artifacts, necrosis and regressive hyalinization were excluded.

Adequate tumor and normal tissue samples were available from 11 patients and underwent targeted massively parallel sequencing (n = 11) using the FDA-cleared MSK-IMPACT assay, which targets all exons and selected introns of 468 (n = 7) or 505 (n = 4) cancer genes as described previously (25). Of the 11 cases, eight had sequencing performed at the time of diagnosis, and three were among the patients identified from the germline mutation database and were sequenced retrospectively at the time of this study.

In brief, somatic single-nucleotide variants (SNV) were detected using MuTect (v1.17; ref. 26) and small insertions and deletions were identified using a combination of Strelka (v1.0.15; ref. 27), VarScan2 (v2.3.7; ref. 28), Lancet (v1.0.0; ref. 29); and/or Scalpel (v0.5.3; ref. 30) with further curation using visual inspection as described previously (31). Copy-number alterations in the tumors were identified using FACETS (32) as described previously (31). Mutational hotspots were classified according to Chang and colleagues (33).

Analysis of MSI status by PCR in tumors (n = 11) was performed using the MSI Analysis System, version 1.2 kit (Promega). In brief, this assay analyzes five mononucleotide microsatellite loci, including NR-21, BAT-25, MONO-27, NR-24, and BAT-26, in both tumor and normal DNA. A ≥ 3 bp shift in the tumor DNA relative to the matched normal tissue constituted instability at one locus. Instability at ≥2 of the five microsatellite loci defines MSI-H status, with fewer than two unstable loci classified as microsatellite stable (MSS), according to methods described previously (34). One microliter of amplified PCR product was applied to the Applied Biosystems 3730 DNA Analyzer. Automatic fragment analysis was carried out with GeneMapper 6.0 Software (Applied Biosystems).

The mutational signatures exposures were obtained using Signature Multivariate Analysis (SigMA), a tool for the decomposition of mutational signatures that can be optimized for the analysis of targeted capture sequencing panel results obtained from FFPE tissue samples. It has been demonstrated that when only a few SNVs per sample are available, SigMA outperforms other state-of-the-art methods (35). SigMA computes signature exposures for samples with at least five SNVs, while other tools (e.g., DeconstructSig) require ≥10 variants (36). The SigMA was set with customized settings (i.e., data type: msk, cancer type: breast, check msi: true). The exposures obtained were converted using the NNLS algorithm by SigMA into percentages per signature as described previously (36). Additional analysis was performed using the tool SigProfiler, another state-of-the-art tool that identifies the probability for each signature to cause a specific mutation type in a cancer sample. The python package SigProfilerPlotting was used for the visualization of the mutational spectrum (Fig. 2G; ref. 37).

ER-positive/HER2-negative (n = 9), ER-positive/HER2-positive (n = 1), and ER-negative/HER2-negative (n = 1) Lynch syndrome breast cancers were compared with 1,918 breast cancers subjected to MSK-IMPACT targeted sequencing from the study by Razavi and colleagues (38). Tumors were matched by hormone receptor (HR) status, tumor type [invasive ductal (n = 9) or lobular carcinoma (n = 2), respectively], age (10-year intervals), and menopausal status at a 1:3 ratio (n = 33), as described previously (31). Comparison of frequencies of genes altered by somatic genetic alterations between Lynch syndrome breast cancers (n = 11) and control-matched breast cancers (n = 33) was performed using Fisher exact test. Multiple testing correction using the Benjamini–Hochberg method was applied to control for the FDR whenever appropriate. Tumor mutational burdens (TMB) and MSISensor scores were compared between Lynch syndrome breast cancers (n = 11) and control-matched breast cancers (n = 33) by the Mann–Whitney U test. Dominant mutational signatures of sequenced tumors with at least five SNVs were compared between Lynch syndrome breast cancers (n = 9) and control-matched breast cancers (n = 16) by Fisher exact test. P values < 0.05 were considered significant. Statistical analysis was performed using R (version 3.1.2).

Data are available in a repository (cBioPortal for Cancer Genomics) and can be accessed via a DOI link.

Two hundred seventy-two (1.4%) MMR gene pathogenic variants indicating Lynch syndrome were identified from the total cohort of 20,110 patients. This included 57 (21%), 96 (35%), 63 (23%), and 56 (21%) variants affecting MLH1, MSH2, MSH6, and PMS2, respectively. Thirteen patients in this cohort (5%) had been diagnosed with primary breast cancers (Table 1), all of whom had confirmatory germline testing using a multigene inherited cancer susceptibility panel comprised of 76 (n = 1) or 88 genes (n = 12). Two of these patients (cases 2 and 3) overlapped with a previous study (14).

Of the 13 patients, pathogenic germline variants were identified in MLH1 (n = 2, 15%), MSH2 (n = 2, 15%), MSH6 (n = 6, 46%), and PMS2 (n = 3, 23%). Seven patients (54%) had breast cancers as their sole malignancy, 5 patients (38%) had other Lynch syndrome–related cancers, and 1 patient was diagnosed with synchronous uterine and breast carcinomas (8%; Table 1). Eleven patients (84%) had a family history of Lynch syndrome–related neoplasia. The median age of breast cancer diagnosis was 47 years (range, 33–74).

Histologically, 10 breast cancers (77%) were invasive ductal carcinomas and three (23%) were invasive lobular carcinomas (Table 2). All breast cancers were of histologic grade 2 (38%) or 3 (62%). The median tumor size was 2.6 cm (range, 1.0–10.0). Of the 13 tumors, 11 tumors (84%) were ER-positive/HER2-negative, one tumor (case 8) was ER-positive/HER2-positive, and the remaining one tumor (case 7) was ER-negative/HER2-negative. Stromal TILs ranged from 1% to 25%. PD-L1 (SP142) staining was positive (>1%) in five of nine tumors (55%; Table 2).

In terms of locoregional disease control, 69% of patients received total mastectomy and 15.5% were treated with breast conservation therapy. The remaining 15.5% had metastatic disease at presentation. The median follow-up is 36 months (range, 3–240; Table 2). At the time of study, the majority of patients were either alive with disease (46%) or dead of disease (15%). The remaining cohort showed no evidence of disease (39%; Table 2).

We next sought to define whether breast cancers developing in the context of Lynch syndrome germline variants would harbor the hallmark features of MMRd. IHC analysis of 12 breast cancers (Table 1) with available material revealed loss of MMR proteins in five cases (42%), all of which are ER-positive/HER-2 negative breast cancers. Case 4, harboring an MLH1 c.954delC frameshift deletion, demonstrated loss of MLH1 and PMS2 with partial loss of MSH6 protein (Fig. 1A–F). Loss of MLH1 and PMS2 protein was also observed in case 13, harboring an MLH1 deletion. Deficiency of MSH2 and MSH6 protein was seen in two cases (cases 10 and 11), with missense mutations in MSH2. Isolated protein loss of MSH6 was observed in case 6, harboring a frameshift mutation resulting in a premature termination codon 12 amino acids downstream of p. Leu1330Valfs*12.

MSK-IMPACT analysis of 11 breast cancers revealed a repertoire of somatic mutations consistent with that reported for invasive breast cancers, including TP53 (54%), PIK3CA (45%), and GATA3 (27%; Fig. 1G; Supplementary Table S1). Three of four breast cancers that demonstrated MMR protein loss by IHC showed two hits in the proposed Lynch syndrome gene. Two cases (cases 4 and 13) showed loss of heterozygosity in an MMR gene, whereas case 6 harbored a second somatic inactivating mutation in MSH6 (p. Phe1088Leufs; Fig. 1G). Case 10 (germline MSH2) lacked evidence of biallelic inactivation despite loss of MSH2 and MSH6 proteins by IHC. Thus, inactivation of the gene may be the result of another inactivating regulatory process such as somatic hypermethylation, which has been reported to occur in Lynch syndrome individuals with MSH2 germline variants (39).

Of the four cases demonstrating MMR protein loss by IHC with adequate materials for sequencing, three showed evidence of MSI by PCR analysis of five microsatellite loci and one yielded an inconclusive result (Fig. 1G; Table 3). Three of four cases displayed high TMB (i.e., > 10 mutations per Mb; ref. 40) and either indeterminate (three cases) or high (one case) MSISensor scores (Fig. 1G; Table 3). The sole case (case 10) deemed MSI-H (score of > 10 by MSISensor) was a 63-year-old female with a missense mutation in the MSH2 gene. The tumor was also MSI by PCR (Table 3). Both patient and family members had history of Lynch syndrome–related cancers. The tumor was a moderately differentiated ER-positive/HER2-negative invasive ductal carcinoma with low TILs (2%). IHC analysis for PD-L1 was negative. The patient underwent breast conservation surgery followed by radiotherapy and hormonal therapy. The patient currently has no evidence of disease after 144 months of clinical follow-up.

Nine cases were found to have a least five SNVs (81%) and were subjected to mutational signature analysis with SigMA. All cases with MMR protein loss and indeterminate or high MSISensor scores demonstrated dominant microsatellite instability signatures. Notably, three of these MMRd breast cancers displayed expression of PD-L1 (>1%) in the tumor-associated inflammatory cells with variable percentages of tumor stromal infiltrating lymphocytes (mean, 14.3; range, 2–25%; Fig. 1G).

Given the higher frequency of MMRd/MSI (4/11, 36%) observed in the Lynch syndrome breast cancer cohort, and prior observations of MMRd in <2% of all breast cancers (10), we compared the TMBs, MSISensor scores, and dominant mutational signatures to a 1:3 cohort of breast cancers (n = 33) matched according to age, tumor type, HR profile, and menopausal status, respectively (Supplementary Table S2). Breast cancers in the Lynch syndrome cohort (n = 11) displayed a higher TMB (P = 0.003, Mann–Whitney U test; Supplementary Fig. S1A) and had MSI scores showing a positive trend toward significance (P = 0.069, Mann–Whitney U test; Supplementary Fig. S1B) when compared with control-matched breast cancers (n = 33). We also observed an enrichment in dominant MSI mutational signatures in tumors harboring at least five SNVs were also enriched in Lynch syndrome breast cancers (n = 11) relative to control-matched breast cancers (n = 16, P = 0.040, Fisher exact test; Supplementary Fig. S1C). We next compared the genomic characteristics of Lynch syndrome breast cancers relative to the control-matched breast cancer cohort (Supplementary Fig. S2). No statistically significant differences in single-gene comparisons were identified. Given the small sample size of Lynch syndrome breast cancers included in this study, further analyses to define differences in mutational frequencies between Lynch syndrome breast cancers and non–Lynch syndrome breast cancers are warranted.

In summary, five of 13 cases (38.5%) with tissue available for IHC and/or MSK-IMPACT displayed features of MMRd or MSI-H (summarized in Table 3).

Given that MSI-H has been approved as a pan-cancer biomarker of benefit from checkpoint inhibitors (41, 42) we queried whether any patients with Lynch syndrome with MMRd tumors had received immunotherapy. A 64-year-old female (case 6), with a germline MSH6 frameshift mutation and remote history of an ER-positive breast cancer 18 years prior, developed a contralateral ER-positive/HER2-negative/MMRd/PD-L1–positive breast cancer with synchronous lymph node and extensive bone metastases involving both the pelvis and lumbar spine (Fig. 2A–F). Anti-PD1 therapy (pembrolizumab) was initiated in the palliative setting with complete pathologic response at 5 months on PET imaging (Fig. 2G). The patient has no radiologic evidence of disease 13 months after initiation of therapy.

We explored the clinical, morphologic, and genomic characteristics of breast cancers occurring in individuals with Lynch syndrome from a single tertiary cancer center. Our study revealed 5 of 13 breast cancers in individuals with Lynch syndrome demonstrating MMRd by either IHC, PCR, and/or multigene panel studies. We were able to identify these breast cancers using a large sequencing pool from >20,000 patients who underwent germline testing at MSKCC (New York, NY). The majority were ER-positive/HER2-negative (85%), moderately to poorly differentiated breast cancers (100%). At a median follow-up of 36 months, the patients were AWD (2/5) or NED (3/5). One patient with advanced MMRd breast cancer and distant metastasis demonstrated durable and complete clinical response to immunotherapy administered in the palliative setting and currently is NED after 13 months of the initiation of immunotherapy.

Our findings provide direct evidence that a subset of primary breast cancers developing in the context of a Lynch syndrome germline pathogenic variant harbors features of MMRd, including MSI-H, in line with the previous observations showing MMRd breast cancers in the context of Lynch syndrome (8, 10). MSI-H has been approved as a pan-cancer biomarker of benefit from immune checkpoint inhibitors (41, 42). Consistent with the notion that some breast cancers developing in the context of Lynch syndrome germline pathogenic variants are bona fide MSI-H cancers, benefit from immune checkpoint inhibition was observed in case 6, as expected. Further studies beyond the case highlighted here are warranted.

MSISensor scores of ≥10 have been shown to reliably identify MSI-H status in solid tumors from various primary sites (34, 43). Applying this criterion in identifying MSI-H breast cancers has been explored. Latham and colleagues (12) identified the Lynch syndrome in 15,045 patients who had matched tumor/normal DNA sequencing encompassing over 50 cancer types through MSK-IMPACT platform in MSKCC from January 1, 2014 to June 30, 2017. They identified Lynch syndrome in 0.3% (7/2,371) of patients with breast cancer and all seven cases were MSS tumors based on MSISensor scores of <3. In our current study, we identified 272 individuals with Lynch syndrome among 20,110 patients with cancer who had consecutive genetic testing performed through MSKCC clinical pipeline with a cut-off date of November 30, 2020. Only 3,583 patients included in our study had the clinical genetic testing performed before the cut-off date of June 30, 2017 employed in Latham's study, representing the overlapping patient population between these two cohorts (up to 3,583 of 20,110; 18%). Two patients in the population included in both studies had developed primary breast cancers (cases 2 and 3 in our study) showing no evidence of MMRd. Including the overlapping 2 patients included in Latham's and additionally identified 11 patients unique to our study, 5 patients were identified to have MMRd breast cancers based on loss of MMR proteins by IHC (5/13) and/or tumor-normal sequencing studies on 11 cases with available tissue materials (4/11) showing dominant MSI mutational signatures, high TMB, and indeterminate or high MSISensor scores. Of the nine cases subjected to mutational signature analysis by SigMA, the four MMRd breast cancers harbored 57, 52, 45, and 17 SNVs, well above the five SNV threshold required for analysis (35). These cases were found to harbor a dominant MSI mutational signature and a mutational spectrum consistent with that of MSI-H cancers (i.e., mutational signatures 20 and 21). Only one MMRd breast cancer (case 10) in our study had a MSISensor score of >10, suggesting that MMRd breast cancers may have a lower MSISensor threshold than other MSI/MMRd cancers, such as colorectal and endometrial cancers, for which the assay was initially intended to identify (34). In line with the study from Latham and colleagues (12), our results suggest that further evaluation might be indicated to determine whether current MSISensor cut-off values can faithfully identify MMRd/MSI-H breast cancers.

This study has limitations. Despite the large cohort of patients subjected to tumor-normal MSK-IMPACT sequencing, the sample size of breast cancers developing in the context of Lynch syndrome was small, consistent with the rarity of MMRd and/or MSI-H in primary breast cancers. MMRd breast cancer is reported to represent around 1.7% of all breast cancers, with a much smaller fraction involving individuals with Lynch syndrome (10). Signature analysis could only be performed in 9 of 11 patients, given that only MSK-IMPACT sequencing was performed. Mutational signature decomposition when the number of SNVs is low (e.g., in the case of cancers subjected to MSK-IMPACT containing 5–10 somatic SNVs) may have suboptimal robustness. Although a substantial proportion of breast cancers occurring in patients with Lynch syndrome likely constitute sporadic breast cancers developing in the context of Lynch syndrome, whole-genome sequencing, to ascertain the frequency of MMRd mutational signatures, as well as MMR IHC analysis, to define the loss of MMR proteins, of consecutive breast cancers from patients with Lynch syndrome is warranted.

Although our study is not intended to address whether breast cancer belongs in the spectrum of Lynch syndrome–related cancers, it is important to note that at the population level, there appears to be no increased risk for breast cancer in Lynch syndrome individuals with MLH1, MSH2, and PMS2 variants (44, 45). Individuals with MSH6 variants may confer a slightly higher risk of developing breast cancer; however, significance was not achieved in a large case–control study (44). Currently, breast cancers are not considered Lynch syndrome–associated cancers per current NCCN guidelines due to confounding data regarding risk in patients with Lynch syndrome (46), and breast cancers developing in the context of Lynch syndrome are currently not included in the special management of hereditary breast cancers (47). It is plausible that the remaining tumors in our cohort likely correspond to sporadic breast cancers arising in Lynch syndrome individuals, as reported previously (48). With recent studies supporting universal genetic testing for patients with breast cancer ≤60 (49), it remains uncertain whether MMR genes should be added to the testing panel. Our results emphasize, however, that in up to 4 in 10 patients with Lynch syndrome who develop breast cancers, their tumors may be etiologically linked with MMRd caused by the pathogenic germline variant affecting MLH1, MSH2, MSH6, or PMS2 genes. These findings support the contention that MMR IHC analysis and/or MSI assessment by PCR or multigene sequencing should be performed in the tumor specimens from patients with Lynch syndrome who develop breast cancer.

In conclusion, we demonstrated pathologic and genomic features of MMRd/MSI-H in a subset of breast cancers arising in patients with Lynch syndrome, establishing a likely etiologic link between the germline alteration and hallmark features of MMRd/MSI in this context. Breast cancers occurring in Lynch syndrome should be tested for MMRd by either IHC or genomic sequencing to identify the patients who may benefit from immunotherapy.

F. Pareja reports grants from NCI/NIH outside the submitted work. P. Razavi reports grants from Grail, ArcherDx, Illumina, and Epic Sciences; grants and personal fees from Novartis; personal fees from AstraZeneca, Foundation Medicine, Tempus Labs, and Natera outside the submitted work. M.E. Robson reports personal fees from Change Healthcare outside the submitted work. J.Z. Drago reports personal fees from Biotheranostics outside the submitted work. H.Y. Wen reports personal fees from Merck and AstraZeneca outside the submitted work. L. Zhang reports other support from Decipher Medicine outside the submitted work. B. Weigelt reports personal fees from Repare Therapeutics outside the submitted work. J.S. Reis-Filho reports personal fees from Paige, Repare Therapeutics, Grupo Oncoclinicas, Roche Tissue Diagnostics, Novartis, Genentech, Roche, In Vicro, and Eli-Lilly outside the submitted work. H. Zhang reports personal fees from Roche/Genentech outside the submitted work. No disclosures were reported by the other authors.

C.J. Schwartz: Conceptualization, resources, formal analysis, methodology, writing–original draft, writing–review and editing. E.M. da Silva: Conceptualization, data curation, methodology. A. Marra: Conceptualization, data curation, formal analysis. A.M. Gazzo: Data curation, methodology. P. Selenica: Data curation, methodology. V.K. Rai: Formal analysis. D. Mandelker: Conceptualization, data curation, supervision, methodology. F. Pareja: Conceptualization, writing–review and editing. M. Misyura: Data curation, formal analysis. T.M. D'Alfonso: Conceptualization, writing–review and editing. E. Brogi: Conceptualization, writing–review and editing. P. Drullinsky: Conceptualization. P. Razavi: Conceptualization, writing–review and editing. M.E. Robson: Conceptualization. J.Z. Drago: Conceptualization, writing–review and editing. H.Y. Wen: Conceptualization, formal analysis, writing–review and editing. L. Zhang: Conceptualization. B. Weigelt: Conceptualization, writing–review and editing. J. Shia: Conceptualization, supervision, writing–review and editing. J.S. Reis-Filho: Conceptualization, resources, data curation, writing–review and editing. H. Zhang: Conceptualization, data curation, formal analysis, supervision, writing–review and editing.

This work was funded in part by the Marie-Josée and Henry R. Kravis Center for Molecular Oncology and the NCI Cancer Center Core grant #P30-CA008748. F. Pareja is partially funded by an NIH K12 CA184746 grant. F. Pareja, B. Weigelt, and J.S. Reis-Filho are funded in part by the NIH/NCI P50 CA247749 01 grant. E.M. da Silva is partially funded by the MSK-MIND grant. J.S. Reis-Filho is funded in part by a grant from the Breast Cancer Research Foundation.

The funders had no role in the design of the study; the collection, analysis, and interpretation of the data; the writing of the article; or the decision to submit the article.

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.