Efficacy of immunotherapy in gastrointestinal (GI) tumors with mismatch repair deficient (MMRd) unusual phenotype: an AGEO real-world study

Abstract

Background

Tumors with mismatch repair deficiency (MMRd) classically display concomitant loss of MLH1/PMS2 or MSH2/MSH6 on immunohistochemistry (IHC) with microsatellite instability-high (MSI-High) status on molecular testing. Nevertheless, a different phenotype can occur in up to 15% of MMRd tumors (unusual phenotype). Data on the efficacy of immunotherapy in this population remain scarce.

Methods

We conducted a retrospective study within the IMMUNODIG MSI cohort including patients with advanced gastrointestinal MMRd tumors treated with immune checkpoint blockade in the real-world setting. We selected patients with both IHC and MSI assays data available. Unusual MMRd tumors were classified into four distinct groups: (1) isolated loss of PMS2 or MSH6 with MSI-H (isolated/MSI-H), (2) complex loss of MMR proteins with MSI-H (complex/MSI-H), (3) loss of one or more MMR proteins without MSI-H (MMRd-IHC/microsatellite stability (MSS) or MSI-low (MSI-L)), and (4) four MMR proteins retained with MSI-H (retained IHC/MSI-H).

Results

Out of 759 patients in the IMMUNODIG-MSI cohort, 571 patients met inclusion criteria. Of these, 90 (15.8%) had an unusual phenotype (47 isolated/MSI-H, 19 complex/MSI-H, 16 MMRd-IHC/MSS or MSI-L and 8 retained IHC/MSI-H). Compared with classical phenotypes, patients with a tumor harboring an unusual phenotype had a younger age at treatment (p=0.005), increased RAS mutation (p=0.005), reduced BRAF p.V600E mutation rates (p<0.001), a higher proportion of Lynch syndrome (p<0.001) and a higher prevalence of non-colorectal cancers (p=0.021). After a median follow-up of 28.1 months (mo), there was a significant difference in progression-free survival, with median values of not reached, 66.4 mo, 37.2 mo, 18.3 mo and 5.5 mo for complex/MSI-H, isolated/MSI-H, classical, retained IHC/MSI-H and MMRd-IHC/MSS or MSI-L subgroups, respectively (p<0.001). Notably, objective response rates were 59.1%, 58.7%, and 63.2% for complex/MSI-H, isolated/MSI-H and classical contrasting with 50% and 25% for retained IHC/MSI-H and MMRd-IHC/MSS or MSI-L, with no complete response observed in the latter two.

Conclusion

Our findings underscore the need for dual testing and advocate for the presence of both MMRd-IHC and MSI-H for optimal immunotherapy response. Of note, complex MMR aberrations and isolated PMS2/MSH6 losses with MSI-H may represent promising candidates for enhanced immunotherapy efficacy.

WHAT IS ALREADY KNOWN ON THIS TOPIC

• Mismatch repair deficiency (MMRd) tumors typically exhibit diffuse loss of MLH1/PMS2 or MSH2/MSH6 on immunohistochemistry (IHC), along with microsatellite instability-high (MSI-H) on molecular testing. However, the increasing use of comprehensive MMR testing has revealed “unusual phenotypes” that deviate from these classical patterns, and there is limited data on the efficacy of immune checkpoint blockade (ICB) in these cases.

WHAT THIS STUDY ADDS

• We found a significant difference in progression-free survival, with median values of not reached, 66.4 mo, 37.2 mo, 18.3 mo and 5.5 mo for complex/MSI-H, isolated/MSI-H, classical, retained IHC/MSI-H and MMRd-IHC/microsatellite stability (MSS) or MSI-low (MSI-L) subgroups, respectively (p<0.001). Notably, objective response rates were 59.1%, 58.7%, and 63.2% for complex/MSI-H, isolated/MSI-H and classical contrasting with 50% and 25% for retained IHC/MSI-H and MMRd-IHC/MSS or MSI-L, with no complete response observed in the latter two (considered as discordant cases).

HOW THIS STUDY MIGHT AFFECT RESEARCH, PRACTICE OR POLICY

• Our data suggest that discordant cases display poor response to ICB and should systematically be reviewed by an expert pathologist. In the case of true discordant cases, tumor mutational burden and POLE mutation should be systematically tested for ICB indication. Conversely, complex MMR protein alterations and isolated PMS2 or MSH6 losses that are MSI-H are potentially promising candidates for enhanced immunotherapy efficacy, and should not be excluded from immunotherapy treatment.

Introduction

Historically, mismatch repair (MMR) testing served only as a genetic marker for the hereditary cancer predisposition Lynch syndrome.^{1 2} In the past decade, immunotherapy has demonstrated frequent and durable responses in tumors harboring defects in the MMR system. The following registration trials of pembrolizumab and nivolumab in MMR-deficient-immunohistochemistry (MMRd-IHC) and/or microsatellite instability (MSI) colorectal cancer (CRC) expanded the scope of MMR testing, making it an agnostic biomarker for immunotherapy indication.3,6 Today, MMR testing is predominantly used as a predictive marker for immune checkpoint blockade (ICB) efficacy.

Currently, the two main diagnostic methods for detecting mismatch repair deficiency are IHC to assess the four MMR protein expression (MLH1, PMS2, MSH2 and MSH6) and PCR-based molecular testing to determine the functional consequence of MMR defect which is MSI. The term “MMRd” generally refers to any tumor exhibiting mismatch repair deficiency, regardless of the detection method used. In contrast, “MMRd-IHC” specifically denotes tumors identified as MMR-deficient through IHC, while “MSI” refers to those detected using molecular biology techniques. MMRd tumors classically display a loss of MLH1/PMS2 or MSH2/MSH6 on IHC, along with MSI-high (MSI-H) on molecular testing. In these classical MMRd cases, the loss of staining is uniform across the tumor. However, the growing use of broad MMR testing has led to the identification of “unusual phenotypes” in MMRd tumors, which deviate from these classic presentations.7,13 These include single MMR protein loss, loss of three or four proteins, atypical dual protein loss (different from MLH1/PMS2 and MSH2/MSH6 loss), subclonal loss, and discordant results between IHC and molecular tests. In a large cohort of 4948 tumors tested for MMRd, Jaffrelot et al reported that up to 15% of MMRd tumors exhibited unusual phenotypes.¹⁴ Most guidelines consider MMR protein IHC as the gold standard test to identify cancers with MMRd and consider the predictive values of IHC and molecular techniques as equivalent.^{15 16} In cases of verified discordant results, international guidelines recommend interpreting any evidence of MMRd (by IHC, PCR, or next-generation sequencing) as sufficient for patients to qualify for ICB therapy.¹⁶ Nevertheless, there is limited data comparing the predictive value of these different techniques for ICB efficacy. It is important to note that IHC does not assess the functional aspect of the mismatch repair system, which is a crucial factor for ICB since MMR defects are thought to increase the tumor mutation load and the number of neoantigens underlying the efficacy of ICB. In addition, it is now well established that single loss of PMS2 or MSH6 is frequently associated with Lynch syndrome (but not exclusively), but data on the efficacy of ICB in this population remain scarce.^{10 17 18}

Globally, the clinical impact of these unusual phenotypes on immunotherapy outcome has not been documented. Besides analyzing the frequency and clinical characteristics of these phenotypes, our aim was to determine the efficacy of immunotherapy across different subgroups of unusual phenotypes.

Material and methods

Patients

We performed a retrospective multicenter study within the national IMMUNODIG MSI cohort including patients with advanced gastrointestinal tumors treated with exclusive ICB (anti-programmed cell death protein-1 (PD-1) or anti-programmed death-ligand 1 (PD-L1)±anti-cytotoxic T-lymphocyte associated protein 4 (CTLA-4) agents) from January 2015 to December 2024. All patients who were diagnosed with MMRd status using both techniques (IHC and molecular biology) were included in the study. The study was conducted in accordance with the Declaration of Helsinki.

MMRd phenotype

MMRd-IHC status was determined by IHC for the mismatch repair proteins MLH1, MSH2, MSH6 and PMS2 and was defined by the loss of at least one protein. MSI phenotype was determined by PCR-based assessment of microsatellite alterations using either an in-house panel of five microsatellite markers (NR21, NR22, NR24, BAT25 and BAT26) or the commercially available kit Promega. Tumors were classified as MSI-H when at least two microsatellite markers showed instability. MSI-low (MSI-L) tumors were included with microsatellite stable (MSS) tumors, as per international guidelines.¹⁵ If normal tissue was available, comparison with normal DNA was performed to eliminate polymorphism. Subclonal loss was defined as only part of the tumor showing complete loss of MMR staining.

Unusual MMRd tumors were classified into four distinct subgroups: (1) isolated loss of PMS2 or MSH6 with MSI-H (isolated/MSI-H), (2) complex loss defined as any loss of MMR protein different from the isolated/MSI-H group and classical diffuse loss of heterodimers MLH1/PMS2 and MSH2/MSH6 (complex/MSI-H) (3) loss of one or more MMR proteins without MSI-H (MMRd-IHC/MSS or MSI-L), (4) four MMR proteins retained with MSI-H (retained IHC/MSI-H). The two latter groups are also commonly defined as discordant cases.⁷ Of note, a slight redefinition of subgroups was made compared with our previous study,¹⁴ aiming to improve their clinical relevance by isolating completely discordant cases. All pathology reports of unusual cases included in the study were centrally reviewed. Additionally, at least 50% of tissue blocks from each unusual subgroup underwent central review. If more than 15% of reviewed cases were found to be misclassified, all patients in that subgroup who had not undergone central review were excluded. If the rate of misclassification after central review was below 15%, and assuming this error rate was representative of the entire subgroup, we estimated that the final overall misclassification rate would not exceed 8.1%. Therefore, in this case, all patients without central review were retained in the analysis.

For cases with available data on tumor mutational burden (TMB) or POLE mutations, assessment was performed using next-generation sequencing (NGS) with the TruSight Oncology 500 (TSO500) panel.

Lynch syndrome

Patients reported as having Lynch syndrome had a confirmed germline mutation.

If a patient had a germline testing negative and/or had no history of other cancers and presented with a BRAF p.V600E mutation or MLH1 promoter hypermethylation, the case was classified as sporadic. For the remaining 11 patients, germline testing was performed, but results were unavailable; these cases were therefore categorized as missing.

Endpoints

Progression-free survival (PFS) was defined as the time from the date of initiation of immunotherapy to disease progression or death from any cause. Overall survival (OS) was defined as the time from the date of immunotherapy initiation to the date of death from any cause. Patients alive at the time of last follow-up were censored at this date. Progression was assessed according to Response Evaluation Criteria in Solid Tumors criteria (RECIST), V.1.1. Objective response rate (ORR) was defined as the percentage of patients with a complete or partial response according to RECIST V.1.1 and disease control rate as the proportion of patients with complete/partial response and stable disease as best response. The cut-off date for the present analysis was January 1, 2025.

Statistical analysis

RStudio (V.12.0, Chicago, Illinois, USA) was used for statistical analysis. Quantitative variables were described using median value (range). Qualitative variables were summarized for the entire population and by cohorts, using: numbers, percentages, number of missing data. PFS and OS were estimated using the Kaplan-Meier method with their respective 95% CIs. Response rates were calculated using percentage.

Parameters with a p value<0.05 were considered statistically significant.

Results

Patient population

Of 759 patients in IMMUNODIG MSI, 591 underwent both IHC and MSI assays, 111 only had IHC testing, 9 only had PCR testing and the information indicating whether a second technique had been performed was missing for 48 patients (figure 1). Among patients with unusual tumors based on local assessment, 76/120 tumor blocks were centrally reviewed. All tumors centrally reviewed were confirmed to be either MMRd-IHC or MSI-H. However, 15 blocks/76 (19.7%) were misclassified, and an additional 4 patients (5.3%) showed IHC misinterpretation that did not alter their assigned initial group (online supplemental table 1). The most frequent errors were found in the complex group (48.1%) where faint nuclear staining of MSH6 or subclonal loss of MSH6 was wrongfully interpreted as retained expression. Finally, 571 patients were included, of whom 90 (15.8%) had an unusual phenotype. 47 had an isolated MSH6 or PMS2 loss (24 and 23, respectively). 19 had a complex loss with either MLH1 or MSH2 isolated loss, loss of three or four proteins, atypical couple loss or subclonal loss. 16 patients displayed MMRd-IHC and MSS/MSI-L tumors (12 MSS and 4 MSI-L). Among this group, 3 tumors had isolated PMS2 loss, 2 isolated MSH6 loss, 1 had a loss of 3 proteins and 10 had a classical heterodimer loss. Among MSI-L tumors, two had rectal cancer with isolated MSH6 loss, and two were pancreatic cancers exhibiting classical heterodimer loss. Finally, eight patients had retained IHC/MSI-H. Among unusual phenotype tumors, 26.7% had discordant MMRd tumor.

Figure 1. Flow chart: distribution of MMRd tumors among the five subgroups: classical, isolated/MSI-H, retained IHC/MSI-H, complex/MSI-H, MMRd-IHC/MSS or MSI-L. †Data on whether a second technique was performed is missing. *See online supplemental table 1 for details. **Patients with subclonal loss (n=7): all had classical heterodimer loss and MSI-H (five with MSH6 subclonal loss and two with MLH1 subclonal loss). Patients in green have discordant MMRd tumors. IHC, immunohistochemistry; MMRd, mismatch repair deficient; MSI-H, microsatellite instability-high; MSI-L, microsatellite instability-low; MSS, microsatellite stable.

Baseline characteristics

Overall, 396 (69.4%) patients had a colon cancer, 54 (9.5%) a gastric cancer, 40 (7.0%) a small bowel cancer, 34 (6%) a rectal cancer, 20 (3.5%) a pancreatic cancer, 16 (2.8%) a cholangiocarcinoma and 11 (1.9%) another cancer. Between the different MMRd phenotype groups, there was a significant difference in terms of age, primary tumor location, RAS and BRAF p.V600E mutation rates and proportion of Lynch syndrome (table 1). In particular, germline mutation was identified in 40.9% and 50% in the isolated/MSI-H and complex/MSI-H groups compared with 15% in the classical phenotype. There was no difference regarding potential prognosis factors at baseline for immunotherapy efficacy, such as Eastern Cooperative Oncology Group – Performance Status (ECOG-PS), stage at diagnosis, primary tumor resection rate, type and number of metastases, line of treatment. Of note, the rate of patients receiving an anti-PD-(L)1 or anti-PD-(L)1+anti-CTLA-4 combination treatment was comparable across all groups.

Table 1. Patient and tumor baseline characteristics per MMRd phenotype.

	Classical	Isolated MSI-H	Complex MSI-H	MMRd-IHC MSS or MSI-L	MMRp-IHC MSI-H	P value
	N=481 (%)	N=47 (%)	N=19 (%)	N=16 (%)	N=8 (%)
Median age (range)	68.3 (31–88)	62.8 (27–89)	61.6 (19–98)	59.6 (29–90)	49.9 (27–72)	0.005
Sex ratio (male/female)	1 (243/238)	1.5 (28/19)	1.1 (10/9)	1.7 (10/6)	3 (6/2)	0.419
ECOG-PS						0.967
ECOG 0	139 (29.1)	11 (23.9)	6 (31.6)	6 (40.0)	4 (57.1)
ECOG 1	238 (49.9)	24 (52.2)	11 (57.9)	8 (53.3)	2 (28.6)
ECOG≥2	100 (21.0)	11 (23.9)	2 (10.5)	1 (6.7)	1 (14.3)
Missing	3	0	0	0	0
Stage at diagnosis						0.789
I	3 (0.6)	0 (0.0)	0 (0.0)	0 (0.0)	0 (0.0)
II	49 (10.2)	8 (17.0)	3 (15.8)	1 (6.25)	0 (0.0)
III	187 (38.9)	15 (31.9)	7 (36.8)	4 (25.0)	4 (50.0)
IV	238 (49.5)	23 (48.9)	9 (47.4)	10 (62.5)	4 (50.0)
Missing	4	1	0	1	0
Primary tumor location						0.021
Colon	339 (70.5)	31 (66.0)	11 (57.9)	8 (50.0)	7 (87.5)
Rectum	25 (5.2)	4 (8.5)	1 (5.3)	3 (18.8)	1 (12.5)
Gastric	50 (10.4)	3 (6.4)	1 (5.3)	0 (0.0)	0 (0.0)
Small bowel	33 (6.9)	4 (8.5)	1 (5.3)	2 (12.5)	0 (0.0)
Pancreas	14 (2.9)	2 (4.3)	2 (10.5)	2 (12.5)	0 (0.0)
Cholangiocarcinoma	12 (2.5)	2 (4.3)	1 (5.2)	1 (6.3)	0 (0.0)
Other*	8 (1.7)	1 (2.1)	2 (10.5)	0 (0.0)	0 (0.0)
Primary tumor resection	302 (64.8)	22 (55.0)	11 (57.9)	6 (40.0)	3 (50.0)	0.214
Metastatic organs
≥2 organs involved	228 (47.4)	24 (51.1)	5 (26.3)	7 (43.8)	5 (62.5)	0.35
Peritoneum	227 (47.5)	25 (54.4)	7 (36.8)	7 (43.8)	4 (57.1)	0.723
Liver	148 (31.0)	15 (32.6)	6 (31.6)	9 (56.3)	2 (28.6)	0.329
Lung	72 (15.1)	7 (15.2)	6 (31.6)	2 (12.5)	0 (0.0)	0.339
Distant lymph nodes	262 (54.8)	19 (42.2)	6 (31.6)	8 (50.0)	5 (71.4)	0.124
Histological subtype						0.766
Conventional adenocarcinoma	355 (73.8)	35 (74.5)	15 (79.0)	10 (62.5)	5 (62.5)
Mucinous	94 (19.5)	6 (12.8)	3 (15.8)	3 (18.8)	1 (12.5)
Other†	32 (6.7)	6 (12.8)	1 (5.3)	3 (18.8)	2 (25.0)
RAS/RAF mutational status
RAS mutation	88 (24.7)	18 (54.6)	5 (41.7)	3 (30.0)	2 (28.6)	0.005
BRAF p.V600E mutation	156 (42.6)	3 (8.3)	4 (30.8)	3 (30.0)	2 (28.6)	<0.001
Missing	125	11	7	6	1
Identified germline mutation	71 (15)	18 (40.9)	9 (50)	3 (18.8)	3 (37.5)	<0.001
Missing	7	3	1	0	0
Mean systemic line of ICB	1.7 (1; 9)	1.7 (1; 5)	2.1 (1; 5)	1.6 (1; 3)	1.8 (1; 4)	0.585
Immunotherapy agent						0.995
Anti-PD-1	395 (82.1)	39 (83.0)	17 (89.5)	15 (93.8)	7 (87.5)
Anti-PD-L1	19 (4.0)	1 (2.1)	0 (0.0)	0 (0.0)	0 (0.0)
Combination with anti-CTLA-4	67 (13.9)	7 (14.9)	2 (10.5)	1 (6.3)	1 (12.5)

P values in bold indicate statistical significance.

^*Other includes esophagus (N=7), appendix (N=1), pseudomyxoma (N=1), adenocarcinoma compatible with colorectal origin but primary tumor not found (N=2).

^†Other includes signet ring-cell, medullary and neuro-endocrine tumors.

CTLA-4, cytotoxic T-lymphocyte associated protein 4; ECOG-PS, Eastern Cooperative Oncology Group – Performance Status; ICB, immune checkpoint blockade; IHC, immunohistochemistry; MMRd, mismatch repair deficient; MMRp, Mismatch repair proficient; MSI-H, microsatellite instability-high; MSI-L, microsatellite instability-low; PD-1, programmed cell death protein-1; PD-L1, programmed death-ligand 1.

Efficacy of immunotherapy

At the data cut-off date, the median follow-up was 28.1 months (mo) (95% CI 25.7 to 30.5). Overall, 247 patients (43.3%) had progressed and 180 patients (31.5%) had died. There was a significant difference in PFS between the groups, with median PFS of not reached (NR) (95% CI 50.5 to NR), 66.4 mo (95% CI 40.1 to NR), 37.2 mo (95% CI 26.8 to 53.3), 18.3 mo (95% CI 4.8 to NR) and 5.5 mo (95% CI 2.4 to NR) for complex/MSI-H, isolated/MSI-H, classical, retained IHC/MSI-H and MMRd-IHC/MSS or MSI-L subgroups, respectively (p=0.00038) (figure 2). There was also a significant difference in terms of OS, with median values of NR (95% CI 50.5 to NR), NR (95% CI 66.4 to NR), 72 mo (95% CI 53.3 to NR), NR (95% CI 14.4 to NR), and 15.9 mo (95% CI 10.1 to NR) for complex/MSI-H, isolated/MSI-H, classical, retained IHC/MSI-H and MMRd-IHC/MSS or MSI-L subgroups, respectively (p=0.016). The survival curves were comparable when looking at the subgroups of CRCs and non-CRCs only as well as in the Lynch and non-Lynch syndrome subgroups, although not always reaching statistical significance, possibly due to lack of power (figure 3 and online supplemental figure 1). However, there was no significant survival difference when comparing patients with Lynch syndrome and sporadic cancer (online supplemental figure 2).

Figure 2. Kaplan-Meier curves for progression-free survival (A) and overall survival (B) curves across different MMRd subgroups. IHC, immunohistochemistry; MMRd, mismatch repair deficient; MSI-H, microsatellite instability-high; MSI-L, microsatellite instability-low; MSS, microsatellite stable.

Figure 3. Kaplan-Meier progression-free survival and overall survival curves across different MMRd subgroups in patients with colorectal cancer (A+B) and patients without colorectal cancers (C+D). IHC, immunohistochemistry; MMRd, mismatch repair deficient; MSI-H, microsatellite instability-high; MSI-L, microsatellite instability-low; MSS, microsatellite stable.

A total of 553 patients were evaluable for tumor response according to RECIST V.1.1 criteria. ORRs significantly differed between groups (p=0.048), with rates of 59.1%, 58.7%, and 63.2% in the complex/MSI-H, isolated/MSI-H, and classical groups, respectively. In contrast, response rates were lower in the retained IHC/MSI-H (50%) and MMRd-IHC/MSS or MSI-L (25%) groups, with no complete responses observed in the latter two (figure 4). In the discordant groups (n=24), eight patients experienced a partial response as best response. Among these eight patients, one had a known pathogenic POLE p.Val411Leu mutation with high TMB and three had high TMB (53, 58.1 and 71 Mut/Mb). These four patients had no progression of disease at the time of data cut-off. Among the four other patients, only two underwent POLE mutation and TMB testing (which were negative). One of these two patients had progressive disease after 10 months of immunotherapy. All the other remaining patients had no progressive disease at the time of data cut-off.

Figure 4. Best response rate per MMRd subgroup. CR, Complete response; IHC, immunohistochemistry; MMRd, mismatch repair deficient; MSI, microsatellite instability; MSS, microsatellite stable; PD, Progression disease; PR, Partial response; SD, Stable disease.

Particularity of subclonal losses

Overall, seven patients had tumors with subclonal loss (five for MSH6 and two for MLH1) and displayed a classical heterodimer loss. All these patients had Lynch syndrome. In terms of response to ICB, there were two complete responses, three partial responses, one stable disease and one progressive disease.

Discussion

ICB has transformed the treatment landscape for MMRd cancers, yielding remarkable and durable responses in both neoadjuvant and advanced settings.¹⁹ However, a significant proportion of MMRd patients exhibit primary resistance, highlighting the heterogeneity of MMRd tumors. With the advent of tumor molecular boards and the ongoing shift towards biomarker-guided, histology-agnostic approaches in cancer management, expanded MMRd testing has revealed unusual MMRd tumor phenotypes, for which immunotherapy is often prescribed despite limited data. To our knowledge, this is the first study investigating the efficacy of ICB in these unusual MMRd phenotypes within gastrointestinal tumors. We distinguished four distinct groups of unusual phenotypes: isolated loss of PMS2 or MSH6 with MSI-H, complex IHC profiles with MSI-H, MMRd-IHC/MSS or MSI-L and retained IHC/MSI-H.

First, in the large real-world IMMUNODIG cohort (n=759), we found that at least 15.8% of patients had only one technique for MMRd tumor diagnosis. This is an important point, as the use of both techniques (PCR and IHC) ensures better diagnostic accuracy for MMRd phenotype and avoids the false-positive results reported in 9% of cases by Cohen et al, which are thought to be the main cause of primary resistance to immunotherapy.²⁰ In the recent CheckMate 8WH study including patients with MMRd CRC, up to 13% of tumors were in fact false positives after central review (MMRp and MSS). Conversely, after central review of cases who had undergone two techniques locally in our study, there were no false positives.

Second, the unusual phenotype represented 15.8% of MMRd tumors. Our findings suggest that the clinicopathologic characteristics of these unusual phenotypes differ significantly from classical MMRd cases. Patients with unusual phenotypes were notably younger, had a lower BRAF p.V600E mutation rate and a higher RAS mutation rate, a higher likelihood of Lynch syndrome, and were more likely to present with non-CRCs. This confirms previous results observed by our group in a large monocentric cohort.^{14 21} These differences may be all linked to Lynch syndrome. Due to the limited number of patients, multivariate analysis could not be performed. Of note, no significant difference in survival was observed between Lynch and sporadic cases, and similar patterns in survival were seen according to MMR groups in both cohorts.

Moreover, our cohort included 4.2% of discordant cases (MMRd-IHC/MSS or MSI-L and retained IHC/MSI-H) and we found that ICB were less effective in these cases, with MSS/MSI-L tumors showing the poorest outcomes close to those with proficient MMR/MSS tumors. There is also scattered clinical evidence suggesting that the efficacy of ICB varies based on whether patient selection was guided by MMRd-IHC or MSI status. Potential factors contributing to these discordances include technical and preclinical variables—such as hypoxia time and fixation issues for both techniques²²—and low tumor-to-normal cell ratios for PCR.²³ However, it is now clear that true cases of discordant tumors exist, likely driven by genetic or biological factors.²⁴

None of the registration trials have directly compared MMRd-IHC and MSI biomarkers for ICB efficacy. Subsequent analysis of trial data using MSI sensor (NGS-based MSI analysis) in CRC revealed that MMRd tumors exhibit a wide spectrum of MSI levels and that higher MSI levels are associated with better response rates, while patients with progressive disease had significantly lower MSI levels.²⁵ In a separate cohort of 33 patients with CRC treated with anti-PD-1 antibodies, treatment efficacy was evaluated based on MSI levels determined by the Bethesda panel (patient selection was based on either MMRd-IHC or MSI). In this study, no response was observed in MSS and MSI-L tumors, while patients with tumors showing MSI≥3/5 marker positivity experienced significantly longer PFS.²⁶ A recent study also confirmed that the majority of MMRd-IHC/MSS patients derive no benefit from ICB.²⁷ Growing data suggest that MSI as a biomarker may offer greater accuracy than MMRd-IHC because MSS tumors, typically associated with lower TMB, may not generate sufficient neoantigens to elicit a strong response to ICB, even if the tumor is MMRd by IHC.

Conversely, discordant retained IHC/MSI-H cases may arise from germline or somatic mutations in MMR genes that impair MMR function without affecting protein antigenicity, or from MMRd caused by genes other than those typically tested via IHC.²⁸ Although ICB appeared more effective in this subgroup compared with MMRd-IHC/MSS or MSI-L tumors, survival and response rates were still significantly poorer than those of other groups.

Currently, regulatory approval for ICB therapy in advanced solid tumors applies to tumors that are either MSI-H or MMRd-IHC.¹⁶ Including discordant cases in these approvals may account for some of the non-responses observed in clinical practice, warranting further investigation. Of note, an analysis of tumors associated with Lynch syndrome revealed that approximately 30% of non-colorectal and non-endometrial cancers exhibited MSI-intermediate status, as determined by NGS and when available tissue were all MMRd by IHC.²⁹ Therefore, we propose that the two test modalities (IHC and MSI molecular testing) should be considered complementary—underscoring the importance of conducting both for ICB eligibility—but not equivalent. In case of discordant results, these tumors should be reviewed by an expert pathologist as the majority of the cases are, in fact, false discordant cases.^{14 20 24} Additionally, most diagnostic tools were originally optimized for CRCs, and it is suggested that PCR-pentaplex molecular testing lacks sensitivity in other tumor types.¹⁴

Finally, for confirmed discordant MMR tumors, routine testing for POLE mutations and TMB should be considered as those tumors derive ICB benefit and were found in 4/6 patients evaluated with discordant MMR status and objective response. This indicates that, on the one hand, a subset of patients in this discordant group can benefit from ICB and should not be overlooked. On the other hand, it suggests that outcomes for discordant cases without POLE mutations or high TMB are worse than reported here. Conversely, patients with tumors in the isolated and complex phenotype groups exhibited comparable—if not superior—sensitivity to ICB than observed in the classical group. These findings highlight the importance of not excluding these unusual subtype phenotypes from immunotherapy considerations, as they appear capable of deriving similar therapeutic benefits.

Interestingly, we also report a series of seven patients with tumors exhibiting subclonal loss of MMR proteins. Traditionally, the College of American Pathologists defined MMRd as “no staining in the tumor anywhere”. As a result, tumors with subclonal loss—where portions of the tumor retain staining with an intensity stronger than the surrounding stroma—were historically classified as MMR-proficient.³⁰ However, subclonal loss, where only part of the tumor shows loss of MMR staining, has gradually been recognized as a form of MMRd. Our findings align with emerging literature,31,33 showing that subclonal loss is often associated with underlying Lynch syndrome, observed in all our cases in our study. While the exact mechanism behind subclonal loss remains unclear, we report that ICB showed efficacy in these tumors (5/7 ORR), supporting their classification as MMRd. The question of whether a cut-off lower limit of the tumor proportion with loss of staining for the tumor to be categorized as MMRd should be defined is unclear. For endometrial carcinomas, the British Association of Gynaecological Pathologists currently recommends a 10% cut-off.

Our study has several limitations, many of which arise from its multicenter nature and retrospective design. First, the lack of centralized review of all cases in the study may have introduced bias in the interpretation of IHC and molecular data. Several studies previously demonstrated that the majority of cases initially diagnosed as discordant were in fact false positives and, as such, should be re-evaluated.^{14 24} However, after centralized review by an expert pathologist, they identified up to 35% of true discordant cases among unusual phenotypes (26.7% in our study) and there is limited data on the clinical implications of these tumors. In our cohort, all cases were assessed by pathologists in cancer centers, which likely minimized intersite variability and 68.9% of unusual cases included in the study were centrally reviewed. Additionally, all pathological reports of unusual phenotypes included in this study were centrally reviewed. Second, we acknowledge that this patient population is rare, and the small sample size within each subgroup of unusual MMRd phenotypes potentially limit the statistical power of our findings. Furthermore, the complex group consisted of a heterogeneous group of tumors, likely with varying prognoses and predictive values for ICB efficacy. Nevertheless, this study represents, to our knowledge, the largest reported series of unusual MMRd phenotypes to date and the first to specifically investigate the efficacy of immunotherapy in this population.

Conclusion

While larger, prospective studies are required to validate our results and provide more precise treatment algorithms, our study provides a foundation for expanding immunotherapy use beyond classical MMRd. Our findings highlight the critical need for comprehensive, dual testing strategies—incorporating both MMRd-IHC and MSI assessments—to ensure the most accurate identification of patients likely to benefit from ICB and to guide Lynch syndrome testing. Our data suggest that discordant cases display poor response to ICB and should systematically be reviewed by an expert pathologist. In the case of true discordant cases, TMB and POLE mutation should be tested. Conversely, complex MMR protein alterations and isolated PMS2 or MSH6 losses that are MSI-H are potentially promising candidates for enhanced immunotherapy efficacy, although further investigation is needed to dissect these different subgroups and clarify the biological underpinnings of these findings. In the long term, tailoring treatment based on both precise MMR protein expression analysis and MSI status could help refine treatment strategies for further improvement of patient outcomes.

Data availability statement

Data are available upon reasonable request.