Mucinous Histology, BRCA1/2 Mutations, and Elevated Tumor Mutational Burden in Colorectal Cancer

Mucinous colorectal carcinomas (MC) constitute 10% of colorectal malignancies. Recently, an increased risk of colorectal cancer has been demonstrated in germline BRCA1/2 mutation carriers. Furthermore, BRCA1/2 germline mutation carriers have exhibited a higher-than-expected frequency of MC tumors. Here, we investigate the relationship between BRCA mutations and mucinous histology in colorectal carcinoma patients, using both an existing cohort of sequenced colorectal tumors and a prospective case-control study comparing MC and conventional adenocarcinoma (AC) patients tested for BRCA mutations. We discovered that MC tumors exhibit a statistically significantly higher incidence of BRCA mutations in addition to a higher average mutation count when compared to AC tumors in the existing cohort. The strongest predictor of the mutation count was mucinous histology, independently of other variables including microsatellite instability. Contrary to our hypothesis, the first association did not recur in the prospective case-control study, likely due to our pathological definition of MC tumors and small sample size. Finally, we observed a higher tumor mutational burden (TMB) in MC tumors compared with AC tumors. We suggest that the association between MC histology, BRCA mutations, and increased TMB may open the door to the utilization of simple tests (such as histopathologic characterization) to detect patients who may benefit from immunotherapy in colorectal cancer.


Introduction
Colorectal cancer (CRC) is the third most common malignancy worldwide, currently accounting for 700,000 deaths worldwide per year. e global burden of CRC, according to recent estimations, is anticipated to rise by 60% by 2030 [1].
While colorectal tumors were previously considered to be a single homogenous entity, it is now known that they are in fact a heterogeneous collection of tumors, each with its own distinct histological and molecular features that vary in their treatment and prognosis. e heterogeneous population of CRC is mainly comprised of two histological subtypes: 10-15% mucinous carcinomas (MC) and 85-90% adenocarcinomas (AC) [2].
MC tumors have a tendency to develop in young patients and are associated with late diagnosis at advanced stages, possibly because their typical location in the proximal colon is associated with less symptomatic presentation and a faster disease progression [3]. Clinically, MC prognosis has proven to be slightly worse than AC, with 2-8% increased hazard of death even when corrected for stage at presentation [4]. A limited response to systemic therapy in metastatic disease has also been reported [5]. MC histology has therefore been considered as an unfavorable prognostic indicator of CRC.
is consensus has been recently challenged due to the identification of the importance of the sidedness (right vs. left colon) in the prognosis. is has led to an understanding that for colonic MC tumors there is no difference in overall survival after correction for stage and sidedness [6]. Yet, for rectal MC tumors, there is a reduced rate of complete response and tumor downstaging following neoadjuvant chemoradiotherapy [7]. e carcinogenesis of MC is not clearly understood, though the higher prevalence of MC in hereditary and acquired conditions such as inflammatory bowel diseases, hereditary nonpolyposis colorectal cancer (HNPCC), and past radiotherapy treatment suggests that MC may derive from an alternative oncogenic pathway [8]. Regarding the genetic and molecular patterns, MC tumors tend to overexpress the MUC2 and MUC5AC genes which are responsible for the formation of excess mucous. Other common molecular aberrations in MC include higher incidence of PI3K, SMAD4, and BRAF mutations [5,[9][10][11][12]. Importantly, MC tumors are associated with microsatellite instability (MSI), which is known to be involved in most cases of HNPCC and in 15% of sporadic CRCs. MSI is caused by inactivation of DNA mismatch repair genes (e.g., MLH1 and MSH2), triggering an uncontrolled tumor growth [5,[13][14][15].
Classically, BRCA1/2 genes encode important proteins responsible for maintenance of genome integrity and response to DNA damage [15,16]. Hereditary mutated BRCA1/2 tumor suppressor genes are key factors for pathogenesis and development of breast and ovarian cancers. BRCA1/2 role in the carcinogenesis of CRC is currently unknown. Recent retrospective study of BRCA1/2 carriers who developed CRC detected a higher-than-expected incidence of left-sided MC tumors [17]. Ending long-lasting debate, a new meta-analysis has clearly shown a statistically significant increased risk of colorectal cancer development in carriers of BRCA1 mutations [18].
In this study, we aim to further investigate the relationship between BRCA mutations and mucinous histology in colorectal cancer patients.

Patients.
Patients were eligible if they were 18 years of age or older and had a colorectal malignancy with valid histology of adenocarcinoma or mucinous features. Patients were considered as MC if the tumor pathology was described as having one of the following features: mucin-producing cells, signet ring cells, a focal mucinous component, or a mucin predominant feature. All patients provided written informed consent for any genetic research. e study was approved by the Institutional Review Board.
Excluded patients were those who did not have available pathology slides or a sufficient quality of material for BRCA analysis.

Database Analysis of TCGA Measurements.
It is important to mention that TMB assessed by WES is usually reported as the total number of mutations per tumor, while TMB outputs from gene panel assays are usually normalized to mutations per megabase (mut/Mb) because they differ in the number of genes and target region size [20].
In our paper, we utilize a measure called "mutation count," defined as somatic nonsynonymous variants in encoding genes by exome sequencing as determined by TCGA [19,20].
An additional measure we utilize is the MSI score. is measure was also derived from the TCGA database and is evaluated by MSIsensor, a software tool that quantifies MSI in paired tumor-normal genome sequencing data and reports the somatic status of corresponding microsatellite sites in the human genome [22].

Statistical Analysis.
In order to compare different variables between the two groups, we used the chi-squared test and Fisher's exact test for categorical variables and the Student t-test and Mann-Whitney U test for quantitative variables. Analysis of more than two groups was conducted by the Kruskal-Wallis test for quantitative variables and by the chi-squared test for categorical variables. Spearman's rank-order correlation was used to compare two quantitative variables. In the MSKCC cohort, a linear regression model was constructed for all variables that were statistically significantly linked to the mutation count. All p values are corrected for multiple hypotheses by the Bonferroni method [23].

BRCA Mutations Are Linked to MC Histology and a Higher Mutation Count in an Existing Database.
To assess whether there is a higher incidence of BRCA1/2 mutations in MC tumors than in AC tumors, we performed an analysis of a cohort of targeted sequencing of 1134 metastatic colorectal cancer samples [19] (hereby the MSKCC database). e database included 128 MC patients and 725 AC patients (conventional adenocarcinoma), while other histological subtypes were excluded. Our analysis showed a significantly higher incidence of BRCA mutations in the MC tumors compared to AC (19/128 MC 14.8%, 30/725 AC 4.1%, p value <0.001, by chi-squared). e MSKCC database also includes the mutation count for each sample, defined as somatic nonsynonymous variants in encoding genes by exome sequencing as determined by TCGA; this feature is known to be prominent among MC tumors and is often linked to MSI [24]. Interestingly, several other variables in the MSKCC database presented a similar behavior, several of which were known features of MC tumors ( Table 1): age at diagnosis, fraction of genome altered, and primary tumor location (average mutation count for right colon tumor was 20.1 versus 9.5 for left colon tumors). For the latter variable, this relation remained even when examining the exact tumor site (cecum-22, ascending colon-19, hepatic flexure-18.6, and no specific location in right colon-19.5 average mutation count). As expected, the MSI score (see Methods) was also statistically significantly correlated with mutation count (p value <0.001). e average mutation count in tumors with MC histology was 24.8, indeed much larger than the average mutation count of 8.9 for tumors with AC conventional histology (p value <0.001). We noticed that BRCA mutations were linked to a higher mutation count in a statistically significant manner. We found a much larger amount of mutations in patients with mutated BRCA somatic genotypes versus patients with the wild-type (WT) somatic BRCA genotype (average of 59.4 versus 9.4, respectively, p value <0.001).

BRCA-Mutated Tumors Can Be Divided into a High Mutation Count Group with Mucinous Histology and a Low Mutation Count Group with Adenocarcinoma Histology.
While tumors with BRCA mutations indeed tended to have higher mutation counts (Figure 1(a)), the analysis revealed two distinct groups of BRCA-mutated tumors that differ significantly in their mutation count: a group with high mutation count and group with low mutation count. While some of the variability between these two groups could be explained by MSI score, some of the BRCA-mutated tumors did not have a high MSI score despite a high mutation count (Figure 1(b)). We decided to employ two parallel strategies in order to further explore this phenomenon. (A) We compared the different variables in the MSK database between the two groups. (B) We studied the relationship of the different variables with the mutation count directly among BRCA-mutated tumors. We suspected that some features would discriminate between the two groups, and, indeed, fraction of genome altered, tumor sample histology, stage at diagnosis, primary tumor locations, and MSI score were significantly different between the two groups (Table 2), a result that was in complete agreement between the two strategies we employed. Finally, we constructed a linear regression model for the mutation count among BRCAmutated tumors, utilizing the features found to be statistically significant in the previous analysis (Table 3). Nearly 0.40 of the variances in the mutation count between the different patients with BRCA mutations could be explained using these variables alone. e strongest predictor of the mutation count was mucinous histology, independently of other variables.  (Figure 2). None of the background features differed significantly between the mucinous histology and adenocarcinoma histology groups, indicating that the two groups were not biased by their background properties (Table 4). Since KRAS, BRAF, and PI3K mutations are known to have a higher frequency in MC patients [5,[9][10][11][12], we performed sequencing tests for those mutations. However, we found no statistically significant differences in the frequencies of these mutations between the two groups, though there was a positive trend in the KRAS   Journal of Oncology mutations towards MC group (p � 0.08). In addition, no association was found between MSI and the MC group. All patients were tested for somatic BRCA1/2 mutations ( Figure 3); among 70 CRC patients, 23 revealed a nonsynonymous BRCA mutation (i.e., 32%). Our cohort presents a trend towards a higher frequency of nonsynonymous mutations in either BRCA1 or BRCA2 in MC tumors compared to AC tumors, but it is not statistically significant (12/30, 40% of MC group, 11/40, 27% of AC group, p value � 0.2705, by chi-squared test). However, when analyzing BRCA2 mutations separately, we did observe a trend towards a higher frequency of mutations in the MC group (9/29, 31% of MC group, 6/40, 15% of AC group). Additionally, two pathogenic mutations of BRCA2 were present only in the MC group (c.7480 C > T and c.1670 T > C). Notably, one common mutation (c.8850 G > T) comprised half of the BRCA2 mutations detected in the AC group. On the BRCA1 gene, the same pathogenic mutation c.68_69delAG was present in both MC and AC groups. e distribution of mutations along the genes by the cBioPortal mutation mapper tool (https://www.cbioportal.org/mutation_mapper) does not indicate a bias for specific or hotspot locations or domains along the proteins between the two groups ( Figure 4).

A Prospective
Lastly, since we observed a higher mutation count in MSKCC data for both MC tumors and BRCA-mutated  Table 5, p � 0.07). In addition, BRCA-mutated tumors had numerically elevated TMB, in comparison to BRCA WT cancers (p � 0.14).

Discussion
In the current work, we have suggested a novel correlation between CRC histology, mutational burden, and BRCA mutations.
Our analysis of the MSKCC database detected a statistically significant higher incidence of BRCA mutations in the MC group as listed above (19/128 MC 14.8%, 30/725 AC 4.1%, p value <0.001, by chi-squared test). Additionally, average mutation counts in tumors with MC histology were higher compared with the AC group (24.8 and 8.9, respectively, p value <0.001). Our analysis might shed a light into the relationship between BRCA mutations and high mutation counts, since the mutated BRCA group has shown higher mutation counts compared with the BRCA WT group (average of 59.4 versus 9.4, respectively, p value <0.001).
Furthermore, we demonstrated two distinct groups of tumors with BRCA mutations: a high-mutation-count group with both mucinous histology and high MSI and a lowmutation-count group with both adenocarcinoma histology and low MSI score.
is finding can be explained by the well-known association between mucinous histology and MSI, suggesting MSI as a reasonable explanation for the high mutation counts in the MC group. Nevertheless, our analysis further revealed a small group of BRCA-mutated tumors with high mutation counts and a low MSI score (Figure 1(b)), possibly implicating BRCA as an independent predictor of high mutation count.
It is interesting to ponder what characterizes these different subgroups and what causes the high mutation   (Table 2, Table 3). Some of the variables were not statistically significantly linked to the mutation count within the regression model, indicating additional correlations between variables within the model which explain the same variance in the mutation count. e strongest predictor of mutation counts was mucinous histology, independently of other variables, possibly suggesting that this feature determines the mutation count in patients with BRCA mutations.
Since the linear regression model indicated that mucinous histology, and not MSI, is the best predictor of mutation counts, it is possible that the BRCA-mutated low-MSI, high-mutation-count group is associated with mucinous histology. Our data also correlates with a previous report by Ciriello et al. [26], who characterized a subset of ultramutated CRC with an altered double-strand break repair mechanism. Notably, >50% of these tumors had somatic mutations in BRCA1/2 genes. However, a further study should be done to validate and establish the existence of this specific subgroup.
With the intention to robustly establish the link between MC histology and BRCA mutations, we tested a cohort of AC and MC patients with similar background features in our medical center ( Figure 3). Unfortunately, we could not reestablish the statistically significant link between BRCA mutations and the MC group (12/30, 40% of MC group and 11/40, 27% of AC group were BRCA-mutated).
Notably, even mutations that are known to be found in significantly higher incidence in MC tumors such as KRAS, BRAF, and PI3K were not seen in our cohort, prompting a suspicion that the lack of association is related to limitations of this specific cohort itself, and may explain our failure to reestablish the link between BRCA mutations and MC histology. Indeed, this analysis was performed on a limited sample size and with a broad definition of MC histology.
is broad definition was linked to the variance between observers and to the MC WHO criteria, which are based on the evaluated amount of mucin, a component that is difficult to define accurately. However, we observed a trend towards a higher frequency of BRCA2 mutations in the MC group (9/ 29, 31% of MC group, 6/40, 15% of AC group).
Lastly, since we observed a higher mutation count in MSKCC data, we have decided to further investigate the implications of this finding and to reestablish it in our local cohort. To link mutation count and TMB, we relied on a previous method described by Chalmers et al. [24], where mutation count was divided by the estimated exome sample size of 38 Mb to calculate mutation count per MB. Mutation count per MB was found equivalent to TMB per MB as both represent the total number of mutations counted divided by the size of the coding region of the targeted territory.
Later, TMB analysis was performed in a prospective cohort ( Figure 5, Table 5). We observed that MC tumors are enriched for intermediate-high TMB tumors ( Figure 5, p � 0.07). A study by Naseem et al. [27] may hint at the importance of this finding; this impressive study presented 6396 CRC tumor samples tested with next-generation sequencing for pathogenic mutations, MSI and TMB. BRCA pathogenic mutations were detected in 1.1% (n = 72) of tumors, while BRCA2 in 2.8% (n = 179). BRCA1/2mutations were associated with higher TMB in all CRCs, including MSI-H and MSS cases (p < 0.001). Among MSS cases with POLE wild-type status, BRCA1 (p � 0.0269) and BRCA2 (p � 0.0151) mutations were associated with high      Journal of Oncology mechanism might be the driver for a higher-mutation load or alternatively that the mutations in the BRCA genes themselves are passenger mutations due to the overall increased mutations load. An intriguing question might be "what is the further impact of our findings on the evaluation of CRC patients of Jewish-Ashkenazi ancestry, for whom the incidence of germline BRCA and Lynch syndrome mutations are higher [28]?" It is important to emphasize that genetic testing for germline mutations involves important ramifications regarding the genetic counseling needed for descendants and the potential cascade testing. us, testing for germline mutations warrants patients' consent and understanding. Moreover, we tried to utilize PARP inhibition approach in one of the patients in our cohort, as PARP inhibition is synthetically lethal in BRCA-deficient tumors (FDA approved for ovarian, pancreatic, and breast tumors with BRCA1/2 mutation [5,25,29,30]). e patient was a 53-year-old male with rectal adenocarcinoma (mismatch repair proficient, KRAS and BRAF wild-type) with pelvic and lung metastases. He underwent somatic tumor analysis that showed pathogenic BRCA1 mutation (c.68_69delAG), later proved to be germline. Following achievement of maximal response to first-line chemotherapy with FOLFOX and anti-EGFR antibody (Panitumumab), the patient started Veliparib (PARP inhibitor, kindly provided by AbbVie) on July 2017. e treatment was well tolerated on 300 mg BID and the patient remained with stable disease (Figure 6) for almost 23 months (June 2019), when new mediastinal and pulmonary lesions appeared. As represented here, PARP inhibitors might serve as a potential future therapeutic approach in BRCA-mutated CRC, especially for challenging MC patients.
In addition, emerging evidence suggests that BRCA1 mutation may even influence the survival outcomes among metastatic CRC patients treated with Oxaliplatin or Irinotecan-based regimens [31].
Taken together, this data imply that BRCA1/2 and MC histology may serve as a potential surrogate marker for tumors with higher TMB. is "low-tech" biomarker can increase the number of patients who may benefit from novel treatment strategies based on immunotherapy for TMB-high tumors [28]. Further studies are required to elucidate the real-world value of TMB analysis in MC colorectal cancer with or without BRCA1/2 mutation.

Data Availability
e data used to support the findings of this study are available from the corresponding author upon request.

Conflicts of Interest
e authors declare that there are no conflicts of interest regarding the publication of this article.