The colorectal cancer immune paradox revisited.

Tumor infiltrating lymphocytes (TILs) represent a strong independent predictor of relapse and overall survival in colorectal cancer (CRC). However, it appears that a majority of CRCs, i.e., microsatellite stable (MSS) tumors, are refractory to immune checkpoint blockers. The results of recent comprehensive analyses of genomic data provide possible answers.

Currently we are witnessing major breakthroughs in cancer therapy by the development of effective immunotherapeutic approaches. Specifically, strategies that use antibodies to block immune checkpoint molecules like CTLA-4, PD-1, and PD-L1 are showing impressive results in a number of cancers including not only melanoma, but also cancers like lung, head and neck, or bladder cancers among others. Paradoxically, patients with CRC, a cancer for which tumor infiltrating lymphocytes (TILs) represent a strong independent predictor of relapse and overall survival, do not benefit from the administration of a PD-1 targeting antibody. A notable exception represents patients with microsatellite-instable (MSI) phenotype.

We have recently provided a high-resolution genomic view on the immunophenotypes and antigenomes (repertoire of tumor antigens) and provide here explanations for this paradox. We developed an analytical strategy and examined genomic data sets from The Cancer Genome Atlas (TCGA) (n = 598). Briefly, we used RNA- and whole-exome NGS data to chart the antigenome comprising two major classes: cancer-germline antigens and neo-antigens. Additionally, we defined a compendium of genes using expression data from purified immune cells in order to derive immune signatures related to specific subpopulations, and then used RNA-sequencing data from the TCGA cohort to identify subpopulations of TILs.

The cohort was analyzed with respect to the three distinct molecular phenotypes: (1) mutational status (hypermutated and non-hypermutated), (2) microsatellite status (MSS and MSI), and (3) methylation status. The quantification of the immune subpopulations showed that TILs were associated with distinct molecular phenotypes or the combinations thereof. As expected, MSI tumors were characterized by enrichments of TILs mostly related to adaptive immunity. A small group of MSS tumors, which were hypermutated were characterized by lower enrichment of effector memory and central memory CD4+ and CD8+ cells. The non-hypermutated MSS tumors were also enriched with TILs, but the levels were reduced compared to the hypermutated tumors.

The genetic and the immunophenotypic variability of the CRC tumors imposed the question if different tumors use different mechanisms of tumor escape. We first analyzed the genetic heterogeneity of the tumors using exome sequencing data and SNP-array data. Based on the cancer cell fractions, the cohort was divided into six groups: MSI group (n = 69), hypermutated MSS group (n = 19), and four groups of MSS tumors with low (n = 63), two intermediate (n = 123, and n = 96) and high heterogeneity (n = 105). We then examined the infiltration of immunosuppressive cell types (myeloid-derived suppressor cells (MDSCs) and regulatory T cells (Tregs)), and the expression of five classes of immunomodulating molecules: immunoinhibitory genes, immunostimulatory genes, MHC class I, MHC class II, and non-classical MHC molecules.

One striking observation was how the genetic basis of the tumors determines the tumor escape mechanisms. Hypermutated tumors (MSI and hypermutated MSS tumors) showed higher intratumor heterogeneity, suggesting that the greater mutational load in these tumors results in a higher load of neo-antigens (635 ± 308 and 1,651 ± 1,455, respectively), and likely promotes T-cell activation and infiltration. Furthermore, in these tumors, Tregs and MDSCs were depleted whereas activated CD8+ and CD4+ cells were enriched. This strong immunological response was counterbalanced by an increased expression of several immunoinhibitors including CTLA-4, PD-1, and IDO1, which might explain why are these tumors progressing. Interestingly, similar effective immune responses were observed also in the group of non-hypermutated tumors with low genetic heterogeneity. This inverse association of the tumor heterogeneity and the immune responses in MSS tumors was evident at several levels: enrichment of immunosuppressive cells, expression of immunostimulators, and expression of MHC class I and MHC class II molecules. The data suggests that in the three groups of MSS tumor with intermediate and high heterogeneity the tumor escape was governed by downregulation MHC I and MHC II molecules, and upregulation of HLA-G, which is associated with worse overall survival.

In summary, the results of our in silico analyses provide possible answer for the CRC immune paradox (Fig. 1). Patients with MSI tumors benefit from therapy with anti PD-1 antibodies due to the favorable genetic (heterogeneous tumors, high load of neo-antigens) and immune characteristics (infiltration of CD8+ and CD4+ T cells, depletion of Tregs and MDSCs, upregulation of MHC class I and MHC class II molecules). In contrast, a majority (84%) of the patients with MSS tumors show unfavorable features: downregulation of MHC class I and MHC class II molecules, increased expression of HLA-G, and enrichment of MDSCs.

Figure 1.

Proposed model that may predict responses to anti-PD-1 blockade. The tumor escape mechanism is determined by the genetic status, i.e., hypermutation (for MSI and MSS tumors) or tumor homogeneity (for non-hypermutated MSS tumors). Left: Tumors escape immunosurveillance by upregulation of immunoinhibitory molecules, are enriched with CD8+ and CD4+ cells, and hence, would benefit from therapy with anti-PD-1 antibodies. Right: tumors escape immunosurveillance by downregulation of MHC I and MHC II molecules, and upregulation of HLA-G. These tumors are enriched with immunosuppressive cells and would not respond to therapy with anti-PD-1 antibodies.

Following these arguments, there are two groups of patients, patients with hypermutated MSS and non-hypermutated MSS but homogeneous tumors (16%) that would likely benefit from anti PD-1 blockers since the immunophenotypes are similar to the MSI tumors. This is further supported by a recent study in which a small group of MSS patients indeed showed response.

Our model is in concordance with recently suggested ones and extends our understanding of the tumor-immune interaction in CRC. We advocate that for the treatment of CRC patients with checkpoint inhibitors not only the characterization of the genetic status (i.e., mutational load and tumor heterogeneity), but also of the immunophenotype (TILs and immunomodulatory molecules) of the tumor is required.