Selective reactivation of STING signaling to target Merkel cell carcinoma

Significance

The immune-dampened status of Merkel cell carcinoma (MCC) tumors presents a major obstacle to developing effective immunotherapies. We propose that STING silencing may cause the immunologically “cold” MCC tumor microenvironments by blocking cytokine production and, consequently, impeding cytotoxic T cell infiltration, activation, and killing of tumor cells. Our findings indicate that targeted activation of STING^{S162A/G230I/Q266I} by DMXAA could be a viable strategy for bolstering antitumor adaptive immunity in STING-silenced cancers. DMXAA does not stimulate human STING activity. Therefore, when combined with AAV delivery of STING^{S162A/G230I/Q266I} to primary tumors, it may achieve tumor-specific STING activation without concomitant pathology caused by systemic inflammation frequently observed with traditional human STING agonists. This approach could synergize with existing immune-checkpoint therapies to improve MCC treatment.

Abstract

Merkel cell carcinoma (MCC) is a lethal skin cancer that metastasizes rapidly. Few effective treatments are available for patients with metastatic MCC. Poor intratumoral T cell infiltration and activation are major barriers that prevent MCC eradication by the immune system. However, the mechanisms that drive the immunologically restrictive tumor microenvironment remain poorly understood. In this study, we discovered that the innate immune regulator stimulator of IFN genes (STING) is completely silenced in MCCs. To reactivate STING in MCC, we developed an application of a human STING mutant, STING^{S162A/G230I/Q266I}, which we found to be readily stimulated by a mouse STING agonist, DMXAA. This STING molecule was efficiently delivered to MCC cells via an AAV vector. Introducing STING^{S162A/G230I/Q266I} expression and stimulating its activity by DMXAA in MCC cells reactivates their antitumor inflammatory cytokine/chemokine production. In response to MCC cells with restored STING, cocultured T cells expressing MCPyV-specific T cell receptors (TCRs) show increased cytokine production, migration toward tumor cells, and tumor cell killing. Our study therefore suggests that STING deficiency contributes to the immune suppressive nature of MCCs. More importantly, DMXAA stimulation of STING^{S162A/G230I/Q266I} causes robust cell death in MCCs as well as several other STING-silenced cancers. Because tumor antigens and DNA released by dying cancer cells have the potential to amplify innate immune response and activate antitumor adaptive responses, our finding indicates that targeted delivery and activation of STING^{S162A/G230I/Q266I} in tumor cells holds great therapeutic promise for the treatment of MCC and many other STING-deficient cancers.

Merkel cell carcinoma (MCC) is a highly aggressive neuroendocrine cancer often found in the skin. It is the second most common cause of skin cancer death after melanoma (1), and the incidence of MCC has increased by 95% in the last decade (2). Approximately 80% of MCC cases can be directly linked to Merkel cell polyomavirus (MCPyV) infection (3, 4). In MCPyV⁺ MCCs, the MCPyV genome is clonally integrated into the tumor genome to express the viral oncogenes, large T (LT) and small T (sT), which together promote MCC growth (5–7). Therefore, MCPyV infection represents a key causal factor for MCC development (6, 7). In addition, immunosuppression (8, 9), HIV infection (10), organ transplants (11), sunlight exposure, and ultraviolet (UV) radiation (12) can increase the likelihood of MCC development.

MCC metastasizes rapidly. Currently, there is no universally effective treatment for MCCs that have reached the metastatic stage (13). The newly developed PD-1/PD-L1 immune checkpoint blockade therapies showed promising results in clinical trials, but responses are often short-lived while a significant portion of MCC patients are resistant to the treatment (14–18). PD-1 treatment also causes highly adverse side effects in ≈28% of patients (14–18). Therefore, alternative strategies are needed to combat MCC.

Significantly increased risk of developing MCPyV-associated MCC has been observed among immunocompromised individuals (5, 19). At the same time, more than 90% of MCC patients have normal immune function but still fail to clear the tumors (19, 20). MCC tumors develop despite patient production of T cells with MCPyV oncoprotein-specific TCRs (20, 21). Presence of CD8⁺ T cells in MCC tumors, however, is positively associated with patient survival (22–25). Unfortunately, T cell tumor infiltration is detected in only 4–18% of MCCs (20, 23, 25–27), and in these rare cases, infiltrating T cells demonstrate significantly suppressed activity (28). These observations suggest that MCPyV-associated MCCs can escape immunological eradication by restricting T cell tumor infiltration and repressing T cell activation. Still, the mechanisms by which MCC evades immune destruction are largely unknown. That MCPyV⁺ MCCs have the capacity to resist the first line of existing immunotherapies underscores the urgency to understand MCC immune escape mechanisms so as to overcome this hurdle. Key characteristics MCCs share with immunologically “cold” tumors (29) suggest that stimulating intratumoral T cell recruitment and activation is a prerequisite for a broadly successful immunotherapeutic strategy.

Stimulator of IFN genes (STING) is a key immune mediator for activating cytokine production (30–33). Cancer cells often maintain a high level of damaged DNA, which could stimulate STING-dependent induction of type I interferons (IFNs) and other proinflammatory cytokines (30, 33, 34). These molecules not only display direct cytotoxic activity against tumor cells but also promote tumor antigen-specific intratumoral T cell infiltration, activation, and tumor repression (30–33, 35, 36). These findings highlight a critical role of STING in antitumor T cell responses.

In this study, we discovered that STING expression is repressed in MCC cell lines and tissues. We hypothesized that STING silencing may underlie the immune suppressive nature of MCCs and that reactivation of STING in MCCs could stimulate antitumor cytotoxicity. To test this idea, we developed an approach to activate STING specifically in the tumor cells. We found that restoring STING function in MCC cells reactivates downstream antitumor cytokine production and stimulates T cell migration, T cell activation, and MCC cell death. Our results suggest that targeting the STING pathway may hold great therapeutic promise for the treatment of MCC and other STING-silenced cancers.

Results

STING Is Silenced in MCC Cell Lines and Tissues.

To identify factors contributing to the immune-dampened phenotype in many MCC tumors, we examined the expression levels of type I IFN response-related innate immune sensors in MCC cells. We first analyzed publicly available RNA-sequencing (seq) datasets for transcript levels of cGAS, STING, TLRs1-10, RIG-I, and MDA5 in an MCC cell line, MKL-1, and human dermal fibroblasts (HDFs) (37, 38). We found that among these genes, STING is the one with the most dramatically reduced (over 3,000-fold) expression in MKL-1 cells compared to HDFs (Fig. 1A). RT-qPCR analysis confirmed that STING mRNA level is lower by over 3,000-fold in MKL-1 relative to HDF cells (Fig. 1B). We also analyzed STING protein levels in MCPyV⁺ MCC cells (MKL-1, MKL-2, MS-1), MCPyV⁻ MCC cells (MCC13, MCC26, UISO), and two primary HDF cell lines. While abundant STING protein is present in MCC13, MCC26, UISO, and HDFs, it is completely undetectable in all three MCPyV⁺ MCC cell lines: MKL-1, MKL-2, and MS-1 cells (Fig. 1C). Based on our observations in MCC cell lines, we probed for the presence of STING in MCC tumor lesions. Costaining for STING protein and the MCC marker CK20 allowed us to demarcate STING protein level in heterogeneous tumor cell populations. In all five of the MCC lesions examined, STING was specifically silenced in CK20⁺ MCC cells while maintaining robust expression in the surrounding CK20⁻ stromal cells (Fig. 1D and SI Appendix, Fig. S1).

Fig. 1.

STING is silenced in MCCs. (A) Differential expression of innate immune sensor genes in MKL-1 and HDFs calculated based on published RNA-seq data (37, 38). (B) The mRNA levels of the indicated innate immune sensors in MKL-1 and HDFs were measured by RT-qPCR and normalized to GAPDH mRNA levels. The levels in HDFs were set to 1. Error bars represent SEM of three independent experiments. (C) Whole-cell lysates of MCC cells and two primary HDF cell lines were immunoblotted using the indicated antibodies. GAPDH was used as a loading control. This experiment was performed three times, and similar results were obtained. (D) MCC tumor lesions were stained for CK20 (red) and STING (green) and counterstained with DAPI. The staining was performed on MCC lesions from five different patients. Shown are the representative images from two patients (*P < 0.05, **P < 0.01, ***P < 0.001). (Scale bars, 10 µm.)

STING silencing in MCC cells may explain their ability to escape immune eradication and T cell-associated checkpoint therapies. We hypothesized that low STING activity in MCC limits the cytokine production necessary for recruitment and activation of cytotoxic T cells. If that is true, reactivation of STING in MCCs may stimulate antitumor immune cytotoxicity.

Ectopic Expression of STING in MCC Cells Stimulates Downstream Cytokine Production.

To determine the functional impact of STING silencing in MCC cells and whether exogenous STING expression reactivates cytokine production, we transfected MKL-1 cells stably expressing RFP or STING with double-stranded DNA (dsDNA) and cGAMP (Fig. 2 and SI Appendix, Fig. S2). Compared to mock-transfected cells, both dsDNA and cGAMP moderately stimulated expression of STING downstream cytokines, such as IFNβ, CCL5, CXCL10, and IL-29, specifically in MKL-1/STING stable cells but not in the MKL-1/RFP cells (Fig. 2D and SI Appendix, Fig. S2). Expression of IRF3, which is not regulated by STING, was not affected by the treatment (Fig. 2D and SI Appendix, Fig. S2). To examine whether reactivating the STING pathway could be used as a therapeutic approach for MCC, we stimulated MKL-1/STING stable cells with small molecular compound etoposide or DMBA. RT-qPCR analysis showed that neither etoposide nor DMBA could significantly activate cytokine gene expression (SI Appendix, Fig. S3). These results suggest that while STING downstream components are functional in MCC cells, an alternative approach is needed to effectively reactivate STING in these cells.

Fig. 2.

Ectopic expression of STING in MCC cells reactivates downstream cytokine production. (A–C) Stable expression of RFP or STING in MKL-1 cells was confirmed by Western blotting (A), RT-qPCR (B), and immunofluorescence staining (C). (D) MKL-1 cells stably expressing RFP or STING were transfected with and without 1 kb of 1 µg/mL dsDNA. At 24 h posttransfection, the mRNA levels of the indicated genes were measured by RT-qPCR and normalized to GAPDH mRNA levels. The values for mock transfected RFP control cells were set to 1. Error bars represent SEM of three independent experiments (ns.: not significant,*P < 0.05, **P < 0.01, ***P < 0.001). (Scale bars, 10 µm.)

An Approach to Reactivate STING and Downstream Cytokine Production in MCC Cells.

Small molecular compound DMXAA (5,6-dimethylxanthenone-4-acetic acid) is a highly potent agonist of mouse STING (mSTING) but has no effect on human STING (hSTING) (39). Substitution mutations of S162A, G230I, or Q266I in hSTING confer strong affinity for DMXAA and render it highly sensitive to DMXAA treatment (40, 41). Thus, we established stable expression of STING^{S162A/G230I/Q266I} triple mutant in MKL-1 and MS-1 MCC cells maintaining natively silenced WT STING. Equivalent stable cells expressing RFP were produced as a negative control (Fig. 3 A–D). When MKL-1 or MS-1 cells stably expressing STING^{S162A/G230I/Q266I} were treated with DMXAA, we observed robust induction of IFNs as well as proinflammatory cytokines and chemokines compared to dimethyl sulfoxide (DMSO)-treated cells and the RFP stable control cells (Fig. 3 E and F). Expression of cytokines including IFNβ, CCL5, CXCL10, TNFα, IL-6, and IL-29 was increased up to 20,000-fold in DMXAA-treated STING mutant stable cell lines (Fig. 3 E and F), which is markedly higher than observed in MKL-1/WT STING stable cells transfected with either dsDNA or cGAMP (Fig. 2D and SI Appendix, Fig. S2). DMXAA also significantly stimulated PD-L1 expression in MCC/STING^{S162A/G230I/Q266I} stable cells (Fig. 3 E and F). The stimulated expression of these genes was also confirmed at the protein level for CXCL10, CCL5, and PD-L1 (Fig. 3 C and D). This inflammatory response is specific to the STING^{S162A/G230I/Q266I}–DMXAA interaction since the RFP control cell line elicited no response to DMXAA treatment. In addition, when IRF3, the key downstream mediator of the STING signaling pathway, was knocked out in MKL-1/STING^{S162A/G230I/Q266I} stable cells, DMXAA treatment showed significantly reduced cytokine production (SI Appendix, Fig. S4 A and B). This finding suggests that the DMXAA-induced cytokine production is largely mediated by canonical IRF3 transcriptional activity downstream of the STING signaling pathway.

Fig. 3.

Ectopic expression of STING^{S162A/G230I/Q266I} in MCC cells reactivates downstream cytokine production. (A and B) MKL-1 or MS-1 cells stably expressing RFP or STING^{S162A/G230I/Q266I} were stained for STING. (Scale bars, 10 µm.) (C–F) MKL-1 (C and E) or MS-1 cells (D and F) stably expressing RFP or STING^{S162A/G230I/Q266I} were treated with DMSO or DMXAA. At 24 h posttreatment, whole-cell lysates were immunoblotted using the indicated antibodies (C and D). GAPDH was used as a loading control. The experiments were performed three times, and similar results were obtained. The mRNA levels of the indicated genes were measured by RT-qPCR and normalized to GAPDH mRNA levels. The mRNA levels in DMSO-treated RFP control cells were set to 1. Error bars represent SEM of three independent experiments (ns.: not significant,*P < 0.05, **P < 0.01, ***P < 0.001).

The chemoattractants CCL5 and CXCL10 are crucial for promoting intratumoral infiltration of CD4⁺ and CD8⁺ T cells, whereas IFNβ, PD-L1, TNFα, and IL-6 are key immune modulators that can alter antitumor responses. Our results therefore suggest that STING^{S162A/G230I/Q266I} could be combined with DMXAA to induce CD8⁺ T cell infiltration and activation in MCC.

STING Reactivation Induces T Cell Migration and Activation In Vitro.

We next tested whether the cytokines induced by DMXAA in MCC cells stably expressing STING^{S162A/G230I/Q266I} could facilitate T cell migration and tumor cell killing. Since CXCL10 is one of the cytokines most highly induced by STING activation (Fig. 3 E and F), we measured its concentration in the culture medium of the MKL-1 stable cells after treatment with DMSO or DMXAA. Only DMXAA-treated MKL-1/STING^{S162A/G230I/Q266I} stable cells produced detectable amounts of CXCL10 protein in the medium (Fig. 4A). Because CXCL10 is a chemoattractant for CD8⁺ T cells (42), we tested whether the amount stimulated by DMXAA treatment was sufficient to promote T cell migration. We found that the CXCL10⁺ medium obtained from DMXAA-treated MKL-1/STING^{S162A/G230I/Q266I} stable cells significantly increased T cell migration in a transwell-based chemotaxis assay (Fig. 4B).

Fig. 4.

Reactivation of the STING pathway by DMXAA stimulation of STING^{S162A/G230I/Q266I} induces T cell migration, activation, and killing of MCC cells in vitro. (A) MKL-1 cells stably expressing RFP or STING^{S162A/G230I/Q266I} were treated with DMSO or 10 μg/mL DMXAA in T cell migration medium. At 72 h posttreatment, the CXCL10 level in the supernatant was measured using ELISA. (B) The media collected from cells described in A was used in a chemotaxis assay. Migrated αCD3/αCD28-stimulated primary CD8⁺ T cells were counted by hemocytometer. T cell migration medium served as the control group. (C and D) MKL-1/RFP/HLA-A2⁺ cells, MKL-1/STING^{S162A/G230I/Q266I}/HLA-A2⁺ cells, or the MKL-1 cell culture media (serving as a no-tumor cell control) were incubated with either DMSO or 10 μg/mL DMXAA for 3 h, and then mixed with HLA-A2 negative primary T cells stably expressing MCPyV or HIV TCR (SI Appendix, Fig. S5). The cell mixtures were incubated with or without 1,000 nM MCPyV TCR-specific peptide for 3 h. IFNγ and IL-2 mRNA expression levels were assessed by RT-qPCR. (E) MKL-1 cells stably coexpressing HLA-A2, CBG, and either RFP or STING^{S162A/G230I/Q266I} were seeded into the bottom wells of the transwell plate and treated with DMSO or DMXAA. A2-LTA_KLL– or A2-HIV–specific CD8⁺ T cells were cultured in the top wells. After 24 h, the remaining viable MKL-1 cells in the bottom well were lysed and luciferase activity was measured using a luciferase reporter system kit (Promega). Error bars represent SEM of three independent experiments (*P < 0.05, **P < 0.01, ***P < 0.001).

Since the cytokines and chemokines induced by the STING^{S162A/G230I/Q266I}–DMXAA interaction in MCCs have the potential to promote T cell migration (Figs. 3 E and F and 4B), we sought to determine if this approach also rescues T cell activation and killing of MCC cells in vitro. To generate effector T cells, we transduced primary human CD8⁺ T cells with the lentiviral vector expressing HLA-A2-MCPyV LT-specific T cell receptor (TCR), A2-LTA_KLL, which specifically recognizes a single MCPyV epitope (KLLEIAPNC, “KLL”) (26). Tetramer staining indicated that ≈30% of the transduced T cells were A2-LTA_KLL positive (SI Appendix, Fig. S5 A and B). In the same manner, we generated the CD8⁺ T cells expressing HLA-A2 restricted HIV-specific TCR, which recognizes HIV_Pol epitope IV9 (43), to serve as a negative control (SI Appendix, Fig. S5 A and B).

We tested the functional activities of each transduced T cell population in an in vitro killing assay by combining them with K562.A2 cells stably expressing GFP fused to the truncated LT antigen (LTT) normally expressed in MCC cells. Effective killing of the GFP-LTT–positive K562.A2 cells was detected for primary CD8⁺ T cells expressing HLA-A2–specific MCPyV TCR, but not untransduced primary CD8⁺ T cells, nor those expressing HIV TCR (SI Appendix, Fig. S5C). In addition, treatment with an MCPyV LT-specific peptide caused a dose-dependent induction of IFNγ and TNFα in the primary CD8⁺ T cells expressing A2-LTA_KLL MCPyV TCR, but not in the untransduced T cells (SI Appendix, Fig. S6). This result confirms that the CD8⁺ T cells expressing HLA-A2–specific MCPyV TCR are functional.

By establishing baseline functionality of these MCPyV and HIV peptide-targeting T cells, we could then measure the impact of MCC STING reactivation on their cytokine production and killing capacity. Because MKL-1 has almost no HLA class-I surface expression (44), we used lentivirus to introduce stable expression of HLA-A*02 (45) in MKL-1/STING^{S162A/G230I/Q266I} and MKL-1/RFP stable cells. To examine T cell activation, we mixed DMSO or DMXAA-treated MKL-1/STING^{S162A/G230I/Q266I}/HLA-A2 cells with primary T cells stably expressing MCPyV TCR or HIV TCR, and then monitor IFNγ and IL-2 produced by the T cells. MKL-1 and other MCC cells ectopically expressing HLA-A2 were known to stimulate very little activation of T cells expressing A2-LTA_KLL MCPyV TCR (46). This observation was confirmed in our experiment (Fig. 4 C and D). To increase the T cell response so we could better discern differences, MCPyV TCR-specific peptide was added to the cells to augment cognate pMHC on the cell surface. We found that DMXAA-treated MKL-1/ STING^{S162A/G230I/Q266I}/HLA-A2 cells were able to stimulate IFNγ and IL-2 production specifically in MCPyV TCR⁺ T cells but not in HIV TCR⁺ T cells (Fig. 4 C and D and SI Appendix, Fig. S7). In contrast, very little IFNγ and IL-2 were produced by the T cells in the presence of a control MKL-1 cell line expressing RFP instead of STING^{S162A/G230I/Q266I} or in the absence of the tumor cells (Fig. 4 C and D and SI Appendix, Fig. S7).

Since the treatment of MCC cells with STING^{S162A/G230I/Q266I} mutant and DMXAA can stimulate both T cell migration and activation, we employed a transwell killing assay to specifically measure the killing capacity of those T cells that have successfully migrated to the tumor cells (Fig. 4E). Expression of green click beetle luciferase (CBG) was introduced in MCC cells by a lentivirus to monitor T cell killing of the tumor cells. These cells were treated with DMXAA or DMSO in the bottom wells while A2-LTA_KLLc– or A2-HIV–specific CD8⁺ T cells were cultured in the top wells (Fig. 4E). We observed that only in MKL-1/STING^{S162A/G230I/Q266I} stable cells did DMXAA treatment induce killing compared to the DMSO control (Fig. 4E). Cotreatment of DMXAA and A2-LTA_KLL–specific CD8⁺ T cells resulted in subtly increased killing of MKL-1/STING^{S162A/G230I/Q266I} cells but not MKL-1/RFP cells (Fig. 4E). These findings support that DMXAA stimulation of STING^{S162A/G230I/Q266I} mostly drives LTA_KLL-specific CD8⁺ T cells migration toward the MCC cells, which could then contribute to the killing of cancer cells in vitro. More importantly, even in the absence of LTA_KLL-specific CD8⁺ T cells, DMXAA activation of STING^{S162A/G230I/Q266I} mutant could result in significant MCC cell death (Fig. 4E).

Activation of the STING Pathway Triggers MCC Cell Death.

DMXAA started to inhibit the proliferation of MKL-1 cells expressing STING^{S162A/G230I/Q266I} at 24 h posttreatment. We also detected vigorous cytokine production after MCC/STING^{S162A/G230I/Q266I} stable cells were treated with DMXAA (Figs. 3 E and F and 4A). When we continued to culture these cells in the presence of DMXAA, more than 99% of the cells were killed by 72 h (Fig. 5 A and B). In contrast, most of the MCC/RFP cells treated with DMXAA and MCC/STING^{S162A/G230I/Q266I} cells treated with DMSO remained healthy throughout the experiment (Fig. 5 A and B). Moreover, proliferation of MCC/STING^{S162A/G230I/Q266I} cells was inhibited by DMXAA in a dose-dependent manner (Fig. 5 C and D), indicating that cell death is directly linked to DMXAA treatment. These data illustrate that STING activity is toxic to MKL-1 cells and that prolonged induction of STING may lead to significant cell death in MCCs.

Fig. 5.

Activation of the STING pathway inhibits MCC cell proliferation. (A and B) MKL-1 (A) or MS-1 (B) cells stably expressing RFP or STING^{S162A/G230I/Q266I} were treated with DMSO or 10 µg/mL DMXAA. At 72 h posttreatment, cell viability was measured by Titer-GLO 3D cell viability assay. (C and D) MKL-1 (C) or MS-1 (D) cells stably expressing STING^{S162A/G230I/Q266I} were treated with DMSO or increasing doses of DMXAA. At 72 h posttreatment, cell viability was measured by Titer-GLO 3D cell viability assay. Error bars represent SEM of three independent experiments (*P < 0.05, ***P < 0.001).

Development of an AAV STING^{S162A/G230I/Q266I} Vector to Activate STING and Induce Cancer Cell Death.

AAV vectors have been approved by the US Food and Drug Administration for therapeutic applications. We therefore tested the potential of an AAV-based approach for MCC treatment. After screening AAV capsids for serotype 1, 2, 7, 8, 9n, rh10, Anc80L65, and AAV2-retrograde, we found that the AAV2-retrograde vector can infect nearly 100% of MKL-1 cells (Fig. 6A). We therefore generated an AAV2 to deliver the STING^{S162A/G230I/Q266I} mutant into MKL-1 and MS-1 MCC cells. Treatment with both STING^{S162A/G230I/Q266I} AAV and DMXAA robustly inhibited the proliferation of MKL-1 and MS-1 cells, whereas cells treated with STING^{S162A/G230I/Q266I} AAV in the presence of DMSO remained viable (Fig. 6B).

Fig. 6.

Development of AAV STING^{S162A/G230I/Q266I} expression vector to activate STING and inhibit cancer cell proliferation. (A) AAV-GFP (serotype AAV2-retrograde) virions were used to infect MKL-1 cells in DMEM/F12, EGF, and bFGF medium. After 4 d, the cells were analyzed using a fluorescent microscope. Shown are representative images from three independent experiments. (Scale bars, 10 µm.) (B) AAV-STING^{S162A/G230I/Q266I} virions were used to infect MKL-1 or MS-1 cells in DMEM/F12, EGF, and bFGF medium. At 2 d postinfection, the cells were treated with DMSO or DMXAA. At 72 h posttreatment, cell viability was measured by Titer-GLO 3D cell viability assay. (C) Lung cancer cell A549, cervical cancer cell HeLa, and midline carcinoma cell HCC2429 were infected with GFP or STING mutant AAV virions. The cells were treated and analyzed as in B. (D) Pancreatic cancer cells BxPC-3, PANC-1, and AsPC-1 were infected with GFP or STING mutant AAV virions and analyzed as in C. Error bars represent SEM of three independent experiments. (E) AsPC-1 cells stably expressing RFP or STING^{S162A/G230I/Q266I} were treated with DMSO or 10 μg/mL DMXAA. At 72 h posttreatment, the CXCL10 level in the supernatant was measured using ELISA. (F) The media collected from cells described in E were used in a chemotaxis assay. Migrated CD8⁺ T cells were counted by hemocytometer. T cell migration medium was used for the control group (*P < 0.05, **P < 0.01, ***P < 0.001).

Besides MCC cells, STING is also silenced in many other types of cancer cell lines, such as those derived from cervical cancer, midline carcinoma, and pancreatic cancers (SI Appendix, Fig. S8). In each of these cancer types, STING protein is either not detected or significantly lower than the level found in the normal primary fibroblasts (SI Appendix, Fig. S8). Importantly, treatment of STING-silent cancer cells with both STING^{S162A/G230I/Q266I} AAV and DMXAA killed almost all of the cancer cells by 72 h (Fig. 6 C and D). However, the same treatment spares most of the normal human cells such as HDFs (SI Appendix, Fig. S9A). Further analysis showed that the low toxicity observed in HDFs is likely because the AAVs transduce HDF at a much lower efficiency as compared to HeLa cells (SI Appendix, Fig. S9B).

DMXAA-induced STING activation is likely the cause of cancer cell death as the same cells treated with STING^{S162A/G230I/Q266I} AAV and DMSO displayed little viability change in the timeframe (Fig. 6 C and D). In addition, DMXAA did not induce significant cell death in cancer cells treated with AAVs encoding WT STING, which lacks the ability to respond to DMXAA (SI Appendix, Fig. S9A). However, IRF3 knockout only modestly rescued the cell viability (SI Appendix, Fig. S4C), indicating that the cell death is partially caused by a cytokine-independent mechanism(s), which could include STING-induced endoplasmic reticulum (ER) stress described in previous studies (47, 48).

Besides promoting significant cancer cell death, CXCL10 was induced after AsPC-1 pancreatic cancer cells were treated with both STING^{S162A/G230I/Q266I} AAV and DMXAA, but not when these components were added individually (Fig. 6E). Furthermore, the CXCL10-containing medium collected from DMXAA-treated AsPC-1/STING^{S162A/G230I/Q266I} stable cells could significantly stimulate T cell migration (Fig. 6F). These data recapitulating our earlier findings in MCC cells suggest that targeted delivery and activation of mutant hSTING could be a viable therapeutic strategy for many cancers in which STING is repressed.

Discussion

The immune-dampened status of highly aggressive MCC tumors presents a major obstacle to developing broadly effective immunotherapies. The mechanism driving this phenotype remains poorly understood. To probe characteristics that might contribute to immune evasiveness in MCCs, we examined the status of several known sensors and regulators of innate responses to pathogen- and damage-associated molecular patterns in MCC cell lines. We found that STING is silenced in both MCC cell lines and patient-derived tumor lesions. The absence of STING is relevant to the observed cases of “cold” MCC tumors since STING function normally inhibits tumor development (36, 49) and its activation stimulates strong antitumor immunity (50–52). We reasoned that silenced STING could be a major contributor to MCC immune escape and a prime candidate for therapeutic intervention.

To reignite STING activity in MCC, we introduced the hSTING mutant, STING^{S162A/G230I/Q266I}, which is highly sensitive to an mSTING agonist: DMXAA. Restoring STING activity by DMXAA in MCC cells greatly stimulated downstream antitumor cytokine production, T cell migration, and T cell activation in vitro. It also bolstered killing of the tumor cells by MCPyV-specific T cells. Our data lead us to propose that STING silencing enables MCC immune escape by limiting cytokine production, thereby impeding cytotoxic T cell infiltration, activation, and killing in the tumor microenvironment. Reintroducing STING functionality therefore represents a viable path to overcoming MCC immune resistance.

Still, dangers exist to systemic activation of innate immune regulators. Overstimulation of STING can cause antiproliferative effects and cell death in T cells and myeloid cells—at least in mouse tumor models (48, 53–56). Furthermore, STING gain-of-function mutations have been reported in multiple human genetic diseases, including STING-associated vasculopathy with onset in infancy, systemic lupus erythematosus-like syndromes, and familial chilblain lupus diseases (57–60). Unsurprisingly, traditional hSTING agonists can induce inflammatory disease and cancers (30, 61). In addition, favorable outcomes cannot be expected with hSTING agonists used in STING-silenced cancers. To avoid the detrimental effects of STING overstimulation and overcome the limitation of hSTING agonists, we have developed a strategy that exploits the responsiveness of STING^{S162A/G230I/Q266I} to DMXAA, an mSTING agonist that does not activate hSTING (39, 62, 63).

Our results illustrate that combined STING^{S162A/G230I/Q266I} and DMXAA treatment effectively represses cancer cell proliferation, without inhibiting T cell functional activity (Fig. 4). Indeed, the safety of DMXAA in humans has been well documented (64–66). Tumor-specific STING activation should be possible via local AAV-mediated delivery of STING^{S162A/G230I/Q266I} combined with DMXAA treatment. Because of its high in vivo gene delivery efficiency and low immunogenic potential, the AAV developed in this study affords an ideal vector for delivering STING^{S162A/G230I/Q266I} in vivo to introduce targetable STING. DMXAA presents little risk of systemic inflammatory activation or cytotoxicity in immune cells because it does not engage endogenous hSTING. Thus, our two-pronged strategy has potential to overcome the limitations of traditional hSTING agonist-based therapies.

A major barrier to effective MCC treatment is poor T cell intratumoral infiltration. DMXAA treatment of MCC cells expressing STING^{S162A/G230I/Q266I} exhibits great potential to overcome this hurdle. The combined factors highly induce chemoattractants such as CCL5 and CXCL10 for T cell recruitment and increase killing by migrated T cells (Figs. 3 and 4). Another characteristic of MCC that underlies poor patient outcomes is its rapid progression to nonresponding metastasis. Importantly, activation of STING^{S162A/G230I/Q266I} by DMXAA stimulates rapid and robust induction of cell death in MCC as well as several other STING-silenced cancers (Figs. 4–6). This is an exciting discovery because, in the in vivo setting, tumor antigens released by dying cancer cells could be engulfed by antigen-presenting cells (APCs) for cross-presentation and priming of tumor-specific T cells. In addition, tumor cell destruction could also generate DNA damage-associated molecular patterns (DAMPs) that could be detected by dendritic cells (DCs), macrophages, and natural killer (NK) cells to amplify the antitumor immune response (67). Together, these molecular events could generate capacity for distal adaptive responses that provide long-lived immunologic memory and protection against metastases. This tumoricidal effect of DMXAA-induced STING^{STINGS162A/G230I/Q266I} activation will be best tested in a humanized mouse model for MCC when it becomes available.

Currently, one of the major treatment modalities for advanced MCC is PD-1 and PD-L1 immune-checkpoint therapies. Still, over half of all patients either fail to respond to these treatments or experience highly adverse side effects (14–18). These clinical observations could be explained at least partially by our finding that STING, which plays a critical role in the efficacy of immune checkpoint therapies (36), is silenced in MCCs. In addition, a recent study observed a trend toward improved overall survival rates in patients with PD-L1–positive MCC tumors under PD-1 antibody treatment (68). PD-L1 expression is also considered a predictor of better outcomes in melanoma, renal cell carcinoma, nonsmall-cell lung carcinoma, metastatic colorectal cancer, and gastric cancer (69). We found that activation of the STING pathway by DMXAA significantly stimulates PD-L1 expression in MCC cells (Fig. 3). Our findings therefore suggest that combining DMXAA-mediated STING activation with PD-1/PD-L1 treatment may achieve synergistic antitumor activity that enhances the efficacy of checkpoint inhibitors in MCC.

Our discovery that STING is silenced in MCCs provides a explanation for MCC resistance to immune eradication and immune checkpoint blockade. We have developed an innovative approach to restore STING function for stimulating T cell tumor-infiltration/activation in MCC and potentially synergizing with existing immune checkpoint inhibitors to improve MCC tumor treatment. Our proposed approach can be applied to many other tumors in which STING is silenced (Fig. 6), or in other contexts where T cell infiltration proves challenging like chimeric antigen receptor T cell therapy.

Materials and Methods

Isolation of primary human dermal cells, cell culture, recombinant plasmid construction, generation of stable cell line, cell proliferation assay, Western blot analysis, AAV production, reverse transcription and qPCR, immunofluorescent staining, chemotaxis assay, ELISA, in vitro killing assay, tetramer staining, and intracellular cytokine staining were performed using standard methods. To generate CD8⁺ T cells stably expressing HIV- or MCPyV-TCR, primary CD8⁺ T cells from healthy donors were cultured with CD3/CD28-activating Dynabeads (Invitrogen) and transduced with lentiviral vectors encoding HIV- (70) or MCPyV-TCR gene (26). A full description of methods and reagents used in this study can be found in SI Appendix, Materials and Methods.

Data Availability Statement.

All data generated in this study are presented in the paper and SI Appendix.