In this study, we identified CT-26 and Colon 26 as different sensitive tumor models to anti-PD-1 therapy by in vivo screening and performed deep tumor-immune characterizations of the two syngeneic models to understand determinants driving anti-tumor activities of ICB. CT-26 is a murine colorectal carcinoma derived from BALB/c mice, and the syngeneic tumor model is one of the most commonly used murine solid tumor models and is categorized as a highly immunogenic tumor model [35]. Colon 26 is also a murine colorectal carcinoma derived from BALB/c mice, but the syngeneic tumor model has not been reported as frequently as CT-26 model. For Colon 26 tumor model, its sensitivity to immunotherapies, immune contexture of the TME and the differences from other commonly used murine tumor models has not been fully documented. We evaluated anti-tumor efficacies of anti-mPD-1 and anti-mCTLA-4 mAbs in the CT-26 and Colon 26 tumor bearing mice. Although the same cohort BALB/c mice were used for the efficacy study, sensitivities to anti-PD-1 were different between the models, suggesting that host genetics and microbiota before tumor inoculation do not affect the sensitivities. Multiple studies in a variety of tumor types have found a positive correlation between tumoral PD-L1 expression and ICB response or overall survival, while others have detected no association [3]. PD-1 blockade with pembrolizumab is approved by FDA for non-small-cell lung carcinoma patients whose tumors are positive for PD-L1 expression [36]. However, it’s an imperfect biomarker since in some tumors, such as renal cell carcinoma and first-line bladder cancer [37], there appears to be no correlation between PD-L1 expression and the likelihood of clinical response. In this study, we also performed immunohistochemistry analysis to determine PD-L1 expression level in the tumor (data not shown), and there was no statistically difference of the expression between CT-26 and Colon 26 tumor tissues. This suggested that tumoral PD-L1 expression may not explain the different sensitivity to anti-mPD-1 mAbs.
In contrast to anti-mPD-1, sensitivities to anti-mCTLA-4 were similar between the models. CTLA-4-blocking antibodies augment the binding of CD80/86 to CD28 rather than to CTLA-4 which triggers T-cell survival and expansion [38]. It also selectively depletes intra-tumoral CTLA-4 expressing Tregs via FcγR dependent mechanisms [39, 40]. At baseline, there was no notable difference of CTLA-4 expressing Tregs in the tumor-bearing mice between the two models. There was also similar expression of Cd80, Cd86, Fcgr1, Fcgr3 and Fcgr4 in the tumor tissues (data not shown). These data may reflect similar sensitivities to anti-mCTLA-4 mAbs between CT-26 and Colon 26 models.
We found that baseline immune population in the tumor was different between CT-26 and Colon 26. Compared to Colon 26 tumors, CT-26 tumors were more infiltrated by not only CD45 + immune cells but also anti-tumor immune cells including CD8 + T cells, NK cells and DCs. Chemokines can determine the distribution of immune cells in the tumor. CXC-chemokine receptor 3 (CXCR3) and its ligands CXC-chemokine ligand 10 (CXCL10) and CXCL11 have a key role in driving the trafficking of Th1 cells, CD8 + T cells and NK cells into the TME [41]. Our data showed that gene expression levels of Cxcl10 and Cxcl11 in the 100 mm3 CT-26 tumors are higher than Colon 26 (data not shown). This may reflect more CD8 + T cells and NK cells infiltration to the CT-26 tumor than Colon 26.
Despite growing literature reports supporting the association between mutational burden and immune checkpoint therapy response, tumor mutational burden had poor predictive power to differentiate complete or partial response from progressive disease as a single variable (42). In this study, Colon 26 has higher TMB than CT-26. Thus, TMB cannot explain the differential anti-mPD-1 sensitivity for our two tumor models. Nevertheless, we cannot ruled out other genetic features beyond mutational burden—such as genetic driver events, intra-tumoral heterogeneity, and mutational signatures—may affect response to immune checkpoint blockade [42].
Colorectal cancer (CRC) is the third most predominant cancer throughout the world [43]. Sonic Hedgehog, Wnt/ß-catenin, TGF-ß/SMAD, EGFR and Notch pathways are the major pathways that could be targeted for CRC therapy [43]. MSI-positive CRC patients have demonstrated positive response with ICB therapy [1]. We examined a few MMR related genes expression and found that MLH1 and MSH6 gene expression were higher in Colon 26 cell and tissue than CT-26, suggesting MMR difference might not explain differential anti-PD1 response. However, further investigation will be needed to clarify the difference of MMR gene expression leads to functional effects on MMR.
Interestingly, our transcriptome analysis highlighted Wnt pathway as one of the most distinct features between CT-26 and Colon 26. Wnt signaling has been shown to play a major role in regulating the immune tolerance against tumors [34]. Excluding immune cell (mainly CTL) infiltration into the TME constitutes a prominent mechanism of Wnt-mediated immuno-evasion in different types of cancer, mainly melanomas [14]. Activation of Wnt/β-catenin pathway in cancer cells inhibits secretion of CCL4 that is required for the recruitment of BATF-3 dependent dendritic cells to the tumor micro-environment, resulting in reduced levels of DC-derived CXCL10 and limited CD8 + CTL infiltration and cross-priming [14]. In consistent with this mechanism, we also observed less infiltration of dendritic cells and CD8 + T cells, as well as low expression of Cxcl10 and Xcr1 (a marker for cross-presenting CD8α + DCs) in Colon 26 tumor tissues (100 mm3) than CT-26.It’s been reported that the lack of effector cells into the TME, as typically observed in tumor with active canonical Wnt signaling, is possibly a main cause of primary resistance to cancer immunotherapies with ICB [34]. In mouse melanoma models, tumor-intrinsic active β-catenin signaling results in T-cell exclusion and resistance to combination of anti-PD-L1 and anti-CTLA-4 mAbs [14]. Conversely, RNAi mediated β-catenin inhibition increased T cell infiltration in ICB-refractory syngeneic mouse melanoma tumors [44]. Our study provides additional evidence and supports the potential resistance mechanism in mouse colon models.
With these observation from the comparison of CT-26 and Colon 26, we hypothesize that Wnt pathway signaling may explain the differential anti-PD-1 antibody response in CT-26 and Colon 26. In our study, we did not test the impact of Wnt signal inhibition on sensitivity to anti-PD-1 in Colon 26. In order to validate our Wnt hypothesis, we will need to generate β-catenin knock-out Colon 26 and compare sensitivity to anti-PD-1 treatment in the knock-out and the parental Colon 26 tumor bearing mice or test inhibitors of Wnt ligand secretion, such as LGK-974 [44], by combining with anti-PD-1 treatment in Colon 26 tumor models. Unlike anti-PD-1, there was no notable different sensitivity to anti-CTLA-4. We also need further investigation on whether Wnt pathway in the tumor will specifically affect sensitivity to anti-PD-1 therapy or not.