Characteristics of Mismatch Repair Deficient Colon Cancer in Relation to MMR Protein Loss, Hypermethylation Silencing, and Constitutional and Biallelic Somatic MMR Gene Pathogenic Variants

Abstract

BACKGROUND:

Mismatch repair deficient colon cancer is heterogeneous. Differentiating inherited constitutional variants from somatic genetic alterations and gene silencing is important for surveillance and genetic counseling.

OBJECTIVE:

Determine the extent to which the underlying mechanism of loss of mismatch repair influences molecular and clinicopathologic features of microsatellite instability-high colon cancer.

DESIGN:

Retrospective analysis.

SETTING:

Comprehensive cancer center.

PATIENTS:

Patients with microsatellite instability-high colon cancer of stage I, II, or III.

INTERVENTION:

Curative surgical resection.

MAIN OUTCOMES AND MEASURES:

Hypermethylation of the MLH1 promoter, biallelic inactivation, constitutional pathogenic variant, and loss of specific mismatch repair proteins.

RESULTS:

Of the 157 identified tumors with complete genetic analysis, 66% had hypermethylation of the MLH1 promoter, 18% had constitutional pathogenic variant (Lynch syndrome), 11% had biallelic somatic MMR gene pathogenic variants, and 6% had unexplained high microsatellite instability. The distribution of mismatch repair loss was as follows: MLH1 and PMS2 co-loss, 79% of the tumors; MSH2 and MSH6 co-loss, 10%; MSH6 alone, 3%; PMS2 alone, 2%; other combinations, 2%; no loss, 2%. Tumor mutational burden was lowest in MLH1- and PMS2-deficient tumors. MSH6-deficient tumors had the lowest levels of tumor-infiltrating lymphocytes, lowest MSIsensor scores, and fewest frameshift deletions. Patients with MLH1 promoter hypermethylation were significantly more likely to be older and female and to have right-colon lesions than patients with biallelic inactivation. Mutation was the most prevalent second hit in tumors with biallelic inactivation and tumors of patients with Lynch syndrome.

LIMITATIONS:

Potential selection or referral bias, missing data for some patients, and relatively small sizes of some subgroups.

CONCLUSIONS:

Clinical characteristics of mismatch repair deficient colon cancer vary with the etiology of microsatellite instability, and its molecular characteristics vary with the affected mismatch repair protein. See Video Abstract at http://links.lww.com/DCR/Bxxx.

INTRODUCTION

The DNA mismatch repair (MMR) system is responsible for recognizing and correcting replication errors in newly synthesized DNA. Tumors with deficient MMR (MMR-D) accumulate abundant mutations and develop into adenocarcinoma along the microsatellite instability-high (MSI-H) pathway, as opposed to the chromosomal instability pathway. MSI-H tumors, which account for 15–20% of colon cancer cases, can be identified by defective MMR proteins, molecular evidence of MSI-H, and high mutational burden (hypermutation).^1,2 Commonly found in the proximal colon, they demonstrate poor differentiation with medullary features and have abundant tumor-infiltrating lymphocytes.³ In the metastatic setting, the hypermutator MSI-H phenotype is particularly responsive to immunotherapy with checkpoint blockade, although only 20–50% of patients with MSI-H colorectal cancer achieve an objective response, with an even smaller fraction achieving durable responses.^4,5

At least 3 molecular mechanisms are responsible for the development of MSI-H colon cancer: MLH1 hypermethylation, constitutional pathogenic variant (Lynch syndrome, LS), and double somatic (DS) mutations. Acquired hypermethylation (HM) of the MLH1 promoter with epigenetic gene silencing, the most common mechanism, is associated with BRAF V600E mutation and sporadic tumor development.^4,6,7 About 25% of MSI-H colon cancers are associated with LS, which results from a constitutional pathogenic variant of MLH1, MSH2 (EPCAM), MSH6, or PMS2.^4,8 The second allele is inactivated by mutation or loss of heterozygosity (LOH). MSI-H can also occur by biallelic somatic inactivation of the MMR genes. This DS MMR loss can occur by mutation or a combination of mutation and LOH. Before the widespread characterization of somatic mutation analysis, these cases were often referred to as Lynch-like.^9–11

The etiology of MMR-D colon cancer is relevant to clinical care for multiple reasons. In patients with MMR-D tumors due to LS, implementation of high-risk cancer surveillance and risk-reducing strategies, as well as predictive genetic testing for at-risk family members, is necessary. MMR-D etiology is also associated with important clinicopathologic and molecular features. Etiology and the pattern of MMR protein loss have also been associated with response to combination cytotoxic and targeted therapy.^5,12 Moreover, in patients with advanced MMR-D tumors, additional biomarkers prognostic of tumor recurrence and/or predictive of tumor response to immunotherapy have yet to be identified. In-depth characterization of MMR-D tumors may help to further differentiate the heterogeneity of this subset of tumors.^1,13,14 This study was aimed at comparing clinical characteristics, pathology findings, and driver mutations among MMR-D colon cancers according to MSI etiology and/or patterns of MMR protein loss.

MATERIALS AND METHODS

Patients

With approval from the institutional review board, institutional databases were queried for patients with MMR-D colon cancer who underwent curative surgical resection at Memorial Sloan Kettering Cancer Center (MSK) between January 2014 and December 2019 and who had next generation sequencing (NGS) results available. Patients who received neoadjuvant therapy, patients who had an incomplete resection, and patients with known metastatic disease at diagnosis were excluded.

MSI-H etiology

MSI-H was identified with the MSK-IMPACT (Integrated Mutation Profiling of Actionable Cancer Targets) assay (see below) using an MSIsensor score threshold of 10, which corresponds to 10% of evaluated microsatellites being unstable. MSIsensor has >99% concordance with MSI PCR and immunohistochemistry (IHC) for MMR proteins.¹⁵ Tumors with hotspot mutations in the exonuclease domain of DNA polymerase ε or δ1 were excluded.

MSI-H etiology was determined by a PCR-based MLH1 methylation test, somatic mutations from NGS, and IHC staining of MMR proteins in the tumor. The tumors were then categorized into 4 groups: HM-MSI (silencing of the MLH1 promoter region and/or a BRAF V600E somatic mutation), LS-MSI (pathogenic germline MMR mutations), DS-MSI (biallelic MMR gene loss), and unexplained MSI etiology. For LS-MSI and DS-MSI, the mechanism of MMR gene loss was further characterized by NGS as mutation or LOH if that characterization was concordant with abnormal IHC MMR staining.

Clinical and pathologic evaluation

Clinicopathologic data were retrieved from electronic medical records. Patient histories of metachronous and synchronous LS-related cancers and family histories of colorectal cancer and LS-related cancers were reviewed according to the Amsterdam II criteria and the revised Bethesda criteria for hereditary colorectal cancer and compared between the groups. Tumors were staged according to the criteria of the AJCC Cancer Staging Manual, 8^th ed., and classified as right-sided (cecum, ascending colon, or transverse colon) or left-sided (descending colon, sigmoid, or rectosigmoid).

Tumor-infiltrating lymphocytes (TIL) were quantified as previously described.^16,17 Briefly, a tumor was classified as TIL high if the number of lymphocytes per high-power field was ≥4, averaged from 5 high-power fields in an area determined (upon examination of the entire tumor) to have the highest TIL concentration.¹⁷

IHC staining for MMR protein (MLH1, MSH2, MSH6, and PMS2) expression was performed as previously described.¹⁸ Tumor samples lacking one or more proteins were categorized as MMR deficient. MMR loss was categorized as MLH1 deficiency (co-loss of MLH1 and PMS2), MSH2 deficiency (co-loss of MSH2 and MSH6), loss of MSH6 alone, or loss of PMS2 alone.

Next-generation sequencing

Pairs of tumor DNA and normal-tissue DNA were analyzed by the MSK-IMPACT assay. MSK-IMPACT is a high-throughput NGS platform that captures somatic mutations, copy number alterations, and structural variations in exons and selected introns for 410–468 oncogenic genes.¹⁹ Tumor mutational burden (TMB) was calculated by dividing the total number of mutations by genomic area sequenced and reported as mutations per megabase (mt/Mb). LOH was determined by FACETS, an allele-specific copy number and clonal heterogeneity analysis open-source tool.²⁰ The MSIsensor score and the numbers of frameshift mutations (insertions and deletions) and somatic mutations for designated genes were obtained from MSK-IMPACT. Briefly, MSIsensor identifies and quantifies microsatellites in the regions of the genome covered by MSK-IMPACT and determines whether the microsatellite is statistically different between the patients sequenced blood and the tumor. The MSIscore is a percentage of the sites that are altered.²¹

Statistical analysis

Statistical analyses were performed using R software version 4.0 and IBM SPSS Statistics 25, with a statistical significance cutoff of P < 0.05. Qualitative variables were analyzed with the chi-square or Fisher exact test. Continuous quantitative variables were compared using the Wilcoxon test for medians. Statistical significance p value was adjusted for multiple comparisons with the Bonferroni correction.

RESULTS

Of the 174 patients who met the inclusion criteria, 9 were missing genetic data and 8 were missing MLH1 methylation status. Of the remaining 157 patients, 103 (66%) had HM-MSI, 28 (18%) had LS-MSI, 17 (11%) had DS-MSI, and 9 (6%) had an unexplained MSI-H etiology (Fig. 1).

FIGURE 1.

Determination of MMR-D etiology based on IHC, NGS, constitutional pathogenic variant analysis. Percentage of each subgroup was based on the total cohort of 174 patients.

Clinicopathologic characteristics associated with the etiology of MSI

The clinicopathologic characteristics of each MMR-D MSI-H etiology group are listed in Table 1 and Supplementary Table S1. Patients with HM-MSI colon cancer were older on average and more likely to be female than patients with LS-MSI or DS-MSI cancer, and their tumors were more likely to be in the right colon than those of patients in the other groups.

Table 1.

Clinical and molecular features of MMR-D colon cancer in relation to the etiology of MSI-H.

Characteristic	No. of patients (%)			P value	Pairwise P values
	HM-MSI (n = 103)	LS-MSI (n = 28)	DS-MSI (n = 17)
Age, yearsa	74 (43–89)	48.5 (25–78)	53 (28–89)	<0.0001	<0.0001,b <0.0001,c 0.536d
Sex
Female	67 (65)	9 (32)	6 (35)	0.002	0.002,b 0.017,c 0.933d
Male	36 (35)	19 (68)	11 (65)
Tumor location
Right colon	94 (91)	22 (79)	12 (71)	0.027	0.089,b 0.010,c 0.495d
Left colon	9 (9)	6 (21)	5 (29)
Tumor grade
Well/moderate	64 (63)	16 (59)	13 (76)	0.489
Poor	37 (37)	11 (41)	4 (24)
LVI
Present	49 (48)	11 (39)	6 (35)	0.497
Absent	53 (52)	17 (61)	11 (65)
Stage
I/II	70 (68)	22 (79)	11 (65)	0.500
III	33 (32)	6 (21)	6 (35)
TIL level
Low	34 (33)	9 (32)	4 (24)	0.738
High	69 (67)	19 (68)	13 (76)
TMB, mt/Mba	54 (20–112.3)	68 (29–361.6)	61 (36–134.3)	0.057
No. of frameshiftsa
Deletions	15 (2–37)	13.5 (1–38)	19 (6–39)	0.320
Insertions	4 (0–15)	4 (2–19)	3 (0–14)	0.125
MSIsensor scorea	36 (10–50.5)	32 (10–51)	34 (18–52)	0.221

Abbreviations: DS, double somatic; HM; MMR-D, mismatch repair deficient; hypermethylation; LS, Lynch syndrome; LVI, lymphovascular invasion; MSI-H, microsatellite instability-high; TIL, tumor-infiltrating lymphocytes; TMB, tumor mutational burden.

^aMedian (range)

^bHM-MSI vs LS-MSI

^cHM-MSI vs DS-MSI

^dLS-MSI vs DS-MSI

Tumor grade, lymphovascular invasion, stage at diagnosis, and TIL level did not differ significantly between the groups. Nor did the median number of frameshift deletions or insertions. LS-MSI and DS-MSI tumors had similar median TMBs (68 and 61, respectively), which were somewhat higher than the median TMB of 54 for HM-MSI tumors (P = 0.057) (Table 1). Median MSIsensor scores did not differ significantly between the groups.

IHC staining analysis for the entire cohort of 174 patients showed that 137 tumors (79%) had loss of both MLH1 and PMS2, 18 tumors (10%) had loss of both MSH2 and MSH6, 5 tumors (3%) had loss of MSH6 only, and 4 tumors (2%) had loss of PMS2 only (Supplementary Table S2). The remaining 10 tumors had loss of multiple MMR proteins, no loss of MMR proteins, or missing data. As expected, HM-MSI was associated primarily with loss of both MLH1 and PMS2 expression, although none of HM-MSI tumors exhibited PMS2-only loss.²² In contrast, DS-MSI was associated with loss of both MLH1 and PMS2 and with loss of both MSH2 and MSH6. The LS-MSI group had all 4 patterns of MMR loss.

Family and personal history data for the 3 MSI-H etiology groups are listed in Table 2 and Supplementary Table S3. Twenty-two (79%) of the 28 patients with LS-MSI had family history of colorectal cancer, compared with 37 (36%) of the 103 patients with HM-MSI (p < 0.001) and 5 (29%) of the 17 patients with DS-MSI (p = 0.002). Eight (29%) of the 28 LS-MSI patients had metachronous colorectal cancer, compared with 4 (4%) of the 103 HM-MSI patients (p = 0.0005) and none of the 17 DS-MSI patients (p = 0.017). Family history of colorectal cancer and incidence of metachronous colorectal cancer did not differ significantly between patients with DS-MSI and patients with HM-MSI. Fourteen (50%) of the 28 LS-MSI patients had family history of LS-related noncolorectal neoplasms, compared with 22 (21%) of the 103 HM-MSI patients (p = 0.004) and 5 (29%) of the 17 DS-MSI patients (p = 0.222). Ten (36%) of the 28 LS-MSI patients met the Amsterdam II criteria for LS, compared with none of the 17 DS-MSI patients (p = 0.007) and none of the 103 HM-MSI patients (p < 0.0001). Twenty-six (93%) of the 28 LS-MSI patients met the revised Bethesda criteria for LS, compared with 40 (39%) of the 103 HM-MSI (p < 0.0001) and 13 (77%) of the 17 DS-MSI patients (p = 0.179). DS-MSI patients were more likely to meet revised Bethesda criteria than HM-MSI patients (p = 0.007).

Table 2.

Patient and family history in relation to the etiology of MMR-D.

History	No. of patients (%)			P value	Pairwise P values
	HM-MSI (n = 103)	LS-MSI (n = 28)	DS-MSI (n = 17)
Met diagnostic criteria
Revised Bethesda	40 (39)	26 (93)	13 (77)	<0.001	<0.0001,a 0.007,b 0.179c
Amsterdam III	0	10 (36)	0	<0.001	<0.0001,a 0.007c
Personal history
Metachronous CRC	4 (4)	8 (29)	0	<0.001	0.0005,a 1.000,b 0.017c
Synchronous CRC	3 (3)	1 (4)	0	0.752
LS-related neoplasmsd	4 (4)	3 (11)	0	0.199
Family history
CRC	37 (36)	22 (79)	5 (29)	<0.001	<0.0001,a 0.785,b 0.002c
LS-related neoplasmsd	22 (21)	14 (50)	5 (29)	0.012	0.004,a 0.532,b 0.222c

Abbreviations: CRC, colorectal cancer; DS, double somatic; HM, hypermethylation; MMR-D, mismatch repair deficient; LS, Lynch syndrome; MSI-H, microsatellite instability-high.

^aHM-MSI vs LS-MSI

^bHM-MSI vs DS-MSI

^cLS-MSI vs DS-MSI

^dNoncolorectal

Molecular characteristics associated with the etiology of MSI

The MSK-IMPACT data on oncogenic mutations are shown in Fig. 2A. With control for multiple comparisons, as expected, LS-MSI and DS-MSI were more likely to harbor driver mutations in MLH1 and MSH2, than HM-MSI patients, as these somatic MMR alterations were contributing to the etiology of the MSI-H. KRAS hotspot mutations were most common in DS-MSI patients, while APC mutations were most common in LS-MSI patients. BRAF driver mutations were found in 84 of the 103 HM-MSI tumors, of which 81 had V600E BRAF alterations and 3 had non-V600E BRAF alterations. Two of the non-V600E alterations were from fusion (AGAP3-BRAF).

FIGURE 2.

Oncogenic mutation and mechanism of biallelic inactivation in MMR-D colon cancer. (A) Statistically significant oncogenic mutations in patients based on etiology of MMR-D (Bonferroni corrected). (B) Mechanisms of biallelic inactivation in patients with DS-MSI and patients with LS-MSI. Abbreviations: DS-MSI, double somatic microsatellite instability; HM, hypermethylation; HOMDEL, homozygous deletion; LOF, loss of function; LOH, loss of heterozygosity; LS, Lynch syndrome.

The data on biallelic inactivation in LS-MSI and DS-MSI tumors are shown in Fig. 2B. Of the 17 DS-MSI tumors, 10 (59%) had 2 acquired somatic loss-of-function mutations and 6 (35%) had 1 somatic hit with LOH. Among the 28 LS-MSI tumors, the second hit was somatic mutation in 12 tumors (43%), LOH in 9 tumors (32%), and indeterminate in 5 tumors (18%). Pathogenic loss-of-function alterations in MMR proteins in the LS-MSI and DS-MSI tumors are listed in Supplementary Tables S4 and S5.

Characteristics associated with type of MMR loss

MMR protein expression as determined by IHC was associated with patient age (Table 3). Consistent with the enrichment of HM-MSI, patients with loss of both MLH1 and PMS2 were older on average than the other groups. Four of the 5 tumors with loss of only MSH6 were in the TIL-low category, while tumors in the other 3 groups were mostly TIL-high.

Table 3.

Clinical and molecular features of MMR-D colon cancer in relation to loss of MMR proteins.

Characteristic	No. of patients (%) with loss of:				P value	Pairwise P values
	MLH1/PMS2 (n = 137)	MSH2/6 (n = 18)	MSH6 alone (n = 5)	PMS2 alone (n = 4)
Age, yearsa	72 (28–89)	51 (25–91)	60 (47–73)	55 (38–70)	0.0003	0.002,b 0.130,c 0.130,d 0.587,e 0.902,f 0.667g
Sex
Female	79 (58)	7 (39)	1 (20)	2 (50)	0.191
Male	58 (42)	11 (61)	4 (80)	2 (50)	0.191
Stage
I/II	97 (71)	14 (78)	4 (80)	2 (50)	0.692
III	40 (29)	4 (22)	1 (20)	2 (50)
TIL level
High	99 (72)	13 (72)	1 (20)	2 (50)	0.066
Low	38 (28)	5 (28)	4 (80)	2 (50)
TMB, mt/Mba	53.5 (20.2–115)	71.6 (20.2–134.3)	70.2 (59.7–361.6)	87.8 (29–218.6)	0.001	0.003,b 0.083,c 0.324,d 0.941,e 0.807,f 0.876g
MSIsensor scorea	35 (10–52)	36.5 (10–48)	20 (18–29)	27 (17–43)	0.036	0.855,b 0.027,c 0.697,d 0.158,e 0.827,f 0.905g
No. of frameshiftsa
Deletions	15 (2–39)	19.5 (8–35)	5 (3–8)	8.5 (1–32)	0.001	0.313,b 0.002,c 0.313,d 0.003,e 0.313,f 0.902g
Insertions	4 (0–15)	4 (0–14)	2 (1–13)	12.5 (8–19)	0.010	0.543,b 0.382,c 0.011,d 0.514,e 0.012,f 0.130g

Abbreviations: MMR, mismatch repair; MSI-H, microsatellite instability-high; TIL, tumor-infiltrating lymphocytes; TMB, tumor mutational burden.

^aMedian (range)

^bMLH1/PMS2 vs MSH2/6

^cMLH1/PMS2 vs MSH6 alone

^dMLH1/PMS2 vs PMS2 alone

^eMSH2/6 vs MSH6 alone

^fMSH2/6 vs PMS2 alone

^gMSH6 only vs PMS2 alone

MMR protein expression was also associated with TMB, MSIsensor score, and frameshift mutations (Fig. 3 and Table 3). Median TMB was 53.5 (range, 20–115) mt/Mb in tumors with loss of MLH1 and PMS2, compared with 71.6 (range, 20–134) mt/Mb in tumors with loss of MSH2 and MSH6 (p = 0.003). MSIsensor scores in tumors with loss of MSH6 alone were significantly lower than in tumors with loss of both MLH1 and PMS2. Tumors with loss of only MSH6 also had significantly fewer frameshift deletions than tumors with loss of both MLH1 and PMS2 and tumors with loss of both MSH2 and MSH6. Tumors with loss of only PMS2 had significantly more frameshift insertions than tumors with loss of both MLH1 and PMS2 and tumors with loss of both MSH2 and MSH6.

FIGURE 3.

TMB (A) and MSIsensor scores (B) for MSI-H tumors with different types of MMR loss. Means (red circles) in panel A and medians and first and third quartiles in panel B are indicated.

DISCUSSION

The 4 major MSI-H etiologies (HM, LS, DS, and unexplained) were associated with patient age but not with tumor grade, AJCC stage, lymphovascular invasion, TIL, or MSI score. Although TMB was somewhat lower in HM-MSI tumors than in LS- and DS-MSI tumors, this was not statistically significant. These findings support the hypothesis that sporadic loss of MMR and inherited loss of MMR result in a similar hypermutator phenotype with oncogenesis along the MSI-H pathway.^10,23 The lack of a significant difference in TIL levels between the groups (consistent with the findings of some studies^14,23 but not others²⁴) is potentially explained by the similar rates of frameshift insertions and deletions, resulting in neoantigen presentation on MHC class I and immune cell recruitment. We recently found that frameshift insertions and deletions were more important determinants of TIL level than TMB in MSI colorectal cancer.²⁵

We also found associations between oncogenic somatic mutations and the etiology of MSI-H. LS-MSI tumors had significantly more mutations in APC, a tumor suppressor gene whose mutation leads to overactivation of the Wnt/β-catenin signaling pathway and subsequent proliferation, migration, invasion, and metastasis.²⁶ DS-MSI tumors had more mutations in KRAS, which leads to constitutive activation of downstream proliferative signaling pathways,²⁶ has been associated with poor prognosis, and is a predictive marker of resistance to anti-epidermal growth factor receptor therapy.^5,26 In the HM-MSI cohort, 2 of the samples with MLH1 hypermethylation non-BRAF V600E changes had fusion alterations of BRAF. Cocco et al previously proposed that fusion testing should be completed in these particular patients, as this kinase fusion might be targetable for kinase inhibitors.²⁷ These differences in mutational landscape may explain, in part, the recently reported MSI etiology-associated variation in tumor response to adjuvant therapy.¹²

Similarly to Pearlman et al,¹³ we found that LS accounted for about one-third of the MSI-H tumors that were not associated with HM. DS loss of MMR accounted for 65% of the MSI-H tumors not associated with either HM or LS. Compared with HM-MSI patients, patients with LS-MSI were more likely to have metachronous colorectal cancer and to have family history of colorectal cancer. LS-MSI patients were also more likely to meet Amsterdam criteria. Although 13 of the 17 DS-MSI patients met the revised Bethesda criteria, only 5 had family history of colorectal cancer and none had personal history of synchronous or metachronous colorectal cancer.

These data support the current guidelines for individuals with DS-MSI, which do not require intensive surveillance but rather surveillance based upon personal and/or family cancer history.^10,13 Compared to the historic screening criteria, the current diagnostic algorithms for MMR-D colon cancer—including MMR IHC, NGS, HM analysis, and germline testing—can determine the MMR-D category for 95% of patients, facilitating appropriate follow-up, surveillance, and screening. Universal colorectal cancer testing using the aforementioned algorithm and tests should be implemented. The remaining 5% of cases are likely associated with pathogenic somatic or constitutional pathogenic variants not detectable by current testing methods.^13,23 The clinical and genetic characteristics of this MSI-H group suggest that it may be intermediate between hereditary MMR-D and sporadic MMR-D, thereby requiring individualized cancer surveillance recommendations.

Molecular features in our cohort were associated with the pattern of MMR protein loss rather than MMR-D etiology. MSIsensor scores and the rate of frameshift deletions were lowest in tumors with loss of MSH6 alone, which may explain the high proportion of TIL-low tumors in this group. Klarskov et al²⁸ also found that MSH6-deficient colon tumors had a relatively low prevalence of TIL. These findings may be related to a functional redundancy in the MMR system responsible for the initial recognition of erroneous bases and mismatches. The system consists of the heterodimers hMutSα (MSH2/MSH6 complex) and hMutSβ (MSH2/MSH3 complex). hMutSα recognizes single base mispairs as well as slippages at mono- and di-nucleotide sequences, and hMutSβ recognizes slippages at dinucleotide or longer repeats. Thus, hMutSβ (the MSH2-MSH3 complex) can preserve at least part of the MMR function when the MSH6 complex is inactivated.^5,28,29 This may explain the lower MSIsensor scores and the lower prevalence of frameshift deletions in tumors with MSH6 deficiency.

No such MMR redundancy exists for MLH1, MSH2, or PMS2, and their loss was associated with greater destabilization, higher MSIsensor scores, and higher rates of frameshift deletions compared with MSH6 loss. So, tumors with higher MSIsensor score generate a higher number of neoantigens via frameshift mutations presumably resulting in enhanced immunogenicity. The reduced TMB in MLH1- and PMS2-absent tumors is consistent with the data reported by Salem et al.¹ The heterogeneity in TMB and MSI between tumors with different types of MMR protein loss may have implications for tumor response to immunotherapy.³⁰ Further studies are needed to further explore those implications.

The second acquired hit following pathogenic somatic mutation in DS-MSI tumors or constitutional pathogenic variant in LS-MSI was most commonly somatic mutation rather than LOH. Intriguingly, 1 patient with LS had multiple pathogenic somatic mutations. This finding supports the importance of performing germline analysis in patients with MSI-H even in the presence of multiple somatic mutations on tumor/normal sequencing, especially for tumors with loss of only MSH6 or PMS2, which is less common in patients with HM- or DS-MSI.^9,13,23

The strengths of our study include the well-curated, granular database and in-depth molecular analysis. Limitations include the potential selection or referral bias associated with analysis of data from a single specialized institution, the exclusion of 10% of the identified patients due to lack of genetic analysis data or unknown MLH1 methylation status, and the relatively small numbers of MSH6-deficient and PMS2-deficient tumors. Unfortunately, there were so few Stage IV patients that underwent upfront surgical resection to include in the analysis. Further analysis with a larger dataset is needed to validate our findings with incorporation of clinical outcomes data.

CONCLUSION

In conclusion, the heterogeneity of MMR-D colon cancers with respect to MSI-H etiology and loss of MMR proteins has potentially significant clinical implications. Determining the etiology of MSI-H can help with selection of appropriate genetic counseling and screening for the patient and family members, and determining the pattern of MMR loss can help with identification of patients whose disease is likely to respond to immunotherapy. MSI-H etiology that is not associated with HM, LS, or DS requires further investigation.