• Gastric Cancer
• Discordance
• Microsatellite Instability
• Mismatch Repair Deficiency

Post-hoc exploratory analyses of the major phase 3 trials have recently shown that Mismatch Repair Deficient (dMMR)/Microsatellite Instable (MSI) non metastatic gastric cancer (GC) have a favourable prognosis with surgical treatment only. Furthermore, neither perioperative nor adjuvant treatments provide a survival benefit when added to surgery. These findings were confirmed by the results of an individual patient data (IPD)-meta-analysis, combining the data of the ITACA-S and the ARTIST trials with those of the previously cited studies (MAGIC and CLASSIC). In our daily clinical practice, we are challenged to directly proceed to surgery in case of locally advanced resectable dMMR/MSI GCs. Indeed, the MMR expression test is performed by immunohistochemistry (IHC) on pre-treatment endoscopic biopsy fragments, whilst in the abovementioned studies the dMMR/MSI status was evaluated on surgical specimens by molecular testing. The ESMO guidelines for MMR deficiency evaluation recommend, however, IHC as the initial method of assessment.
Moving from this evidence, the aims of the study are the following: first, to compare the results of MMR proteins IHC staining performed on GC endoscopic biopsy with those performed on surgical specimen either in chemotherapy naïve patients or after neoadjuvant treatment; second, to assess the reliability of MMR testing on endoscopic biopsies; third, in case of discordance between biopsy and resection specimen, to characterize this recognised subgroup and identify the potential impact on prognosis or response to preoperative chemotherapy of dMMR/MSI GCs.
We retrospectively selected FFPE tumour samples of resected GCs, analysed by IHC for MMR expression at two Italian institutions, Azienda Ospedaliero-Universitaria Pisana (AOUP) and Istituto Oncologico Veneto (IOV), from 2005 to 2021. Only non-metastatic cases at the time of diagnosis with their paired preoperative endoscopic diagnostic tissue biopsy available were included. If lacking, IHC for MMR expression was performed on biopsy specimens. In case of debatable interpretation, molecular PCR (protein chain reaction) testing as confirmatory test was performed either on biopsy or on resected specimen. At AOUP, MMR protein analysis (MLH1, PMS2, MSH2, and MSH6) was performed by an IHC approach with the fully automated BenchMark ULTRA system (Roche-Ventana Medical Systems, Tucson, AZ) platform and the antigen–antibody reaction was visualized using the Optiview DAB IHC VENTANA Detection Kit, while at IOV MMR expression was evaluated by IHC with the automated platform Autostainer Link 48 (Dako, Carpinteria, CA, USA) and the antigen-antibody reaction was inspected with the EnVision FLEX kit with diaminobenzidine as chromogen. The staining interpretation was MMR deficiency in case of loss of protein expression of at least one marker. If required, molecular testing was performed after PCR amplification by the EasyPGX melting analysis at AOUP and by the Titano MSI test (Sanger test) at IOV.
We aimed to include at least 30 dMMR cases with full clinical and pathologic parameters. We would have considered preoperative biopsies reliable for MMR/MSI assessment if concordance between results achieved on biopsies and surgical specimens would have been equal or higher than 95% (i.e., 29 out of 30 cases). Fisher’s exact test and Pearson's χ2 test were applied to evaluate the association of concordant (dMMR) and discordant (pMMR) cases on biopsy tissue with baseline clinicopathological characteristics. Survival analyses were performed by comparison of survival percentages among different subgroups, at prespecified cut-offs, with Pearson’s χ2 test. Statistical significance was set at p ≤ 0.05 for a two-sided test. The statistical software Medcalc version 14.8.1 (Ostend, Belgium) was used for analyses.
Seventy-four dMMR out of 926 (8%) early and locally advanced GC at the time of diagnosis were identified after revision of the electronic medical records. Only 51 were suitable for our analysis (corresponding presurgical biopsies available for assessment). Clinicopathological characteristics and outcome data were consistent with those reported in literature: older age at onset, gastric antrum tumours as predominant primary location and the intestinal subtype as the most frequent. After 2-years follow-up, 89.5% of patients were alive. Disease relapse was observed in thirteen patients (pts) (25%), mainly within the first year after surgery (10 out of 13 pts).
Forty-five (88%) of the dMMR resected specimens exhibited a loss of MLH1/PMS2, whereas 4% displayed a loss of MSH2/MSH6, in agreement with previous reports. Interestingly, retained expression of MMR proteins (pMMR phenotype) was found in 9 (18%) out of 49 biopsies analysed (“discordant cases”). All the discordant cases presented a loss of MLH1/PMS2 in the corresponding surgical sample. Loss of MLH1/PMS2 was found in the biopsies with a lower frequency (66%) compared to that of the operated tumours. Heterogeneous expression of each MMR protein was observed. Isolated losses of PMS2 and MSH6 were described in 2% of cases. The PCR-based molecular test was performed on 6 out of 9 discordant cases (in 2 cases, the analysis was not feasible due to sample inadequacy). In 3 cases, the MSI status was confirmed; only 1 case, that underwent to neoadjuvant treatment, was found to be stable (MSS), but its low sample cellularity (equal to the sensibility threshold) compromises the test reliability. A heterogeneous pattern of MMR protein loss between biopsy and resected specimen was found in 7 pts (13.7%). No difference in terms of outcome or clinicopathological features was detected between discordant/heterogeneous and concordant samples.
The cases defined as discordant because of the presence of nuclear staining of MMR proteins on biopsy could be considered as a subset of heterogeneous tumours, provided that the proficiency of the MMR proteins is not attributable to misdiagnosis. False negative results are generally due to technical issues (pre-analytical errors, inadequate cellularity of the sample or of the control); on the contrary, false positive results might be connected to non-functional but antigenic competent MMR proteins (generally due to missense mutations). The pathologist’s expertise is crucial to go beyond these limits through the recognition of cases that require a confirmatory PCR molecular test. The notable disagreement between biopsy and resected specimen found in our study might be interpreted as an expression of MMR/MSI heterogeneity. This intra-tumour mosaic of pMMR and dMMR areas in GC has already been described in the literature, even though with a lower prevalence (8.9%). In one case of advanced GC its presence was predictive of resistance to immunotherapy; in another one of non-metastatic resected GC, pMMR regions were described especially in the lymph-nodes, suggesting this subclone expansion as leader of tumour progression.
The absence of a centralized review by an expert blinded pathologist and the lack of a confirmatory PCR in all heterogeneous cases represent possible limitations of our study. However, the assessment of MMR expression by two reference centres for pathological and molecular diagnostics strengthens our results. Moreover, our data collection of 51 dMMR GC is undoubtedly impactful.
In conclusion, the remarkable MMR protein expression disagreement between biopsy and surgical specimen demonstrated in our work suggests IHC on biopsy fragments as an insufficiently reliable method of assessment of MMR/MSI status in GC. An urgent need for standardisation is demanded for MMR/MSI status evaluation on GC biopsy: multiple sampling providing an adequate number of fragments and/or a complementary molecular biology could be reasonable strategies to improve the quality of the analyses. Moreover, MMR/MSI heterogeneity represents a possible confounder and should be considered for its putative predictive role in terms of response to immunotherapy, which is progressively been anticipating in earlier lines of treatment in dMMR/MSI GC. Further NGS approaches as well as MSI detection on liquid biopsy might be cutting-edge strategies to explore heterogeneity of MMR defects in GC.