Published OnlineFirst August 31, 2018; DOI: 10.1158/1541-7786.MCR-18-0619

Genomics Molecular Cancer Research Genomic Alterations Associated with Recurrence and TNBC Subtype in High-Risk Early Breast Cancers Timothy R. Wilson1, Akshata R. Udyavar2, Ching-Wei Chang3, Jill M. Spoerke1, Junko Aimi1, Heidi M. Savage1, Anneleen Daemen2, Joyce A. O'Shaughnessy4,5,6, Richard Bourgon2, and Mark R. Lackner1

Abstract

The identification of early breast cancer patients who may sequencing panel was used to compare tumor specimens benefit from adjuvant chemotherapy has evolved to include from patients who had a recurrence event with a matched set assessment of clinicopathologic features such as tumor size who did not. The somatic mutation and copy-number and nodal status, as well as several -expression profiles alteration landscapes of high-risk early breast cancer for ER-positive, HER2-negative cancers. However, these patients were characterized and alterations associated with tools do not reliably identify patients at the greatest risk relapsewereidentified. Tumor mutational burden was eval- of recurrence. The mutation and copy-number landscape of uated but was not prognostic in this study, nor did it triple-negative breast cancer (TNBC) subtypes defined by correlate with PDL1 or CD8 . However, gene expression is also largely unknown, and elucidation of TNBC subtypes had substantial genomic heterogeneity with this landscape may shed light on novel therapeutic oppor- a distinct pattern of genomic alterations and putative under- tunities. The USO01062 phase III clinical trial of standard lying driver mutations. chemotherapy (with or without capecitabine) enrolled a cohort of putatively high-risk patients based on clinical Implications: The present study uncovers a compendium of features, yet only observed a 5-year disease-free survival genomic alterations with utility to more precisely identify event rate of 11.6%. In order to uncover genomic aberra- high-risk patients for adjuvant trials of novel therapeutic tions associated with recurrence, a targeted next-generation agents.

Introduction been shown to provide additional prognostic information beyond traditional IHC-based classification. More recently, Curtis Breast cancer is a highly heterogeneous disease which for and colleagues subtyped breast cancer into 10 distinct integrative decades has been subtyped and treated based on the IHC staining subtypes based on whole-genome analysis of copy-number altera- of three receptors: estrogen receptor (ER), progesterone receptor tions and gene expression, again showing distinct prognostic (PR), and epidermal growth factor receptor 2 (ERBB2, HER2). implications across subtypes (6). With the seminal paper by Perou and colleagues (1), and follow Within the triple-negative subtype of breast cancer (TNBC; i.e., on work from other groups (2–4), the breast cancer community those that stain negative by IHC for ER, PR, and HER2), four to six began to appreciate the molecular heterogeneity that exists within distinct biological subtypes have been defined at the transcrip- breast cancer at the transcriptional level. Gene-expression–based tional level (7–9). For example, in the study by Lehmann and classifiers, such as PAM50 (5), MammaPrint (3), and others, have colleagues, the authors grouped TNBC into six subtypes: Basal- like 1 (BL1), Basal-like 2 (BL2), Immunomodulatory (IM), Mes- enchymal (M), Mesenchymal Stem-like (MSL), and Luminal AR 1Department of Oncology Biomarker Development, Genentech, Inc., South San (androgen receptor, LAR; ref. 7). Interestingly, TNBC cell lines Francisco, California. 2Department of Bioinformatics and Computational classified according to Lehmann and colleagues displayed differ- 3 Biology, Genentech, Inc., South San Francisco, California. Department of ent sensitivities to chemotherapeutics and/or targeted therapies, 4 Biostatistics, Genentech, Inc., South San Francisco, California. US Oncology, suggesting that different genomic alterations may drive each Dallas, Texas. 5Baylor University Medical Center, Dallas, Texas. 6Texas Oncology, Dallas, Texas. subtype. Additional work from other groups, including our own, has shown that the TNBC subtypes have implications for path- Note: Supplementary data for this article are available at Molecular Cancer ologic complete response rates following neoadjuvant therapy Research Online (http://mcr.aacrjournals.org/). (10, 11) and for disease-free survival (DFS) following adjuvant T.R. Wilson and A.R. Udyavar contributed equally to this article. chemotherapy (12). To date, the mutational and copy-number Corresponding Authors: Timothy R. Wilson, Genentech, Inc., 1 DNA Way, MS 422A, profiles associated with the TNBC subtype have not been fully South San Francisco, CA 94080. Phone: 650-467-8872; Fax: 650-225-5770; characterized. E-mail: [email protected]; and Mark R. Lackner, Phone: 650-225-1000; Surgical resection of the tumor followed by adjuvant therapy to Fax: 650-225-5770; E-mail: [email protected] eradicate micrometastatic lesions is potentially curative in doi: 10.1158/1541-7786.MCR-18-0619 patients with early breast cancer. Early screening, incorporation 2018 American Association for Cancer Research. of hereditary risks and treatment improvements have dramatically

www.aacrjournals.org 97

Downloaded from mcr.aacrjournals.org on September 23, 2021. © 2019 American Association for Cancer Research. Published OnlineFirst August 31, 2018; DOI: 10.1158/1541-7786.MCR-18-0619

Wilson et al.

+ A HER2 N = 57 PAM50 ER_IHC PR_IHC 17q ERBB2 17q CDK12 ERBB2 ERBB2 TP53 CDK12 CDK12 PIK3CA TP53 TP53 SPOP PIK3CA PIK3CA TP53 CDK12 SPOP SPOP GPR124 GPR124 GPR124 MYC MYC MYC PIK3CA RARA SPOP CD79B RARA RARA RNF43 MYC CD79B CD79B GPR124 RNF43 RNF43 BRIP1 RARA BRIP1 BRIP1 17q GNA13 GNA13 CD79B 17q GNA13 RAD51C

RNF43 RAD51C RAD51C

PRKAR1A BRIP1 PRKAR1A PRKAR1A ARID1B ARID1B ARID1B GNAS GNA13 GNAS GNAS MLL2 MLL2 RAD51C MLL2 MLL3 MLL3 MLL3 ARID1B PIK3C2B GNAS

PRKAR1A PIK3C2B PIK3C2B PRKDC MLL2 PRKDC PRKDC ZNF703 MLL3 ZNF703 ZNF703 FGF19 FGF19 FGF19 FGF3 FGF3 FGF3

11q PIK3C2B 11q CCND1 PRKDC CCND1 CCND1

FGF4 ZNF703 FGF4 FGF4 FGF19 BRCA2 FGF3 BRCA2 BRCA2 FGFR1 FGFR1 FGFR1 FGF4 HOXB13 CCND1 HOXB13 HOXB13 TOP2A TOP2A TOP2A FANCM BRCA2 FANCM FANCM FGFR1 FAT3 XB13 FAT3 FAT3

LRP1B OP2A LRP1B LRP1B HO RANBP2 T RANBP2 RANBP2 FAT3

CCNE1 FANCM CCNE1 CCNE1 CDKN2A CDKN2A CDKN2A PREX2 LRP1B PREX2 PREX2 RUNX1T1 RUNX1T1 RUNX1T1 RANBP2 SPTA1 CCNE1 SPTA1 SPTA1 BLM BLM BLM A1 CDKN2A EP300 PREX2 EP300 EP300 FLT4

FLT4 BLM FLT4 SPT UNX1T1 MAP3K1 MAP3K1

MAP3K1 R NBN NBN NBN EP300 NOTCH2 FLT4 NOTCH2 NOTCH2 TP53BP1 TP53BP1

TP53BP1 NBN ZNF217 ZNF217 MAP3K1 TCH2 0 255075100 O

N Prevalence TP53BP1 ZNF217 HR+ N = 178 B PAM50 ER_IHC PR_IHC

PIK3CA PIK3CA PIK3CA MLL3 MLL3 MLL3 MAP3K1 MAP3K1 MAP3K1

CCND1 MLL3 CCND1 CCND1 FGF19 FGF19 FGF19 11q 11q FGF4 FGF4

MAP3K1 FGF4

FGF3 CCND1 FGF3 FGF3 FGF19

GPR124 FGF4 GPR124 GPR124

8p ZNF703 FGF3 8p ZNF703 ZNF703 FGFR1 FGFR1 FGFR1

GATA3 GPR124 GATA3 GATA3 ZNF703 MLL2 A3 MLL2 MLL2 FGFR1 BRCA2 AT BRCA2 BRCA2 G

TP53 MLL2 TP53 TP53 MYC MYC MYC BRCA2 NBN TP53 NBN NBN

MYC 8q 8q RUNX1T1 RUNX1T1 RUNX1T1 PREX2 NBN PREX2 PREX2 ARID1B ARID1B ARID1B CDH1 CDH1 CDH1 PREX2 GNAS RUNX1T1 GNAS GNAS ARID1B

MYST3 CDH1 MYST3 MYST3 PIK3C2B PIK3C2B GNAS PIK3C2B PRKDC PRKDC PRKDC MLL MYST3 MLL MLL SPTA1 SPTA1 SPTA1 PIK3C2B MLL PRKDC ZNF217 A1 ZNF217 ZNF217 T FAT3 FAT3 FAT3 SP

IGF1R T3 IGF1R

A IGF1R F RANBP2 ZNF217 RANBP2 RANBP2 ARID1A ARID1A ARID1A IGF1R FAT1 FAT1 FAT1

T1 POLE

RANBP2 POLE FA ARID1A 0255075100

POLE Prevalence C TNBC N = 162 PAM50 TP53 TP53 TP53 NOTCH1 NOTCH1 NOTCH1 GATA3 GATA3 GATA3

PTEN A3 PTEN PTEN T MYC TCH1 MYC MYC

PRKDC GA NO PRKDC PRKDC PTEN 8q PREX2 MYC 8q PREX2 PREX2 RUNX1T1 RUNX1T1 RUNX1T1 NBN NBN NBN BRCA1 PRKDC

PREX2 BRCA1 BRCA1

RB1 NBN RB1 RB1 FAT3 FAT3 FAT3 RUNX1T1 RB1 KDM5A T3 KDM5A KDM5A BRCA1

PIK3C2G FA PIK3C2G PIK3C2G 12p LRP6 12p LRP6 LRP6 RAD52 RAD52 RAD52 SPEN KDM5A SPEN SPEN PIK3CA LRP6 PIK3CA PIK3CA BRCA2 PIK3C2G RAD52 BRCA2 BRCA2 MLL3 SPEN MLL3 MLL3 SPTA1 SPTA1 SPTA1 SDHC PIK3CA SDHC SDHC BRCA2 1q NTRK1 MLL3 1q NTRK1 NTRK1

DDR2 SPTA1 DDR2 DDR2 GNAS SDHC GNAS GNAS LRP1B LRP1B LRP1B NTRK1 TSC2 DDR2 TSC2 TSC2 GPR124 GNAS GPR124 GPR124

MLL2 LRP1B MLL2 MLL2 NF1 TSC2 NF1 NF1 NOTCH2 NOTCH2 NOTCH2 MLL2 NF1 ARID1B GPR124 ARID1B ARID1B IRS2 IRS2 IRS2 MCL1 TCH2 MCL1 MCL1

PTPRD IRS2 PTPRD NO PTPRD ROS1 ARID1B ROS1 ROS1 ZNF703 MCL1 ZNF703 ZNF703 TRRAP TRRAP TRRAP PTPRD CCNE1 ROS1 CCNE1 CCNE1 CREBBP CREBBP CREBBP MAP3K1 ZNF703 MAP3K1 MAP3K1 TRRAP PIK3C2B PIK3C2B CCNE1 CCNE1 0255075100 CREBBP CREBBP MAP3K1 MAP3K1 Prevalence PIK3C2B PIK3C2B

Alteration type HER2_IHC Amplification Positive Log2 odds ratio Loss Negative Alteration type Short variant Co-occurence Short variant 8 Rearrangement ER_IHC Amplification Positive 4 PAM50 Negative Loss 0 Basal Rearrangement Mutual exclusivity Her2 LumA PR_IHC LumB Positive Normal Negative

98 Mol Cancer Res; 17(1) January 2019 Molecular Cancer Research

Downloaded from mcr.aacrjournals.org on September 23, 2021. © 2019 American Association for Cancer Research. Published OnlineFirst August 31, 2018; DOI: 10.1158/1541-7786.MCR-18-0619

Genomic Profiling of High-Risk Early Breast Cancers

ABHER2+ vs. HR+ HER2+ vs. TNBC C HR+ vs. TNBC 35.0 35.0 35.0 TP53 ERBB2 ERBB2 20.0 20.0 CDK12 20.0 CDK12 10.0 10.0 10.0 PIK3CA SPOP CCND1 5.0 TP53 5.0 5.0 KDM5A PIK3C2G CD79B PIK3CA FGF3 FGF19 FDR FDR GNA13 BRCA1 RAD51C FDR FGF4 NOTCH1 RB1 TP53 KRAS

2.5 SPOP 10 10 2.5 RNF43 2.5 CDH1 MDM4 PTEN 10 LRP6 BRIP1 SDHA CD79B FGF19 ZNF703 LRP1B PRKAR1A AURKA RAD52 RARA RAD51C TOP2A RB1 KDM5A MAP3K1 FGF6 CCND3 −log FGF3 POLD1

−log CBFB

−log NF1 1.0 TOP2A GNA13 1.0 FGF4 NOTCH1 1.0 LY N CDKN1A AURKA BRCA1 BRIP1 MYC HOXB13 RNF43 CCND1 CHD4 GPR124 CUL4A PDCD1LG2 FANCM

−5 0 5 −5 0 5 −5 0 5 log2 Odds ratio log2 Odds ratio log2 Odds ratio

FDR significance Not significanta Significant FDR significance Not significanta Significant FDR significance Not significanta Significant DE F + + HER2 vs. HR HER2+ vs. TNBC HR+ vs. TNBC 10.0 P53 PATHWAY 10.0 WNT BETA CATENIN SIGNALING 10.0 ALLOGRAFT REJECTION E2F TARGETS

6.0 PI3K AKT MTOR SIGNALING 6.0 DNA REPAIR 6.0 EPITHELIAL MESENCHYMAL 4.0 MYOGENESIS 4.0 TRANSITION OXIDATIVE PHOSPHORYLATION 4.0 COMPLEMENT UV RESPONSE DN MYOGENESIS APICAL SURFACE APICAL SURFACE DNA REPAIR APOPTOSIS 2.0 PI3K AKT MTOR SIGNALING HEME METABOLISM INTERFERON GAMMA RESPONSE 2.0 ALLOGRAFT REJECTION WNT BETA CATENIN SIGNALING 2.0 ESTROGEN RESPONSE LATE UV RESPONSE DN UNFOLDED PROTEIN E2F TARGETS EPITHELIAL MESENCHYMAL TRANSITION FATTY ACID METABOLISM MYC TARGETS V1

FDR RESPONSE OXIDATIVE PHOSPHORYLATION FDR FDR

10 P53 PATHWAY HEDGEHOG SIGNALING

10 HEME METABOLISM

PEROXISOME 10 ESTROGEN RESPONSE LATE XENOBIOTIC METABOLISM −log −log −log

0.2 0.2 0.2

−5.0 −2.5 0.0 2.5 5.0 −5.0 −2.5 0.0 2.5 5.0 −5.0 −2.5 0.0 2.5 5.0 log2 Odds ratio log Odds ratio 2 log2 Odds ratio FDR a Not significanta Significant FDR a Not significanta Significant FDR a Not significanta Significant

Figure 2. Enrichment of genomic alterations and hallmark pathways by IHC subtype. Volcano plots show the enrichment of overall alterations (short variants, copy-number alterations and rearrangements) for single (A–C) or MSigDB (18) Hallmark pathway gene sets (D–F). The x axis represents the log2 odds ratio that a gene would be altered in one IHC subtype over the other as indicated by each title. Title subtype A versus subtype B indicates that highlighted genes/pathways on the left-hand side of the graph are higher expressed in subtype A, while genes/pathways highlighted on the right-hand side of the graph are higher expressed in subtype B. The y axis represents the negative log10 Benjamini and Hochberg FDR adjusted P value, applied to each panel independently. The horizontal gray line denotes adjusted P value of 0.2 and two vertical gray lines denote log2 odds ratio of 1. Only the genes and pathways that have log2 odds ratio > 1 and adjusted P value < 0.2 are annotated. improved survival (13, 14); however, a subset of patients will still M0, and both ER- and PR-negative). Despite this enrichment recur with metastatic disease. Defining the subset at very high risk strategy, the 5-year DFS rate was 88.4% (15), highlighting the for recurrence remains challenging and slows the development of need for additional means to define truly high-risk patients. investigational agents that are attempting to show improvement In the current study, we molecularly profiled tumors from the in 5-year DFS, because the large majority of patients will not USO01062 trial using the FoundationOne next-generation develop disease recurrence within 5 years. In order to enrich for sequencing (NGS) platform. We focused on those patients who patients most in need of novel therapies, adjuvant clinical trial experienced a DFS event and compared them with a demograph- attempt to enroll high-risk patients largely based on clinical ically matched set of patients from this trial who did not expe- features such as tumor size and nodal status. The USO01062 rience a DFS event. We identify high-risk molecular traits that may adjuvant phase III trial of standard chemotherapy, with or without be used to more accurately select patients at a high-risk of capecitabine, enrolled 2,611 patients based on high-risk clinical recurrence at 5 years. We also report the mutational, copy-number features (T1–3, N1–2, M0; or T > 2 cm, N0, M0; or T > 1 cm, N0, alteration, and rearrangement landscape of TNBC subtypes.

Figure 1. Genomic landscape within IHC breast cancer subtypes. Figure displays the genomic landscape of (A) HER2þ,(B)HRþ and (C) TNBC denoting the most frequently mutated genes in each IHC subtype. Left, log2 odds ratio for co-occurrence of mutation in gene pairs, with mutually exclusive events corresponding to a negative log2 odds ratio (violet color), and co-occurrence events corresponding to a positive log2 odds ratio (green color). Only gene pairs with adjusted P value < 0.2 are colored in the plots. Middle, Tile plots showing the detailed genomic landscape of the most frequently altered genes, within each IHC subtype. PAM50 status and HER2/ER/PR status measured by IHC are shown as relevant at the top of each plot. Commonly coamplified genes on same loci are grouped and denoted by black brackets. Right, Bar plots denoting the overall prevalence of genes delineating the individual prevalence of the four types of alterations—short variant (green), amplification (red), loss (blue), and rearrangement (orange).

www.aacrjournals.org Mol Cancer Res; 17(1) January 2019 99

Downloaded from mcr.aacrjournals.org on September 23, 2021. © 2019 American Association for Cancer Research. Published OnlineFirst August 31, 2018; DOI: 10.1158/1541-7786.MCR-18-0619

Wilson et al.

Materials and Methods either docetaxel or docetaxel plus capecitabine as adjuvant ther- apy for female patients with high-risk breast cancer (clinicaltrials. USO01062 (NO17629) study gov/show/NCT00089479; ref. 15). Tissue samples were collected Patients were enrolled onto the parent study USO01062, A and analyzed following approval by the US Oncology Institu- randomized, open-label, multicenter, phase III trial comparing tional Review Board and appropriate confirmation of written regimens of doxorubicin plus cyclophosphamide followed by informed consent.

Targeted NGS profiling + A HER2 : DFS Samples were submitted to a CLIA-certified, New York State- AR * accredited, and CAP-accredited laboratory (Foundation Medi- MCL1 * cine) for NGS-based genomic profiling using the Foundation HSD3B1 Mut_with_events Mut HR P FDR Medicine FoundationOne comprehensive genomic panel (16). TNFAIP3 AR 3 3 11.1 0.002 0.15 * MCL1 3 3 8.81 0.002 0.15 * Tumor mutational burden (TMB) was determined by Chalmers BCL6 HSD3B1 3 3 9.73 0.004 0.2 TNFAIP3 3 3 5.08 0.016 0.5 ZR and colleagues (17). See Supplementary Methods for more BCL6 2 2 7.15 0.021 0.5 INSR INSR 2 2 6.56 0.023 0.5 JUN 2 2 6.56 0.023 0.5 detail. JUN EPHA7 2 2 5.54 0.034 0.57 FGFR4 2 2 8.91 0.038 0.57 EPHA7 KIT 2 2 5.32 0.038 0.57 LTK 2 2 5.24 0.041 0.57 Prevalence of alterations within IHC subtypes LRP6 2 2 5.47 0.048 0.59 FGFR4 Prevalence for each type of alteration was computed by calcu- KIT lating the sum of alterations in a given gene for all patients divided LTK by the total number of patients in each IHC subtype. Relative LRP6 prevalence in Fig. 6; Supplementary Fig. S8 for TNBC subtypes was 0.1 1.0 10.0 200.0 Hazard Ratio calculated by subtracting the prevalence of a given gene within each TNBC subtype by the prevalence of that gene in all TNBC + B HR : DFS Mut_with_events Mut HR P FDR patients. WT1 2 2 20.7 0.0002 0.032 * WT1 * DIS3 3 3 9.13 0.0003 0.032 * DIS3 * CDK8 2 2 17.1 0.0003 0.032 * CDK8 * IGF2R 5 5 5.38 0.001 0.032 * IGF2R * MERTK 4 4 6.35 0.001 0.032 Mutual exclusivity, co-occurrence, and enrichment of MERTK * * ATM * ATM 12 15 3.07 0.001 0.032 * MITF * MITF 2 2 10.8 0.003 0.1 * alterations within IHC subtypes GATA1 * GATA1 3 3 6.37 0.003 0.1 * EPHB6 * EPHB6 8 10 3.11 0.003 0.1 * For each type of alteration (short variant, amplification, loss, ERCC4 * ERCC4 8 10 3.18 0.003 0.1 * FLT4 * FLT4 2 2 8.9 0.004 0.11 * APCDD1 * APCDD1 4 4 4.4 0.006 0.14 * rearrangement), we considered a patient as "altered or 1" if the TSHR TSHR 3 3 4.93 0.009 0.2 GATA6 GATA6 2 2 7.85 0.009 0.2 patient had an alteration in a particular gene, and "wild-type or 0" ALK ALK 8 10 2.7 0.011 0.22 BRAF BRAF 3 3 4.52 0.012 0.22 otherwise. Two or more alterations in the same gene in the same CDKN2B CDKN2B 5 6 3.22 0.015 0.26 CDH5 CDH5 2 2 5.24 0.028 0.43 VEGFA VEGFA 3 3 3.74 0.029 0.43 patient were considered as 1 for simplicity. Similarly, for pathway PARK2 PARK2 3 3 3.63 0.031 0.43 BACH1 BACH1 3 3 3.63 0.032 0.43 analysis, if any gene in a given pathway was altered, the patient NUP93 NUP93 3 3 3.69 0.032 0.43 MLH1 MLH1 2 2 4.76 0.034 0.44 CD79A CD79A 2 2 5.79 0.037 0.46 was considered as "altered" or 1 for that pathway, and "wild-type H3F3A H3F3A 8 10 2.4 0.039 0.46 CDK4 CDK4 2 2 4.45 0.043 0.49 or 0" otherwise. For pathway analysis, we used the MSigDB 50 PARP1 PARP1 5 6 2.63 0.045 0.5 PTCH2 PTCH2 4 5 2.82 0.049 0.51 hallmark gene sets (18) and filtered all gene sets to only include 0.1 1.0 10.0 200.0 Hazard Ratio the 401 genes that were part of the FoundationOne panel. C For each pair of genes within each IHC subtype, Fisher exact test TNBC: DFS was used to compute mutual exclusivity (negative log2 odds ratio) GNA13 * and co-occurrence (positive log odds ratio). The enrichment of EMSY * Mut_with_events Mut HR P FDR 2 IKBKE * GNA13 2 2 20.2 0.0002 0.055 * alterations (given by log2 odds ratio) for individual genes or CDK6 * EMSY 4 5 5.71 0.001 0.15 * INHBA IKBKE 4 5 5.58 0.001 0.15 * * CDK6 2 2 10.5 0.002 0.16 * pathways across IHC subtypes was calculated by applying a Fisher MST1R * INHBA 2 2 9.33 0.003 0.18 * MAP3K1 * MST1R 5 5 4.1 0.003 0.18 * exact test to a 2-by-2 contingency table. In all cases, raw P values TSHR MAP3K1 10 14 2.81 0.004 0.18 * TSHR 2 2 7.66 0.008 0.3 – ERBB2 ERBB2 5 6 3.44 0.010 0.36 were corrected for multiple testing using the Benjamini LYN LY N 11 14 2.28 0.014 0.44 q < BRCA1 BRCA1 3 17 0.235 0.015 0.44 Hochberg method. Only log2 odds ratios with values 0.2 are DOT1L 6 6 2.95 0.017 0.46 DOT1L IKZF1 3 3 4.18 0.020 0.46 shown or annotated in Figs. 1 and 2. IKZF1 NKX2−1 2 2 5.6 0.020 0.46 NKX2−1 DNMT3A 4 6 3.44 0.024 0.52 DNMT3A CREBBP 3 14 0.287 0.036 0.68 HSP90AA1 8 10 2.34 0.036 0.68 fi CREBBP PRKDC 18 28 1.81 0.039 0.68 Identi cation of alterations associated with high-risk of HSP90AA1 FBXW7 5 6 2.65 0.042 0.68 PRKDC FGF3 4 5 2.84 0.048 0.68 recurrence GPR124 12 15 1.92 0.048 0.68 FBXW7 FGF3 For the purpose of this analysis, we considered a patient as GPR124 "altered or 1" if the patient had any type of alteration in a 0.1 1.0 10.0 200.0 Hazard Ratio particular gene (mutation/copy number/rearrangement), and "wild-type or 0" otherwise. We first assessed the prognostic Figure 3. significance of all clinical covariates listed in Supplementary Table Association of genomic alterations with DFS. Forest plots displaying hazard S2, in the 291 event-matched patients (in DFS) using the Cox- ratios (with unadjusted 95% confidence intervals) of genes whose alteration proportional hazards model and identified lymph node status to (mutation, copy number, and rearrangement) was associated with DFS in be highly prognostic. Next for each IHC subtype, we computed HER2þ (A), HRþ (B), and TNBC (C). For each gene, the table shows the number hazard ratio for DFS using the Cox-proportional hazards model of recurring patients with alterations, the total number of patients with alterations, the hazard ratio point estimate, and raw and adjusted P values. All with lymph node status included as a covariate. We report only the genes with raw P values < 0.05 are shown in the plots, and those with adjusted P prognostic alterations in the forest plot in Fig. 3 that are altered in value < 0.2—corresponding to the most robust signal—are denoted by stars. at least 2 patients (i.e., 2 or more patients) and have a raw P value

100 Mol Cancer Res; 17(1) January 2019 Molecular Cancer Research

Downloaded from mcr.aacrjournals.org on September 23, 2021. © 2019 American Association for Cancer Research. Published OnlineFirst August 31, 2018; DOI: 10.1158/1541-7786.MCR-18-0619

Genomic Profiling of High-Risk Early Breast Cancers

ABTMB in IHC−based subtypes TMB vs. PAM50 subtypes C TMB vs. TNBC subtypes

15 15 15 KW P : 0.0029 KW P : 7.6e−05 KW P : 0.97

10 10 10 TMB TMB TMB

5 5 5

0 0 0 + + HER2 HR TNBC Basal Her2 LumA LumB BL1 BL2 IM LAR M MSL (N = 41) (N = 128) (N = 120) (N = 92) (N = 30) (N = 46) (N = 33) (N = 27) (N = 28) (N = 11) (N = 8) (N = 11) (N = 5) IHC subtypes PAM50 TNBC subtypes

DE F 25 25

20 20 2 cor = 0.038 cor = 0.026 cor = 0.51 P = 0.55 P = 0.68 P = 5.4e−18 15 15 1

10 10 0 TMB TMB PDL1

5 5 −1

0 0 −2

−5 −5 −2 −1 0 1 2 −4 −2 0 2 −4 −2 0 2 CD8 CD8 PDL1

Figure 4. Correlation of TMB, prognosis, and immune markers. Tumor mutational burden (TMB, mutations per megabase) stratified by IHC (A), PAM50 (B), and Lehmann et al. TNBC (C) subtypes. The P values denote significance by the Kruskal–Wallis test. Association of TMB with CD8 gene expression (D), TMB with PDL1 gene expression (E), and correlation of CD8 with PDL1 gene expression (F), across all breast cancer patients. Each dot is a patient. Pearson correlation coefficient, linear model fit (blue line), and associated P values (for nonzero slope) are shown on each plot.

<0.05. The raw P values were corrected for multiple testing using breast cancers of patients who experienced a DFS event following the Benjamin–Hochberg method after filtering for genes that had adjuvant chemotherapy within the USO01062 study (15). Of the alterations occurring in at least 2 patients. The types of alterations 2,611 enrolled patients, 1,181 patients had tumor tissue available in each gene are described in the oncoprint to prevalence plots in for genomic profiling (12). Of the 1,181 patients, 145 patients Supplementary Fig. S3. experienced a DFS event and were selected for NGS profiling (Supplementary Fig. S1). We also profiled a demographically Other statistical analyses matched control cohort of 146 samples from patients in this The P values for multigroup comparisons in the boxplots in Fig. study who did not experience a DFS event, and a further 108 4 were computed using the Kruskal–Wallis test. The log-rank test tumor samples from patients who did not have a DFS event to was used to detect survival differences in the Kaplan–Meier curves increase the statistical precision of mutation prevalence estimates for DFS. The oncoprint prevalence plots were generated using the within IHC subtypes. As shown in Supplementary Table S1, the ComplexHeatmap package in R. The lollipop plots in Supple- demographic and clinical characteristics were well balanced mentary Fig. S5 were generated with the MutationMapper tool on between the DFS event and control groups. Supplementary Fig. www.cbioportal.org. All other plots were generated using the base S2 shows the overall genomic landscape of the clinically defined or ggplot2 package in R. This study has been reported according to high-risk early breast cancer population. The most prevalent the Reporting Recommendations for Tumor Marker Prognostic alterations were consistent with previously published findings in Studies (REMARK) criteria (19). TCGA (20). To help interpret and organize this complex landscape, we next Results explored the genomic landscape within the IHC-defined sub- Targeted next-generation profiling of high-risk early breast types, in all the molecularly characterized tumors, regardless of cancer patients patients' clinical outcomes (Fig. 1 and Supplementary Table S2). þ þ In order to define the genomic landscape of a clinically defined, Within HER2 ,HR disease (hereafter referred to as HER2 , Fig. high-risk early breast cancer population, we profiled the primary 1A, n ¼ 57), the most frequently somatically mutated genes were

www.aacrjournals.org Mol Cancer Res; 17(1) January 2019 101

Downloaded from mcr.aacrjournals.org on September 23, 2021. © 2019 American Association for Cancer Research. Published OnlineFirst August 31, 2018; DOI: 10.1158/1541-7786.MCR-18-0619

Wilson et al.

þ TP53 (68%), PIK3CA (39%), and ARID1B (19%) and the most enrichment was noted in HR disease in comparison with þ þ frequently copy-number–altered genes were ERBB2 (72%), HER2 . Comparing HER2 breast cancer with TNBC (Fig. CDK12 (58%), MYC (23%), and genes within the 11q13.3 2E, Supplementary Fig. S3) showed an enrichment of pathways þ (21%) and 17q (19%) loci. Within HR , HER2 disease (here- in TNBC, namely, the allograft rejection and apical surface þ after referred to as HR , Fig. 1B, n ¼ 178), the most frequently pathways. Other pathways enriched in TNBC compared with þ somatically mutated genes were PIK3CA (45%), TP53 (30%), and HER2 breast cancer were the DNA repair (e.g., homologous MLL3 (24%), and the most frequently copy-number–altered recombination), p53 pathway, PI3K/AKT/mTOR pathway, as genes were coamplified genes within the 11q13.3 (21%), well as fatty acid metabolism and oxidative phosphorylation. þ 8p11–12 (20%), and 8q (13%) loci. Within TNBC (Fig. 1C, Comparing HR with TNBC tumors (Fig. 2F; Supplementary þ n ¼ 162), the most frequently somatically mutated genes were Fig. S3), we identified an enrichment of pathways in HR TP53 (93%), NOTCH1 (19%), and BRCA1 (18%), and the most disease such as the complement and estrogen response late frequently copy-number–altered genes were GATA3 (19%) and pathways. Conversely, E2F targets, myogenesis, interferon gam- coamplified genes within the 8q (19%) and 12q (11%) loci. ma response, apical surface, PI3K/AKT/mTOR, and hedgehog We next assessed the co-occurrence or mutual exclusivity of pathways were upregulated in TNBC tumors. somatically altered genes within each IHC-defined subtype in all þ the molecularly characterized cancers. Within HER2 disease (Fig. Association of genomic alterations with DFS 1A), significantly co-occurring events included ERBB2 with The ability to identify patients who are most likely to experience CDK12; 17q22–24 genes with 11q13.3 genes as previously a DFS event through genomic analysis could pave the way for described (21) or with SPOP; 8p11–12 genes and BRCA2 with designing adjuvant studies in specific high-risk populations and þ CCNE1 and TP53BP1. Within HR disease (Fig. 1B), co-occur- could also be the basis for identification of therapeutically rele- rence events included 8p11–12 genes with MYST3 or with vant targets in these patients. We directly compared the genomic 11q13.3 genes; MLL2 with MLL3 and TP53 mutations with 8q landscape of patients who experienced a DFS event with a genes. Within TNBC disease (Fig. 1C), co-occurrence events matched set of control patients (total 291 patients). As IHC status included: 1q23 genes; 8q genes; 1q23 genes with 8q genes, and correlates with outcome in early breast cancer, we controlled for 12p genes. These amplification events are in general agreement its impact by specifically looking within each IHC subtype for with previous reports (22–24). Few genes were found to exhibit enrichment of genomic alterations (Fig. 3). We also controlled for statistically significant patterns of mutually exclusive mutation, lymph node status as this was found to associate with worse DFS with the only examples being TP53 with PRKAR1A, RAD51C, outcomes in our data set. þ þ BRIP1, RNF43, or SPOP in HER2 disease; and TP53 with PIK3CA Within HER2 disease (Fig. 3A, n ¼ 41), we identified 12 þ or CDH1 in HR disease. genomic alterations (mutations, copy-number alterations and rearrangements) that produced a raw P value <0.05 for association Enrichment of genomic alterations and pathways by IHC with DFS, though only two genes (AR and MCL-1) yielded an subtype adjusted P value under our 0.20 FDR cutoff. These 12 genes were þ We next assessed whether any mutational or copy-number each altered in 5% to 7% of HER2 tumors (Supplementary alterations were enriched within the different IHC subtypes Fig. S4). We conducted a co-occurrence analysis on the 12 genes þ þ (Fig. 2). Directly comparing HER2 to HR disease (Fig. 2A) with a raw P value <0.05 for association within disease; only two þ showed enrichment in HER2 disease of alterations (short var- gene pairs were found to exhibit significant co-occurrence of iants, copy-number alterations and rearrangements) in RNF43, alteration in the same tumor, HSD3B1 and LRP6, and INSR and þ GNA13, and 17q genes. No enrichment of alterations was JUN (Supplementary Fig. S4). Within HR disease (Fig. 3B, n ¼ þ þ observed in HR compared with HER2 disease. Comparing 128), we identified 28 genomic alterations that produced a raw þ HER2 disease with TNBC (Fig. 2B) showed enrichment in P value <0.05 for association with DFS, 12 of which were asso- þ HER2 disease of PIK3CA, AURKA, RNF43, TOP2A, and ampli- ciated with poor survival post-FDR correction. Most prevalent fied genes on 17q and 11p13.3, whereas enrichment of alterations among the prognostic alterations were mutations in ATM (12%), in TP53, BRCA1, KDM5A, RB1, and NOTCH1 were observed in EPHB6 (8%), ALK (8%), and ERCC4 (8%, Supplementary Fig. þ TNBC disease. Lastly, comparing HR disease with TNBC (Fig. S4). Among the genes shown in Fig. 3B, two pairs of genes were 2C) showed many statistically significant differences, including found to mutually co-occur in the same tumor, these being H3F3A an enrichment of alterations in 11p13, 8p, PIK3CA, AURKA, and PARP1, and EPHB6 and PARK2 (Supplementary Fig. S4). þ CDH1, and MAP3K1 in HR disease, whereas alterations in TP53, Lastly, within TNBC (Fig. 3C, n ¼ 120), we identified 21 genomic BRCA1, RB1, PIK3C2G, PDCD1LG2, MYC, NOTCH1, PTEN, alterations that produced a raw P value <0.05 for association with FGF6, RAD52, LYN, CCND3, CDKN1A, and KRAS were prefer- DFS, seven of which were associated with poor prognosis in TNBC entially seen in TNBC. post-FDR correction. Most prevalent among the poor prognostic As many individual genes can cause activation of a common alterations were alterations in MAP3K1 (12%), MST1R, IKBKE, pathway, we next grouped genes by pathway and assessed and EMSY (4%, Supplementary Fig. S4). Conversely, genes asso- which pathways were differentially altered within the IHC ciated with good prognosis at a raw P value <0.05, though not subtypes. For pathway analysis, we used the Broad Molecular significant after multiple testing correction, were CREBBP (HR, Signatures Database (MSigDB) 50 hallmark gene set collection 0.29) and BRCA1 (HR, 0.24). Of the genes identified in Fig. 3C, þ þ (18). Comparing HER2 to HR disease (Fig. 2D; Supplemen- only one pair was found to significantly co-occur in the same þ tary Fig. S3) showed enrichment in HER2 disease of the p53 tumor, namely, LYN and PRKDC (Supplementary Fig. S4). Sup- pathway, Wnt-bcatenin signaling, epithelial mesenchymal tran- plementary Fig. S5 depicts the spatial distribution of single- sition, peroxisome and pathways associated with cell survival nucleotide variants associated in the genes that were prognostic including E2F targets, apoptosis and UV response. No pathway post-FDR correction and present in 5 or more samples.

102 Mol Cancer Res; 17(1) January 2019 Molecular Cancer Research

Downloaded from mcr.aacrjournals.org on September 23, 2021. © 2019 American Association for Cancer Research. Published OnlineFirst August 31, 2018; DOI: 10.1158/1541-7786.MCR-18-0619

Genomic Profiling of High-Risk Early Breast Cancers

ABBL1 N = 38 Counts BL2 N = 33 Counts 25 0 5 10 15 20 30 35 10 25 15 20 30 PAM50 PAM50 0 5 100% TP53 37% MYC 97% TP53 32% NOTCH1 48% MYC 36% PREX2 24% PRKDC 33% RUNX1T1 21% ARID1B 30% PRKDC 21% BRCA1 30% NBN 18% BRCA2 24% LRP6 21% GATA3 21% PIK3C2G 18% GNAS 21% GATA3 21% KDM5A 21% SPTA1 21% MLL3 21% AXL 21% CDKN1B 21% PTEN 21% FAT3 16% BLM 21% LY N 18% MCL1 18% AR 18% NOTCH2 18% BRCA1 18% NTRK1 18% GNAS 16% PTPRD 18% GPR124 18% RB1 18% NCOR1 15% RB1 16% LRP6 18% SPEN 16% MLL2 18% ZNF703 13% RAD52 18% CREBBP 13% CCNE1 18% PTEN 16% GPR124 15% ARID1B 16% SPTA1 15% BRCA2 13% CDKN1A 15% BRD4 11% DDR2 15% CCNE1 15% CDKN2A 13% FGFR1 15% CDKN2B 13% NF1 15% FANCM 11% PDCD1LG2 12% FGFR2 13% PRDM1 15% LRP1B 13% PRKCI 15% NF1 13% RICTOR 15% NOTCH2 13% ROS1 15% NTRK1 11% SPEN 12% PARP4 15% SDHA 11% TSC2 15% SDHC 11% ABL2 12% AKT2 11% ATR 12% ALK 11% CD274 12% DDR2 11% CHD2 12% EPHB4 11% CHD4 12% FANCA 11% FANCE 12% FGFR1 12% FLT4 11% FANCL 12% INSR 11% FANCM 9% IRS2 11% FAT3 12% KDM5A 11% GRM3 12% KRAS 11% LRP1B 12% MAP3K1 11% MAP3K1 12% MCL1 11% MED12 12% MLL 11% MLL 12% MLL3 6% NOTCH1 11% MYST3 12% NOTCH3 11% NBN 12% NOTCH4 11% PDGFRA 12% PARK2 11% PIK3CA 12% PBRM1 11% POLE 12% RANBP2 11% RUNX1T1 9% STAT3 11% SDHC 12% TNKS 11% TERC 12% TRRAP 11% TSC1

Counts CDIM N = 18 LAR N = 11 Counts 0 2 4 6 8 10 12 14 2 4 PAM50 PAM50 0 6 8 83% TP53 33% CREBBP 73% TP53 28% BRCA1 45% PIK3CA 28% FAT3 36% PTEN 22% LRP1B 27% RAD54L 22% NOTCH1 22% PIK3C2B 18% TSC2 17% APC 27% FANCI 17% AR 27% SPEN 17% GATA3 18% CASP8 17% INPP4B 17% KDM6A 18% CDH1 17% LRP6 18% DOT1L 17% MLL3 18% ERBB2 17% MUTYH 18% FAS 17% NOTCH2 18% FAT3 17% SPTA1 11% AKT3 18% FGFR1 11% BRCA2 18% FLT1 11% CRLF2 18% GNAS 11% DAXX 18% GRIN2A 11% FANCD2 11% FLT4 18% IKZF1 11% GATA4 18% IRS2 11% HGF 18% KEAP1 11% KDM5A 18% MLL2 11% KDR 11% MLL2 18% MTOR 11% MRE11A 18% MYC 11% NOTCH3 18% NF1 11% NTRK1 18% NOTCH1 11% PALB2 18% PIK3C2G 11% PIK3CA 11% PIK3R2 18% PIK3R1 11% PTEN 9% PNRC1 11% RAD50 9% PRDM1 11% RNF43 18% PRKDC 11% ROS1 11% SDHC 18% RICTOR 11% SPEN 18% ROS1 11% TNF 18% SDHA 11% TOP2A 18% SETD2 11% TP53BP1 11% TRRAP 18% SOX10 11% TSC2 18% TET2 11% ZNF217 18% TRRAP 11% ZNF703 18% ZBTB2 EF M N = 14 Counts MSL N = 7 Counts 0 2 4 6 10 12 8 14 3 4 6 0 1 5 PAM50 PAM50 2 86% TP53 100% TP53 57% LRP1B 57% ROS1 36% IRS2 57% SPTA1 29% FAT3 43% ATR 43% BRCA1 21% CCND1 43% DDR2 43% FLT4 21% CCNE1 43% MYC 43% NOTCH1 14% CUL4A 43% PARP1 43% PIK3C2G 21% GATA3 43% PREX2 43% PRKDC 21% GPR124 43% SDHC 29% AKT3 21% IL7R 29% BCORL1 29% EPHB6 21% LZTR1 29% FAT1 29% GATA3 21% MLL2 29% H3F3A 29% KDM5A 21% NOTCH1 29% MED12 29% MET 14% PIK3CA 29% NBN 29% NF1 21% PTEN 29% PIK3C2B 14% PIK3R2 14% SDHA 29% RB1 29% RUNX1T1 21% SPEN 29% TSC2

Alterations PAM50 Amplification Basal Loss Her2 Short variant LumA Rearrangement

Figure 5. Genomic landscape of TNBC molecular subtypes. Plots showing the genomic landscape of the frequently mutated genes within each Lehmann et al. TNBC subtypes. A, Basal-like 1 (BL1), (B) Basal-like 2 (BL2), (C) Immunomodulatory (IM), (D) Luminal AR (androgen receptor, LAR), (E) Mesenchymal (M), and (F) Mesenchymal Stem-like (MSL). The frequency of the combined alterations and the individual prevalence of the four types of alterations—short variant (green), amplification (red), loss (blue), and rearrangement (orange) for each gene is shown as percentages and bar plots of counts.

www.aacrjournals.org Mol Cancer Res; 17(1) January 2019 103

Downloaded from mcr.aacrjournals.org on September 23, 2021. © 2019 American Association for Cancer Research. Published OnlineFirst August 31, 2018; DOI: 10.1158/1541-7786.MCR-18-0619

Wilson et al.

Figure 6. Distribution of prevalent TNBC genes in the Lehmann et al. TNBC subtypes. Figure shows significant genes that are frequently mutated in TNBC from Fig. 1C, that have a significant difference in prevalence within the TNBC subtypes by Fisher test (adjusted P value < 0.2). Dot size denotes prevalence of mutation of a gene within a given subtype. Color shows the relative prevalence of a given gene across Lehmann et al. TNBC subtypes, that is, the degree to which mutation rate in a given TNBC subtype is less or greater than the average rate across all TNBC subjects. The adjusted P values for the genes are indicated on the right side.

Association of TMB, prognosis, and immune markers where high expression was associated with good prognosis (Sup- High TMB has been shown to correlate with clinical benefit plementary Fig. S7), and this association was most pronounced in from PD-1/PD-L1 checkpoint inhibition (25). Additionally, TMB TNBCs (P ¼ 0.041), as previously observed (12). High CD8 gene þ þ derived by the FoundationOne-targeted gene panel has been expression weakly trended with better DFS in the HR and HER2 shown to act as a suitable surrogate for TMB derived via subgroups, but this effect was not statistically significant. Within whole-exome sequencing (26), and to correlate with outcomes the Lehmann and colleagues TNBC subgroups, CD8 gene expres- for atezolizumab in urothelial cancer (27). Therefore, we utilized sion was highest in the IM and MSL groups and lowest in the BL2 data generated by the FoundationOne assay in our current study group (Supplementary Fig. S7B). Together, these data suggest that to determine whether TMB was associated with disease subtype, high CD8 gene expression, rather than TMB, may represent an immune gene expression, and DFS. Within IHC subtypes, median immune-activated tumor environment in breast cancer, particu- þ TMB was 1.5-fold higher in TNBC compared with HER2 and larly in various TNBC subtypes. þ HR (4.05 vs. 2.7 and 2.7, respectively, Fig. 4A). We next assessed TMB across the PAM50 subtypes (Fig. 4B). Similar to TNBC, the Genomic landscape of TNBC molecular subtypes basal-like subtype had the highest median TMB (4.5), along with TNBC is a heterogeneous disease and can be subtyped by gene the luminal B (4.5), followed by luminal A (2.7) and HER2- expression into as many as six distinct molecular subtypes (7). We enriched (2.25) subtypes. TMB did not correlate significantly with previously reported the prevalence and prognostic implications of Lehmann and colleagues TNBC subtypes (Fig. 4C). the Lehmann and colleagues gene-expression subtypes in this TMB did not correlate with CD8 or PD-L1 gene expression (Fig. adjuvant population (12). However, the genomic landscape and 4D and E) and was not associated with DFS in the entire NGS underlying mutational drivers within each subtype are largely population, or within the IHC-defined subtypes (Supplementary unknown. Therefore, we assessed the somatic mutation, copy Fig. S6). However, CD8 gene expression, a marker for cytotoxic number, and rearrangement landscape within the Lehmann and effector T cells, strongly correlated with PD-L1 gene-expression colleagues defined subtypes (Fig. 5; Supplementary Fig. S8A–F). levels (Fig. 4F). Although median CD8 gene expression was All TNBC subtypes had a high prevalence of TP53 mutations similar across the IHC subtypes (Supplementary Fig. S7A), it was ranging from 73% to 100%. In the BL1 subgroup, mutations in significantly prognostic in the entire breast cancer population, NOTCH1, ARID1B, BRCA1, and BRCA2 were common, as were

104 Mol Cancer Res; 17(1) January 2019 Molecular Cancer Research

Downloaded from mcr.aacrjournals.org on September 23, 2021. © 2019 American Association for Cancer Research. Published OnlineFirst August 31, 2018; DOI: 10.1158/1541-7786.MCR-18-0619

Genomic Profiling of High-Risk Early Breast Cancers

copy-number amplifications in GATA3, KDM5A, and RAD52 (Fig. alteration, and they occurred throughout the coding region of 5A). Within the BL2 subgroup, which has a poor outcome, the gene. Loss-of-function mutations in TP53 result in cells being genomic amplifications were common in 8q locus genes (PREX2, more reliant on the Chk1 pathway to repair DNA breaks (30). RUNX1T1, NBN, LYN) and 12q locus genes (LRP6 and Preclinical studies suggest that Chk1 inhibitors can potentiate the PIK3C2G, Fig. 5B). The IM subtype, which has the best outcome, cytotoxic effects of chemotherapies, such as gemcitabine (31), and harbored mutations in CREBBP, PIK3C2B, and APC, and ampli- such combinations are currently being assessed clinically (32). fications in GATA3 and AKT3 (Fig. 5C). As the IM subgroup Also in TNBC, we identified DDR2, RAD52, and KDM5A ampli- associated with a favorable outcome and contained a high prev- fication together with PTEN loss, as well as BRCA1 and RB1 alence of CREBBP and BRCA1 mutations as well as high CD8 gene mutations, which suggest increased reliance on other DNA repair expression, we next tested whether mutations in CREBBP and mechanisms and an impaired ability to repair DNA following BRCA1 were independent correlates of outcome. When control- certain therapeutics. These results are particularly intriguing in ling for CD8 gene expression, only BRCA1 retained its association light of recent reports suggesting that patients with higher geno- with prognosis (raw P value ¼ 0.046). The LAR subgroup was mic instability and TMB levels have an increased likelihood of mostly nonbasal by PAM50 analysis and enriched for the HER2 responding to checkpoint inhibition (33). PAM50 subtype. The LAR subgroup was largely driven by PI3K Cancer pathways can be regulated by many genes to achieve the signaling, as tumors had mutations in PIK3CA, PIK3R1, mTOR, same biological effect. Using the MSigDB hallmark pathway gene PTEN, IRS2, and TSC2 along with copy-number amplifications in sets, we found that distinct pathways were altered within the three ERBB2 and FGFR1 (Fig. 5D). The M subgroup, which had the IHC subtypes. Myogenesis, oxidative phosphorylation, and worst outcome, contained mutations in IRS2 and NOTCH1 and PI3K–AKT–MTOR pathways were altered with significantly higher þ þ copy-number amplifications in CCNE1, CCND1, and IL7R (Fig. frequency in TNBCs, compared with HER2 and HR tumors. 5E). Lastly, the MSL subgroup, which had a favorable outcome, Activation of PI3K–AKT–MTOR signaling was also shown in displayed mutations in ROS1, ATR, and MET and copy-number TCGA (20), which reported higher PI3K pathway activity in the amplifications in SPTA1, DDR2, and MYC (Fig. 5F). basal subtype of breast cancer, a large majority of which are We next assessed how the prevalence patterns of the most TNBCs. PI3K pathway-targeted therapies, such as the AKT inhib- abundant alterations identified in Fig. 1C are represented across itor ipatasertib, have been shown to prolong DFS in TNBC in the six TNBC subtypes. Out of the 42 most frequently altered combination with chemotherapy (34). genes in the entire TNBC cohort (Fig. 1C), only seven displayed a By comparing the somatic mutation, copy-number alteration significant (FDR < 0.2) differential mutation rate across the six and rearrangement landscapes between patients who experienced TNBC subtypes (Fig. 6). The MSL and BL2 subtypes more fre- a recurrence event and those who did not, we identified numerous quently harbor six of these seven TNBC genes. The remaining alterations that were associated with clinical outcome. Although subtypes frequently show alterations in only two of the seven in some cases the mutations were of low prevalence, our findings identified genes. Lastly, we assessed the prevalence of the TNBC show several key alterations that may be targetable and associated þ prognostic genes associated with DFS from Fig. 3C (Supplemen- with relapse in this analysis. For example, in HR disease, we tary Fig. S8G). Out of the 21 prognostic genes in TNBC (Fig. 3C), found that alterations in cell-cycle/DNA response and repair only two genes CREBBP and PRKDC showed differential mutation genes, such as CDK8, CDK4, CDKN2B, ATM, and ERCC4, were rates across the six TNBC subtypes (raw P value < 0.05); however, associated with increased rates of recurrence. These results are þ neither was significant post-FDR correction. timely given the beneficial effects of CDK4/6 inhibitors in HR metastatic breast cancer (35), with the results from adjuvant studies eagerly anticipated. þ Discussion In HER2 disease, we found that AR and MCL-1 were signif- A number of elegant studies describing the genomic and icantly associated with poor prognosis. Agents that target AR are transcriptomic profiles of breast cancer and showing the spectrum routinely used for the treatment of men with prostate cancer (36), of mutations and copy-number alterations within IHC- and and small-molecule inhibitors targeting MCL-1 are under clinical PAM50-defined subtypes have been published over the past evaluation (ClinicalTrials.gov NCT02675452), both of which þ several years (20, 28, 29). However, little is known about how perhaps could be utilized in these genomically altered HER2 these alterations affect clinical outcome. Using a targeted NGS breast cancers. panel, we profiled the genomic landscape of a clinically defined Within TNBC, protumorigenic inflammatory cytokine sig- high-risk patient population that was enrolled onto the naling genes, such as IKBKE and MAP3K1, were significantly USO01062 phase III adjuvant capecitabine trial. Specifically, we associated with poor prognosis (37, 38). MAP3K1,which chose patients who had a recurrence event and we matched them regulates JNK activation and cell migration (39), has recently demographically to a control set of samples from patients who did been identified as a driver gene in metastatic breast cancer not have a recurrence event in order to uncover genomic traits that samples from the SAFIR01 clinical trial (40). Within TNBC, may be used to select high-risk patients for future adjuvant trials of alterations in BRCA1 and CREBBP were associated with a better novel agents and to potentially uncover therapeutic targets. DFS, although these findings were not statistically significant At the population level, our genomic analysis has commonal- after multiple testing correction. However, amplification of þ ities with previous findings. Within HER2 breast cancer, 72% of EMSY, a nuclear protein that binds and represses BRCA2 and the samples showed ERBB2 copy-number amplification by the increases genomic instability (41), was significantly associated FoundationOne assay, suggesting significant but nonetheless with poor prognosis in TNBC. This finding raises the question imperfect concordance between NGS-based amplification assess- of whether patients whose early-stage TNBCs harbor an EMSY ment and traditional scoring by IHC. Within TNBC, mutations in amplification might benefit from the addition of carboplatin to the tumor suppressor TP53 were the most common genetic standard chemotherapy.

www.aacrjournals.org Mol Cancer Res; 17(1) January 2019 105

Downloaded from mcr.aacrjournals.org on September 23, 2021. © 2019 American Association for Cancer Research. Published OnlineFirst August 31, 2018; DOI: 10.1158/1541-7786.MCR-18-0619

Wilson et al.

Cancer immunotherapy checkpoint inhibitors, such as those carried out many of our analyses within IHC-defined subtypes, that target PD-L1 and PD-1, can unlock the patient's own immune further reducing our power to detect differences between patients system to unleash an anticancer response, particularly in cancers who experienced a progression event versus those who did not. In that have a high mutational load such as lung cancer and mel- addition, our study was not powered to formally statistically test anoma (42, 43). We found that TMB was highest in the TNBC the differences in genomic alterations within the six TNBC sub- subtype, and when classifying samples by PAM50 subtype, we types nor to correlate these alterations with outcomes. In spite of further showed that the luminal B tumors have a similar TMB to these limitations, our study does have detailed clinical outcomes the basal-like subtype. We found that gene expression of the on the patients enrolled onto the USO01062 trial, and it identifies activated T-cell marker, CD8, did not correlate with TMB, suggest- multiple, novel genomic associations that warrant further testing ing that TMB may not be a predictive marker of benefit from in an independent data set. immune activity in breast cancer, which may be in part due to the In conclusion, using a targeted NGS approach, we characterized low TMB levels in breast cancer. the somatic mutation and copy-number alteration landscape of Using unbiased microarray gene-expression technologies, the high-risk early breast cancer patients in a prospective phase III trial TNBC subtype has emerged as the most heterogeneous of breast where all patients received state of the art adjuvant chemotherapy. cancers with the claudin-low (44, 45), Basal-A/B and HER2- We describe the mutational landscape and enrichment patterns enriched subtypes (46). Lehmann and colleagues identified six for certain alterations at the single-gene and pathway levels, and molecular TNBC subtypes, which they subsequently modified to a within IHC-defined subtypes. We discovered genomic traits asso- four-subtype classifier by removing the tumor-associated micro- ciated with disease recurrence that may be used to select high-risk environment, including the stromal TILs, by microdissection of patients in future studies. We show that TMB did not correlate the tumor specimens (7, 8). We previously applied the original with clinical outcome overall, nor within any IHC subtype. Lehmann and colleagues gene classifier to our adjuvant Finally, we show that the Lehmann and colleagues TNBC subtypes USO01062 population and showed its relationship to clinical have distinct mutational landscapes, and we uncovered several outcome (12). In the current analysis, we report the distinct previously unrecognized alterations that may be therapeutically mutation and copy-number alterations within the six TNBC relevant in this patient population. subtypes. The LAR subtype, which has previously been reported to contain the majority of PIK3CA mutations present within Disclosure of Potential Conflicts of Interest TNBC (12, 47), appears almost completely driven by PI3K sig- J.M. Spoerke, H.M. Savage, and T.R. Wilson have ownership interest naling, suggesting that the LAR population may be potentially (including stock, patents, etc.) in Roche. J.A. O'Shaughnessy is a consultant/ responsive to PI3K–mTOR inhibitors, perhaps in the presence of advisory board member for AstraZeneca, Novartis, and Lilly. R. Bourgon has antiandrogen receptor therapies. Within the M subtype, IRS2 was ownership interest (including stock, patents, etc.) in Hoffmann-La Roche, Genentech. No potential conflicts of interest were disclosed by the other authors. a frequently altered gene. IRS2 is a substrate for insulin receptor kinase 1, inhibitors of which have been tested in clinical trials, with unfavorable results, albeit in biomarker unselected popula- Authors' Contributions tions (48). It is possible that the IRS2-altered M subpopulation Conception and design: T.R. Wilson, A.R. Udyavar, J.A. O'Shaughnessy, fi M.R. Lackner could bene t from inhibition of the IGFR pathway. Recently, Development of methodology: A.R. Udyavar, R. Bourgon Barecehe and colleagues utilized the METABRIC and TCGA data Acquisition of data (provided animals, acquired and managed patients, sets to analyze the mutational and copy-number landscape in provided facilities, etc.): T.R. Wilson, J. Aimi, H.M. Savage, J.A. O'Shaughnessy TNBC and made similar observations (49). Specifically, they Analysis and interpretation of data (e.g., statistical analysis, biostatistics, noted that LAR tumors were associated with higher TMB and computational analysis): T.R. Wilson, A.R. Udyavar, C.-W. Chang, A. Daemen, PI3K pathway activation, M tumors had activated EGFR and J.A. O'Shaughnessy, R. Bourgon, M.R. Lackner Writing, review, and/or revision of the manuscript: T.R. Wilson, A.R. Udyavar, Notch signaling and BL1 tumors showed copy-number losses C.-W. Chang, J.M. Spoerke, H.M. Savage, A. Daemen, J.A. O'Shaughnessy, BRCA1/2 RB1 for and . Given that each Lehmann subtype may R. Bourgon, M.R. Lackner comprise approximately 10% to 15% of TNBC, and that TNBC Administrative, technical, or material support (i.e., reporting or organizing itself represents 15% to 18% of all breast cancers, the develop- data, constructing databases): A.R. Udyavar, J.M. Spoerke, H.M. Savage ment of targeted therapies in these patient populations will likely require phase II testing in small subgroups of molecularly selected The costs of publication of this article were defrayed in part by the payment of advertisement patients whose breast cancers have been screened for multiple page charges. This article must therefore be hereby marked in accordance with 18 U.S.C. Section 1734 solely to indicate this fact. clinical trial-qualifying alterations. Although the total number of patients in the USO01062 trial is large, the statistical power in our study was nonetheless limited by the low number of progression events and our selection of only a Received June 11, 2018; revised July 26, 2018; accepted August 17, 2018; subset of patients' tumors for genetic analysis. Furthermore, we published first August 31, 2018.

References 1. Perou CM, Sorlie T, Eisen MB, van de Rijn M, Jeffrey SS, Rees CA, et al. 3. van 't Veer LJ, Dai H, van de Vijver MJ, He YD, Hart AA, Mao M, et al. Gene Molecular portraits of human breast tumours. Nature 2000;406: expression profiling predicts clinical outcome of breast cancer. Nature 747–52. 2002;415:530–6. 2. Sorlie T, Perou CM, Tibshirani R, Aas T, Geisler S, Johnsen H, et al. Gene 4. van de Vijver MJ, He YD, van't Veer LJ, Dai H, Hart AA, Voskuil DW, et al. A expression patterns of breast carcinomas distinguish tumor subclasses with gene-expression signature as a predictor of survival in breast cancer. N Engl J clinical implications. Proc Natl Acad Sci U S A 2001;98:10869–74. Med 2002;347:1999–2009.

106 Mol Cancer Res; 17(1) January 2019 Molecular Cancer Research

Downloaded from mcr.aacrjournals.org on September 23, 2021. © 2019 American Association for Cancer Research. Published OnlineFirst August 31, 2018; DOI: 10.1158/1541-7786.MCR-18-0619

Genomic Profiling of High-Risk Early Breast Cancers

5. Parker JS, Mullins M, Cheang MC, Leung S, Voduc D, Vickery T, et al. 25. Rizvi NA, Hellmann MD, Snyder A, Kvistborg P, Makarov V, Havel JJ, et al. Supervised risk predictor of breast cancer based on intrinsic subtypes. J Clin Cancer immunology. Mutational landscape determines sensitivity to PD-1 Oncol 2009;27:1160–7. blockade in non-small cell lung cancer. Science 2015;348:124–8. 6. Curtis C, Shah SP, Chin SF, Turashvili G, Rueda OM, Dunning MJ, et al. The 26. Campesato LF, Barroso-Sousa R, Jimenez L, Correa BR, Sabbaga J, Hoff PM, genomic and transcriptomic architecture of 2,000 breast tumours reveals et al. Comprehensive cancer-gene panels can be used to estimate muta- novel subgroups. Nature 2012;486:346–52. tional load and predict clinical benefit to PD-1 blockade in clinical practice. 7. Lehmann BD, Bauer JA, Chen X, Sanders ME, Chakravarthy AB, Shyr Y, et al. Oncotarget 2015;6:34221–7. Identification of human triple-negative breast cancer subtypes and pre- 27. Balar AV, Galsky MD, Rosenberg JE, Powles T, Petrylak DP, Bellmunt J, et al. clinical models for selection of targeted therapies. J Clin Invest 2011; Atezolizumab as first-line treatment in cisplatin-ineligible patients with 121:2750–67. locally advanced and metastatic urothelial carcinoma: a single-arm, multi- 8. Lehmann BD, Jovanovic B, Chen X, Estrada MV, Johnson KN, Shyr Y, et al. centre, phase 2 trial. Lancet 2017;389:67–76. Refinement of triple-negative breast cancer molecular subtypes: implica- 28. Stephens PJ, Tarpey PS, Davies H, Van Loo P, Greenman C, Wedge DC, et al. tions for neoadjuvant chemotherapy selection. PLoS One 2016;11: The landscape of cancer genes and mutational processes in breast cancer. e0157368. Nature 2012;486:400–4. 9. Burstein MD, Tsimelzon A, Poage GM, Covington KR, Contreras A, 29. Banerji S, Cibulskis K, Rangel-Escareno C, Brown KK, Carter SL, Frederick Fuqua SA, et al. Comprehensive genomic analysis identifies novel AM, et al. Sequence analysis of mutations and translocations across breast subtypes and targets of triple-negative breast cancer. Clin Cancer Res cancer subtypes. Nature 2012;486:405–9. 2015;21:1688–98. 30. Harada N, Watanabe Y, Yoshimura Y, Sakumoto H, Makishima F, Tsuchiya 10. Masuda H, Baggerly KA, Wang Y, Zhang Y, Gonzalez-Angulo AM, M, et al. Identification of a checkpoint modulator with synthetic lethality to Meric-Bernstam F, et al. Differential response to neoadjuvant chemo- p53 mutants. Anticancer Drugs 2011;22:986–94. therapy among 7 triple-negative breast cancer molecular subtypes. 31. Xiao Y, Ramiscal J, Kowanetz K, Del Nagro C, Malek S, Evangelista M, et al. Clin Cancer Res 2013;19:5533–40. Identification of preferred chemotherapeutics for combining with a CHK1 11. Jovanovic B, Mayer IA, Mayer EL, Abramson VG, Bardia A, Sanders ME, et al. inhibitor. Mol Cancer Ther 2013;12:2285–95. A randomized phase II neoadjuvant study of cisplatin, paclitaxel with or 32. Infante JR, Hollebecque A, Postel-Vinay S, Bauer TM, Blackwood EM, without everolimus in patients with Stage II/III triple-negative breast Evangelista M, et al. Phase I study of GDC-0425, a checkpoint kinase 1 cancer (TNBC): responses and long-term outcome correlated with inhibitor, in combination with gemcitabine in patients with refractory increased frequency of DNA damage response gene mutations, TNBC solid tumors. Clin Cancer Res 2017;23:2423–32. subtype, AR status, and Ki67. Clin Cancer Res 2017;23:4035–45. 33. Goodman AM, Kato S, Bazhenova L, Patel SP, Frampton GM, Miller V, 12. Wilson TR, Yu J, Lu X, Spoerke JM, Xiao Y, O'Brien C, et al. The molecular et al. Tumor mutational burden as an independent predictor of landscape of high-risk early breast cancer: comprehensive biomarker response to immunotherapy in diverse cancers. Mol Cancer Ther analysis of a phase III adjuvant population. NPJ Breast Cancer 2016; 2017;16:2598–608. 2:16022. 34. Kim SB, Dent R, Im SA, Espie M, Blau S, Tan AR, et al. Ipatasertib plus 13. Park JH, Anderson WF, Gail MH. Improvements in US breast cancer paclitaxel versus placebo plus paclitaxel as first-line therapy for metastatic survival and proportion explained by tumor size and estrogen-receptor triple-negative breast cancer (LOTUS): a multicentre, randomised, double- status. J Clin Oncol 2015;33:2870–6. blind, placebo-controlled, phase 2 trial. Lancet Oncol 2017;18:1360–72. 14. Smith BD, Jiang J, McLaughlin SS, Hurria A, Smith GL, Giordano SH, et al. 35. Finn RS, Martin M, Rugo HS, Jones S, Im SA, Gelmon K, et al. Palbociclib Improvement in breast cancer outcomes over time: are older women and letrozole in advanced breast cancer. N Engl J Med 2016;375:1925–36. missing out? J Clin Oncol 2011;29:4647–53. 36. Scher HI, Fizazi K, Saad F, Taplin ME, Sternberg CN, Miller K, et al. 15. O'Shaughnessy JA, Koeppen H, Xiao Y, Lackner MR, Paul D, Stokoe C, et al. Increased survival with enzalutamide in prostate cancer after chemother- Patients with slowly proliferative early breast cancer have low 5-year apy. N Engl J Med 2012;367:1187–97. recurrence rates in a phase III adjuvant trial of capecitabine. Clin Cancer 37. Barbie TU, Alexe G, Aref AR, Li S, Zhu Z, Zhang X, et al. Targeting an IKBKE Res 2015;21:4305–11. cytokine network impairs triple-negative breast cancer growth. J Clin Invest 16. Frampton GM, Fichtenholtz A, Otto GA, Wang K, Downing SR, He J, et al. 2014;124:5411–23. Development and validation of a clinical cancer genomic profiling test 38. Kyriakis JM, Avruch J. Mammalian MAPK pathways based on massively parallel DNA sequencing. Nat Biotechnol 2013;31: activated by stress and inflammation: a 10-year update. Physiol Rev 1023–31. 2012;92:689–737. 17. Chalmers ZR, Connelly CF, Fabrizio D, Gay L, Ali SM, Ennis R, et al. 39. Pham TT, Angus SP, Johnson GL. MAP3K1: genomic alterations in cancer Analysis of 100,000 human cancer genomes reveals the landscape of tumor and function in promoting cell survival or apoptosis. Genes Cancer mutational burden. Genome Med 2017;9:34. 2013;4:419–26. 18. Liberzon A, Birger C, Thorvaldsdottir H, Ghandi M, Mesirov JP, Tamayo P. 40. Lefebvre C, Bachelot T, Filleron T, Pedrero M, Campone M, Soria JC, et al. The Molecular Signatures Database (MSigDB) hallmark gene set collection. Mutational profile of metastatic breast cancers: a retrospective analysis. Cell Systems 2015;1:417–25. PLoS Med 2016;13:e1002201. 19. McShane LM, Altman DG, Sauerbrei W, Taube SE, Gion M, Clark GM, et al. 41.Hughes-DaviesL,HuntsmanD,RuasM,FuksF,ByeJ,ChinSF,etal. REporting recommendations for tumor MARKer prognostic studies EMSY links the BRCA2 pathway to sporadic breast and ovarian cancer. (REMARK). Nat Clin Pract Oncol 2005;2:416–22. Cell 2003;115:523–35. 20. Cancer Genome Atlas Network. Comprehensive molecular portraits of 42. Weber JS, D'Angelo SP, Minor D, Hodi FS, Gutzmer R, Neyns B, et al. human breast tumours. Nature 2012;490:61–70. Nivolumabversuschemotherapyinpatients with advanced melanoma 21. Staaf J, Jonsson G, Ringner M, Vallon-Christersson J, Grabau D, Arason A, who progressed after anti-CTLA-4 treatment (CheckMate 037): a - et al. High-resolution genomic and expression analyses of copy number domised, controlled, open-label, phase 3 trial. Lancet Oncol 2015;16: alterations in HER2-amplified breast cancer. Breast Cancer Res 2010;12: 375–84. R25. 43. Rittmeyer A, Barlesi F, Waterkamp D, Park K, Ciardiello F, von Pawel J, et al. 22. Latham C, Zhang A, Nalbanti A, Maner S, Zickert P, Blegen H, et al. Frequent Atezolizumab versus docetaxel in patients with previously treated non- co-amplification of two different regions on 17q in aneuploid breast small-cell lung cancer (OAK): a phase 3, open-label, multicentre rando- carcinomas. Cancer Genet Cytogenet 2001;127:16–23. mised controlled trial. Lancet 2017;389:255–65. 23. Giltnane JM, Hutchinson KE, Stricker TP, Formisano L, Young CD, Estrada 44. Prat A, Parker JS, Karginova O, Fan C, Livasy C, Herschkowitz JI, et al. þ MV, et al. Genomic profiling of ER breast cancers after short-term estrogen Phenotypic and molecular characterization of the claudin-low intrinsic suppression reveals alterations associated with endocrine resistance. subtype of breast cancer. Breast Cancer Res 2010;12:R68. Sci Transl Med 2017;9:eaai7993. 45. Taube JH, Herschkowitz JI, Komurov K, Zhou AY, Gupta S, Yang J, et al. 24. Gelsi-Boyer V, Orsetti B, Cervera N, Finetti P, Sircoulomb F, Rouge C, et al. Core epithelial-to-mesenchymal transition interactome gene-expression Comprehensive profiling of 8p11–12 amplification in breast cancer. signature is associated with claudin-low and metaplastic breast cancer Mol Cancer Res 2005;3:655–67. subtypes. Proc Natl Acad Sci U S A 2010;107:15449–54.

www.aacrjournals.org Mol Cancer Res; 17(1) January 2019 107

Downloaded from mcr.aacrjournals.org on September 23, 2021. © 2019 American Association for Cancer Research. Published OnlineFirst August 31, 2018; DOI: 10.1158/1541-7786.MCR-18-0619

Wilson et al.

46. Neve RM, Chin K, Fridlyand J, Yeh J, Baehner FL, Fevr T, et al. A collection of 48. Qu X, Wu Z, Dong W, Zhang T, Wang L, Pang Z, et al. Update of IGF-1 breast cancer cell lines for the study of functionally distinct cancer subtypes. receptor inhibitor (ganitumab, dalotuzumab, cixutumumab, teprotumu- Cancer Cell 2006;10:515–27. mab and figitumumab) effects on cancer therapy. Oncotarget 2017;8: 47. Gonzalez-Angulo AM, Stemke-Hale K, Palla SL, Carey M, Agarwal R, 29501–18. Meric-Berstam F, et al. Androgen receptor levels and association with 49. Bareche Y, Venet D, Ignatiadis M, Aftimos P, Piccart M, Rothe F, et al. PIK3CA mutations and prognosis in breast cancer. Clin Cancer Res Unravelling triple-negative breast cancer molecular heterogeneity using an 2009;15:2472–8. integrative multiomic analysis. Ann Oncol 2018;29:895–902.

108 Mol Cancer Res; 17(1) January 2019 Molecular Cancer Research

Downloaded from mcr.aacrjournals.org on September 23, 2021. © 2019 American Association for Cancer Research. Published OnlineFirst August 31, 2018; DOI: 10.1158/1541-7786.MCR-18-0619

Genomic Alterations Associated with Recurrence and TNBC Subtype in High-Risk Early Breast Cancers

Timothy R. Wilson, Akshata R. Udyavar, Ching-Wei Chang, et al.

Mol Cancer Res 2019;17:97-108. Published OnlineFirst August 31, 2018.

Updated version Access the most recent version of this article at: doi:10.1158/1541-7786.MCR-18-0619

Supplementary Access the most recent supplemental material at: Material http://mcr.aacrjournals.org/content/suppl/2018/08/31/1541-7786.MCR-18-0619.DC1

Cited articles This article cites 48 articles, 15 of which you can access for free at: http://mcr.aacrjournals.org/content/17/1/97.full#ref-list-1

Citing articles This article has been cited by 1 HighWire-hosted articles. Access the articles at: http://mcr.aacrjournals.org/content/17/1/97.full#related-urls

E-mail alerts Sign up to receive free email-alerts related to this article or journal.

Reprints and To order reprints of this article or to subscribe to the journal, contact the AACR Publications Department at Subscriptions [email protected].

Permissions To request permission to re-use all or part of this article, use this link http://mcr.aacrjournals.org/content/17/1/97. Click on "Request Permissions" which will take you to the Copyright Clearance Center's (CCC) Rightslink site.

Downloaded from mcr.aacrjournals.org on September 23, 2021. © 2019 American Association for Cancer Research.