The role of cytosolic nucleic acid sensing in immune regulation is highly complex, involving multiple sensors, effector pathways, and downstream modulators. We have shown here that cytosolic dsDNA introduced into malignant cells can upregulate expression of the chemokine CCL22. CCL22 binds the CCR4 receptor on Tregs and recruits Tregs to the tumor microenvironment (31–35). Importantly, CCL22 secreted directly from malignant cells has been shown to promote Treg recruitment (37), and reducing CCL22 production in malignant cells decreased Foxp3 mRNA in tumors in mice (85). Our finding that cytosolic dsDNA can robustly upregulate CCL22 in some cancers but not others may have clinical implications for treatments with oncolytic viruses, STING agonists, or plasmid DNA tumor vaccines.
Of special interest was our finding that dsDNA upregulated CCL22 predominantly through IRF3, while NF-κB had a more minimal role. NF-κB, not IRF3, is well-characterized to have multiple pro-tumor effects (96, 97) and to regulate CCL22 expression in other contexts (54, 85, 90–95). However, recent studies have also begun to evaluate potential adverse effects of IRF3 on cancer outcomes (106–108). In our study, the malignant cell lines that increased CCL22 the most, HeLa and MCF7, appeared to have intact IRF3 and IFN-β activation, suggesting that this axis may be critical for CCL22 upregulation in some cancer cells. However, a limitation to the current study is the size of the dataset tested. A much larger sampling of human cancer cells will be needed to determine the prevalence of dsDNA-mediated activation of CCL22 in malignant cells, as well as to determine whether IRF3 is a widespread mechanism. Notably, IRF3 activation triggers expression of type I interferons, yet two prior studies have reported that type I interferons inhibit CCL22/MDC (36, 98).
Our finding that dsDNA treatment of MCF7 cells resulted in IRF3 phosphorylation and concomitant activation of both IFN-β and CCL22, despite a lack of detectable phosphorylation on STING S366, remains to be explained. Apropos to this question is that, although a majority of studies report a requirement for S366 phosphorylation in STING-mediated activation of IRF3, one study reported that S366 phosphorylation can prevent the interaction between IRF3 and STING required for IRF3 phosphorylation (109, 110). While we cannot exclude the possibility that a STING-independent pathway activated IRF3 in MCF7 cells, we concluded that, in the cell line used for pathway analysis, STING was required for CCL22 upregulation.
The profound differences in CCL22 upregulation observed between the two strains of HeLa cells indicate that cancer cells can gain, or lose, the capacity to upregulate CCL22 in response to dsDNA. Cancer cell lines, particularly HeLa cells, have been well-documented to evolve in culture in response to myriad selection pressures that vary between laboratories (111). Notably, contradictory reports regarding the ability of HeLa cells to induce interferons have been reported as early as 1961 (112–114). It is unknown whether the differences observed in this study arose from a single, large-effect mutation or multiple smaller-effect mutations that collectively produced the phenotypic change, but such evolution and clonal expansion in vivo could conceivably contribute to acquired immune evasion.
Previous studies investigating CCL22 regulation indicate specificity based on species and cell type. For example, cell-type specific regulation of CCL22 in humans can be seen in the effects of interferon-gamma (IFN-γ), with studies reporting that IFN-γ increased CCL22 in keratinocytes (115–117), had no effect in fibroblasts (118) or airway smooth muscle cells (119), inhibited CCL22 in monocytes and macrophages (120), and was inversely correlated with CCL22 production in T cells (121). Interestingly, another nucleic acid sensing pathway, toll-like receptor 9 (TLR9), has also been implicated in CCL22 regulation, again in an apparently context-dependent manner. Unlike the cGAS-STING pathway, which detects cytosolic dsDNA, TLR9 detects single-stranded, CpG-containing DNA in endosomal compartments of sentinel immune cells. Previous reports show that the single-stranded TLR9 ligands, CpG-oligodeoxynucleotides (CpG-ODN), strongly enhanced CCL22 expression in murine dendritic cells (65), but another study concluded that CpG-ODN inhibited CCL22 in tumor-associated murine dendritic cells (36). Moreover, a potential species-specific regulation has also been suggested by studies showing that CpG-ODN repressed CCL22 across a range of murine bulk tumor samples (36) and in a murine asthma model (122) but increased CCL22 in cell isolates from human ovarian tumors (36). Taken together, these studies clearly show that our understanding of the variables influencing CCL22 regulation remains incomplete, and this is especially true with respect to human cancers and nucleic acid sensing. Indeed, CCL22 expression was previously thought to be absent in malignant cells, but several studies including our own have shown that CCL22 can be expressed in these cells (39, 54, 84, 85, 123–126).
Multiple studies have reported deleterious effects of elevated CCL22 in cancer (32, 36–84). A recent study also showed that in cervical cancer, intrinsic STING signaling in T cells promoted induction of Tregs (iTregs) (127). Such a scenario combined with release of CCL22 from malignant cells could conceivably retain Tregs in the tumor microenvironment, contributing to immune evasion. However, it should also be pointed out that CCL22 upregulation may not be detrimental in all cancers. Indeed, some studies have shown a protective role for CCL22, most notably in colon cancers and head and neck squamous cell carcinomas (HNSCC) (128–144). In HNSCC, studies have been divided about whether CCL22 correlates with reduced survival and metastasis (81–84) or increased survival (136–138). Although reasons for these conflicting reports remain unknown, it is interesting to note that the effect of CCL22 in HNSCC in Kaplan Meier survival analyses appears to vary with sex (e.g. Fig. S1F). It may also be possible that CCL22 and Tregs confer a protective effect in cancers in which onset and progression are associated with inflammation, such as colorectal.
Efforts to inhibit CCL22-mediated Treg recruitment in cancer have prompted recent clinical trials of the anti-CCR4 antibody mogamulizumab in advanced solid tumors, but with mixed results (32, 145–148). CCL22 recruits Tregs by binding to the receptor CCR4, and mogamulizumab has been used successfully to deplete Tregs in refractory adult T cell leukemia/lymphoma and cutaneous T cell lymphoma (31, 149, 150). Several preclinical studies showed that blocking the CCL22-CCR4 interaction also reduced Tregs in other cancers (37, 38, 151). However, a potential drawback noted for a total CCR4 blockade by agents such as mogamulizumab is that the chemokine CCL17 also binds CCR4, and CCL17 is reported to have nonredundant and even opposing functions to CCL22 (32), with CCL17 tending to promote inflammatory responses while CCL22 induces immune tolerance (65, 152). It has also been reported that an anti-CCL17 antibody failed to block Treg migration to tumors, while an anti-CCL22 antibody in the same study successfully blocked migration (38). Thus, specific antagonism of CCL22 rather than CCR4 has been suggested as a superior strategy (124). Another option would be to inhibit expression of CCL22, but this approach requires more knowledge of CCL22 regulation in cancer cells. The findings presented here offer a foundation for future studies investigating the molecular pathways regulating CCL22 in response to dsDNA in human cancer cells.