Clinical relevance of pathogenic germline variants in mismatch repair genes in Chinese breast cancer patients

The prevalence and clinical relevance of pathogenic germline variants in MMR genes have not been investigated in large series of breast cancers. In this study, we screened the germline variants in MMR genes in 8085 consecutive Chinese breast cancer patients, and investigated the MMR/PD-L1 protein expression and tumor mutation burden (TMB) of breast tumors from MMR variant carriers. We found that 15 of 8085 patients (0.19%) carried a pathogenic germline variant in MMR genes. Compared with non-carriers, MMR variant carriers might have worse recurrence-free survival (unadjusted hazard ratios [HR] = 2.70, 95% CI: 1.12–6.49, P = 0.027) and distant recurrence-free survival (unadjusted HR = 3.24, 95% CI: 1.45–7.22, P = 0.004). More importantly, some of the breast cancers from MMR carriers displayed MMR protein loss (5/13), TMB-high (2/10), and PD-L1 positive expression (9/13). This study showed that MMR variant carriers were rare in breast cancer. They might have worse survival and part of them might benefit from immunotherapy.


INTRODUCTION
Pathogenic germline variants in four mismatch repair (MMR) genes (MLH1, MSH2, MSH6, and PMS2) have been found to result in hereditary nonpolyposis colorectal carcinoma (HNPCC, also called Lynch syndrome), which is an autosomal dominant inherited disease primarily associated with colorectal, endometrial, gastric, small intestinal, hepatobiliary, renal pelvic, and ureteral cancers 1 . Currently, colorectal cancers from MMR variant carriers are associated with poor differentiation, extensive lymphocytic infiltration and a superior survival compared with sporadic cases 2,3 . More importantly, disfunction of MMR genes in HNPCC results in MMR protein loss, and induces errors of DNA polymerase during DNA replication, causing the accumulation of mutations in the genome, especially in the repetitive DNA sequences called microsatellites 4 . As a consequence, HNPCCs typically exhibit loss of MMR protein expression, microsatellite instability (MSI) and high tumor mutation burden (TMB), which make them sensitive to immunotherapy 5,6 .
A recent clinical trial has highlighted the significant responses of solid cancers with MMR deficiency to immunotherapy, regardless of the cancer type 7 , which caused considerable interest in investigating MMR in various cancer types. Whether breast cancer can be recognized as part of the spectrum of HNPCC has been debated. A previous study concluded that breast cancer is not associated with HNPCC 8 . However, multiple studies reported that women from HNPCC families had an increased risk of breast cancer ranging from 2-to 4-fold compared with the general population [9][10][11][12][13][14] , and some of the breast cancers from HNPCC families exhibited MMR protein loss or MSI [15][16][17][18][19][20][21] . These results suggested a potential role for MMR deficiency in breast tumorigenesis. Notably, these previous studies focused on breast cancers from HNPCC families, while MMR genes have not been investigated in a large series of consecutive breast cancers.
In this study, we aimed to investigate the frequency of pathogenic germline variants in MMR genes (MLH1, MSH2, MSH6, and PMS2) in a large series of Chinese breast cancer patients. Then we compared the clinicopathological characteristics and survival between MMR variant carriers and non-carriers. Finally, we explored whether these breast cancer patients with germline MMR variants can potentially benefit from immunotherapy.
Somatic mutation profile, TMB and MSI status of breast cancers from MMR variant carriers Ten of the 15 MMR variant carriers had FFPE tissues with enough tumor purity (>20%) for DNA extraction and depth sequencing on 654 cancer-related genes. To ensure the accuracy of sequencing of FFPE tissue, target region sequencing was performed on FFPE tumor tissues and fresh-frozen tumor tissues from two MMR variant carriers (P8 and P22). Similar somatic mutation profiles (exactly the same oncogenic mutations) and somatic copy number changes were detected in FFPE tumor tissues and freshfrozen tumor tissues from the same case (Supplementary Table 1). The TMB estimated from FFPE tumor tissues was slightly lower than that from fresh-frozen tumor tissues (Supplementary Table 1).

DISCUSSION
In this study, 0.19% of the Chinese breast cancer patients were found to carry a pathogenic germline variant in the four MMR genes. Compared with non-carriers, MMR variant carriers showed distinct histology and poor survival. In addition, 5 of the 13 breast cancers from MMR variant carriers showed MMR protein loss consistent with the underlying germline variants, and the breast cancers with germline MMR variants showed a relatively high rate of TMB-high (2/10) and PD-L1 positive expression (9/13).
We found that pathogenic germline variants in MMR genes were rare in consecutive breast cancers (15/8085, 0.19%), and germline variants were more common in PMS2/MSH6 genes than in MLH1/MSH2 genes in breast cancers. In addition, we found that MMR germline variants might affect the tumor phenotype and somatic mutation profile. Our study revealed a potential association between MMR germline variants and medullary/papillary histology in breast cancer, which was also reported in breast cancers from HNPCC families in previous studies 16,17,19 . In addition, PTEN somatic mutations were more common in breast cancers from MMR variant carriers than in those from non-carriers. A similar observation was also reported in endometrioid cancers, in which PTEN loss/somatic mutation were closely associated with MMR deficiency 22,23 . In contrast, the effect of MMR germline variants on the survival of breast cancer was different from that of colorectal cancer. MMR germline variants predict a better survival in colorectal cancers 2 , while our study and another recent study both found that breast cancer patients carrying MMR germline variants might have worse survival than non-carriers 24 . Nevertheless, independent breast cancer cohorts are needed to validate the association between MMR germline variants and breast tumor phenotype because of the limited number of samples in our study.
The poor prognosis of breast cancer patients carrying MMR germline variants urged us to explore whether these patients could benefit from immunotherapy. In a previous clinical trial 7 , MMR protein loss was established as a biomarker predicting the response to immunotherapy, regardless of the cancer type. A recent study also reported that one metastatic breast cancer patient with an MMR germline variant achieved a robust and durable response upon immunotherapy 25 . In this study, we found that one-third of breast cancers (5/13, 38.5%) with MMR germline variants showed MMR protein loss. This rate was similar to the rate

METHODS Patients
A total of 8085 consecutive breast cancer patients who were treated at the Breast Center of Peking University Cancer Hospital from October 2003 to May 2015 were included in this study 35 . The cohort was unselected for age at diagnosis and family history. Detailed demographic information and tumor characteristics of each patient were collected from medical records and/or telephone interviews. Estrogen receptor (ER), progesterone receptor (PR), and human epidermal growth factor receptor 2 (HER2) status were determined using the breast tumor tissue obtained from a core needle biopsy or taken from surgery. ER or PR immunostaining was considered positive when >1% of the TCs showed positive nuclear staining. HER2 positivity was defined as a score of 3+ via immunohistochemical staining or HER2 gene amplification via fluorescence in situ hybridization. Informed written consent was obtained from all participants. This study was approved by the Research and Ethics Committee of Peking University Cancer Hospital.

MMR germline variant classification
Panel sequencing (including four MMR genes: MLH1, MSH2, MSH6, and PMS2) was performed on genomic DNA extracted from the peripheral blood of the 8085 unselected breast cancer patients 35 . In this study, we reanalyzed the MMR germline variants detected in our previous report. Germline variations were called with GATK (version 3.6). Annotations were defined using ANNOVAR. Only variants with <1% population frequency in the population databases including gnomAD (v3.1.2) and TOPMed (version 20210514) were collected (Supplementary Data 1). Among these, truncating variants (nonsense and frameshift variants) were included in this study, but truncating variants in the last 55 base pairs of the penultimate exon or last exon that potentially avoid nonsense-mediated messenger RNA decay and do not influence known functional domains were excluded. For splice-site, synonymous, nonsynonymous, in-frame, and stop-loss variants, only variants classified as pathogenic or likely pathogenic by ClinVar (version 20210501) were included in the analysis. Variants with conflicting interpretations of pathogenicity in ClinVar were further annotated according to the ACMG/AMP standards and

H&E staining
Hematoxylin and eosin (H&E) staining was performed on 4 μm sections of formalin-fixed, paraffin-embedded breast cancer tissue from each MMR variant carrier. The staining procedures were as follows: dewaxing, dehydration, hematoxylin, differentiation, bluing, eosin, dehydration, clearing, and cover-slipping. The pathologists histopathologically evaluated the tumor cell area in each H&E section.

Immunohistochemistry (IHC)
Immunohistochemistry was performed on 4 μm sections of formalin-fixed, paraffin-embedded tissue from MMR variant carriers. MMR protein detection was carried out using primary antibodies against mlh1 (clone GM002, mouse monoclonal antibody, catalogue numbers: GT230407, working solution, Gene Tech), msh2 (clone RED2, rabbit monoclonal antibody, catalogue numbers: GT231007, working solution, Gene Tech), msh6 (clone EP49, rabbit monoclonal antibody, catalogue numbers: GT219507, working solution, Gene Tech), and pms2 (clone EP51, rabbit monoclonal antibody, catalogue numbers: GT215907, working solution, Gene Tech). MMR protein loss was defined as nuclear immunostaining negative for one or several MMR proteins in the TCs on whole section slides. PD-L1 IHC staining was performed using the primary antibody for the pd-l1 protein (clone SP142, rabbit monoclonal antibody, catalogue numbers: ab228462, dilution ratio 1:400, Abcam). The percentage of PD-L1 was scored as a continuous variable. The threshold for PD-L1 positivity was set at ≥1% positive TCs and/or ICs. At least one positive control (embryo tissue section) and one negative control (PBS instead of primary antibody) were used in each IHC assay ( Supplementary Fig. 1 Sequenced reads were aligned to the human reference genome (NCBI Build 37) by the Burrows-Wheeler Aligner (version 0.1.22). Somatic indels and single nucleotide variations were called by MutLoc with an additional filter to exclude artificial mutations introduced by FFPE tissue. In brief, duplicates and soft clipped reads removed data was analyzed in MutLoc with these parameters (align quality: 30; strand bias: 0.05; keep the mutation site with highest align quality if more than one mutation sites were examined within 11 bp; keep the mutation sites supported by at least three different reads). In addition, we filtered out single strand bias based on a read pair orientation of larger than 20:1. Somatic copy number variations were called by GATK (version 3.6). Function annotations were defined using ANNOVAR. All the somatic mutations detected in breast cancers from the ten MMR variant carriers were listed in Supplementary Data 2. Tumor mutation burden (TMB) was defined as the number of nonsynonymous somatic mutations (single nucleotide variants and small insertions/deletions) per mega-base in coding regions. The TMB of each tumor was determined on 1.6 Mb of sequenced DNA and reported as mutations/Mb. TMB ≥ 10 Mut/Mb was considered as TMB-high.

PCR-based MSI detection
Genomic DNA extracted from the FFPE sections of one case and freshfrozen tumor tissues of two cases had sufficient quality for MSI detection. Genomic DNA extracted from matched blood samples served as controls to filter germline variants in microsatellite loci. Five mononucleotide loci (NR-21, BAT-26, BAT-25, NR-24, MONO-27) were used for MSI analysis (Promega, USA). Tumors were classified MSI-H (≥2 loci exhibit instability), MSI-L (only one locus exhibits instability) or microsatellite stable (MSS, all five loci exhibit stability).  Fig. 3 Somatic mutation profiles of breast cancers from MMR variant carriers. a Landscape of frequently mutated genes in breast cancers from MMR variant carriers. b Comparison of the somatic mutation rates of PIK3CA, TP53, PTEN, and ARID1A between breast cancers carrying MMR germline variants and general breast cancers. Data are analyzed for statistical significance using Fisher's exact test. Two-sided P values < 0.05 were considered to be statistically significant.