Discordant DNA mismatch repair protein status between synchronous or metachronous gastrointestinal carcinomas: frequency, patterns, and molecular etiologies

Abstract

The widespread use of tumor DNA mismatch repair (MMR) protein immunohistochemistry (IHC) in gastrointestinal tract (GIT) carcinomas has unveiled cases where the MMR protein status differs between synchronous/metachronous tumors from the same patients. This study aims at examining the frequency, patterns and molecular etiologies of such inter-tumoral MMR discordances. We analyzed a cohort of 2,159 colorectal cancer (CRC) patients collected over a 5-year period and found that 1.3% of the patients (27/2159) had ≥2 primary CRCs, and 25.9% of the patients with ≥2 primary CRCs (7/27) exhibited inter-tumoral MMR discordance. We then combined the 7 MMR-discordant CRC patients with 3 additional MMR-discordant GIT carcinoma patients and evaluated their discordant patterns and associated molecular abnormalities. The 10 patients consisted of 3 patients with Lynch syndrome(LS), 1 with polymerase proofreading-associated polyposis[PAPP], 1 with familial adenomatous polyposis[FAP]), and 5 deemed to have no cancer disposing hereditary syndromes. Their MMR discordances were associated with the following etiologies: 1) PMS2-LS manifesting PMS2-deficient cancer at an old age when a co-incidental sporadic MMR-proficient cancer also occurred; 2) microsatellite instability-driven secondary somatic MSH6-inactivation occurring in only one - and not all - PMS2-LS associated MMR-deficient carcinomas; 3) "compound LS" with germline mutations in 2 MMR genes manifesting different tumors with deficiencies in different MMR proteins; 4) PAPP or FAP syndrome-associated MMR-proficient cancer co-occurring metachronously with a somatic MMR-deficient cancer; and 5) non-syndromic patients with sporadic MMR-proficient cancers co-occurring synchronously/metachronously with sporadic MMR-deficient cancers. Our study thus suggests that inter-tumoral MMR discordance is not uncommon among patients with multiple primary GIT carcinomas (25.9% in patients with ≥2 CRCs), and may be associated with widely varied molecular etiologies. Awareness of these patterns is essential in ensuring the most effective strategies in both LS detection and treatment decision-making. When selecting patients for immunotherapy, MMR testing should be performed on the tumor or tumors that are being treated.

INTRODUCTION

DNA mismatch repair (MMR) deficiency, detectable by MMR protein immunohistochemistry (IHC) and/or PCR- or sequencing-based microsatellite instability (MSI) testing, is a molecular phenotype that characterizes the vast majority of Lynch syndrome (LS)-associated tumors as well as subsets of sporadic malignancies (1–4). Colorectal cancer (CRC) represents the prototypical tumor type that may harbor this molecular defect. In LS-associated tumors, the MMR deficiency occurs as a result of a deleterious germline mutation in an MMR gene (MLH1, MSH2 / EPCAM, MSH6 or PMS2) as the first hit, and a somatic event inactivating the wild type allele as the second hit. In sporadic CRCs, the MMR deficiency is primarily secondary to MLH1 promoter hypermethylation or biallelic somatic MMR mutations/alterations (5, 6).

MMR protein IHC is widely used as the frontline test for the detection of MMR deficiency given its easy availability. Accurate IHC detection of tumor MMR protein status bears important clinical implications. For LS detection, MMR IHC not only facilitates the identification of patients who should undergo further LS work-up, the IHC staining patterns may also help determine the pathogenicity of gene mutational variants or how the different MMR proteins interact. For therapeutic purposes, MMR IHC is being increasingly used to guide the selection of patients for therapy with immune checkpoint inhibitors.

As the clinical application of MMR deficiency continues to expand and the use of MMR protein IHC continues to increase(7), various seemingly infrequent but potentially pertinent scenarios emerge and call for attention. A particular scenario we have encountered in our practice relates to the complexities brought on by the discordant MMR protein status among synchronous or metachronous carcinomas of the gastrointestinal tract (GIT). This study therefore aims at providing an in-depth evaluation of these discordant cases. In the study, we first evaluated the frequency of inter-tumoral MMR discordance in CRC patients with 2 or more primary CRCs, then analyzed their discordant patterns and the underlying molecular mechanisms. Our data suggest that inter-tumoral MMR discordance is not uncommon and often portends complex molecular alterations. A heightened awareness of this phenomenon may serve to bring a higher level of precision into the detection and clinical application of MMR deficiency in GIT carcinomas.

MATERIALS AND METHODS

The study was approved by the Institutional Review Board. A cohort of 2,159 CRC cases was utilized for frequency analysis. This cohort was retrieved from a prospectively maintained database by the Pathology department of a tertiary care cancer center between January 2013 and the end of 2017. These cases were accrued as part of the institutional protocol to perform MMR IHC on colorectal cancers that either fulfilled certain criteria (the Bethesda guidelines, 2013; aged 70 approach, 2014–2016) or were tested universally (all newly diagnosed tumors, 2017). A second cohort of 12 selected GIT carcinoma patients with ≥ 2 primary GIT cancers (from years 2018–2019) was also examined to identify patients with inter-tumoral MMR discordance in an attempt to enrich such cases for the study of discordant patterns and molecular etiologies. This cohort was collected in a non-systematic fashion.

MMR IHC was performed on whole sections. Antibodies included MLH1 (clone G168–728, diluted 1:250; BD PharMingen or clone M1, ready-to-use, Ventana), MSH2 (clone G219.1129, ready-to-use, Cell Marque), MSH6 (clone 44, ready-to-use, Ventana), and PMS2 (clone A16.4, diluted 1:100, BD Biosciences). As guided by the individual patient’s clinical scenario, selected tumors then underwent further testing as necessary: 1) PCR based MSI testing using the Promega Analysis System; 2) MLH1 promoter methylation via a clinically validated pyrosequencing platform, the PyroMark (Qiagen, Hilden, Germany) platform (the system tests 5 CpG sites within the MLH1 promoter in region −209 to −181 from the transcription start site, and methylation status was graded as "present" when all 5 CpG sites were methylated above 10%)(1)(8–10); 3) Tumor sequencing for detection of tumor somatic mutations and MSI status, using MSK-IMPACT, a hybridization capture based next-generation sequencing(NGS)assay that assesses the coding regions of multiple cancer genes, along with selected promoters, introns, and copy number status, and determines MSI status according to an MSIsensor score (11–13); and 4) Germline mutation testing, performed as part of the clinical diagnostic work-up and according to standard methodologies including MSK-IMPACT. Additionally, MSH6 methylation analysis was performed in the context of this study via the Infinium methylation EPIC/850k platform (Illumina, San Diego, CA)(14).

Among patients with multiple primary synchronous or metachronous GIT carcinomas, those exhibiting discordant MMR IHC results between the different tumors were further analyzed to determine: 1) patient clinical characteristics (personal and family history of malignancy; clinical outcomes - retrieved from the hospital information system); 2) tumor pathological features (tumor histology type; tumor differentiation; increased tumor-infiltrating lymphocytes [TILs] - defined as >=4 TILs/HPF averaged from 5 consecutive fields in areas with the highest TILS); and, 3) the most likely etiology of the abnormal MMR IHC of each individual tumor (utilizing available clinical, pathologic, and molecular data- including germline- and somatic- mutational data).

RESULTS

For frequency analysis, the cohort of 2,159 CRC patients yielded 27 patients (1.3%) with 2 or more primary CRCs (Fig. 1). Of the 27 patients, 20 (71.4%) had concordant IHC MMR protein status between the tumors: 5 had normal MMR protein expression in all tumors and none of the 5 had clinical or molecular evidence of LS; 15 had abnormal MMR protein expression in all tumors and the abnormality patterns were consistent across tumors; 7 of the 15 were LS.

Figure 1.

Identification of patients with multiple primary colorectal carcinomas and inter-tumoral discordance in mismatch repair protein status utilizing a database of colorectal carcinomas tested by mismatch repair protein immunohistochemistry.

The remaining 7 of the 27 patients showed MMR protein discordance between their tumors, resulting a frequency of 25.9% (7/27) for inter-tumoral MMR discordance in patients with 2 or more CRCs.

To enrich cases with inter-tumoral MMR discordance, we examined 12 more patients with 2 or more primary GIT carcinomas collected non-systematically, and found 3 with inter-tumoral MMR discordance.

The 7 MMR-discordant CRC cases and the 3 additional MMR-discordant GIT carcinoma cases constituted our study subjects for the assessment of discordant patterns and molecular etiologies. The 10 patients were classified into 3 groups: LS, non-LS hereditary syndromes, and sporadic/likely sporadic. Their clinical, pathological, and molecular findings are summarized in Table 1 and Figure 2, and described below.

Table 1.

Clinical, pathological, and molecular characteristics of MMR status-discordant synchronous or metachronous gastrointestinal tract carcinomas^*

Case ID	Sex	Tumor ID	Age at dx	Tumor site	Family/personal history	MMR IHC/MSI testing**	BRAF mutation /MLH1 methylation	Tumor somatic mutations in major MMR genes	Pathogenic germline mutation	Presumed molecular basis for the discordant MMR status
GROUP 1_Lynch syndrome (LS)
1	M	1a	81	Sigmoid colon	Brother and sister with cancer, type unknown	MMR-P / MSS	ND	None detected	PMS2 c.641_644dupTGGT (p.Cys216Glyfs*34)	Sporadic MMR-P tumor + PMS2 LS tumor
		1b	83	Cecum		PMS2-D/MSH6 clonal loss / MSI-H	ND	MSH6 c.3261dupC (p. F1088Lfs*5)
2	M	2a	66	Ascending colon	Sister: cervical cancer in 60s; nephew (sister’s son): CRC at 40	PMS2-D / MSI-H	ND	ND	PMS2 IVS9–2A>G	Both PMS2 LS tumors, one with secondary loss of MSH6
		2b	73	Sigmoid colon		PMS2/MSH6-D / MSI-H	ND	ND
3	M	3a	43	Right colon	Father and Paternal grandfather: CRC in their 40’s; Mother: uterine cancer.	MSH2/MSH6-D	ND	ND	MLH1 c.1246A>T (p.K416X) AND MSH2 c.2038C>T (p. R680X)	Dual MLH1 and MSH2 LS manifesting as either MLH1D or MSH2-D tumors
		3b	63	Stomach fundus		MLH1/PMS2-D	ND	ND
GROUP 2_Non-LS hereditary colorectal cancer predisposing syndrome
4	M	4a	45	Sigmoid colon	Father: colon polyps and sarcoma	MMR-P	ND	ND	POLE c.1270C>G (p.Leu424Val)	Both may be POLE -driven, one potentially with somatic inactivation of MSH2
		4b	47	Rectum		MSH2/MSH6-D	ND	ND
5	F	5a	28	Rectum	Maternal grandmother: brain tumor probably metastasis of lung cancer.	MMR-P	ND	ND	APC p.R283*	FAP associated colonic carcinoma + MLH1 methylated jejunal tumor (15 years post total colectomy)
		5b	43	Jejunum		MLH1/PMS2-D / MSI-H	MLH1 methylated	MLH1 c.335A>G (p.H112R)
GROUP 3_Likely sporadic
6	M	6a	52	Stomach	No family history. Personal history of long-standing UC.	MLH1/PMS2-D / MSI-H	ND	MSH2 c.78G>A (p.M26I)***	Negative (expanded gene panel)	Likely sporadic MLH1 methylated tumor + Sporadic MMR-P tumor
		6b	52	Rectum		MMR-P / MSS	ND	None detected
7	M	7a	82	Rectum	None known	MMR-P	ND	ND	ND	Sporadic MLH1 methylated tumor + Sporadic MMR-P tumor
		7b	85	Cecum		MLH1/PMS2-D	BRAF V600E positive by VE1 IHC	ND
8	M	8a	77	Ascending colon	Paternal grandfather: probably died of colon cancer; nephew: “had his colon removed at 24”	MMR-P		ND	Negative (expanded gene panel)	Sporadic MLH1 methylated tumor + Sporadic MMR-P tumor
		8b	77	Transverse colon		MLH1/PMS2-D	BRAF V600E positive by sequencing, MLH1 methylated	ND
		8c	77	Rectosigmoid colon		MMR-P / MSS	ND	None detected
9	M	9a	76	Descending colon	None known	MMR-P / MSS	ND	None detected	Negative (expanded gene panel)	Likely sporadic somatic mutation driven MMR-D tumor + Sporadic MMR-P tumor
		9b	76	Rectosigmoid colon		MSH2/MSH6-D / MSI-H	ND	MSH2 c.2493delA (p.E832Sfs*9)
10	F	10a	75	Sigmoid colon	None known	MMR-P	ND	ND	Negative (targeted testing****)	Sporadic MMR-P tumor + likely sporadic MMR-D tumor
		10b	79	Ascending colon		MSH2/MSH6-D / MSI-H	ND	ND

MMR, mismatch repair; Dx, diagnosis; MMR-P, mismatch repair proficient; MSS, microsatellite stable; ND, not done; D, deficient; MSI-H, microsatellite instable - high; CRC, colorectal cancer; UC, ulcerative colitis.

^*All tumors were adenocarcinomas except tumor 3b which was adenocarcinoma with a minor component of squamous cell carcinoma.

^**All MSI testing was done by MSK-IMPACT except for cases #2 which was by PCR with 5/5 marker instable.

^***This is likely a passenger mutation, this tumor has normal MSH2 staing on immunohistochemistry.

^****Included MSH2 sequencing, MSH2 deletion and duplication, MSH6 sequencing, and MSH6 rearrangement analysis.

Figure 2.

Distribution of inter-tumoral mismatch repair protein immunohistochemistry discordances among Lynch syndrome patients, patients with non-Lynch syndrome colorectal cancer predisposing syndromes, and patients with sporadic or likely sporadic gastrointestinal tract cancers.

Group 1_LS patients

As outlined in Table 1, all 3 patients in this group had a positive family history of malignancy. Of the 3 patients, two were PMS2-LS. These 2 patients each had 2 metachronous colonic adenocarcinomas, and all tumors were late onset, occurring at age ≥ 66 years.

In patient 1, the MMR discordance was in the form of MMR-proficient (tumor 1, asigmoid colon cancer, Fig. 3A–C) versus MMR-deficient colon tumors (tumor 2, a cecal cancer, with loss of PMS2 as well as clonal loss of MSH 6, Fig. 3D–F). Both tumors had histological features concordant with the MMR IHC status (with tumor 1 having no increased TILs and tumor 2 having high TILs). MSK-IMPACT assay on the 2 tumors both showed findings concordant with the IHC results: tumor 1 had a low MSIsensor score at 0.8 (indicating microsatellite stable or MSS) and a low tumor mutation burden (TMB) at 4.4 mutations/megabase (mt/Mb), and tumor 2 had a high MSIsensor score at 34.6 (i.e. MSI-H) and a high TMB at 56.2 mt/Mb. Notably, in tumor 2, a somatic MSH6 frameshift mutation involving the C8 tract in exon 5 (coding microsatellite) (p.F1088Lfs*5, allele frequency at 0.25 and tumor content at ~50%) was detected. This was believed to be a manifestation of the MSH6 coding microsatellite being a target of MSI (2,15) and helped explain the clonal loss of MSH6 on IHC,However, the MSH6 variant was consistent with single allele loss only; we codld not identify the mechanism for the inactivation of the 2^nd allele: FACETS did not reveal evidence of loss of heterozygosity and no MSH6 methylation was detected No shared somatic mutations were detected between tumors 1 and 2.

Figure 3.

Inter-tumoral mismatch repair (MMR) protein immunohistochemistry (IHC) discordance in Lynch syndrome (LS). A-F illustrate a PMS2-LS patient with one colonic carcinoma showing normal MMR IHC (A, H&E morphology; B, normal MSH6 IHC; C, normal PMS2 IHC) and another colonic carcinoma showing abnormal MMR IHC (D, H&E morphology; E, clonal loss of MSH6; F, complete loss of PMS2). G-L illustrate another PMS2-LS patient with one colonic carcinoma showing loss of PMS2 only (G, H&E morphology; H, normal MSH6 IHC; I, complete loss of PMS2) and another colonic carcinoma showing loss of both MSH6 and PMS2 (J, H&E morphology; K, complete loss of MSH6; L, complete loss of PMS2). M-R illustrate two small sigmoid colon adenomas from a “compound LS” patient with germline mutation in both MLH1 and MSH2; one adenoma shows abnormality in MLH1/PMS2 (M, H&E morphology; N, normal MSH6 IHC; O, loss of PMS2) and one shows abnormality in MSH2/MSH6 (P, H&E morphology; Q, loss of MSH6; R, normal PMS2 IHC in adenoma). Note that in R, the MSH2/MSH6-deficient adenoma also exhibits a cluster of 3 PMS2-deficient crypts (outlined by small white box and magnified in inset).

In patient 2, the 2 colon tumors (1 an ascending colon cancer and 1 a sigmoid colon cancer) both had histological features suggestive of MMR deficiency with high TILs, and both manifested PMS2 deficiency on IHC (Fig. 3G–L) and exhibited MSI-H by the PCR based method. However, the sigmoid colon cancer had, in addition, complete loss of MSH6 (Fig. 3K). Tumor sequencing could not be performed in this case. Nonetheless, given that the only pathogenic mutation detected on germline testing involved PMS2 (Table 1), i.e., no germline mutation in MSH6 (or other MMR gene), we speculated that the loss of MSH6 was likely related to secondary somatic mutation of the MSH6 coding microsatellites, similar to that seen in patient 1 and akin to previous reports(2, 15). Different from patient 1, the loss of MSH6 in patient 2 was complete (in the tumor section examined), suggesting that the MSH6 deficient clone had essentially overtaken the entire tumor with no MSH6 wild type clone remaining.

Patient 3 was an unusual dual MLH1/MSH2 LS case ("compound LS"). The two metachronous GIT cancers occurred 20 years apart (Table 1). Both tumors were MMR deficient but involved different MMR proteins: the first tumor, a colon carcinoma, had loss of MSH2/MSH6 while the second tumor, a gastric carcinoma, had loss of MLH1/PMS2. Both tumors exhibited high TILs. Remarkably, during the 20 years between the 2 GIT carcinomas, the patient also developed numerous sebaceous neoplasms and cutaneous squamous cell tumors, involving both the head and neck region and the limbs and trunk, as well as multiple (about 20) colon adenomas (ranging from "diminutive" to 1cm in size). Systematic MMR IHC evaluation of all tumors with available tissue (n=52) revealed striking heterogeneity in MMR protein status as outlined in Table 2, with some tumors losing MLH1/PMS2 and others losing MSH2/MSH6. This heterogeneity was also present among lesions at the same location or of the same histology type, and seen already in morphologically normal-appearing MMR-deficient crypts(16) and early colon adenomas. Fig. 3M–R illustrate 2 sigmoid colon adenomas, one (4mm in size) losing MLH1/PMS2 (Fig. 3M–O) and one (10mm in size) losing MSH2/MSH6 (Fig. 3P–R); the MSH2/MSH6-deficient colon adenoma also exhibited a minute focus of 3 MLH1/PMS2-deficient crypts (Inset of Fig. 3R), highlighting the remarkable mosaic pattern in the manifestation of deficiency in MLH1 versus MSH2 in this compound LS patient, beginning at stages preceding morphological alteration.

Table 2.

Mismatch repair protein immunohistochemistry staining results of tumors from a Lynch syndrome patient with 2 pathogenic germline mutations, 1 in MLH1 and 1 in MSH2 (case 3).

Tumor type (number of tumors tested)	MLH1/PMS2 loss	MSH2/MSH6 loss	MMR retained
Sebaceous adenoma (n=25)	6	19	0
Atypical/borderline sebaceous neoplasms (n=6)	2	4	0
Sebaceous carcinoma (n=4)	0	4	0
Cutaneous squamous cell carcinoma (n=5)	1	3	1
Colorectal adenoma (n=10)	1	2	7
Colorectal carcinoma (n=1)	0	1	0
Gastric carcinoma (n=1)	1	0	0
Total (n=52)	11	33	8

Group 2_ non-LS hereditary colorectal cancer predisposition syndromes

Two patients belonged to this group, and they represented 2 syndromic conditions: polymerase proofreading-associated polyposis (PPAP) and familial adenomatous polyposis (FAP) (Table 1, Figure 1). As listed in Table 1, both patients had a positive family history of malignancy, and the diagnosis of the hereditary syndrome in both was established by positive germline mutation testing.

The PPAP patient developed 2 metachronous colorectal carcinomas before age 50. One is MMR-proficient, but with morphological features resembling MMR-deficient carcinomas including increased TILs (Fig. 4A–C). The second tumor also had morphological features suggestive of MMR deficient and indeed showed loss of MSH2/MSH6 (Table 1, Fig. 4D–F).

Figure 4.

Polymerase proofreading-associated polyposis (PPAP) and familial adenomatous polyposis (FAP) manifesting mismatch repair (MMR)-deficient gastrointestinal tract carcinoma in addition to syndrome-associated MMR-proficient carcinoma. A-F illustrate a PPAP patient having two colorectal carcinomas. One tumor shows morphological features similar to MMR-deficient colonic carcinomas (A-C, H&E morphology: A, conspicuous tumor-infiltrating lymphocytes; B, histologic heterogeneity with gland-forming component juxtaposed with mucinous component that contains signet ring cells; C, non-signet ring mucinous component). But this tumor has normal MMR IHC (not illustrated). The other tumor also shows increased tumor infiltrating lymphocytes, and in this one, it is accompanied by abnormal MMR IHC (D, H&E morphology; E, complete loss of MSH6; F, normal PMS2 IHC). G-L illustrate an MMR-deficient small bowel adenocarcinoma in an FAP patient (G, focal serrated changes seen on mucosal surface in the tumor; H-J, different regions showing different degrees of differentiation with H being moderately differentiated and I and J being poorly differentiated; K, normal MSH6 IHC; L, complete loss of PMS2).

Tumor somatic mutation testing could not be performed in this patient. Nonetheless, based on the current knowledge about PPAP (17–19) and the lack of germline mutation in the MMR genes, it was speculated that the MMR deficiency in the 2^nd tumor was most likely secondary to somatic inactivation of MSH2 probably causally related to the POLE mutation.

In the FAP patient, the first cancer, a rectal adenocarcinoma, was detected at age 28. This was accompanied by the detection of hundreds of colon polyps, and prompted a total proctocolectomy. The carcinoma was MMR proficient, with concordant morphological patterns. The patienťs second cancer, a metachronous jejunal adenocarcinoma, occurred 15 years after the total proctocolectomy. This cancer showed mixed histological patterns (Fig. 4G–J). Surprisingly, although the patient had been found to have multiple ileal and duodenal adenomas of conventional intestinal type, no pre-existing adenoma of this type was detected in association with the jejunal carcinoma. Instead, focal surface changes reminiscent of colonic serrated lesions were noted towards the periphery of the tumor (Fig. 4G), and the tumor also showed increased TILs (albeit only focally and mildly). By IHC, this jejunal carcinoma showed loss of MLH1/PMS2 (Fig. 4K–L). It was further proven that the MLH1/PMS2 deficiency was associated with MLH1 promoter methylation, with a positive tumor MLH1 promoter methylation testing. By MSK-IMPACT, this tumor had a MSIsensor score of 35.5 (i.e., MSI-H) and a TMB of 36 mt/Mb.

Group 3_Likely sporadic GIT carcinomas

Of the 5 patients in this group, 3 underwent germline mutation testing by MSK-IMPACT with negative results; one of the 3 had long-standing inflammatory bowel disease. A 4^th patient underwent targeted germline mutation guided by the MMR IHC results (Table 1) and result was also negative. The 1 patient (patient #7) who did not have germline mutation testing had cancers that was considered likely sporadic based on clinical features (including old age of cancer onset, negative family history of malignancy, and MMR-deficient tumor proven to harbor MLH1-promoter hypermethylation).

All 5 patients in this group shared a common pattern of inter-tumoral MMR discordance: sporadic MMR-proficient colorectal cancer (or cancers) co-occurring synchronously (3 cases) or metachronously (2 cases) with a likely sporadic MMR-deficient cancer (Table 1, Fig. 5). In 3 patients, the MMR-deficient cancer had loss of MLH1/PMS2: one of the 3 was tested for MLH1 methylation and BRAF V600E mutation (by sequencing), and was found to be positive on both tests; one showed positive BRAF V600E IHC (clone VE1); and one did not undergo testing for methylation or BRAF mutation. In 2 patients, the MMR-deficient cancer had loss of MSH2/MSH6, one of the two patients was tested by MSK-IMPACT, the MMR deficient tumor in this patient had a likely pathogenic somatic MSH2 mutation.

Figure 5.

An apparently sporadic MLH1-methylated MLH1/PMS2-deficient colonic adenocarcinoma coexisting with two mismatch repair (MMR)-proficient primary colonic carcinomas in a 77-year-old patient with negative germline mutation testing. A and B, H&E morphology of the 2 MMR-proficient colonic adenocarcinomas with conventional type histology and no increased tumor-infiltrating lymphocytes. C-E illustrate the patient’s MMR-deficient colonic carcinoma (C, H&E morphology with mucinous features and increased tumor-infiltrating lymphocytes; D, normal MSH6 IHC; E, complete loss of PMS2).

DISCUSSION

In this study, we found that 25.9% of the patients with 2 or more primary CRCs had discordant MMR status between their tumors, suggesting that inter-tumoral MMR discordance is not an uncommon phenomenon in patients with multiple CRCs. We note, however, that our observation is based on a CRC population that is limited to those who had MMR IHC done following varied guidelines over time. Our baseline finding of 1.3% of CRC patients having 2 or more primary CRCs may therefore not reflect the general population. In the literature, the incidence of synchronous CRC is estimated to be 1–4% among patients with CRC in general (20–22) and the absolute risk of metachronous colorectal cancer is found to be age-dependent: ≥10% in patients below age 65 and 1.0%–8.0% in patients above age 65 in one study (21).

At the current time, while studies exist that recognize the existence of inter-tumoral MMR discordance in patients with multiple GIT cancers particularly CRCs (23, 24), systematic profiling of the molecular mechanisms underlying the discordances is scarce. Thus, a major goal of our study is to systematically analyze the patterns and molecular etiologies in these cases. Our analysis found a spectrum of intricate clinical scenarios where complex molecular abnormalities led to varied MMR protein expression in the different tumors within the same patients. Awareness of these scenarios carry important diagnostic and management implications as discussed below.

PMS2-LS cases may manifest their first syndrome-associated carcinoma at an old age, a setting permissive for co-incidental sporadic cancer to occur

Detection of mutational variants in PMS2 has been challenging due to the presence of PMS2 pseudogenes and the occurrence of gene conversion. In recent years, however, technological advancements have allowed better detection and consequently a much-improved understanding of PMS2-LS. A number of characteristics of this condition are becoming clearer now including: 1) a low cancer penetrance but still an increased cumulative lifetime cancer risk(with a cumulative CRC risk to age 70 estimated at 12–17%)(25–31); 2) a cancer risk highest at an early age but remains increased in old individuals(30); and,3) a population-based prevalence highest among all types of LS(32). The finding of PMS2-LS being the most prevalent in the general population is of particular significance. Earlier estimates that indicated MLH1-LS and MSH2-LS as the most prevalent were mostly based on CRC patients, and did not reflect the prevalence in the general population.

Our case #1 is an example of late-onset PMS2-associated CRC in PMS2-LS (age at diagnosis being 83). The case also illustrates a scenario in which LS individuals can also develop seemingly sporadic CRC unrelated to the LS. Our patienťs 1^st CRC at age 81 had a typical histological and molecular profile of APC-driven sporadic CRC, and shared no common mutations with his PMS2-deficient cancer. In clinical practice, when it comes to LS detection, there should be awareness of such occurrences and vigilance that tumor MMR status is not perfectly sensitive(33); in patients with pertinent positive family cancer history, it may be prudent to pursue germline mutation testing despite normal MMR IHC results or old age.

MSI-driven somatic MSH6 inactivation due to frameshift mutation of its coding microsatellites can complicate MMR IHC interpretation

Previous work has indicated that MSH6 is a target of MSI; in tumors that are deficient in MLH1 and/or PMS2, the MSH6 gene is susceptible to secondary inactivation due to frameshift mutations in its coding microsatellites(15, 34). This phenomenon is exemplified in our cases 1 and 2, both PMS2-LS. In case 1 (tumor 1b), tumor mutational analysis indeed revealed a somatic frameshift variant in MSH6 affecting the coding C8 tract in exon 5 (MSH6_p.F1088Lfs*5). However, this variant had an allele frequency that was consistent with a mono-allelic event only. We were not able to demonstrate a mechanism for the inactivation of the second allele (no evidence of LOH or MSH6 methylation was detected). Possibilities exist that the second allele was inactivated by a somatic or germline mutation that was not covered by our test panel (e.g. a deep intronic splice site mutation or an intronic translocation). In case 2 (tumor 2b), although we could not perform tumor somatic mutation testing in this case, given the presence of PMS2 germline mutation, PMS2 deficiency in the tumor, and lack of MSH6 germline mutation, it seems plausible to speculate that the MSH6 loss in this tumor is also a secondary event akin to what has been reported in the literature(15, 34).

Our cases thus serve to highlight the scenario where MLH1- and/or PMS2- deficient tumors show secondary loss of MSH6. When this happens, especially when the loss is complete (as is the case in tumor 2b),the resultant IHC pattern can create confusion. Awareness of this phenomenon can facilitate the interpretation of IHC results, and help achieve the most effective work-up for the detection of LS patients.

"Compound LS" with two germline MMR gene mutations is rare but an important source of inter-tumoral MMR discordance.

Double germline MMR gene mutation involving two different genes, i.e., "compound LS" (also referred to as "digenic LS"), is rare, but has been documented in the literature. In 2005, Lee et al.(35) described a case with two germline mutations – MLH1_c.350C>T, p.T117M and MSH2_c.965G>A, p.G322B, both deemed to be deleterious. This patient developed colorectal carcinoma. In 2011, Wei et al.(36) described a case with germline MLH1_c.2157dupT and MSH2_c.1168C>T. Most recently, Yilmaz et al. (37) reported another case where two germline mutations coexisted involving MSH6 (c.3261delC, p.F1088fs*2) and PMS2 (c.187G>A, p.V63M), and the patient developed metastatic colorectal carcinoma.

Our case 3 (with dual MLH1 and MSH2 germline mutations) represented yet another example of compound LS. We speculated that the two germline mutations in our patient were likely inherited from the father and mother respectively, as both parents had a history positive for LS-associated cancers. Unfortunately, however, germline testing in the parents or other relatives could not be performed. In our case, a particularly interesting finding is the development of a variety of LS-type tumors including sebaceous neoplasms, squamous cell neoplasms, and colorectal adenomas and gastric carcinoma, in numbers surpassing that commonly seen in typical LS patients (Table 2). This is likely a reflection of the additive or combined risk from the two different MMR mutations. Indeed, our observation that the various tumors exhibited a mosaic pattern of MMR deficiency (some losing MLH1/PMS2 and some losing MSH2/MSH6) even when the tumors are of the same histology type - suggests that the two mutations are both exerting tumorigenic effects at susceptible tissues. The presence of MLH1/PMS2-deficient crypts in an early MSH2/MSH6-deficient colon adenoma (Fig. 2R) is further evidence of their "competing" deleterious effects, in keeping with what has been reported previously(16, 38, 39). It is also interesting to note that among the 10 colonic adenomas tested (Table 2), the 3 with MMR protein loss were also the biggest (4–10mm versus "diminutive"). This is in keeping with the prior observation that the prevalence of MMR deficiency increases with the size of the adenoma and the suggestion that MMR deficiency is a late event in LS-associated colorectal neoplasia(40). In addition to such biological implications, our findings should also serve to inform our clinical LS work-up strategies; in patients with marked inter-tumoral discordance in MMR results (particularly MLH1 loss versus MSH2 loss), compound LS should be included in the considerations for potential etiologies.

Somatically MMR inactivated tumors occurring in PPAP or FAP is another biologically significant etiology for inter-tumoral MMR discordance.

The importance of polymerase proofreading of DNA replication in suppressing tumorigenesis has been demonstrated by the recent discovery of the "ultra-mutated" CRCs occurring in association with somatic(41) or germline (17,42) POLE mutations. Although most POLE mutation-associated CRCs are MMR-proficient, MMR-deficient tumors (alone or co-existing with other MMR-proficient tumors) have been reported to occur, presumably mechanistically related to the deficiency in polymerase proofreading (18,19). Inactivation of the MMR has been reported to affect MSH2/MSH6 (due to somatic mutation/inactivation) or MLH1 (due to MLH1 methylation (43)). Our case 4 is one such example, demonstrating MMR inactivation involving MSH2 and an MMR-deficient tumor (a rectal carcinoma) accompanied by a metachronous MMR-proficient colon cancer.

While our case 4 represents a rare but reported phenomenon, our case 5 appears novel. It reflects the occurrence of an MLH1-methylated MSI small bowel carcinoma 15 years after total proctocolectomy (for adenomatous polyposis plus one MMR-proficient rectal carcinoma) in an FAP patient. The small bowel (jejunum) carcinoma had a typical histological and molecular profile of an MMR-deficient intestinal adenocarcinoma. A particularly striking feature is the presence of focal surface changes resembling colonic serrated lesions in the tumor. It is tempting to speculate that the removal of the large bowel (for a duration of 15 years) may have potentially resulted in an environmental change in the small bowel (including microbiome change) to simulate the environment of the large bowel (particularly right colon), and it is this change that perhaps served to facilitate the development of the MLH1-methylated carcinoma via a pathway that approximates the colonic serrated neoplasia pathway(44).

Molecular mechanisms notwithstanding, both cases highlight important clinical scenarios where inter-tumoral MMR discordance can occur. This is particularly pertinent in young patients, and can serve to better inform effective strategies in working up hereditary cancer predisposition syndromes in this patient population.

Sporadic MMR-deficient tumors coexisting with sporadic MMR-proficient tumors appears to be the most common scenario for inter-tumoral MMR discordance, especially in elderly patients.

It has been well established that both MLH1 promoter methylation and biallelic somatic mutation/inactivation can result in sporadic MMR-deficient GIT carcinomas(1). As is the case with other sporadic carcinomas, these tumors tend to occur in older individuals. It is therefore not surprising to see elderly patients with synchronous or metachronous GIT cancers that are either MMR-proficient or deficient. In our series, this scenario seems to be the most common among all cases with inter-tumoral MMR discordance. As with all the scenarios described above, awareness of such occurrences is important in achieving the most effective interpretation of the MMR results, and consequently the most effective patient management. Notably, when deciding 5-FU based chemotherapy or immunotherapy - two scenarios where MMR status is particularly informative in predicting response, it is important that the MMR testing is performed on the tumor that is to be treated.

Caveats and conclusion

Our study has limitations. First, as has been pointed out, our frequency estimates may not reflect the actual frequency in the general population. Additionally, not all discordant MMR cases had a complete IHC, methylation, and tumor/germline mutational work-up; some conclusions were inferred from indirect evidence and literature data. Nonetheless, our cases highlight the complexity of inter-tumoral discordant MMR status and illustrate various intricate clinical scenarios where varied molecular mechanisms led to the MMR discordance. Given that such discordances can potentially affect patient management strategies, and as more of them may emerge due to the increasingly wider-spread use of MMR IHC(7) as well as molecular/genetic testing(45), a heightened awareness of their occurrence and potential etiologies will become more and more relevant.