Cutting Edge: Activation of STING in T Cells Induces Type I IFN Responses and Cell Death

STING expression and IFN production

We confirmed that murine T cells robustly expressed STING protein at levels comparable to, or higher than, macrophages (Fig. 1A) before testing their response to the STING-specific agonist DMXAA, which readily diffuses across the cell membrane and is a useful tool for experiments with primary T cells. Initially identified as an anti-vascular, pro-IFN cancer therapeutic (8), it was later shown to bind murine but not human (3) STING, activating the tank-binding protein 1 (TBK1)–IFN regulatory factor 3 (IRF3) axis and inducing IFN-I responses. Treatment with DMXAA induced STING-dependent expression of representative ISGs in B6 T cells (Fig. 1B). Because DMXAA is a small molecule and not a CDN, we also treated cells with thiol-modified CDN bisphosphorothioate analogue of cGAMP (R’S’cGAMP), which is known to activate macrophages without lipofection. Electroporation (Fig. 1D) with R’S’cGAMP robustly increased B6 T cell IFIT2 expression versus electroporation alone; lipofection had no effect (data not shown). Electroporation alone resulted in large amounts of cell death, so DMXAA was used in all subsequent experiments.

• Download figure
• Open in new tab
• Download powerpoint

FIGURE 1.

T cells exhibit a functional IFN-I response to STING agonists: (A) STING protein levels in unstimulated T cells and macrophages; ISG mRNA expression in T cells (B) after 4 h treatment with DMXAA, or (D) 5 h after electroporation with R’S’cGAMP. (C, E, and F) DMXAA-induced IFN-β and IFN-γ production, with and without anti-CD3 and anti-CD28, in total CD3⁺ or isolated CD4⁺ and CD8⁺ T cells. (G) TBK1-IRF3, NF_KB, and MAPK phosphorylation in DMXAA-activated T cells. Results are representative of three to four independent experiments.

In addition to increasing ISG expression, DMXAA induced STING-dependent IFN-β and IFN-γ production (Fig. 1C), a striking result because IFN-I production is not typically associated with T cells, whereas IFN-γ is a major T cell cytokine. Moreover, anti-CD3/CD28 activation substantially increased IFN-β versus DMXAA alone (Fig. 1E). Although both CD4⁺ and CD8⁺ T cells produced STING-dependent IFN-β, only the former secreted IFN-γ (Fig. 1F).

To our knowledge, together these data provide evidence that activation of STING in T cells elicits an IFN-I response.

T cell signaling in response to DMXAA

As reported in other cell types (9), DMXAA triggered TBK1 and IRF3 phosphorylation in B6 but not STING^−/− T cells; it also increased p-p38 and p-p65 (Fig. 1G). Changes in phosphorylation of the NF_KB inhibitory protein p105 have not been reported following STING activation, but because TCR activation triggers p105 phosphorylation and degradation to the active transcription factor p50 (10), we were curious whether DMXAA affected this pathway too. Although we detected an increase in p-p105 (Fig. 1G), there was no change in total p105 or p50 (Supplemental Fig. 1A). However, constitutive processing of p105 can obscure p50 detection even when it is transcriptionally active.

In contrast to macrophages, autocrine IFN appeared to have no effect on T cell STING signaling: T cells from mice lacking the IFN-I receptor (IFNAR^−/−) phenocopied B6 T cells (Supplemental Fig. 1B).

Thus, although STING activation in T cells triggers many of the same pathways described in innate cells, it also appears to have unique effects that might result in T cell–specific outcomes.

STING activation is independent of, but augmented by, simultaneous TCR stimulation

The effect of TLR ligands on T cells requires prior or simultaneous TCR stimulation (11, 12), so we investigated whether concurrent STING and TCR activation had synergistic effects on signaling. When we activated T cells with plate-bound anti-CD3 and anti-CD28 and added DMXAA, we found only DMXAA increased STING-dependent p-TBK1 or p-IRF3 (Supplemental Fig. 1C). Other NF_KB and MAPK responses overlapped, with apparent additive and possibly synergistic effects: p105 phosphorylation was weakly induced by TCR stimulation at 60 min in both B6 and STING^−/− cells but was stronger and more rapid in B6 cells when DMXAA was added. p-p38 was similarly enhanced by DMXAA beyond TCR alone only in B6 T cells. But although both B6 and STING^−/− T cells robustly increased p-ERK in response to anti-CD3 and -CD28, which in addition to the p-p65 findings indicated a functional TCR in STING^−/− T cells, no additive effect was detected with DMXAA and TCR activation (Supplemental Fig. 1C).

Characterization of STING^−/− mouse T cell compartment

STING^−/− and B6 T cells exhibited similar levels of p-ERK and p-p65 following TCR stimulation, suggesting no activation defects occur with STING deletion. However, in vivo infection and immunization studies have described altered T cell responses in STING^−/− mice, although this has been interpreted as an effect on APC that alters their interactions with T cells (13).

To rule out major differences in T cell development, we compared thymic and pLN CD4⁻CD8⁻, CD4⁺CD8⁺, and single positive CD4 and CD8 T cell populations, as well as levels of CD4 and CD8 TCR expression, and found no significant differences between B6 and STING^−/− mice (Supplemental Fig. 2A, 2C). We likewise found no differences in naive, memory, or regulatory T cell populations (Supplemental Fig. 2B, 2E), and B6 and STING^−/− T cells expressed comparable levels of CD69 and CD25 following TCR activation (Supplemental Fig. 2D). Coupled with the absence of TCR signaling differences, these data imply B6 and STING^−/− T cells are functionally equivalent.

Effect of DMXAA on T cell proliferation

Our signaling data suggested possible synergy between TCR and STING-mediated signaling in B6 T cells, so we next tested whether adding DMXAA to TCR stimulation altered T cell activation or proliferation. DMXAA had no effect on CD69 expression after 16 h but after 40 h we noted reduced B6 CD25 expression (Supplemental Fig. 2D); we suspect this is due to DMXAA-induced cell death (see below). Although pLN T cells from both strains proliferated equally in response to TCR stimulation, DMXAA blocked this expansion in B6 but not STING^−/− T cells (Fig. 2A). IFNAR^−/− T cells also failed to expand in the presence of DMXAA, dismissing any contribution of autocrine IFN-I (14).

• Download figure
• Open in new tab
• Download powerpoint

FIGURE 2.

DMXAA inhibits TCR-induced proliferation independent of IFN-I. (A) Anti-CD3 and anti-CD28 induced proliferation of B6, KO, and IFNAR^−/− CD3⁺ T cells in the presence of 10 μg/ml DMXAA. (B) Effect of adding DMXAA concurrent with TCR stimulation (day 0), or after 24 (day 1) or 48 (day 2) h. Results shown are representative of three independent experiments.

We then tested whether DMXAA added after initial TCR stimulation continued to block proliferation. DMXAA added on day 0 or day 1 completely inhibited T cell expansion, but if added on day 2 some early proliferation occurred (Fig. 2B). These findings suggested DMXAA might not alter early T cell activation but might initiate a STING-dependent anti-proliferative pathway.

STING-dependent activation of IFN, cell death, and ER stress pathways

To take an unbiased approach to STING activation in T cells, we performed RNA-sequencing on naive B6 and STING^−/− CD3⁺ T cells stimulated with anti-CD3 and anti-CD28, DMXAA alone or both. In agreement with our RT-PCR results, DMXAA alone increased expression of a suite of ISG by B6 but not STING^−/− T cells (Fig. 3C), with a particularly striking effect on IFN-α2, 4, and 5; transcripts that are not highly upregulated in myeloid cells by DMXAA. Adding TCR stimulation further increased expression of most ISGs, reinforcing the possibility of synergy between these pathways.

• Download figure
• Open in new tab
• Download powerpoint

FIGURE 3.

STING activation increases B6 T cell expression of apoptosis, cell stress-related genes in addition to dramatic increases in ISG expression. (A) Gene pathways activated in response to TCR or TCR+DMXAA stimulation as determined using RNA sequencing data clustered based on similarity of samples using GenomePattern software (Broad Institute). (B) Changes in T cell expression of anti- and proapoptotic and ER stress-related genes in T cells in response to TCR or TCR+DMXAA activation. (C) STING-dependent upregulation of IFNs and selected ISGs in T cells. (D) STING-dependent activation of proapoptotic pathways in T cells versus macrophages. TCR: anti-CD3 and anti-CD28. The RNA-sequencing experiment was done once.

Pathway analysis of these data unexpectedly highlighted STING-dependent increases in apoptotic and caspase cascade pathways and decreases in IL-2 and cell cycle pathways with DMXAA (Fig. 3A). Although TCR activation alone in B6 and STING^−/− T cells elicited comparable increases in anti-apoptotic (e.g., BCL2) and decreases in proapoptotic (e.g., BAX) gene expression, adding DMXAA reversed this trend in a STING-dependent manner (Fig 3B), substantially downregulating BCL2 and upregulating BAX only in B6 T cells. Strikingly, none of these prodeath pathways were activated by DMXAA in macrophages: comparison of pathway activation between macrophages and T cells showed greater similarities between STING^−/− T cells and B6 macrophages than STING^−/− and B6 T cells (Fig 3D). CDN-induced lethality has not been noted in innate immune cells and we accordingly saw no increase in proapoptotic gene expression by macrophages. We reasoned this might be because innate immune cells must survive STING-triggering threats to initiate further immune responses, but if STING is primarily detecting viral infection or internal DNA damage in T cells, the host might be better served by cell death.

Although STING-induced death is one explanation for the block we observed in B6 T cell proliferation, it might also be related to reduced expression of the cell cycle and IL2 pathways in response to DMXAA (Fig. 3A), with IL2 in particular having a critical role in T cell expansion. Both B6 and STING^−/− cells sharply increased IL2 expression with TCR activation, but DMXAA blocked this only in B6 cells (Fig. 3A).

Our pathway analysis unexpectedly revealed strong upregulation of genes involved in the unfolded protein (UPR) response, notably Bip/HSPA5 and GADD34, in parallel with ISG upregulation (Fig. 3B). The build-up of unfolded or misfolded proteins within the ER leads to ER stress and blocked protein translation that can kill the cell. UPR activation resolves this stress and prevents death by increasing the folding capacity of the ER (15, 16). TCR stimulation by itself can induce the UPR as activated cells ramp up protein production for proliferation; at the same time, TCR signaling induces Ca²⁺ flux and perturbation of ER Ca²⁺ concentration is known to chemically interfere with protein folding and trigger the UPR. Activation-induced translocation of STING from the ER membrane could further perturb the Ca²⁺ balance and overwhelm the capacity of the UPR to resolve stress, resulting in cell death.

STING activation triggers T cell death

Based on our RNA-sequencing data, we hypothesized that DMXAA might activate proapoptotic pathways in a STING-dependent manner, possibly by inducing a doomed UPR. To test this, TCR-activated T cells treated with DMXAA were stained with annexin V and propidium iodide (PI) to track cell death. The addition of DMXAA to anti-CD3 and anti-CD28 activated B6 T cells resulted in large numbers of annexin V⁺PI⁺ cells compared with their nonDMXAA counterparts but STING^−/− T cells were largely annexin V⁻PI⁻ regardless of DMXAA. We then treated cells with the pan-caspase inhibitor zVAD-fmk and/or the necroptosis inhibitor Nec1 to characterize the type of death (Fig. 4A). By itself neither was able to rescue DMXAA-treated B6 T cells but together they partially rescued STING-dependent T cell death, suggesting that denying DMXAA-treated T cells the apoptotic pathway may instead force death by necroptosis. This rescue appeared to be limited by the half-life of the inhibitors: neither zVAD, Nec1, nor a combination of the two rescued B6 cell proliferation when DMXAA was present (Fig. 4B).

• Download figure
• Open in new tab
• Download powerpoint

FIGURE 4.

DMXAA induces a STING-dependent ER stress response and cell death. zVAD and Nec1 together rescue DMXAA (10 μg/ml)-induced cell death after 24 h (A), but not proliferation at 72 h (B), in anti-CD3 and anti-CD28–activated T cells. (C) Cytotoxicity in CD3⁺ T cells of *DMXAA alone after 12 and 24 h is dose dependent. (D) The DMXAA-induced, STING-dependent increase in IFIT2 mirrors (E) the increased ratio of XBP-1s:XBP-1u and (F) decreased expression of anti-apoptotic BCL2 following DMXAA treatment. Results are representative of three independent experiments. (G) In vivo DMXAA treatment has no effect on pLN CD4⁺ and CD8⁺ T cell populations but (H) increases CD3⁺ T cell ISG expression independent of non-T cell IFN sources. Symbols represent individual mice.

To clarify whether DMXAA-induced cell death requires simultaneous TCR activation (17) we tested the effect of DMXAA alone at varying doses. Naive B6 T cells were largely dead after 12 h treatment with 10 μg/ml DMXAA, whereas 5 μg/ml DMXAA induced an equal degree of cell death after 24 h. DMXAA concentrations below 5 μg/ml had no effect on viability even after 24 h, whereas STING^−/− T cells showed no increase in cell death at any dose or either time point (Fig. 4C).

We next investigated whether there is a link between STING, the UPR, and T cell death. In macrophages IFN production following PRR activation of the UPR requires activation of the transcription factor XBP-1 by unconventional splicing (18), so we examined whether STING activation by DMXAA induced XBP-1 splicing as part of the UPR in T cells. CD3⁺ T cells from B6 and STING^−/− mice expressed comparable levels of unspliced XBP-1 (XBP-1u) that were unchanged by TCR and DMXAA activation, but DMXAA alone increased expression of spliced XBP-1 (XBP-1s) only in B6 T cells, resulting in a higher ratio of XBP-1s:XBP-1u that grew with the addition of anti-CD3 and anti-CD28 activation (Fig. 4E). Concurrent with this STING-dependent increase in the ratio of XBP-1s:XBP-1u was increased expression of IFIT2 and decreased expression of BCL-2 in B6 cells (Fig. 4D, 4F), confirming our RNA-sequencing data showing a correlation between increased ISG and UPR and decreased anti-apoptotic gene expression. Together these data support a link between STING activation in T cells, UPR induction, and subsequent cell death, possibly due to failed ER stress resolution.

Despite the dramatic evidence of ER stress linked T cell death in vitro, when mice were injected i.v. with DMXAA on two successive days we saw no changes in pLN or spleen-naive and memory T cell populations indicative of cell death (Fig. 4G, data not shown). However, we did note dramatic increases in ISG expression in response to DMXAA. To ensure these changes were independent of IFN-I production from non-T cells, we transferred STING^+/+ CD3⁺ T cells expressing CD45.1 into STING^−/− recipients that express CD45.2. After DMXAA administration only CD3⁺CD45.1⁺ T cells exhibited increased ISG expression (Fig. 4H). The absence of T cell death in our in vivo model could be due to a dosage effect, with T cells being exposed to a lower local concentration of DMXAA than in our in vitro experiments. Alternatively, T cells may be receiving survival signals from other cell populations in vivo that are unavailable in vitro.

We describe in this study two novel findings: first, analogous to its effects in innate immune cells, the STING pathway induces a IFN-I response in T cells. This adds to a growing body of evidence that PRR are functional in T cells, though unlike other PRR our data uniquely shows a TCR activation–independent response to STING agonists. In addition, we report STING-dependent UPR induction and cell death in T cells that has not been described in myeloid cells.

The functionality and role of PRR in T cells has been a subject of study and some debate for many years. T cells are known to express various PRR that provide costimulatory signals (19, 20), but TCR-independent cytosolic DNA-sensing could be uniquely useful to allow surveillance of infection or DNA damage in T cells and enable self-destruction, preventing infected or damaged and potentially cancerous T cells from proliferating and causing disease. These findings complement a recent report by Tang et al. (21) demonstrating CDN-induced apoptosis in B cells and B cell lymphomas. But although they concluded this response was B cell–specific because no effect on T cells was observed with in vitro cGAMP treatment, our DMXAA and R’S’cGAMP experiments conclusively demonstrate T cells are responsive to STING activation.

Why the T cell response to STING activation in vitro results in cell death but in vivo leads to IFN-I responses requires further investigation.

These initial findings caution that careful study of STING agonist effects on T cells will be necessary whenever these agonists are tested for therapeutic effects. Boosting T cell IFN-I production with STING agonists, particularly when chimeric Ag receptor T cells are transferred to patients, could dramatically improve anti-tumor responses. Likewise, the inclusion of a DMXAA analog as a vaccine adjuvant may be particularly useful when the desired outcome is an IFN-dominated response. However, more long-term effects of STING activation on T cells remain to be studied. B cell lymphoma death in response to injected cGAMP was studied after weeks of multiple injections (21) and a similar timeline might result in T cell stress and death more similar to our in vitro experiments instead of the IFN-I response we found in our short-term in vivo experiment.

Overall, the work presented in this study strongly suggests that STING has its own unique functional outcomes in an adaptive immune cell type, and it may yet have tricks up its sleeve awaiting our discovery.