Oncolytic viruses possess a unique “duality” of action against cancerous cells in comparison with other therapeutic modalities. Not only can they exert cytotoxic effects on cancer cells (via infection, replication and release of viral progeny), but they simultaneously cause activation and proliferation of hitherto dormant immune cells in the TME. This provides a much-needed “jump start” for the immune system, allowing it to recognize and destroy cancer cells, including those which have acquired the ability to thwart host immunity (e.g., through expression of PD-L1, IL-23 and IL-10 receptors resulting in T cell exhaustion [14,15]).
Reovirus has been shown to preferentially replicate in and be cytopathic to colorectal cancer cells possessing an activated KRAS-signaling pathway [7–10]. In this paper, we have demonstrated that the administration of reovirus against a background of FOLFIRI and bevacizumab therapy in patients with KRAS-mutated mCRC results in multiple anti-tumorigenic alterations at the genomic, protein and immune cell distribution levels.
At the genomic level, we showed that reovirus administration results in statistically significant reductions in the exosomal expression of miR-29a-3p, as early as 48 hours and sustaining through day 15 (Figure 1a). This miRNA has been shown to be upregulated in CRC, and is postulated to contribute to CRC pathophysiology via inhibition of the WWOX tumor suppressor gene [16]. It is interesting that in patients treated with background therapy only, a reduction in miR-29a-3p was also seen at day 15. This may be due to suppression from bevacizumab and/or FOLFIRI treatment during the 2-week cycle. Additionally, the difference in the relative mean Cq values at day 15 between patient groups was not statistically significant. However, the sustained reductions in miR-29a-3p following reovirus administration may provide additional benefit via increased IFN-y expression, as miR-29a-3p is known to suppress IFN-y expression in bacterial infections [17]. As shown in Figure 4, serum increases in IFN-γ in patients treated with reovirus were observed after 72 hours. These increases were not seen in patients treated with background therapy only (data not shown).
The results of the transcriptome analysis (Tables 1 and 2) highlight additional anti-tumor effects of reovirus. The 4-fold and 2-fold increases at 48 hours and at day 15 (respectively) for TAP1 demonstrate reovirus’ protective effect, as TAP1 encodes a protein critical for the expression of peptides on the surface of MHC Class I, and down-regulation of this protein has been shown to promote immune evasion and poor prognosis in colorectal cancer [18].
FCGR2A and IFNAR1 genes encode receptors for antibody-binding and Type I interferon-binding, respectively. The observed fold increases in these genes (23-fold at Day 8 for FCGR2A; 20-fold at 48 hours for IFNAR1) are supportive of increased immunogenic activity following reovirus administration, particularly when coupled with the cytokine expression data (Figure 2), which show increased anti-tumor cytokines (GM-CSF, IL-12p40, IL-12p70 and IL-15 [19]) and decreased pro-tumorigenic cytokines (IL-8, RANTES, VEGF), over a 15-day period.
Of particular interest in the transcriptome analysis is the 33-fold increase in KRAS expression at 48 hours (Table 1), and the fold-reductions observed for VEGFA (2-fold, day 8), CXCR2 (2-fold, day 15), ITGAM (3-fold, day 15; Table 2). Reovirus infection in normal cells is known to trigger double-stranded RNA activated protein kinase (PKR; inhibits translation of viral proteins) phosphorylation [20]; constitutive expression of KRAS inhibits PKR phosphorylation, explaining the preferential replication of reovirus in KRAS-mutated tumor cells. As it is the phosphorylation (and not expression) of PKR that is inhibited by KRAS, the fold increase in KRAS seen following reovirus administration may represent an increased feedback inhibition of PKR protein produced in response to reovirus.
The reduction in VEGFA (a pro-angiogenic molecule [21]) transcript at day 8 is consistent with the observed reduction of serum VEGF over the preceding time points (Figure 2b). While the serum reductions are likely due to the effect of bevacizumab, the transcriptome results are due to reovirus, as an examination of the VEGFA expression changes in the patients who did not receive reovirus (but did receive FOLFIRI and bevacizumab), did not show any reduction (data not shown). Furthermore, an additional analysis of genes that are up-regulated by 2-fold and down-regulated by 0.5-fold at a p-value <0.05 showed that VEGFA is reduced across 48 hours, day 8 and day 15 timepoints (Supplementary Figure 1b).
A similar reduction at day 15 was observed for CXCR2 (the ligand for IL-8, another pro-angiogenic cytokine [22]). Statistically significant reductions in IL-8 were observed across several time points (Figure 2b). In summary, the reductions in VEGFA and CXCR2 demonstrate anti-tumorigenic effects by reovirus at the genomic level.
Lastly, ITGAM encodes CD11b, an integrin which combines with CD18 to form a leukocyte adhesion receptor; bone marrow CD11b+ cells have been shown to promote epithelial-to-mesenchymal transition and metastasis in colorectal cancer [23]. Thus, reductions at day 15 may signify a dampening of metastatic growth of tumor cells by reovirus.
While the aforementioned changes are compelling, it is well known that there is a “tug-of-war” of sorts in the TME [24], between pro- and anti- tumorigenic factors. Thus, the fold increases in STAT3, KLRD1 (CD94) and CD244 (Table 1) also deserve consideration and comment. STAT3 is part of the IL-6/JAK/STAT3 pathway, which is hyperactive in many cancers and is known to suppress the anti-tumor immune response [25]. The increase in STAT3 is thus likely responsible for the increase observed in serum IL-6 at 24 hours post-reovirus administration (Figure 2b). KLRD1 (CD94) is known to suppress NK activity against tumor cells via ligation with the NKG2A receptor on the tumor cell surface, followed by interaction with HLA-E receptor on the NK cell [26]. As shown in the flow cytometry data in this paper, no change in the population of NK cells was observed after administration of reovirus (Figure 4). The fold increases seen of CD244 (responsible for NK and T cell exhaustion [27]), at days 8 and 15 may also be contributing to this finding. While it is clear these changes follow reovirus administration, whether they are driven by reovirus in order to support continued viral propagation, versus being a true counter-response by the tumor cell to avoid destruction (following activation of immune cells in the TME by reovirus infection), is beyond the scope of this study and warrants further investigation.
In summary, the data presented in this paper highlight the potential of reovirus to function as an immunomodulatory and cytotoxic adjuvant to standard chemotherapy in patients with mCRC and KRAS-mutations. We have shown that over a 15-day period, reovirus modulates several anti-tumor changes, across the genomic, protein and immune cell distribution levels. Figure 4 presents the temporal and dynamic effects of reovirus between exosomal miR-29-3p, IFN-y and several immune cell types. As mentioned previously, the decrease in miR-29a-3p may contribute to the increase in IFN-y [17]. Additionally, this increase may be attributed to the rising population of activated CD4+ and CD8+ T-cells, as these predominantly secrete IFN-y [28]. The increased expression of Granzyme B post reovirus treatment in a patient biopsy sample (documented by immunohistochemistry) also demonstrates activation of CD8+ T cells, and highlights reovirus-directed tumor cell-specific destruction.
Limitations of this study include a small sample size of patients analyzed who received reovirus (n=5 for all analyses performed, with the exception of the cytokine analysis, in which data from one patient was excluded, and in the immunohistochemistry analysis, in which one patient was biopsied). Additionally, the pooling of patient samples to present (relative) mean changes over time may be biased towards “responder” patients, which the authors have acknowledged by using the standard error of the mean (appropriate in this small sample set to show the variance within the group). However, inter-patient variability across oncologic therapies is not uncommon – indeed, it is one of the main drivers for the discovery of new investigational agents, like reovirus. The results of this study should therefore be appreciated in the context of the known complexity of genomic, protein and cellular interactions within the TME.