Colorectal cancer (CRC) is one of the main causes of morbidity and mortality, representing approximately 10% of all cancers diagnosed worldwide (1). Mismatch repair deficient (dMMR)/microsatellite instability-high (MSI-H) colorectal tumours are identified in approximately 15% of CRC patients. Due to relatively favourable prognosis in comparison to the mismatch repair proficient (pMMR) CRC, the proportion of these tumours decreases to approximately 4-5% in patients with metastatic CRC (mCRC) (2).

Deficiency in protein products of MSH2, MLH1 and MSH6 genes results in impaired detection of mismatched and unpaired bases, leading to the expression of abnormal proteins that may be recognized as neoantigens by the immune system. Additionally, most likely due to Darwinian selection pressure, the deficiency causes failure of apoptosis upon detection of critical DNA damage (2, 3).

dMMR/MSI-H status has been associated with the lack of benefit from fluoropyrimidines in stage II and III CRC (4–6). Nevertheless, the benefit from oxaliplatin-containing regimens remains unchanged making them the standard of care in patients with CRC requiring adjuvant chemotherapy (6–8). In mCRC the prognosis and response to chemotherapy in patients with dMMR/MSI-H tumours is significantly influenced by the presence of somatic BRAF mutations which are associated with poor prognosis (9, 10). BRAF V600E mutations are causally linked to MLH1 promoter hypermethylation leading to deficiency in MLH1 and PMS2 proteins, the most common cause of dMMR/MSI-H phenotype in patients without Lynch syndrome (11–13).

Monoclonal antibodies against the vascular endothelial growth factor (VEGF) or, in patients with KRAS/NRAS wild-type tumours, against the epidermal growth factor receptor (EGFR) are commonly added to chemotherapy in mCRC. dMMR/MSI-H patients receiving the VEGF antagonist bevacizumab achieved longer overall survival (OS) compared to patients treated with the EGFR inhibitor cetuximab in the CALGB/SWOG 80405 clinical trial (14). Similar findings were recently reported in a retrospective study (15).

The marked difference in the proportion of dMMR/MSI-H tumours between early and advanced CRC stages points to a strong role of immunoediting in the progression of these tumours. The activation of immunosuppressive pathways linked to PD-1 and CTLA4 protects these tumours from destruction by the immune system and could explain the high efficacy of checkpoint inhibitors in this setting (25). However, only 40-50% of patients with dMMR/MSI-H mCRC derive long-term benefit from the treatment. Possible mechanisms of intrinsic and acquired resistance have been comprehensively reviewed by Sahin et al. (26).

PD-L1 expression was not correlated with benefit from immunotherapy in patients with dMMR/MSI-H mCRC or in the tumour-agnostic trial by Le et al. (16, 18, 19).

The efficacy of checkpoint inhibitors in dMMR/MSI-H patients was not associated with BRAF V600E mutation which is present in approximately 30-40% of dMMR/MSI-H mCRC (19, 23). On the other hand, the apparent lower efficacy of immunotherapy in patients with KRAS/NRAS mutated tumours treated with anti-PD-1 monotherapy in the KEYNOTE -177 and CheckMate-142 studies, contrasted with a preserved benefit of combined immunotherapy in CheckMate-142, deserved further study as the subgroups were too small for a valid conclusion (19–21, 23).

Tumour mutation burden (TMB) was a better biomarker for the efficacy of immunotherapy than the absence of an MMR-related protein. It directly correlated with the number of neoantigens (27). Although the great majority of dMMR/MSI-H tumours had high TMB, some dMMR/MSI-H tumours with low TMB have been identified, and they exhibited resistance to therapy with checkpoint inhibitors (28, 29).

Salem et al. compared the TMB according to the type of MMR deficiency. In a heterogeneous cohort of cancers, the largest subgroups comprised endometrial and colorectal tumours. The loss of MSH2/MSH6 was associated with approximately double rate of mutations compared to loss of MLH1/PMS2 (30).

Approximately 3% of colorectal tumours classified as microsatellite stable (MSS) have high TMB, mostly associated with the mutation of polymerase epsilon (POLE) or variant MSH2 (28, 31, 32). These mutations resulted in a high sensitivity to checkpoint inhibitors. Clinical trials enrolling patients with POLE-mutated mCRC along those with dMMR/MSI-H mCRC are ongoing (NCT03150706 and NCT03435107, Table 1 ).

There is growing evidence that the genomic instability in tumours with mismatch-repair deficiency may result in the formation of gene fusions. Cocco and collaborators found that potentially targetable fusions of oncogenes such as NTRK, BRAF, RET, FGFR, ROS1 and ALK were present in 5% of dMMR/MSI-H CRC as compared to only 0.4% of MSS CRC cases. Particular enrichment was seen in the subgroup of dMMR/MSI-H BRAFwt/RASwt CRC cases with MLH1 promoter hypermethylation where the occurrence of these aberrations was as high as 42% (13). Vaňkova et al. confirmed these findings identifying oncogenic gene fusions in as many as 9 of 23 colorectal tumours that were dMMR/MSI-H due to MLH1 promoter hypermethylation and simultaneously BRAFwt/KRASwt. The fusions involved NTRK, ALK and BRAF genes (33). These studies suggest that BRAFwt/KRASwt tumours with MLH1 promoter hypermethylation should also be tested for the presence of oncogene fusions.

It has been shown that dMMR-MSI-H colorectal cancer was associated with specific changes in gut microbiome. Fusobacterium nucleatum was enriched in dMMR/MSI-H CRC and associated with lower number of tumour-infiltrating lymphocytes while the opposite has been suggested for MSS colorectal cancer (34). However, somewhat counterintuitively, F. nucleatum has been found to be positively associated with response to checkpoint inhibitors, enhancing the expression of PD-1 and PD-L1 and activating the stimulator of interferon genes (STING) signalling (35). The activity of interferon-γ pathway was also associated with response in dMMR/MSI-H CRC in a recent transcriptomics study while VEGF-A expression characterised poor responders (36).

Other putative molecular markers of response to checkpoint inhibitors have been proposed, in particular β2-microglobulin mutations (37). Nevertheless, a recent study failed to corroborate their association with treatment outcomes (38).

Finally, peritoneal involvement manifesting with ascites is associated with resistance to immune checkpoint blockade in dMMR/MSI-H gastrointestinal cancers and markedly inferior patient outcomes, even compared to peritoneal disease without ascites (39).

The search for reliable predictors is ongoing and detailed reviews of potential biomarkers for immunotherapy of colorectal cancer are available (40, 41).

Immunotherapy is the preferred treatment option for dMMR/MSI-H CRC resulting in long-term treatment responses in approximately 40-50% of patients. As most non-responders will experience progression within 6-8 months of the treatment initiation, the obvious clinical strategies that have been successfully applied in other tumours with similar early treatment characteristics include the combinations of checkpoint inhibitors with chemotherapy and the use of checkpoint inhibitors as a maintenance therapy in non-progressing patients after the induction (42, 43) In both situations, inclusion of a VEGF-targeted therapy in the regimen appears beneficial. Multiple clinical trials are ongoing and our knowledge on the use of immunotherapy in this setting will keep expanding in the following years ( Table 1 ).

TB has designed and written the manuscript. The author confirms being the sole contributor of this work and has approved it for publication.

The work was supported by grant TN-00064190 from the Ministry of Health, Czech Republic to the Thomayer University Hospital. Publication fee was covered by unrestricted grants from Servier, Roche, AstraZeneca, and Bristol Myers Squibb to the First Faculty of Medicine, Charles University, Prague, Czech Republic. The funder was not involved in the study design, collection, analysis, interpretation of data, the writing of this article or the decision to submit it for publication.

TB received research support and honoraria from Roche, Bristol Myers Squibb, Merck Sharp Dohme, Merck, and AstraZeneca, all unrelated to the present paper.

All claims expressed in this article are solely those of the authors and do not necessarily represent those of their affiliated organizations, or those of the publisher, the editors and the reviewers. Any product that may be evaluated in this article, or claim that may be made by its manufacturer, is not guaranteed or endorsed by the publisher.