High microsatellite instability in CRC tumors is already used to screen patients to identify those more likely to benefit from ICI, but this phenotype is found in only about 5% of patients with metastatic CRC. This highlights the need to identify additional markers that might predict good response to immunotherapy. CRC tumors that show microsatellite stability but have a hypermutation phenotype, often involving POLE mutations, tend to respond strongly to ICI, which has been attributed to strong tumor infiltration by immune cells and high neoantigen burden 36. Tumors with microsatellite stability and high TMB may also respond well 22. The present results suggest that HRR mutations may be useful for predicting response, which could improve the screening of CRC patients for ICI.
More than 70% of CRC cases appear to develop first through mutation in APC, followed by mutations in KRAS, PIK3CA, SMAD4 and TP53, as well as loss of heterozygosity of chromosome 18. Such tumors usually show high CIN expression, microsatellite stability, and negligible CpG island methylation. Consistent with this pathway, APC, KRAS and TP53 were very often mutated in our TCGA, CN and HL cohorts (Figure 1), implying that these mutations co-occurred with HRR mutations. HRR mutations may result in chromosomal instability through deletions, frameshifts, aneuploidy and chromosome aberrations 37,38. Research on other cancers has shown that HRR mutations in the same genes as in our cohorts, such as BRCA2 and POLD1 (Figure 2), can influence response to immunotherapy. In patients with hereditary breast and ovarian cancer 39, BRCA2 mutations may be associated with stronger response to immunotherapy in clinical trials 40,41. Mutations in POLD1 or POLE have been linked to stronger response to ICI in endometrial and non-small cell lung cancers 42,43. Consistent with our findings, mutations in POLD1 or POLE may be useful for identifying patients with CRC who respond to ICI even though they lack high microsatellite instability 44. We found significantly longer survival among HRRmut patients than HRRwt patients after ICI therapy (Figure 4), supporting the idea that HRR mutations can recognize CRC patients with microsatellite stability who may respond well to ICI.
HRR mutations may influence response to ICI via their association with TMB: such mutations were associated with higher TMB in all our cohorts, regardless of whether tumors showed microsatellite instability or stability (Figure 3). The higher TMB may translate to presentation of more altered proteins as neoantigens, which induce antitumoral immune responses. In other words, the higher TMB may primarily alter the tumor immune microenvironment, which might explain why neither TMB nor HRR mutation was associated with overall survival (Figure 3B-C). Indeed, HRRmut tumors contained greater neoantigen load and more abundant CD8+ T cells than HRRwt tumors (Figure 5). In patients with early-stage CRC, increased tumor infiltration by CD8+ PD-1+ T cell is associated with better response to ICI 45. Interestingly, microsatellite instability in CRC is associated with greater tumor infiltration by CD68+ macrophages, CD8+ cytotoxic lymphocytes and CD45+ RO+ T memory cells, which may help explain the better survival of these patients than those with microsatellite stability 46. These considerations imply that the tumor immune microenvironment may strongly influence response to immunotherapy, and our findings indicate that HRR mutations can influence that microenvironment.
In anecdotal support of our proposal that HRR mutations can help predict response to ICI, we describe here two patients at our hospital with metastatic CRC, one of whom had HRR mutations and the other did not, and who responded quite differently to immunotherapy (Figure 6). Our findings justify systematic studies, preferably with larger samples from multiple centers, to validate and optimize HRR mutations as predictors of response to immunotherapy in patients with metastatic CRC.