Our clinical trial is the first to our knowledge, that evaluates the safety and efficacy of autologous dendritic cell vaccine loaded with tumor lysate in combination with CBI in pre-treated mismatch repair-proficient (MSS) metastatic colorectal cancer patients. From this trial we conclude that this strategy is safe and can be developed in a multicenter setting. Regrettably, our study, as has been previously reported with allogeneic vaccines combined with Pembrolizumab (15), or other recent combinations with avelumab and cetuximab (CAVE) (16) or avelumab plus regorafenib (REGOMUNE) (17) did not achieve pre-planned clinical benefit for phase III development.
Although primary-end point of the study was not reached, we observed a median OS of 12.2 months that compares better than the 5 months remarked for trials using anti-PD-1 as monotherapy (5) or between 10–11 months with other strategies with avelumab with approved drugs in mCRC (5,16,17),. In addition, the safety profile with the combination of avelumab and DC vaccine do not differ with single agent therapies separately (5,10). We cannot rule out, that because we use a stringent inclusion criterium (pe. patients with LDH > 1.5ULN were excluded), intrinsic tumor characteristics instead of treatment efficacy, would influence on survival.
The rational for combining vaccines with CBI came from extensive pre-clinical research, that elegantly demonstrated that immunologically cold tumors, generated from the murine colorectal cell lines MC38 and CT26, acquired a hot phenotype after various vaccination approaches (11,12). It should be emphasized that these two murine models show a much higher tumor mutational burden (TMB) than MSS colorectal cancer patients or mouse colorectal tumor organoid models (18). Because of scarcity of clinical data addressing whether vaccines could increase antigen presentation in mCRC, we have taken several approaches to address this important issue with different approaches. First, we did a pre-clinical study that shows that the co-culture of PBMCs with DCs pulsed lysate but not with the PBMCS and the anti-PD-L1 therapy increased T-cell proliferation and IFNγ. Second, by analyzing circulating cytokine levels, we have found that, while CCL5, TGFβ and MMP9 did not display significant variations after therapy, CCL2 and SDF1a indeed showed decreased levels. Interestingly, the only long-term survivor patient showed a > 240 fold-increase of CCL5 levels post-therapy. Third, the numbers of peripheral blood CD8 effector T cells and CD8−PD1+ blood cells did not show significant variations after therapy. Finally, the pro-immune GEP signature did not show significant gains in post-therapy tumor tissue liver samples compared with pre-therapy samples. All these translational data, fail to support at least clinically, a strategy to use ADC vaccines in order to inflame immunological “cold” tumors.
Hyper-progressive disease (HPD) was seen in 4/18 (22%) cases and it was of special concern because we did not observe this type of progression in our previous randomized ADC study (10), in spite of a better clinical profile of patients included in the current study (ECOG PS of 0,1 and without poor biological features (basal LDH < 1.5ULN)). We have seen HPD in IMMETCOLS Cluster 1 patients, whose tumor displays characteristics of a mesenchymal phenotypes (1/3) (33%) but also in patients classified as Cluster 3, with epithelial phenotypes (3/12) (25%). HPD has been extensively described in mesenchymal tumors which are usually enriched with M2- polarized, macrophages and Tregs (19,20). Fully humanized monoclonal antibodies such as Avelumab, have been found to exert at least part or their antitumor activity through IgG1 antibody-dependent cell mediated cytotoxicity (ADCC) (21). However, impaired antibody dependent cellular phagocytosis (ADCP) due to M2 polarization, a recently described mechanism of resistance, probably also contributes to immunosuppression in Cluster 1 (22,23). We were surprised that HPD occurred also in Cluster 3 patients, characterized by an epithelial-glycolytic phenotype endowed with a reductive carboxylation ability, with M1-polarized macrophage infiltration.
Metabolic competition between cancer cells and tumor microenvironment are tightly intertwined with increased aerobic glycolysis and glutamine consumption (24–26). Nevertheless, in our analysis of pre- and post-therapy metastatic samples, the most significantly de-regulated genes were related to lipid metabolism and transport. Interestingly, many of the genes down-regulated post-therapy (APOA1, APOH, APOC3, PLG, SERPINC1, C9, ANGPTL3, CFHR5) have been found specifically associated with colorectal liver metastases, suggesting that the observed changes are clinically meaningful (27,28). Importantly, some of the lipid metabolism and inflammation genes that we found down-regulated in post-therapy samples play an important role in tumor control and immunomodulation (29–33) and protection from oxidative stress (34,35). We speculate that CRC tumors treated with Avelumab and ADC downregulate APOA1 and CCL16, among other genes, but preserve their viability despite increased oxidative stress, thanks to the activation of compensatory reductive pathways (such as pentose phosphate, malic enzyme or isocitrate dehydrogenases). Indeed, these pathways are upregulated in IMMETCOLS Cluster 3. It has been previously described that colorectal liver cancer cells use cholesterol and poly-unsaturated fatty acids to promote ROS and cancer progression (37,38). Metabolic rewiring after glutaminase inhibition, with lipid consumption instead of glutamine uptake, has been described pre-clinically (39). However, to our knowledge, the current study provides the first clinical evidence that a similar metabolic adaptation is also used by colorectal cancer cells after Avelumab and ADC therapy, to boost oxidative stress and contribute to cancer progression (Fig. 5). We propose that a two-pronged metabolic inhibition consisting in 1) to reduce glycolysis (40), glutamine or lipid supplies (41) in order to minimize oxidative stress and 2) or pentose phosphate inhibitors (42,43) to minimize anti-oxidant defenses, merits consideration in combination with vaccines and CBI to achieve clinical efficacy.
We are aware that our study has several limitations. First, the number of recruited patients was small and second, the selected primary end-point (PFS instead of OS) conveys a narrower scope in our observations. Third, the lack of enhanced PFS contrasts with a greater OS observed; this long-term survival effect may be explained due to the limitations of RECIST1.1 interpreting progressive disease in the presence of a immune response. We are also conscious that tumor biopsies are not the most desirable source of antigens in order to prepare a vaccine because of their high heterogenicity compared to synthetic peptides or neoantigens. Finally, the lack of significant post-therapy changes in peripheral blood CD8+ effector T cells numbers should not preclude future analyses of these and other immune populations in pre- and post-therapy tumor tissue biopsies. That said, our additional observations of lack of increased GEP scores observed in post-treatment samples suggests that the combined Avelumab plus ADC therapy may fail to boost the cytolytic efficacy of the immune environment in CRCs.