Prevalence and mutational determinants of high tumor mutation burden in breast cancer

ORIGINAL ARTICLE

Prevalence and mutational determinants of high tumor mutation burden in breast cancer R. Barroso-Sousa1,2yz, E. Jain2,3z, O. Cohen2,3, D. Kim2,3, J. Buendia-Buendia2,3, E. Winer1,4,5, N. Lin1,4,5, S. M. Tolaney1,4,5 & N. Wagle1,2,3,4,5* 1

Department of Medical Oncology; 2Center for Cancer Precision Medicine, Dana-Farber Cancer Institute, Boston; 3Broad Institute of MIT and Harvard, Cambridge; Harvard Medical School, Boston; 5Department of Medicine, Brigham and Women’s Hospital, Boston, USA

4

Available online XXX

Background: High tumor mutation burden (TMB) can benefit immunotherapy for multiple tumor types, but the prevalence of hypermutated breast cancer is not well described. The aim of this study was to evaluate the frequency, mutational patterns, and genomic profile of hypermutated breast cancer. Patients and methods: We used de-identified data from individuals with primary or metastatic breast cancer from six different publicly available genomic studies. The prevalence of hypermutated breast cancer was determined among 3969 patients’ samples that underwent whole exome sequencing or gene panel sequencing. The samples were classified as having high TMB if they had 10 mutations per megabase (mut/Mb). An additional eight patients were identified from a Dana-Farber Cancer Institute cohort for inclusion in the hypermutated cohort. Among the patients with high TMB, the mutational patterns and genomic profiles were determined. A subset of patients was treated with regimens containing PD-1 inhibitors. Results: The median TMB was 2.63 mut/Mb. The median TMB significantly varied according to the tumor subtype (HR/HER2 >HER2þ >HRþ/HER2, P < 0.05) and sample type (metastatic > primary, P ¼ 2.2  1016). Hypermutated tumors were found in 198 patients (5%), with enrichment in metastatic versus primary tumors (8.4% versus 2.9%, P ¼ 6.5  1014). APOBEC activity (59.2%), followed by mismatch repair deficiency (MMRd; 36.4%), were the most common mutational processes among hypermutated tumors. Three patients with hypermutated breast cancerdincluding two with a dominant APOBEC activity signature and one with a dominant MMRd signaturedtreated with pembrolizumab-based therapies derived an objective and durable response to therapy. Conclusion: Hypermutation occurs in 5% of all breast cancers with enrichment in metastatic tumors. Different mutational signatures are present in this population with APOBEC activity being the most common dominant process. Preliminary data suggest that hypermutated breast cancers are more likely to benefit from PD-1 inhibitors. Key words: breast cancer, tumor mutational burden, APOBEC, mutational signatures, immunotherapy, mismatch repair deficiency

INTRODUCTION Despite the success of immune checkpoint inhibitors (ICI) across several tumor types, only a small fraction of patients with metastatic breast cancer (MBC) have benefited from PD-1/PD-L1 inhibitors.1 Recently, the IMPASSION130 study led to the approval of

*Correspondence to: Dr Nikhil Wagle, Department of Medical Oncology, Dana-Farber Cancer Institute, 450 Brookline Ave, Dana 820A, Boston, MA 02215, USA. Tel: þ1-617-632-6419 E-mail: [email protected] (N. Wagle). y Present address: Oncology Center, Hospital Sírio-Libanês Brasília, Brasília, Brazil. z These authors contributed equally to this work. 0923-7534/© 2019 European Society for Medical Oncology. Published by Elsevier Ltd. All rights reserved.

Volume xxx

-

Issue xxx

-

2019

atezolizumab plus nab-paclitaxel for patients with metastatic triple-negative breast cancer (mTNBC) with 1% PD-L1 expression on tumor-infiltrating immune cells.2 Evaluating other predictive biomarkers may help increase the number of breast cancer patients who benefit from ICI. Somatic mutations are known to be the main source of tumor-specific neoantigens, which are a key target of antitumor immunity.3e5 High tumor mutational burden (TMB) is associated with high neoantigen burden, high T-cell infiltration, and high response rates to ICI across different tumor types.6e12 In this study, we evaluated the prevalence, pathological characteristics, mutational patterns, and genomic profiles of hypermutated breast cancers by analyzing publicly available genomic data from 3969 breast https://doi.org/10.1016/j.annonc.2019.11.010

1

R. Barroso-Sousa et al.

Annals of Oncology cancer tumor samples. We also present several patients with hypermutated breast cancer who achieved prolonged clinical benefit following PD-1/PD-L1 inhibitor-based therapies. METHODS Patients and samples We analyzed tumor whole exome sequencing (WES) or gene panel sequencing data from 3969 de-identified individuals with primary or metastatic breast cancer from six publicly available studies (supplementary Table S1, available at Annals of Oncology online). For the subsequent hypermutated cohort analysis, we also included eight patients (from a cohort of 222 patients) from an ongoing study of estrogen receptor (ER)-positive MBC at the Dana-Farber Cancer Institute (DFCI).13 Assessment of TMB TMB was calculated as the total number of mutations in a sample divided by the length of the genomic target region captured by WES or gene panel assay. Samples with TMB of 10 mut/Mb were classified as hypermutated. Statistical analysis of clinical annotations TMB was correlated with available clinical annotations. Patients with complete clinical annotations were considered. The analysis was carried out using R (version 3.5.2). Immune cytolytic score calculation The immune cytolytic score was calculated as described in Rooney et al.6 using RNA-sequencing data from TCGA-BRCA. Neoantigen prediction analysis The MBCproject dataset was utilized for neoantigen prediction analysis using Topiary, PolySolver, and NetMHC (v4.0). Mutational signature analysis The contributions of different mutation signatures were identified for each sample according to the distribution of the six substitution classes (C>A, C>G, C>T, T>A, T>C, T>G) and the bases immediately 50 and 30 to the mutated base, producing 96 possible mutation subtypes, which were compared to the known 30 COSMIC signatures. A sample was determined to have a dominant signature based on the maximum signature score attributed. The analysis was carried out using the maftools package in R (version 3.5.2). Mutation enrichment analysis Mutation rates for known cancer genes were calculated for hypermutated tumors with and without a dominant APOBEC signature. Fisher’s exact test was used to calculate significance. Multiple test correction was done using the 2

https://doi.org/10.1016/j.annonc.2019.11.010

p.adjust() function with false discovery rate method in R (version 3.5.2). RESULTS Hypermutation across breast cancers Genomic and clinical data from three WES (France Study 2016, MBCproject, and TCGA-BRCA) and three targeted panel studies (DFCI-ONCOPANEL, MSK-IMPACT, and VICC) were combined to analyze 3969 patients with breast cancer. The frequency of hypermutation in breast cancers from each dataset (Figure 1A) varied from 2.3% (MSKIMPACT468) to 14.0% (DFCI-OncoPanel). The median TMB across the entire cohort was 2.63 mut/Mb (0.2e290.8; Figure 1B). Overall, 5% (198 patients) of breast cancers analyzed were hypermutated based on the calculated cutoff of 10 mut/Mb (see Methods). Metastatic tumors (see supplementary Methods, available at Annals of Oncology online), had a higher median TMB compared to primary tumors (3.8 versus 2.0, P < 2.2  1016) (Figure 1C). There was a weak correlation between TMB and age at diagnosis (R2 ¼ 0.13, P ¼ 3.6  105) and no significant difference in median TMB across histology types (P ¼ 0.074; supplementary Figure S1AeB, available at Annals of Oncology online). TNBC had significantly higher median TMB (1.8 mut/Mb) compared with HR-positive (1.1 mut/Mb, P ¼ 2.8  108) or HER2-positive cancers (1.3 mut/Mb, P ¼ 0.003) (supplementary Figure S1C, available at Annals of Oncology online). Similar results were seen when comparisons were carried out within each individual cohort, for formalin fixed paraffin embedded (FFPE) tumors only and for frozen tumors only (supplementary Figures S2eS4 and supplementary Table S2, available at Annals of Oncology online). The hypermutated tumors (median TMB ¼ 14.46 mut/ Mb) were analyzed according to clinical and pathological characteristics (supplementary Table S3, available at Annals of Oncology online). The frequency of hypermutation in breast cancer was higher for metastatic samples compared to primary samples (8.4% versus 2.9%, P ¼ 6.529  1014). Additionally, 8.7% of invasive lobular carcinomas (ILC) were hypermutated compared to 4.0% of invasive ductal carcinomas (IDC), but this difference in prevalence was not statistically significant (P ¼ 0.074). Among the metastatic samples, however, we observed a significant enrichment of hypermutation in metastatic ILC (17.0%) compared to metastatic IDC tumors (7.8%) (P <0.002). The frequency of hypermutated breast cancer was similar among different receptor subtypes (3.7%e3.9%). To compare the differences in mutation rate between distant metastatic biopsies, we carried out several exploratory comparisons of primary biopsies from patients who eventually developed metastases and primary biopsies from patients who did not develop metastases, although the number of samples used for these comparisons was small. The frequency of hypermutation in

Volume xxx

-

Issue xxx

-

2019

R. Barroso-Sousa et al.

Annals of Oncology

A Whole exome sequencing

Targeted panel sequencing

Tumor mutation burden (log10 scale)

500 300 200 150 100 75 50 25

4.2%

3.2%

2.6%

14.0%

2.3%

7.6%

6.7%

10 5 4 3 2 1

*France_Study_2016FF (N = 213)

*MBCProject (N = 126)

*TCGA−BRCAFF DFCI−ONCOPANEL−3 MSK−IMPACT410 (N = 977) (N = 301) (N = 1009)

B

MSK−IMPACT468 (N = 1071)

VICC−01−T5A (N = 92)

VICC−01−T7 (N = 180)

C median: 2.63 mutations/Mbs

Wilcoxon, P < 2.2e−16

cutoff: ≥ 10 mutations/Mbs Tumor mutation burden (log10 scale)

400

Number of patients

7.3%

300

200

100

300 200 100 80 40

8.4%

2.9%

10 5 4 3 2 1

0 1 2 4 6 10 50 100 300 Tumor mutation burden (log10 scale)

Metastasis (1496)

Primary (2455)

Figure 1. Tumor mutation burden (TMB) across 3969 primary and metastatic breast cancers. (A) TMB (y-axis) distribution for each dataset (x-axis) used in the analysis. The sample points above the black dotted line at 10 mutations/megabase represents the hypermutated tumors. The percentage of hypermutation is indicated for each dataset. Datasets marked with * were whole exome sequenced and the remaining datasets were sequenced targeted exome panels. Datasets marked with FF on the x-axis used fresh frozen tissue biopsies; the remaining used formalin fixed paraffin embedded (FFPE) tumor tissue. Numbers in parentheses represent the total number of patients included in this analysis from each dataset. (B) Histogram indicating the TMB across 3969 samples, with median TMB (red line) and hypermutation cutoff value (black line). (C) Plot representing median TMB for metastatic versus primary tumors.

distant metastatic tumors biopsies (France Study 2016 and MBCproject; 4/123) was similar compared to primary tumors (TCGA-BRCA, 25/977) (3.1% versus 2.6%, P ¼ 0.7). Volume xxx

-

Issue xxx

-

2019

The frequency of hypermutation among primary tumors that eventually became metastatic (MBCproject; 3/126) was similar to that of primary tumors overall, most of https://doi.org/10.1016/j.annonc.2019.11.010

3

R. Barroso-Sousa et al.

Annals of Oncology which did not recur (TCGA-BRCA, N ¼ 25), (2.4% versus 2.6%, P ¼ 1). Hypermutated breast cancers have a higher cytolytic score and higher neoantigen burden Neoantigen burden was evaluated in tumor samples along with corresponding germline samples from the MBCproject (N ¼ 157).The four hypermutated tumors in this cohort had a higher neoantigen burden compared to 153 non-hypermutated tumors (supplementary Figure S5A and supplementary Methods, available at Annals of Oncology online). We evaluated whether hypermutation correlates with an increased immune cytolytic activity, a surrogate of tumorinfiltrating lymphocytes.6 Using RNA-sequencing data from the TCGA-BRCA dataset (N ¼ 974), we observed that hypermutated tumors (N ¼ 25) had higher cytolytic activity compared to non-hypermutated tumors (P < 0.005; supplementary Figure S5B, available at Annals of Oncology online). Together, these data suggest that hypermutated breast tumors may have a higher neoantigen burden, which may result in increased T-cell infiltration. APOBEC is the dominant mutational process among hypermutated breast cancers We next investigated the potential drivers of hypermutation in breast cancer by assessing the mutational signatures present in these hypermutated tumors (Figure 2). Mutational processes in cancer can arise via diverse mechanisms, including defective DNA repair, enzymatic modification of DNA, ultraviolet light damage, or tobacco smoke and generate unique patterns of mutation types called mutational signatures. We found that most hypermutated breast cancers (59.2%) have a dominant APOBEC activity signature (signature 2 and 13). The APOBEC signature is attributed to the AID/APOBEC family of cytidine deaminases that, when dysregulated, are a major source of mutations in several cancers. In another 36.4% of the hypermutated breast cancers, we found the dominant signature signified a mismatch repair deficiency (MMRd) (signature 6, 15, and 20); w1.0% of tumors had a signature of homologous recombination deficiency (signature 3) associated with BRCA1/2 mutations and 3.4% of tumors exhibited a dominant signature associated with dysregulation of DNA polymerase epsilon (POLE) (signature 10) (Figures 2AeB). The median TMB was higher for samples with dominant APOBEC and homologous recombination deficiency signatures (17.1 mut/Mb and 59.4 mut/Mb, respectively), followed by tumors with dominant POLE and MMRd signatures (12.2 mut/Mb and 12.9 mut/ Mb, respectively) (Figure 2C). The genomic landscape of hypermutated breast cancers with and without a dominant APOBEC signature We determined the differences in the genomic landscape between hypermutated tumors with or without a dominant APOBEC signature (Figure 2D). PIK3CA was found to be mutated in 68.6% of hypermutated tumors with a dominant 4

https://doi.org/10.1016/j.annonc.2019.11.010

APOBEC signature versus 37.6% of hypermutated tumors without a dominant APOBEC signature (P ¼1.63  105, q ¼ 0.015). The proportion of PIK3CA mutations in the helical domain were significantly enriched within hypermutated tumors with a dominant APOBEC signature (47.3% versus 25.0%; P ¼ 0.01; Figure 2E) and mutations in the kinase domain were enriched in hypermutated tumors without a dominant APOBEC (P ¼ 0.19). Detailed PIK3CA alterations counts in different gene domains in hypermutated tumors are presented in supplementary Table S4 (available at Annals of Oncology online). Response to anti-PD-1/PD-L1 based therapies in hypermutated breast cancer Previous studies have demonstrated a correlation between hypermutation and response to ICI.14e16 However, MMRd is the only hypermutated phenotype that has been approved as a marker for ICI use. We hypothesized that hypermutated breast cancers may respond to ICI independent of the underlying mutational process. To test this, we examined the treatment histories of 222 patients with MBC from a prospective metastatic biopsy cohort at DFCI.13 We identified eight patients (3.6%) with hypermutated breast cancer, of whom four had received treatment with anti-PD-1/PD-L1based therapies. Notably, three of these patients achieved an objective response to therapy and a prolonged progression-free interval (Figure 3). Previous systemic treatment details for these three patients are presented in supplementary Table S5 (available at Annals of Oncology online). The fourth patient died 2 weeks after the start of therapy with an anti-PD-1 antibody and hence response to therapy could not be assessed. Analysis of mutational signatures in the metastatic biopsies from the three responding patients demonstrated dominant APOBEC activity signature in two patients and a dominant MMRd signature in the third. DISCUSSION To our knowledge, this work represents the largest study evaluating the prevalence and the mutational processes driving hypermutation in breast cancers. Using sequencing data of 3969 patients from six different cohorts, we found that the prevalence of hypermutation was 5%. The prevalence of hypermutation was significantly higher in metastatic tumors than in primary tumors (8.4% versus 2.9%) and metastatic ILC compared to metastatic IDC (17.0% versus 7.8%). Subsets of these hypermutated tumors revealed a higher neoantigen burden and higher cytolytic activity compared to non-hypermutated tumors. APOBEC activity was the most common dominant mutational process associated with hypermutation in breast cancer (59.2%) followed by the MMRd signature (36.4%). We identified three patients with hypermutated breast cancersdtwo with a dominant APOBEC activity signature and one with a dominant MMRd signaturedwho achieved objective and durable responses following pembrolizumabbased regimens. Volume xxx

-

Issue xxx

-

2019

R. Barroso-Sousa et al.

A

B

Tumor samples

APOBEC MMRd HRD POLE

0.00

0.25

0.50 Signature score

0.75

60

59.2

40

36.4

20

3.4

1.0

0

1.00

APOBEC

MMRd

POLE

HRD

Dominant signature

C

D 300

PIK3CA

200

4

150 −log10 P value

Tumor mutation burden (log10 scale)

Percentage of patients with dominant signature

Annals of Oncology

100 75 50

3 P ≤ 0.01

2

1 25 0 −30 10

−20

−10

0

10

Difference in mutation rate (%) APOBEC

HRD

MMRd

POLE

Dominant signature

E

Samples with dominant APOBEC

APOBEC Signature category

Proportion of mutations

P = 0.01

Samples without dominant APOBEC

P = 0.19

0.4 0.3 0.2 0.1 0.0 Helical

Kinase PIK3CA domains

Figure 2. Mutational signatures prevalent in hypermutated breast cancer. (A) The signature score proportions (x-axis) for 206 hypermutated patients (y-axis). Light green represents apolipoprotein B mRNA-editing enzyme catalytic polypeptidelike 3 (APOBEC); burgundy represents DNA mismatch repair deficiency (MMRd); blue represents homologous recombination deficiency (HRD); and dark green represents the error-prone DNA polymerase epsilon (POLE) signature. (B) The bar plot represents the percentage of patients across four dominant signatures; 59.2% of patients have a dominant APOBEC signature. (C) Tumor mutation burden (TMB) (y-axis, log scale) distribution across four dominant signatures. (D) Volcano plot indicating mutational rate differences (x-axis) for each gene (represented as dot), and significance (y-axis, negative-log scale). Green dots are genes with a higher mutation rate in hypermutated tumors with dominant APOBEC activity. Burgundy dots are genes with a higher mutation rate in hypermutated tumors without dominant APOBEC activity. PIK3CA is significantly enriched in APOBEC dominant hypermutated tumors. (E) PIK3CA alteration proportions for helical and kinase domain in hypermutated tumors with dominant APOBEC activity (green) and hypermutated tumors without dominant APOBEC activity (burgundy).

Volume xxx

-

Issue xxx

-

2019

https://doi.org/10.1016/j.annonc.2019.11.010

5

R. Barroso-Sousa et al.

Annals of Oncology TMB: 13.3 Mut/Mb Dominant signature: APOBEC

Patient 1, female, 57 years

Liver Bx Invasive carcinoma ER–/PR–/HER2–

Liver Bx* Invasive carcinoma ER+/PR–/HER2– Letrozole + palbociclib (7 months)

Clinical trial (2 months)

Pembrolizumab + eribulin NCT02513472 (4 months; complete response)

Ongoing complete response

No therapy (30 months; no evidence of disease)

TMB: 93.8 Mut/Mb Dominant signature: APOBEC Chest wall Bx Invasive carcinoma ER+/PR–/HER2–

Patient 2, female, 62 years Liver Bx* Invasive carcinoma ER+/PR–/HER2– Fulvestrant + capecitabine (17 months)

Exemestane + everolimus (8 months)

Clinical trial (13 months)

Eribulin (6 months)

Liposomal doxorubicin (6 months)

Letrozol + palbociclib (9 months)

Carboplatin Gemcitabine (3 months) (5 months)

Disease progression

Pembrolizumab + nab-paclitaxel (12 months; partial response)

TMB: 21.3 Mut/Mb Dominant signature: MMRd Skin Bx Invasive carcinoma ER–/PR–/HER2–

Patient 3, female, 60 years Lymph node Bx* Invasive carcinoma ER–/PR–/HER2– Capecitabine (10 months)

Navelbine (13 months)

Paclitaxel (17 months)

Eribulin (9 months)

Cisplatin Clinical trial (3 months) (5 months)

Disease progression

Pembrolizumab NCT02447003 (12 months; partial response)

Figure 3. Details of the treatment received in the metastatic setting of patients with advanced breast cancer treated with pembrolizumab-based therapy. APOBEC, apolipoprotein B mRNA-editing enzyme catalytic polypeptide-like 3; Bx, biopsy; Dx, diagnosis; ER, estrogen receptor; MMRd, mismatch repair deficiency; PR, progesterone receptor; TMB, tumor mutation burden.

We hypothesize that the increased prevalence of hypermutation in metastatic samples may be due to acquisition during tumor evolution and could be associated with resistance to previous systemic therapy or with the development of metastases. Indeed, the APOBEC signature is observed late in breast cancer evolution.17,18 An alternative possibility is that hypermutation is enriched in primaries that ultimately become metastatic and thus may be a poor prognostic factor. In an exploratory analysis, we did not find significant enrichment of hypermutation in patients’ primary tumors, which ultimately became metastatic when compared with primary tumors in general, most of which did not recur. The frequency of hypermutation was also similar in distant metastatic biopsies when compared to primary tumors. Additional studies with sufficient sample size evaluating hypermutated metastatic biopsies paired with primaries from the same individuals will be necessary to clarify this issue further. Although PD-L1 expression on immune cells is now established as a predictive biomarker of benefit to atezolizumab plus chemotherapy in mTNBC,2 there are concerns surrounding the broad utility of PD-L1 expression for selecting patients for ICI, including the fact that PD-L1 is a dynamic marker with varying expression over time, use of different detection assays, and discordance among pathologists in determining PD-L1 positivity. In addition, some patients who test positive for PD-L1 may not respond to therapy, while some patients who test negative may still respond.19 This has led to the investigation of additional biomarkers to predict a benefit or resistance to ICI. High TMB has been associated with improved clinical benefit to ICI in multiple cancers,14e16 and TMB and PD-L1 expression 6

https://doi.org/10.1016/j.annonc.2019.11.010

are independent predictive markers of response to ICI and have low correlation across multiple tumors.14 Since MMR defects are an important mechanism associated with hypermutation, we investigated whether hypermutated breast tumors were also associated with MMRd. We found that while 36.4% of these hypermutated tumors have a dominant MMRd signature, the majority (59.2%) had a dominant APOBEC mutational signature. Importantly, this suggests that a majority of hypermutated breast tumors may be missed if we only search for markers of MMRd or microsatellite instability. Although the clinical utility of TMB as a predictive biomarker to ICI in solid tumors is not yet known, in nonsmall-cell lung cancer, the APOBEC signature was specifically enriched in patients with durable clinical benefit after immunotherapy.18 Additionally, APOBEC upregulation correlates with high levels of PD-L1 expression.20 Therefore, it is conceivable that such genomic alteration works as a mechanism of immune escape from an endogenous immune response in tumors with APOBEC dysregulation. In addition, and in agreement with other studies,21,22 we found a relationship between APOBEC-induced mutagenesis and PIK3CA mutations, especially with mutations in the helical domain (Figure 2E). Miao et al. reported that PIK3CA mutations were associated with a complete or partial response to ICI in microsatellite-stable solid tumors.23 In our DFCI MBC cohort, four patients with hypermutated tumors received ICI-based therapies and all three evaluable patients achieved objective and durable responses (Figure 3). Interestingly, two of them had a dominant APOBEC activity signature while the other had a dominant MMRd signature. To better evaluate whether hypermutated Volume xxx

-

Issue xxx

-

2019

R. Barroso-Sousa et al.

Annals of Oncology

breast cancers are responsive to ICI, we have launched a multicenter, single arm, phase II trial (NIMBUS) of nivolumab plus ipilimumab in metastatic hypermutated HER2negative breast cancers (NCT03789110). Given that more than 250 000 women and men are diagnosed with breast cancer in the United States annually, a frequency of 5% implies that w13,000 patients with hypermutated breast cancer are diagnosed annually in the United States, a number comparable to other cancer subtypes like non-small-cell lung cancers with anaplastic lymphoma kinase gene-rearrangements or ROS1 translocation, in which targeted therapy is successfully applied. Furthermore, the increased frequency of hypermutation among metastatic ILC raises the question of whether all ILC should be tested for hypermutation. The strengths of this study include the large sample size, inclusion of subsets of patients with WES and RNAsequencing, the substantial number of patients with metastatic biopsies, and mutational signature analysis. Nevertheless, our study has some limitations. First, clinical annotation data (e.g. receptor subtypes) were unavailable in some studies. Secondly, there might be differences in the definition of clinical annotations (e.g. metastasis versus primary) across different cohorts (see supplementary Methods, available at Annals of Oncology online). Thirdly, we carried out a combined analysis of different datasets and batch effect and cohort bias are possible. Although previous studies have indicated concordance between findings of similar studies using different technological tools,24 we acknowledge the caveats of comparing TMB across platforms. TMB is influenced by tumor purity, ploidy, sequencing depth of coverage, and analysis methodologies. Since we used previously analyzed publicly available data, the reanalysis of each dataset to recalculate TMB in a standardized manner was not feasible. In addition, the definition of high TMB is still not optimized across cancer types.11,25 The cutoff used to define hypermutation in this study is consistent with the one used in large pan-cancer analysis conducted by Campbell et al.26 Our study used a combination of targeted gene panel and WES to determine the TMB cutoff with a majority of samples coming from gene panels. Although gene panels tend to estimate higher mutation burden, we selected larger gene panels, which show good correlation with WES with respect to TMB calculation.26e28 There are multiple ongoing initiatives to standardize TMB assessment, and further work is necessary to establish the best cutoff for using TMB as a predictive biomarker of response to immunotherapy.29,30 In summary, this study demonstrates that 5% of breast cancers have high TMB, with enrichment in metastatic tumors. These tumors are associated with a higher neoantigen burden and increased T-cell infiltration. Different mutational signatures are present in this population with APOBEC activity being the most common. Preliminary data suggest that hypermutated breast cancers are likely to benefit from ICI, independent of underlying mutational process. Volume xxx

-

Issue xxx

-

2019

ACKNOWLEDGEMENTS We thank Karla Helvie, Laura Dellostritto, Lori Marini, Nelly Oliver, Shreevidya Periyasamy, Colin Mackichan, and Max Lloyd for assistance with the DFCI patient sample collection and annotation. We thank Dr Elizabeth Mittendorf for helpful discussions and comments on the manuscript. We thank Kaitlyn Bifolck for her editorial support with this manuscript. We are grateful to all the patients who volunteered for research protocols and generously provided the tissue analyzed in this study. FUNDING This work was supported by the NCI Breast Cancer SPORE at DF/HCC #P50CA168504 (NW, NUL, and EPW), Susan G. Komen CCR15333343 (NW), The V Foundation (NW), The Breast Cancer Alliance (NW), The Cancer Couch Foundation (NW), Twisted Pink (NW), Hope Scarves (NW), Breast Cancer Research Foundation (NUL and EPW), ACT NOW (to the Dana-Farber Cancer Institute Breast Oncology Program), the Fashion Footwear Association of New York (to Dana-Farber Cancer Institute Breast Oncology Program), and the Friends of the Dana-Farber Cancer Institute (to NUL). DISCLOSURE RB-S has served as an advisor/consultant to Eli Lilly and Roche and has received honoraria from Eli Lilly, Roche, Bristol-Myers Squib, Novartis, and Pfizer. RB-S received travel, accommodations, or expenses from Roche. SMT receives institutional research funding from Novartis, Genentech, Eli Lilly, Pfizer, Merck, Exelixis, Eisai, Bristol Meyers Squibb, AstraZeneca, Cyclacel, Immunomedics, Odonate, and Nektar. SMT has served as an advisor/consultant to Novartis, Eli Lilly, Pfizer, Merck, AstraZeneca, Eisai, Puma, Genentech, Immunomedics, Nektar, Tesaro, and NanoString. EPW receives consulting fees from InfiniteMD and Leap Therapeutics, honoraria from Genentech, Roche, Tesaro, Lilly, and institutional research funding from Genentech. NUL has received institutional research funding from Genentech, Cascadian Therapeutics, Array Biopharma, Novartis, and Pfizer (all institutional). N.W. was previously a stockholder and consultant for Foundation Medicine, has been a consultant/advisor for Novartis and Eli Lilly, and has received sponsored research support from Novartis and Puma Biotechnology. None of these entities had any role in the conceptualization, design, data collection, analysis, decision to publish, or preparation of the manuscript. All remaining authors have declared no conflicts of interest. REFERENCES 1. Adams S, Gatti-Mays ME, Kalinsky K, et al. Current landscape of immunotherapy in breast cancer: a review. JAMA Oncol. 2019;5(8): 1205e1214. 2. Schmid P, Adams S, Rugo HS, et al. Atezolizumab and nab-paclitaxel in advanced triple-negative breast cancer. N Engl J Med. 2018;379(22): 2108e2121. 3. Matsushita H, Vesely MD, Koboldt DC, et al. Cancer exome analysis reveals a T-cell-dependent mechanism of cancer immunoediting. Nature. 2012;482(7385):400e404.

https://doi.org/10.1016/j.annonc.2019.11.010

7

Annals of Oncology 4. DuPage M, Mazumdar C, Schmidt LM, Cheung AF, Jacks T. Expression of tumour-specific antigens underlies cancer immunoediting. Nature. 2012;482(7385):405e409. 5. Kreiter S, Vormehr M, van de Roemer N, et al. Mutant MHC class II epitopes drive therapeutic immune responses to cancer. Nature. 2015;520(7549):692e696. 6. Rooney MS, Shukla SA, Wu CJ, Getz G, Hacohen N. Molecular and genetic properties of tumors associated with local immune cytolytic activity. Cell. 2015;160(1-2):48e61. 7. Giannakis M, Mu XJ, Shukla SA, et al. Genomic correlates of immunecell infiltrates in colorectal carcinoma. Cell Rep. 2016;15(4):857e865. 8. Snyder A, Makarov V, Merghoub T, et al. Genetic basis for clinical response to CTLA-4 blockade in melanoma. N Engl J Med. 2014;371(23):2189e2199. 9. Rizvi NA, Hellmann MD, Snyder A, et al. Cancer immunology. Mutational landscape determines sensitivity to PD-1 blockade in non-small cell lung cancer. Science. 2015;348(6230):124e128. 10. Van Allen EM, Miao D, Schilling B, et al. Genomic correlates of response to CTLA-4 blockade in metastatic melanoma. Science. 2015;350(6257):207e211. 11. Hellmann MD, Ciuleanu TE, Pluzanski A, et al. Nivolumab plus ipilimumab in lung cancer with a high tumor mutational burden. N Engl J Med. 2018;378(22):2093e2104. 12. Campesato LF, Barroso-Sousa R, Jimenez L, et al. Comprehensive cancer-gene panels can be used to estimate mutational load and predict clinical benefit to PD-1 blockade in clinical practice. Oncotarget. 2015;6(33):34221e34227. 13. Cohen O, Kim D, Oh C, et al. Whole exome and transcriptome sequencing of resistant ERþ metastatic breast cancer [abstract]. Cancer Res. 2017;77(4_suppl):Abstract nr S1-01. 14. Cristescu R, Mogg R, Ayers M, et al. Pan-tumor genomic biomarkers for PD1 checkpoint blockade-based immunotherapy. Science. 2018;362(6411). 15. Ott PA, Bang YJ, Piha-Paul SA, et al. T-cell-inflamed gene-expression profile, programmed death ligand 1 expression, and tumor mutational burden predict efficacy in patients treated with pembrolizumab across 20 cancers: KEYNOTE-028. J Clin Oncol. 2019;37(4):318e327. 16. Samstein RM, Lee CH, Shoushtari AN, et al. Tumor mutational load predicts survival after immunotherapy across multiple cancer types. Nat Genet. 2019;51(2):202e206. 17. Nik-Zainal S, Davies H, Staaf J, et al. Landscape of somatic mutations in 560 breast cancer whole-genome sequences. Nature. 2016;534(7605): 47e54.

8

https://doi.org/10.1016/j.annonc.2019.11.010

R. Barroso-Sousa et al. 18. Wang S, Jia M, He Z, Liu X-S. APOBEC3B and APOBEC mutational signature as potential predictive markers for immunotherapy response in non-small cell lung cancer. Oncogene. 2018;37(29):3924e3936. 19. Ribas A, Hu-Lieskovan S. What does PD-L1 positive or negative mean? J Exp Med. 2016;213(13):2835e2840. 20. Boichard A, Tsigelny IF, Kurzrock R. High expression of PD-1 ligands is associated with kataegis mutational signature and APOBEC3 alterations. Oncoimmunology. 2017;6(3):e1284719. 21. McGranahan N, Favero F, de Bruin EC, Birkbak NJ, Szallasi Z, Swanton C. Clonal status of actionable driver events and the timing of mutational processes in cancer evolution. Sci Transl Med. 2015;7(283): 283ra254. 22. Temko D, Tomlinson IPM, Severini S, Schuster-Bockler B, Graham TA. The effects of mutational processes and selection on driver mutations across cancer types. Nat Commun. 2018;9(1):1857. 23. Miao D, Margolis CA, Vokes NI, et al. Genomic correlates of response to immune checkpoint blockade in microsatellite-stable solid tumors. Nat Genet. 2018;50(9):1271e1281. 24. Van Allen EM, Robinson D, Morrissey C, et al. A comparative assessment of clinical whole exome and transcriptome profiling across sequencing centers: implications for precision cancer medicine. Oncotarget. 2016;7(33):52888e52899. 25. Legrand FA, Gandara DR, Mariathasan S, et al. Association of high tissue TMB and atezolizumab efficacy across multiple tumor types [abstract]. J Clin Oncol. 2018;36(15_Suppl):12000. 26. Campbell BB, Light N, Fabrizio D, et al. Comprehensive analysis of hypermutation in human cancer. Cell. 2017;171(5):1042e1056. 27. Garofalo A, Sholl L, Reardon B, et al. The impact of tumor profiling approaches and genomic data strategies for cancer precision medicine. Genome Med. 2016;8(1):79. 28. Rizvi H, Sanchez-Vega F, La K, et al. Molecular determinants of response to anti-programmed cell death (PD)-1 and anti-programmed death-ligand 1 (PD-L1) blockade in patients with non-small-cell lung cancer profiled with targeted next-generation sequencing. J Clin Oncol. 2018;36(7):633e641. 29. Stenzinger A, Allen JD, Maas J, et al. Tumor mutational burden standardization initiatives: recommendations for consistent tumor mutational burden assessment in clinical samples to guide immunotherapy treatment decisions. Genes Chromosomes Cancer. 2019;58(8):578e588. 30. Chan TA, Yarchoan M, Jaffee E, et al. Development of tumor mutation burden as an immunotherapy biomarker: utility for the oncology clinic. Ann Oncol. 2019;30(1):44e56.

Volume xxx

-

Issue xxx

-

2019