Research ArticleCancer

cGAS-STING–mediated DNA sensing maintains CD8+ T cell stemness and promotes antitumor T cell therapy

See allHide authors and affiliations

Science Translational Medicine  24 Jun 2020:
Vol. 12, Issue 549, eaay9013
DOI: 10.1126/scitranslmed.aay9013

Intuitive CD8+ T cells sense DNA

Stimulator of interferon genes (STING) agonism is an area of active exploration for cancer immunotherapy. Li et al. examined the cGAS-STING DNA sensing cascade in antitumor CD8+ T cells. They observed dampened STING activity in CD8+ T cells from patients with cancer or mice implanted with tumors. STING signaling supported a stem-like memory phenotype in the T cells, which is known to be beneficial for responses to immunotherapy. Cytosolic DNA was enriched in activated T cells, and STING agonism improved efficacy of adoptive cell therapy in multiple mouse models. These results highlight that CD8+ T cell DNA sensing could be exploited for therapeutic benefit in immunotherapy.

Abstract

Although cGAS-STING–mediated DNA sensing in tumor cells or phagocytes is central for launching antitumor immunity, the role of intrinsic cGAS-STING activation in T cells remains unknown. Here, we observed that peripheral blood CD8+ T cells from patients with cancer showed remarkably compromised expression of the cGAS-STING cascade. We demonstrated that the cGAS-STING cascade in adoptively transferred CD8+ T cells was essential for antitumor immune responses in the context of T cell therapy in mice. Mechanistically, cell-autonomous cGAS and STING promoted the maintenance of stem cell–like CD8+ T cells, in part, by regulating the transcription factor TCF1 expression. Moreover, autocrine cGAS-STING–mediated type I interferon signaling augmented stem cell–like CD8+ T cell differentiation program mainly by restraining Akt activity. In addition, genomic DNA was selectively enriched in the cytosol of mouse CD8+ T cells upon in vitro and in vivo stimulation. STING agonism enhanced the formation of stem-like central memory CD8+ T cells from patients with cancer and potentiated antitumor responses of CAR-T cell therapy in a xenograft model. These findings advance our understanding of inherent cGAS-STING activation in T cells and provide insight into the development of improved T cell therapy by harnessing the cGAS-STING pathway for cancer immunotherapy.

INTRODUCTION

Cyclic guanosine monophosphate (GMP)–adenosine monophosphate (AMP) synthase (cGAS) is a central cytoplasmic DNA sensor that catalyzes the generation of the second messenger cyclic GMP-AMP (cGAMP), which binds to stimulator of interferon genes (STING), an adaptor protein on the endoplasmic reticulum membrane (1, 2). STING, in turn, traffics to the Golgi body and interacts with the downstream TANK-binding kinase 1 (TBK1) and interferon regulator factor 3 (IRF3) (1, 2). The nuclear translocation of activated IRF3 leads to the induction of type I interferons (IFNs) (1, 2). The cGAS-STING pathway has emerged as a dominant pathway that responds to cytosolic DNA in the context of tumor immunity, cellular senescence, and inflammatory diseases (1). Nucleic acid recognition by cGAS bridges antitumor immunity and tumor cell DNA damage provoked by radiotherapy or poly (adenosine diphosphate–ribose) polymerase (PARP) inhibition (36). Likewise, the cGAS-STING pathway in phagocytes dictates antitumor adaptive immunity upon “do-not-eat-me” signaling CD47 blockade or LC3-associated phagocytosis inhibition (7, 8). So far, it has been considered that cGAS-STING pathway, as a promising therapeutic target, is capable of improving current cancer immunotherapy regimens by turning immunologically “cold” to “hot” tumor (1, 9, 10). Gain-of-function STING mutation either causes impaired proliferation of T lymphocytes from individuals carrying a constitutively active STING mutation or results in decreased survival of T cells from mutant STING knock-in mice due to mitotic errors and calcium homeostasis disruption, respectively (11, 12). Consistently, pharmacological hyperactivation of STING by the prolonged exposure of agonists harms T cells through the induction of apoptosis (13). However, the role of naturally cell-inherent cGAS-STING activation in T cells in tumor environments remains poorly understood. Therefore, understanding the direct role of cGAS-STING pathway in T cells will support the rational design of new therapeutic strategies.

T cell–based immunotherapy encounters multiple challenges in eradicating solid tumors (1416). Undoubtedly, unique targetable tumor antigens are a prerequisite for successful T cell therapy (17). Presumably, it is impracticable to reinvigorate terminally exhausted T cells hallmarked with the loss of robust effector functions and the expression of multiple inhibitory receptors (18). Thus, improving T cell quality including persistence and reprogramming has become a priority (16). Manipulation of the differentiation of T cell memory while avoiding exhaustion is a promising avenue to enhance the efficacy of T cell therapy. For instance, addition of 4-1BB signaling domains or a single immunoreceptor tyrosine-based activation motif enhanced central memory differentiation of CAR-T cells, leading to persistence and reduced exhaustion (19, 20). Similarly, insertion of Janus kinase (JAK)–signal transducer and activator of transcription (STAT) signaling domain or silence of NR4A transcription factors prevented terminal differentiation of CAR-T cells with superior antitumor effects (2123). Therefore, it is attractive to define the potential targets available for T cell reprogramming with longer persistence.

Stem cell–like T cells, characterized by self-renewal, expansion, and multipotent capacity, are critical for achieving durable antitumor responses during T cell therapy and immune checkpoint blockade (24). A T cell subpopulation expressing the transcription factor TCF1 (encoded by Tcf7) has the properties of stem cell–like T cells (2528). Stem cell–like exhausted PD-1+CD8+ T cells, but not terminally exhausted PD-1+ CD8+ T cells, are responsible for antitumor effects of immune checkpoint blockade or therapeutic vaccines (2830). Understanding the mechanism underlying the generation and maintenance of stem cell–like T cells offers an avenue to optimize T cell therapy and immune checkpoint blockade. For example, c-Myb overexpression promoted CD8+ T cell stemness and antitumor immunity through restraining T cell differentiation (31). In addition, extracellular potassium treatment preserves CD8+ T cell stemness before transfer and, in turn, sustains CD8+ T cell persistence (32). However, not much is known about intrinsic nontranscriptional regulators of T cell stemness in the tumor. Here, we investigated the role of the cGAS-STING pathway within antitumor T cells.

RESULTS

Deficient cGAS-STING cascade in CD8+ T cells interferes with antitumor immunity

To interrogate the expression profile of the cGAS-STING pathway signaling in circulating T cells of human peripheral blood, we isolated CD8+ T cells from peripheral blood of patients with cervical cancer, ovarian cancer, or endometrial cancer. The expression of cGAS was hardly detected in peripheral blood CD8+ T cells of patients with cancer (Fig. 1, A and B). Likewise, the expression of STING was greatly diminished in peripheral blood CD8+ T cells of patients with cancer compared to those of healthy volunteers, whereas TBK1 expression was largely decreased in CD8+ T cells of some patients (Fig. 1, A and B). To confirm our initial observation, we intracellularly stained additional samples for the detection of STING in T cells by flow cytometry. The expression of STING in peripheral blood CD8+ T cells was significantly decreased in patients with cancer with stratification on tumor types compared to healthy volunteers (P = 0.0001, 0.0009, and 0.0352, respectively) (Fig. 1C). Meanwhile, the expression of STING in peripheral blood CD4+ T cells was also greatly down-regulated in patients with cancer with stratification by tumor types compared to healthy volunteers (Fig. 1D). In addition, the decrease of STING expression was observed in CD14+ monocytes, but not in CD19+ B cells, in peripheral blood of patients with cancer versus healthy volunteers (fig. S1, A and B). Together, these results indicate that the cGAS-STING cascade in circulating T cells is substantially impaired in patients with cancer. These findings raise the possibility that attenuation of the cGAS-STING cascade in circulating T cells from patients with cancer may influence antitumor responses of T cell therapy.

Fig. 1 cGAS-STING cascade in CD8+ T cells dictates antitumor effects of T cell therapy.

(A and B) Representative immunoblots (A) and quantification (B) of cGAS, STING, and TBK1 expression in CD8+ T cells from peripheral blood of healthy volunteers (n = 8) and patients with cancer (n = 9), respectively. (C) Representative histogram (left) and mean fluorescence intensity (MFI) summary (right) showing STING expression in CD8+ T cells from peripheral blood of healthy volunteers (n = 10) and patients [cervical cancer (n = 7), ovarian cancer (n = 10), and endometrial cancer (n = 6)]. (D) Representative histogram (left) and MFI summary (right) showing STING expression in CD4+ T cells from peripheral blood of healthy volunteers (n = 10) and patients [cervical cancer (n = 7), ovarian cancer (n = 10), and endometrial cancer (n = 6)]. (E) Experimental schematic of B16-SIY tumor growth after 2C-CD8+ T cell therapy. s.c., subcutaneous; i.v., intravenous. (F) B16-SIY tumor growth after the therapy with 2C-CD8+ T cells or cGAS−/− 2C-CD8+ T cells. (G) B16-SIY tumor growth after the therapy with 2C-CD8+ T cells or STING−/− 2C-CD8+ T cells. (H) EG7 tumor growth in CD4Cre+-STINGflox/flox and WT mice. (I) GL261 tumor growth in CD4Cre+-STINGflox/flox and WT mice. Representative data are shown from two (A and B, H and I) and three (F and G) independent experiments, (F to I) conducted with three to seven mice per group. Data are represented as means ± SEM. Each dot represents one donor in (C) and (D). (B), (H), and (I) were analyzed with unpaired Student’s t test. (C), (D), and (F) were calculated with one-way ANOVA with Bonferroni’s multiple comparison tests. (G) was analyzed with Kruskal-Wallis test and Dunn’s multiple comparisons test. *P < 0.05, **P < 0.01, ***P < 0.001, and ****P < 0.0001.

To examine whether cGAS is necessary to maintain the function of adoptively transferred tumor-specific T cells, we crossed cGAS knockout (KO) mice with 2C mice (in which CD8+ T cells specifically recognize SIY peptide). Accumulating evidence has shown that stem cell–like T cells are superior to terminal effector T cells in terms of tumor destruction due to much longer T cell persistence and greater proliferative potential (28, 29). Thus, we activated CD8+ T cells with anti-CD3/CD28 in the presence of interleukin-15 (IL-15) and IL-7 (for induction of stem cell–like T cells) before transfer as shown in experimental schematic (Fig. 1E). B16-SIY tumor growth was significantly delayed after activated 2C-CD8+ T cell therapy compared to nontreatment group (P = 0.0003). In contrast, the loss of cGAS in adoptively transferred activated 2C-CD8+ T cells resulted in impaired antitumor effects, which were equivalent to antitumor effects in the untreated group (Fig. 1F). To further investigate whether STING in infused T cells is required for restricting tumor growth, we crossed STING KO mice with 2C mice. Consistently, deficiency of STING in adoptively transferred activated 2C-CD8+ T cells significantly abolished antitumor effects (P = 0.011) (Fig. 1G). These results indicate that cGAS and STING in infused CD8+ T cells are pivotal for the antitumor efficacy of T cell therapy. In addition, these findings raise a possibility that normalization of the cGAS-STING cascade is feasible for improving therapeutic efficacy of engineered human T cells in cancer treatment.

To validate whether the engagement of cGAS-STING pathway affects multiple T cell receptor (TCR) clonotypes, we crossed CD4Cre+ mice with STINGflox/flox mice to accurately deplete STING in T cells. In naïve CD4Cre+-STINGflox/flox mice, about 80% of STING was depleted in splenic CD8+ T cells compared to that in wild-type (WT) mice (fig. S2). We established two immunogenic tumor models, EG7 and GL261, in CD4Cre+-STINGflox/flox mice. In agreement with adoptive transfer experiments, conditional depletion of STING in T cells accelerated tumor progression in both EG7 and GL261 models (Fig. 1, H and I). To rule out the possibility that the deficiency of cGAS-STING pathway may induce developmental defects in CD8+ T cells, we examined T cell percentages in the thymus and spleen of WT, cGAS−/−, STING−/−, STINGgt/gt, IFNAR1−/−, STINGflox/flox, and CD4cre+-STINGflox/flox mice. We found that both CD8+ T cell and CD4+ T cell percentage in the KO mice were comparable to those in WT mice in the thymus and spleen (fig. S3, A to D), indicating that normal thymopoiesis and T cell homeostasis occur in the mice with the deficiency of cGAS-STING pathway. Together, these results indicate that cGAS-STING pathway generally regulates antitumor T cell responses in multiple tumor models.

The cGAS-STING cascade increases the expansion and cytokine production of therapeutic CD8+ T cells

To assess the impact of cGAS and STING on adoptively transferred T cells, we first adoptively transferred Thy1.1+ 2C-CD8+ T cells into the host and measured the proportion of Thy1.1+ in total CD8+ T cells in B16-SIY tumors after therapy as shown in experimental schematic (Fig. 2A). The absence of cGAS resulted in decreased expansion of Thy1.1+CD8+ T cells in the tumor after therapy (Fig. 2B). Likewise, STING deficiency impeded the expansion of Thy1.1+CD8+ T cells in the tumor after therapy (Fig. 2C). Next, to test whether the impaired accumulation of Thy1.1+CD8+ T cells in the absence of cGAS is due to decreased proliferative potential after transfer, we analyzed Ki-67 expression gated on PD-1+CD44+Thy1.1+CD8+ T cells in the tumor after therapy. Ki-67 expression in cGAS-deficient PD-1+CD44+Thy1.1+CD8+ T cells was much lower than in cGAS-sufficient PD-1+CD44+Thy1.1+CD8+ T cells in the tumor (Fig. 2D). To validate whether STING is required for the proliferative potential of CD8+ T cells, we adoptively transferred STING-deficient 2C-CD8+ T cell into B16-SIY–bearing mice. Consistently, STING deficiency resulted in decreased expression of Ki-67 in PD-1+CD44+Thy1.1+CD8+ T cells in the tumor after therapy (Fig. 2E). These results suggest that cGAS-STING cascade is required for the expansion of adoptively transferred CD8+ T cells.

Fig. 2 The cGAS-STING cascade increases the expansion and function of therapeutic CD8+ T cells.

(A) Schematic of analysis of tumor-bearing mice after 2C-CD8+ T cell therapy. Days 10 to 14 after T cell therapy, tumors and tumor-draining lymph nodes (TDLNs) were analyzed. (B) Representative flow plots (left) and quantification (right) of the percentage of 2C-CD8+ T cells and cGAS−/− 2C-CD8+ T cells in total CD8+ T cells in the tumor after therapy. (C) Representative flow plots (left) and quantification (right) of the percentage of 2C-CD8+ T cells and STING−/− 2C-CD8+ T cells in total CD8+ T cells in the tumor after therapy. (D) Representative flow plots (left) and quantification (right) of the percentage of Ki-67+ cells in 2C-CD8+ T cells and cGAS−/− 2C-CD8+ T cells in the tumor after therapy. (E) Quantification of the percentage of Ki-67+ cells in 2C-CD8+ T cells and STING−/− 2C-CD8+ T cells in the tumor after therapy. (F) Representative flow plots (left) and quantification (right) of IFN-γ+ and IFN-γ+TNF-α+ cells in 2C-CD8+ T cells and cGAS−/− 2C-CD8+ T cells in the TDLN after therapy. (G) Quantification of IFN-γ+ and IFN-γ+TNF-α+ cells in 2C-CD8+ T cells and STING−/− 2C-CD8+ T cells in the TDLN after therapy. Representative data are shown from three independent experiments conducted with three to six mice per group. Data are represented as means ± SEM. All data in this figure were analyzed with unpaired Student’s t test. *P < 0.05, **P < 0.01, and ***P < 0.001.

To delineate the role of the cGAS-STING pathway in T cell function, we analyzed cytokine production of adoptively transferred 2C-CD8+ T in tumor-draining lymph nodes (TDLNs) by intracellular staining. When T cells differentiate into terminal effectors, they lose the capacity to produce tumor necrosis factor (TNF) but still maintain the production of IFN-γ (33). The absence of cGAS resulted in the decrease of IFN-γ production in Thy1.1+CD8+ T cells in TDLNs (Fig. 2F). Similarly, the percentage of Thy1.1+CD8+ T cells producing both TNF-α and IFN-γ was also reduced in TDLNs because of cGAS deficiency (Fig. 2F). STING deficiency resulted in decreased frequency of IFN-γ+Thy1.1+CD8+ T cells and IFN-γ+TNF-α+Thy1.1+CD8+T cells in TDLNs (Fig. 2G). Together, these results reveal that cGAS and STING in CD8+ T cells are necessary for sustaining proliferative potential and prolonged functions of CD8+ T cells, thereby enabling the enhanced destruction of established tumors.

The cGAS-STING cascade promotes the maintenance of stem cell–like CD8+ T cells after T cell therapy

Recent studies have shown that stem cell–like exhausted T cells instead of terminally exhausted T cells account for antitumor effects of immune checkpoint blockade (2830). To address whether cGAS is required for the maintenance of stem cell–like CD8+ T cells, we adoptively transferred IL-15–stimulated 2C-CD8+ T cells into mice bearing established B16-SIY tumors and analyzed the features of 2C-CD8+ T cells in TDLNs and tumors. First, we dissected memory subsets and effector subsets among Thy1.1+CD8+ T cells in TDLNs with CD62L and CD44, a lymphoid homing marker and a memory marker, respectively. The central memory subset is characterized with stronger self-renewal and longer persistence compared to other subsets. We found that cGAS deficiency resulted in reduction of the central memory–like subset (CD62L+CD44+) and accumulation of the effector subset (CD62LCD44) in adoptively transferred 2C-CD8+ T cells in TDLNs after therapy (Fig. 3A). Likewise, cGAS deficiency dampened the formation of central memory–like subset (CD62L+CD44+) in adoptively transferred 2C-CD8+ T cells in the tumor after therapy (Fig. 3B). Furthermore, the loss of STING in therapeutic CD8+ T cells led to reduced formation of central memory–like subset (CD62L+CD44+) in adoptively transferred 2C-CD8+ T cells in the tumor after therapy (Fig. 3C). These results suggest that intrinsic cGAS-STING cascade is responsible for retention of CD8+ T cell central memory–like phenotype.

Fig. 3 The cGAS-STING cascade promotes the maintenance of stem cell–like CD8+ T cells after T cell therapy.

(A) Representative flow plots (left) and quantitative data (right) of the percentage of CD44+CD62L+ and CD44CD62L in 2C-CD8+ T cells and cGAS−/− 2C-CD8+ T cells in the TDLN after therapy. (B) Quantitative data of the percentage of CD44+CD62L+ subset in 2C-CD8+ T cells and cGAS−/− 2C-CD8+ T cells in the tumor after therapy. (C) Quantitative data of the percentage of CD44+CD62L+ subset in 2C-CD8+ T cells and STING−/− 2C-CD8+ T cells in the tumor after therapy. (D) Representative flow plots (left) and quantitative data (right) of PD-1 expression in 2C-CD8+ T cells and cGAS−/− 2C-CD8+ T cells in the tumor after therapy. (E) Quantitative data of PD-1 expression in 2C-CD8+ T cells and STING−/− 2C-CD8+ T cells in the tumor after therapy. (F) Quantitative data of the percentage of CD62L+ in PD-1+CD44+Thy1.1+ CD8+ T cells with or without cGAS in the tumor after therapy. (G) Quantitative data of the percentage of CD62L+ in PD-1+CD44+Thy1.1+ CD8+ T cells with or without STING in the tumor after therapy. (H and I) Representative histogram (left) and MFI summary (right) show TCF1 (H) and Slamf6 (I) expression in PD-1+CD44+TIM3lowThy1.1+CD8+ T cells with or without cGAS in the tumor after therapy. (J and K) Representative histogram (left) and MFI summary (right) show TCF1 (J) and Slamf6 (K) expression in PD-1+CD44+TIM3lowThy1.1+CD8+ T cells with or without STING in the tumor after therapy. Representative data are shown from three independent experiments conducted with three to six mice per group. Data are represented as means ± SEM. All data in this figure were analyzed with unpaired Student’s t test. *P < 0.05, **P < 0.01, and ***P < 0.001.

To rule out the possibility that the deficiency of cGAS and STING may have a developmental defect in spontaneous memory formation, we analyzed the memory phenotype of CD8+ T cells in spleens from naïve WT, cGAS−/−, and STING−/− mice, respectively. The percentage of central memory CD8+ T cells was equivalent among WT, cGAS−/−, and STING−/− mice (fig. S3E). We found that tumor-infiltrating transferred Thy1.1+CD8+ T cells had a higher frequency of PD-1 expression in the absence of cGAS or STING, validating that cGAS/STING deficiency leads to impaired functions of T cells after transfer (Fig. 3, D and E). We found that CD62L expression in PD-1+CD44+ 2C-CD8+ T was greatly down-regulated in the absence of cGAS and STING in tumor after therapy (Fig. 3, F and G). To further assess whether cGAS enables sustainment of stem cell–like exhausted features, we examined expression of TCF1, a transcription factor marker of stem cell–like T cells, and Slamf6, a newly identified cell surface marker that distinguishes progenitor exhausted CD8+ T cells from terminally exhausted CD8+ T cells (29). We analyzed tumor-infiltrating PD-1+CD44+TIM3low population of 2C-CD8+ T cells after therapy with the gating strategy shown in fig. S4. Recently, stem cell–like CD8+ T cells have been recognized as progenitor exhausted CD8+ T cells in the tumor, whereas several studies have shown that stem cell–like CD8+ T cells express intermediate amounts of PD-1 (28, 34, 35). We found that either the expression of TCF1 or the coexpression of TCF1 and Slamf6 was higher in PD-1+ populations than that in PD-1 populations (fig. S5, A and B), indicating that stem cell–like CD8+ T cells are mainly distributed in PD-1+ populations in our system. Thus, we examined the T cell stemness in PD-1+CD44+TIM3lowThy1.1+CD8+ T cell population. We observed that cGAS loss led to reduced mean fluorescence intensity (MFI) of TCF1 and Slamf6 gated on PD-1+CD44+TIM3lowThy1.1+ CD8+ T cell population (Fig. 3, H and I), indicating that cGAS presence in CD8+ T cells efficiently retains stem cell–like T cell properties. We next sought to assess whether STING is required for the maintenance of stem cell–like T cells in the tumor after transfer. STING deficiency resulted in the decreased intensity of TCF1 and Slamf6 gated on PD-1+CD44+TIM3lowThy1.1+CD8+ T cells in the tumor (Fig. 3, J and K), revealing that STING is also involved in the maintenance of stem cell–like T cell behaviors. To further confirm that TCF1 is a crucial T cell stemness marker in our system, we adoptively transferred 2C-CD8+ T cells in which TCF1 was reduced by short hairpin RNA (shRNA) knockdown. TCF1 was efficiently down-regulated in CD8+ T cells by shRNA targeting Tcf7 (fig. S6, A and B). Consistent with published literature (26), our data showed that the decrease of TCF1 resulted in lower proliferation potential (Ki-67) in the tumor and reduced frequency of adoptively transferred of 2C-CD8+ T cells in TDLNs (fig. S6, C and D). In addition, the decrease of TCF1 led to reduced expression of stemness markers (Slamf6 and CXCR5) (fig. S6, E and F), validating that TCF1 is an important functional marker of CD8+ T cell stemness in our systems. Collectively, these results indicate that cGAS-STING cascade is required for specific steps toward stem cell–like status in the differentiation program of CD8+ T cells upon chronic stimulation of tumor antigen in vivo, presumably leading to the formation of a central memory–like subset.

Type I IFN signaling facilitates the maintenance of stem cell–like CD8+ T cells after T cell therapy

To address whether the canonical downstream signaling of cGAS/STING is activated in CD8+ T cells, we collected CD8+ T cells stimulated with anti-CD3/CD28 at different time points to detect TBK1 phosphorylation (p-TBK1). The expression of p-TBK1 was greatly increased in CD8+ T cells upon stimulation over time, whereas the increase of p-TBK1 was compromised when either cGAS or STING was deficient (Fig. 4A). To validate whether T cell activation is accompanied with type I IFN induction, we stimulated isolated CD8+ T cells with anti-CD3/CD28, which led to the production of IFN-β in a cGAS/STING-dependent manner (Fig. 4B). These findings suggest that cGAS-STING cascade is responsible for type I IFN induction in CD8+ T cells upon TCR stimulation. To assess the impact of type I IFN signaling on therapeutic T cells, we adoptively transferred anti-IFNAR1–treated 2C-CD8+ T cells into mice bearing established B16-SIY tumors. The antitumor responses of T cell therapy were impaired by the pretreatment of IFNAR1 blockade (Fig. 4C). Similar to the observation with loss of cGAS and STING, IFNAR1 blockade decreased the proliferative potential of 2C-CD8+ T cells after therapy (Fig. 4D). Consistently, the impaired cytokine production in adoptively transferred CD8+ T cells was also detected in the presence of IFNAR1 blockade after therapy (Fig. 4E). In agreement with the absence of cGAS and STING, anti-IFNAR1 treatment diminished the central memory–like subset (CD62L+CD44+) and induced more frequent PD-1 expression in adoptively transferred CD8+ T cells after therapy (Fig. 4, F and G).

Fig. 4 Type I interferon signaling facilitates the maintenance of stem cell–like CD8+ T cells after T cell therapy.

(A) Representative immunoblots (left) and quantitative data (right) of phosphorylated (p-) TBK1 (Ser172) and total TBK1 in CD8+ T cells from WT, cGAS−/−, and STING−/− mice stimulated with anti-CD3/CD28 at different time points. (B) Quantitative PCR analysis of IFN-β mRNA in CD8+ T cells isolated from the spleen of WT, cGAS−/−, and STING−/− mice after stimulation with anti-CD3 /CD28. (C) B16-SIY tumor growth after the therapy of 2C-CD8+ T cells with or without IFNAR1 blockade. (D) Quantification of the percentages of Ki-67+ cells in CD44+Thy1.1+CD8+ T cells with or without IFNAR1 blockade in the spleen after therapy. (E) Quantification of the percentage of IFN-γ+TNF-α+ subset in 2C-CD8+ T cells with or without IFNAR1 blockade in the TDLN after therapy. (F) Quantification of the percentage of CD62L+CD44+ subset in 2C-CD8+ T cells with or without IFNAR1 blockade in the tumor after therapy. (G) Quantitative data of PD-1 expression in 2C-CD8+ T cells with or without IFNAR1 blockade in the tumor after therapy. (H) Quantitative data of the percentage of CD62L+ in PD-1+CD44+Thy1.1+CD8+ T cells with or without IFNAR1 blockade in the tumor after therapy. (I) Representative histogram (left) and MFI summary (right) show TCF1 expression in PD-1+CD44+TIM3lowThy1.1+CD8+ T cells with or without IFNAR1 blockade in the tumor after therapy. Representative data are shown from three independent experiments conducted with three to six mice per group. Data are represented as means ± SEM. (A) and (B) were calculated with one-way ANOVA with Bonferroni’s multiple comparison tests. (C) to (I) were analyzed with unpaired Student’s t test. *P < 0.05 and **P < 0.01.

We then adoptively transferred anti-IFNAR1–treated 2C-CD8+ T cells and 2C-CD8+ T cells into mice bearing established B16-SIY tumors to measure TCF1 expression. The neutralization of IFNAR1 decreased the central memory–like subset in the PD-1+CD44+Thy1.1+CD8+ T cell population and TCF1 expression in the tumor-infiltrating PD-1+CD44+TIM3lowThy1.1+CD8+ T cell population after therapy (Fig. 4, H and I). To precisely investigate the CD8+ T cell–intrinsic role of type I IFN signaling, we crossed 2C-TCR mice with IFNAR1−/− mice. Flow cytometry and real-time polymerase chain reaction (PCR) confirmed that the protein expression and the mRNA expression of IFNAR1 were greatly decreased in CD8+ T cells from IFNAR1+/− 2C-TCR mice compared to 2C-TCR mice (fig. S7, A and B). We next sought to examine the role of IFNAR1 in CD8+ T cells by adoptive transfer of IFNAR1+/− 2C-CD8+ T cells into tumor-bearing mice. Consistent with IFNAR1 blockade, the partial deficiency of IFNAR1 resulted in less frequent Thy1.1+2C-CD8+ T cells in total CD8+ T cells in DLNs and impaired TCF1 expression in PD-1+CD44+TIM3low populations of Thy1.1+2C-CD8+ T cells in tumors (fig. S7, C and D). These results indicate that autocrine type I IFN signaling mediated by cGAS-STING pathway dictates CD8+ T cell differentiation program toward stem cell–like exhausted subset to sustain the potency of T cell therapy.

The cGAS-STING cascade enhances TCF1 expression by inhibiting Akt activity in CD8+ T cells

To further explore whether the cGAS-STING cascade is essential for the maintenance of TCF1 expression in CD8+ T cells, we stimulated purified CD8+ T cells from the spleen and lymph nodes of WT, cGAS−/−, and STINGgt/gt (which harbor a point mutation that confers a null phenotype) mice using anti-CD3 and anti-CD28 in the presence of IL-15. The cell viability was equivalent for CD8+ T cells from WT, cGAS−/−, and STINGgt/gt mice upon stimulation at the indicated time points (fig. S8). Loss of either cGAS or STING impaired TCF1 expression as shown with MFI gated on PD-1+CD44+CD8+ T cell population (Fig. 5, A and B). This was confirmed with mRNA expression of Tcf7 in CD8+ T cells from WT and STING−/− mice upon in vitro stimulation with anti-CD3/CD28, showing that the deficiency of STING decreases Tcf7 mRNA (fig. S9). We next sought to address whether autocrine type I IFNs promote TCF1 expression in CD8+ T cells. We measured TCF1 expression in CD8+ T cells upon stimulation in the presence or absence of IFNAR1 blockade. In agreement with the observation in cGAS- or STING-deficient CD8+ T cells, TCF1 expression in PD-1+CD44+CD8+ T cells was reduced in the presence of IFNAR1 blockade (Fig. 5C). Genetic deficiency of IFNAR1 inhibited TCF1 expression in PD-1+CD44+CD8+ T cells as well (Fig. 5D). To further assess whether the impaired TCF1 expression in CD8+ T cells with cGAS loss is due to the lack of type I IFN production, we conducted a bypassing experiment in which exogenous IFN-β was added into the culture of cGAS-deficient T cells upon stimulation. IFN-β treatment was able to rescue TCF1 expression in cGAS-deficient CD8+ T cells upon stimulation (Fig. 5E). In contrast, 10 times higher amount (50 IU/ml) of IFN-β failed to rescue the expression of TCF1 in cGAS-deficient CD8+ T cells upon stimulation (fig. S10). TCF1 expression in cGAS-sufficient CD8+ T cells was inhibited by IFN-β treatment in a dose-independent manner, probably due to the disturbance of endogenous IFN-β production (fig. S10). This observation suggests that the endogenous IFN-β production mediated by cGAS/STING could precisely regulate TCF1 expression by meeting the basic needs of TSCM differentiation. Together, these results indicate that cGAS-STING–driven type I IFN signaling contributes critically to the maintenance of stem cell–like CD8+ T cell differentiation by regulating TCF1 expression in a cell-autonomous manner.

Fig. 5 The cGAS-STING cascade enhances TCF1 expression by inhibiting Akt activity in CD8+ T cells.

(A to D) Representative histogram (left) and MFI summary (right) of TCF1 expression in PD-1+CD44+CD8+ T cells from WT and cGAS−/− mice (A), WT and STINGgt/gt mice (B), Ctrl and anti-IFNAR1 (C), and WT and IFNAR1−/− mice (D) stimulated with anti-CD3/CD28 and IL-15 for 48 hours. (E) Representative histograms (left) and MFI summary (right) showing TCF1 expression in PD-1+CD44+CD8+ T cells from WT and cGAS−/− mice stimulated with anti-CD3/CD28 and IL-15 in the presence or absence of IFN-β for 72 hours. (F) Representative histogram (left) and MFI summary (right) showing p-AktT308 expression in 2C-CD8+ T cells and STING−/-2C-CD8+ T cells in the tumor after therapy. (G) Representative immunoblots of p-AktT308 and total Akt in CD8+ T cells from WT, cGAS−/−, and STINGgt/gt mice stimulated with anti-CD3/CD28 for the indicated times. (H) Representative histograms (left) and MFI summary (right) showing p-AktT308 expression in CD8+ T cells from WT, cGAS−/−, STING−/−, and IFNAR1−/− mice stimulated with anti-CD3/CD28 for 5 hours. (I) Representative histogram (left) and MFI summary (right) showing p-AktT308 expression in CD8+ T cells from WT mice stimulated with anti-CD3/CD28 in the presence or absence of anti-IFNAR1 antibody for 5 hours. (J) Representative histograms (left) and MFI summary (right) of TCF1 expression in PD-1+CD44+CD8+ T cells from WT and cGAS−/− mice stimulated with anti-CD3/CD28 and IL-15 in the presence or absence of Akt inhibitor (MK2206, 0.05 μM) for 72 hours. Representative data are shown from three independent experiments with at least three culture replicates. Data are represented as means ± SEM. (A) to (D), (F), and (I) were analyzed with unpaired Student’s t test. (E), (H), and (J) were calculated with one-way ANOVA with Bonferroni’s multiple comparison tests. *P < 0.05, **P < 0.01, ***P < 0.001, and ****P < 0.0001.

Recent studies have demonstrated that CD8+ T cells undergo an ordered program involving activation and differentiation into stem cell–like cells upon stimulation (32, 36). This raises a possibility that the cGAS-STING cascade may skew CD8+ T cells toward stem cell–like differentiation based on our above observations. We propose that activation of the cGAS-STING cascade may prevent CD8+ T cell from overactivation. To test this hypothesis, we stimulated CD8+ T cells purified from spleen and lymph nodes from the abovementioned genotypes with anti-CD3/CD28 and stained for activation markers. In the absence of cGAS, STING, or IFNAR1, expression of CD69, CD25, and CD44 on CD8+ T cells upon stimulation was greatly increased at different time points (fig. S11). It has been shown that Akt–mammalian target of rapamycin (mTOR) signaling promotes CD8+ T cell effector differentiation while inhibiting memory-precursor generation (3740). Chronic Akt activation is associated with T cell dysfunction in the tumor, whereas Akt inhibition reinvigorates the expansion of tumor-infiltrating lymphocytes (40, 41). To explore whether cGAS-STING cascade regulates Akt signaling to determine stem cell–like CD8+ T cell fate, we detected Akt activation (as measured by phosphorylation at Thr308) in 2C-CD8+ T and STING−/− 2C-CD8+ T cells in B16-SIY tumor after therapy and observed that the loss of STING increased the chronic activation of Akt (Fig. 5F). Consistently, we found that the absence of either cGAS or STING augmented Akt phosphorylation (p-Akt) in CD8+ T cells upon stimulation in vitro by Western blot (Fig. 5G). To confirm that the cGAS-STING pathway regulates p-Akt in CD8+ T cells upon stimulation, we performed the flow cytometry staining of p-AktT308. The deficiency of cGAS, STING, or IFNAR1 resulted in the increase of p-AktT308 in CD8+ T cells upon stimulation by flow cytometry (Fig. 5H). This phenomenon was also observed in anti-IFNAR1–treated CD8+ T cells upon stimulation as well as by measuring MFI of p-AktT308 (Fig. 5I). To further assess whether Akt overactivation impedes the maintenance of TCF1 expression in CD8+ T cells with loss of cGAS, we added lower amount of Akt inhibitor (MK2206, 0.05 μM) into the culture of cGAS-deficient T cells upon stimulation. The inhibition of Akt was able to rescue TCF1 expression in cGAS-deficient CD8+ T cells upon stimulation (Fig. 5J). Especially, the ability to rescue TCF1 expression depends on the concentration of Akt inhibitor. The higher concentration of Akt inhibitor (5 μM) could greatly increase TCF1 expression in cGAS-deficient CD8+ T cells compared to the lower concentration of Akt inhibitor (0.05 μM) (fig. S12A). Compared to Akt inhibitor, an mTOR inhibitor (rapamycin) failed to rescue TCF1 expression in cGAS-deficient CD8+ T cells (fig. S12B), indicating that mTOR inhibition is independent of the maintenance of T cell stemness in our system. Together, these findings suggest that the cGAS-STING cascade augments the maintenance of stem cell–like CD8+ T cells mainly via the inhibition of chronic Akt signaling.

Genomic DNA is selectively enriched in the cytosol of CD8+ T cells upon stimulation

To assess whether cytosolic DNA is elevated in CD8+ T cells upon stimulation, we imaged double-stranded DNA (dsDNA) and single-stranded DNA (ssDNA). Both dsDNA and ssDNA accumulated in the cytosol of mouse CD8+ T cells upon stimulation (Fig. 6, A to C). The released cytosolic DNA in activated T cells could be removed by deoxyribonuclease (DNase) I (fig. S13A), demonstrating that the antibody staining for cytosolic DNA is specific. We also stained cytosolic RNA with or without ribonuclease (RNase) digestion in T cells after activation. The amount of cytosolic RNA is comparable between activated T cells and naïve T cell, and the staining of cytosolic RNA is diminished after the pretreatment with RNase (fig. S13B).

Fig. 6 Genomic DNA is selectively enriched in the cytosol of CD8+ T cells upon stimulation.

(A) Representative images of cytosolic double-stranded DNA (dsDNA) and nucleic DNA [4API (c DNA uble-stranded DNA (DAPI)] in CD8+ T cells from naïve mice stimulated with anti-CD3/CD28 for the indicated times. Scale bar, 5 μm. (B) Representative images of cytosolic single-stranded DNA (ssDNA) and nucleic DNA (DAPI) in CD8+ T cells stimulated with anti-CD3/CD28 for the indicated times. Scale bar, 5 μm. (C) Quantitative analysis of dsDNA (left) and ssDNA (right) in CD8+ T cells by antibody staining. (D) Quantitative PCR analysis of genomic DNA (gDNA; Tert) and mitochondrial DNA (mtDNA; Dloop1) in CD8+ T cells stimulated with anti-CD3/CD28 at different time points. (E) Quantitative PCR analysis of gDNA and mtDNA in naïve (CD62L+CD44) and activated (CD44+) splenic CD8+ T cells sorted from MC38-SIYhi–immunized mice. (F) Quantitative PCR analysis of gDNA and mtDNA in splenic CD8+ T cells sorted from MC38-bearing mice. (G) Quantitative PCR analysis of gDNA and mtDNA in splenic CD8+ T cells of B16-SIY–bearing mice after anti–PD-L1 treatment. Representative data are shown from three independent experiments with at least three replicates using pooled spleen cells from three mice. Data are represented as means ± SEM. (C), (D), and (G) were calculated with one-way ANOVA with Bonferroni’s multiple comparison tests. (E) and (F) were analyzed with unpaired Student’s t test. *P < 0.05, **P < 0.01, ***P < 0.001, and ****P < 0.0001.

To address the source of DNA fragments, purified cytosolic extracts were analyzed by real-time PCR assay. We found a higher abundance of genomic DNA (gDNA) fragments than mitochondrial DNA (mtDNA) fragments inside the cytosol of CD8+ T cells after stimulation over time (Fig. 6D). To further validate whether activated CD8+ T cells have a higher abundance of gDNA compared to naïve T cells in vivo, we immunized mice with highly immunogenic MC38-SIYhigh tumor cells or inoculated with MC38 tumor cells. The cytosolic gDNA was increased in activated CD8+ T cells (gated on CD44+ population) compared to naïve CD8+ T cells (gated on CD62L+CD44 population) regardless of the immunization model or in tumor-bearing model (Fig. 6, E and F). In contrast, cytosolic mtDNA was barely detected in CD8+ T cells with the stimulators of tumor cells in vivo (Fig. 6, E and F).

Because immune checkpoint blockade reinvigorates the stem cell–like exhausted CD8+ T cell subset (28, 29), we next sought to address whether anti–PD-L1 treatment was able to enrich cytosolic gDNA inside endogenous CD8+ T cells from tumor-bearing mice. Immune checkpoint blockade elevated the amount of cytosolic gDNA in splenic CD8+ T cells compared to the nontreatment group (Fig. 6G). These results suggest that gDNA fragments are enriched in the cytosol of CD8+ T cells encountered with stimulators and that the increased gDNA may trigger nucleic acid recognition by cGAS for maintaining the differentiation of stem cell–like CD8+ T cells.

STING agonism maintains human stem-like central memory CD8+ T cells and improves human CAR-T cell therapy in mice

The phenotype of stem-like central memory CD8+ T cells (TSCM) has been characterized with CCR7+CD62L+CD45ROCD95+ in patients with cancer (24, 40, 42, 43). Because cGAS/STING expression was greatly decreased in CD8+ T cells from patients with cancer, we next sought to assess whether a STING agonist is capable to rescue the differentiation of TSCM cells from these individuals. Isolated naïve CD8+ T cells from patients with cancer were stimulated with anti-CD3/CD28 and IL-15 in the presence of a STING agonist [diABZI compound 3 (C3)] and analyzed by flow cytometry after the gating strategy in fig. S14. STING stimulation increased the frequency of subsets with TSCM-associated surface markers in total CD8+ T cells from patients with cancer (Fig. 7A) or healthy volunteers (Fig. 7B). In contrast, STING stimulation resulted in the impaired formation of subset with effector-associated surface markers in total CD8+ T cells (TEFF; gated on CD45ROCCR7CD62LCD95+) from patients with cancer and healthy volunteers, respectively (Fig. 7, C and D). These results suggest that STING agonism may potentiate the programming of stem-like central memory while avoiding terminal differentiation in human CD8+ T cells. Unexpectedly, STING agonist stimulation was unable to impair the cell viability of human CD8+ T cells from patients with cancer compared to CD8+ T cells from naïve mice (fig. S15, A and B).

Fig. 7 STING agonism augments the differentiation of stem-like central memory CD8+ T cells from patients with cancer and healthy volunteers and antitumor responses of engineered human T cells.

(A to D) Naïve CD8+ T cells were isolated from peripheral blood of nine patients with cancer and six healthy volunteers stimulated with anti-CD3/CD28 and rhIL-15 in the presence or absence of a STING agonist (C3; 100 nM for 4 hours at the beginning) for 72 hours. Representative flow plots (left) and summary (right) of frequency of stem-like central memory CD8+ T cells (TSCM; CD45ROCCR7+CD62L+CD95+) in total CD8+ T cells from patients with cancer (A) and healthy volunteers (B). Representative flow plots (left) and summary (right) of frequency of effector CD8+ T cells (TEFF; CD45ROCCR7CD62LCD95+) in total CD8+ T cells from patients with cancer (C) and healthy volunteers (D). Each dot represents one donor in (A) to (D). (E) NOD-PrkdcscidIL2rγtm1 mice were subcutaneously inoculated with 5 × 105 A549-hCD20 cells and then treated with 1 × 107 human CD20-CAR T cells 7 days later, as shown in experimental schematic. A549-hCD20 tumor growth after the therapy of PBS versus CD20-CAR T cells from a healthy volunteer (left) (n = 6). A549-hCD20 tumor growth after the therapy of CD20-CAR T cells versus STING agonist C3-stimulated CD20-CAR T cells from a patient with cervical cancer (right) (n = 6). Representative data in (A) to (D) are shown from two independent experiments. Quantitative data in (A) and (C) are pooled together from two independent experiments. Representative data in (E) was shown from two independent experiments conducted with CAR-T cells from the same two donors. Data are represented as means ± SEM. (A) to (D) were analyzed with paired Student’s t test. (E) was calculated with unpaired Student’s t test. *P < 0.05 and **P < 0.01; ns, no significant difference.

Next, we sought to measure the phosphorylation of STING in T cells from both healthy donors and patients with cancer after STING agonist stimulation. Our results show that STING agonism could notably enhance the expression of STING phosphorylation in T cells from both populations (fig. S15, C and D), indicating that STING agonists may have a rational application for enhanced human engineering T cell therapy. To evaluate the antitumor activity of engineered anti-human CD20 CAR-T cells treated with STING agonist, we used a xenograft mouse model with subcutaneous A549-CD20 tumors. Anti-hCD20 CAR-T cell therapy alone failed to delay tumor growth compared to the phosphate-buffered saline (PBS) treatment group. In contrast, STING agonist C3 could significantly potentiate antitumor responses of anti-hCD20 CAR-T cell therapy in solid tumors (P = 0.0228) (Fig. 7E), indicating that STING agonism is a promising avenue to improve human T cell therapy in solid tumors. Our proposed mechanism for the differentiation of stem cell–like CD8+ T cells mediated by cGAS-STING pathway in T cell therapy is displayed by a schematic (fig. S16).

DISCUSSION

Accumulating evidence has shown the importance of the cGAS-STING pathway in dictating antitumor immunity through potentiating type I IFN production (1, 44, 45). The process of nucleic acid recognition by cGAS and cascade signaling occurs not only in antigen-presenting cells but also in tumor cells upon various stresses, including radiotherapy, PARP-1 inhibition, and CD47 blockade (47). Recently, it has been demonstrated that STING agonists or gain-of-function mutations facilitate T cell death via proapoptotic transcriptional program or disruption of calcium homeostasis, respectively (12, 13). In addition, the STING agonist DMXAA or constitutively active STING mutations reduce both the proliferative capacity and memory formation of T cells (11). However, our findings indicated that cell-intrinsic cGAS-STING pathway signaling augments the persistence and expansion of adoptively transferred T cells. To reconcile these potentially conflicting observations, our interpretation is that the degree and duration of activation of the endogenous cGAS-STING pathway in T cells upon TCR ligation might be much weaker and shorter compared to those mounted by STING agonists or constitutive gain-of-function mutations. Here, we identified that intrinsic cGAS and STING are capable of facilitating potent antitumor immune responses by increasing the differentiation of stem cell–like TCF1+PD-1+CD8+ T cells after transfer. Furthermore, gDNA is selectively enriched in the cytosol of CD8+ T cells upon stimulation, which leads to the activation of the cascade. Our study reveals that the cGAS-STING pathway plays an unanticipated role in the maintenance of stem cell–like T cells in antitumor immunity.

There are two subsets of exhausted T cells in the tumor based on molecular and functional features (18). One subset is terminally exhausted T cells with superior cytotoxicity and short-lived survival, and another is stem cell–like exhausted T cells with greater multipotency and persistence driven by TCF1 (18). Immune checkpoint blockade acts on stem cell–like exhausted T cells to give rise to pool of differentiated T cells, but not on terminally exhausted T cells, leading to protective antitumor immune responses (28, 29). Likewise, we found that the cGAS-STING pathway not only increased the frequency of TCF1+PD-1+ stem cell–like exhausted T cells but also increased the frequency of effector T cells because of the effector functions that predominantly stem from the stem cell–like exhausted T cells in tumor sites and TDLNs of mice. Correspondingly, cytokine production by adoptively transferred T cells without the cGAS-STING cascade was decreased in tumor-bearing mice, whereas the production of cytokines in cGAS/STING-deficient T cells was increased with the stimulation of anti-CD3/CD28 in vitro. Of note, we detected that transferred T cells without cGAS-STING signaling were more likely to express PD-1. This further validates that cGAS-STING pathway prevents T cell terminal exhaustion. Using the role of STING signaling in sculpting tumor microenvironments, STING agonism has been eagerly anticipated to broaden the scope of checkpoint blockade immunotherapy (1). The synergy of anti–PD-1 and STING agonists is typically attributed to STING pathway activation in tumor cells or antigen-presenting cells (46, 47). Nevertheless, our results reveal that STING’s contribution in other cell types needs to be reappraised in terms of efficacy of combination of anti–PD-1 and STING agonists in clinical trials.

TCF1+PD-1+CD8+ T cells with stem cell–like properties including self-renewal, expansion, and persistence are responsible for tumor control after immunotherapy (28, 29, 31). With recent progress in single-cell profiling, it has been recognized that TCF1 is a hallmark of the stem cell–like T cell subset (25). On the basis of this concept, we found that cGAS-STING pathway–mediated type I IFNs maintained TCF1 expression in CD8+ T cells through Akt inhibition. On the other hand, the treatment of exogenous IFN-β rescued TCF1 expression in cGAS-deficient CD8+ T cells. Conversely, virus-induced type I IFNs have been demonstrated to suppress TCF1 expression in CD8+ T cells (48). High doses of type I IFN may generate stronger inflammatory signals that enable TCF1 silencing, whereas lower type I IFN induction in T cells upon stimulation may cause totally different inflammatory spectrums that enhance TCF1. Approaches to improve the differentiation of stem cell–like T cells are diverse, such as 4-1BB domain insertion, potassium stimulation, and transcription factor reprogramming (19, 22, 31, 32). To our surprise, we observed that cGAS-STING cascade is remarkably down-regulated in peripheral blood CD8+ T cells of patients with cancer compared to healthy volunteers. Our findings provide potential avenues to augment CAR- and TCR-T cell therapy by expanding stem cell–like T cells using STING agonists without causing cell death before transfer or by sustaining phenotypes of stem cell–like T cells using cGAS/STING replenishment.

Cytosolic DNA is enriched in cells suffering from DNA damage and mitochondrial disruption (4952). We observed that genomic dsDNA and ssDNA were selectively enriched in the cytosol of T cells upon TCR ligation. As a further confirmation, more cytosolic DNA was observed in activated T cells compared to naïve T cells obtained either from tumor-bearing mice or from tumor-immunized mice. In addition, anti–PD-L1 treatment enables reinvigorated stem cell–like exhausted T cells to expand with greater proliferative capacity. This process, in turn, would then lead to more cytosolic DNA enrichment in T cells. However, the mechanism underlying gDNA release into the cytosol of T cells was not investigated in this study. To rule out the possibility that cell death may accelerate the release of gDNA into the cytosol, we traced the kinetics of cell death with annexin V and propidium iodide (PI) staining. There was no obvious cell death when the accumulation of cytosolic DNA reached its peak in CD8+ T cells stimulated with anti-CD3/CD28 in vitro, excluding that gDNA release is due to cell death. It has been observed that migrating dendritic cells with the exchange of nucleocytoplasmic content during interphase require efficient nuclear envelope and DNA damage repair for cell survival (53). Thus, the enriched cytosolic DNA may act as a by-product in proliferating T cells, which have efficient nuclear envelope and DNA repair machineries to prevent cell death. Recently, it has been demonstrated that cGAS is anchored to plasma membrane to defend against DNA virus infection, whereas cytosolic cGAS is sensitive in response to genotoxic stress (54). On the other hand, T cell cytoplasm accounts for a relatively small fraction of whole cell area compared to phagocytes, more easily allowing cGAS localized at plasma membrane to encounter cytosolic DNA.

Our study does have certain limitations, such as the use of engineered tumor antigen and Tg T cells to follow stem cell–like CD8+ T cell behaviors and the reason of cytosolic DNA accumulation after T cell activation. Whether cGAS localizes to the plasma membrane of T cells upon stimulation also needs to be further determined. The number of patients and stratification by tumor type should be largely increased to investigate whether STING can be an indicator of tumor prognosis. Although we demonstrated that human engineered CAR-T cells treated with a STING agonist generated a stronger antitumor response in xenograft solid tumor model, it needs to be further determined whether cGAS-STING signaling affects the persistence of adoptive CAR-T cells in patients with cancer in clinical trials. In addition, human CD8+ T cell differentiation with in vitro stimulation is not sufficient to monitor T cell stemness.

Recently, much attention has been paid to the understanding of T cell stemness in immune checkpoint blockade. Our findings reveal a positive role of cGAS-STING pathway in the differentiation of stem cell–like T cells and point out an avenue to enhance T cell therapy through manipulating cGAS-STING pathway. In addition, this study may provide new insight into probing the mechanisms of STING agonist in cancer therapy and rational design of the relevant clinical trials.

MATERIALS AND METHODS

Study design

The objective of this study was designed to identify potential mechanism of the cGAS-STING cascade in T cells after cancer T cell therapy in immunocompetent mouse cancer models. T cell therapy in the tumor was evaluated using adoptive transfer of activated CD8+ T cells from 2C-Tg mice. We monitored tumor growth and analyzed the percentage, function, and stemness of Thy1.1+ 2C-CD8+ T cells in the tumor, spleens, and TDLNs. To investigate the direct role of cGAS-STING in T cell stemness in vitro, mouse CD8+ T cells from lymph nodes and spleen and human CD8+ T cells from healthy volunteers and patients with cancer were isolated and stimulated with anti-human CD3/CD28. To determine whether cytosolic DNA was released in activated CD8+ T cells, cytosolic DNA was extracted and measured by real-time PCR in both in vitro and in vivo experiments. Last, to determine whether STING agonist could enhance the efficacy of CAR-T cells against solid tumors, we used a xenograft model.

In our study, for cell-based experiments, at least biological triplicates were performed in each single experiment, unless otherwise stated. Animals were randomized into different groups after tumor cell inoculation and at least three to six mice were used for each group, unless otherwise indicated. Animals that failed to develop tumors from the beginning were excluded from the analysis. Fixed time points are shown in the figures and indicate the time elapsed from the start of treatment. Whenever possible, the investigators were blinded to group allocation during the experiments and when assessing outcomes. In some cases, selected samples were excluded from specific analyses because of technical flaws during sample processing or data acquisition. Analytical studies were typically performed two to three times in independent experiments, implementing fixed time points of analysis for all experimental groups, unless indicated. All primary data are reported in data file S1.

Tumor growth and treatments

B16-SIY tumor cells (0.5 × 106 to 1 × 106) gifted from Y.-X. Fu (UT Southwestern) were subcutaneously implanted into female C57BL/6J mice. Transferred T cells were isolated from gender-matched 2C, 2C-cGAS−/−, and 2C-STING−/− mice using total CD8+ isolation kit (STEMCELL). Cells were activated by anti-CD3 (2 μg/ml) and anti-CD28 (1 μg/ml) for 48 hours in 96-well round-bottom plates. Then, cells were cultured with fresh medium containing murine IL-15 (10 ng/ml; PeproTech) and murine IL-7 (10 ng/ml; PeproTech) for another 72 hours. On days 14 to 16 after tumor inoculation, 1.2 × 106 activated CD8+ T cells were transferred into mice via retro-orbital intravenous injection. For anti-IFNAR1 blockade experiments, 2C-CD8+ T cells were additionally blocked with anti-IFNAR1 antibody (clone MAR1-5A3; Bio X Cell) at a final concentration of 200 μg/ml during the activation culture, and mice were administrated anti-IFNAR1 antibody (500 μg per mouse) by intraperitoneal injection immediately after the transfer, followed by another treatment 3 days later. EG7 (3 × 105 to 4 × 105) or GL261 tumor cells (5 × 105) were subcutaneously injected into the flank of WT and CD4cre+-STINGflox/flox mice. Tumor volumes were measured along three orthogonal axes (a, b, and c) twice weekly and calculated as tumor volume = a × b × c/2.

CAR-T cell engineering and xenograft model

Lentiviral plasmids containing CAR constructs were generated by standard molecular cloning methods. Briefly, a DNA fragment containing the anti-human CD20 single-chain variable fragment (scFv) derived from rituximab, CD8 hinge and transmembrane domain, 4-1BB, and CD3 ζ intracellular domain was generated by overlap PCR and cloned into the pCDH-EF1a vector (System Biosciences). Lentivirus was produced by transient transfection of Lenti-X 293T with a four-plasmid system. Supernatants containing lentivirus particles were collected at 48 and 72 hours after transfection and concentrated by ultracentrifugation. Human peripheral blood mononuclear cells (PBMCs) from patients with cancer were isolated using Ficoll-Paque PREMIUM sterile solution (GE). Total T cells were purified with an EasySep Human T Cell Isolation kit (STEMCELL). Plate-bound anti-CD3 (Bio X Cell) and soluble anti-CD28 (Bio X Cell) were used to activate T cells in culture medium. Two days after activation, various concentrated CAR-containing lentiviruses were used to transduce T cells at a multiplicity of infection of 10. To evaluate the efficacy of CD20-targeted CAR-T cells against solid tumors, female NOD-PrkdcscidIL2rγtm1 mice were subcutaneously inoculated with 5 × 105 A549-hCD20 cells. About 7 days later, 1 × 107 CAR-T cells were stimulated with 30 nM STING agonist C3 or vehicle control for 4 hours, washed with PBS, and transferred intravenously through the retro-orbital sinus to tumor-bearing mice. Tumor volumes were measured along three orthogonal axes (a, b, and c) twice weekly and calculated as tumor volume = a × b × c/2.

Statistical analysis

No statistical method was used to predetermine sample size. Mice were assigned at random to treatment groups for all mouse studies. Experiments were repeated two to three times. Statistical analysis was performed using GraphPad Prism8 software (GraphPad Software Inc.) and presented as stated in individual figure legends. For comparisons between two groups in animal experiments, P values were calculated using unpaired Student’s t tests. For comparisons between two groups in the tests of human samples, P values shown in Fig. 1B were calculated using unpaired Student’s t tests, whereas P values shown in Fig. 7 (A to D) were calculated using paired Student’s t tests. For the comparisons between three or more groups, P values were calculated using one-way analysis of variance (ANOVA) with Bonferroni’s multiple comparison tests. Data that were determined to be nonparametric were calculated by Kruskal-Wallis test (more than two groups) with Dunn’s multiple comparisons. Data were presented as means ± SEM. We indicated significance corresponding to the following: *P < 0.05, **P < 0.01, ***P < 0.001, and ****P < 0.0001; ns, no significant difference.

SUPPLEMENTARY MATERIALS

stm.sciencemag.org/cgi/content/full/12/549/eaay9013/DC1

Materials and Methods

Fig. S1. STING expression in CD14+ monocytes and CD19+ B cells in PBMCs from healthy volunteers and patients with cancer.

Fig. S2. STING expression in CD8+ T cells of CD4cre+-STINGflox/flox mice.

Fig. S3. T cell development and homeostasis in the KO mice.

Fig. S4. Representative flow cytometry gating strategy of the cells isolated from the tumor of one mouse after transfer.

Fig. S5. The expression of TCF1 and the frequency of TCF1+Slamf6+ cells in PD-1+ or PD-1 tumor-infiltrating 2C-CD8+ T cells.

Fig. S6. The impact of TCF1 on maintaining CD8+ T cell stemness in vivo.

Fig. S7. The frequency and magnitude of TCF1 expression in IFNAR1+/− 2C-CD8+ T cells after transfer.

Fig. S8. Stimulated CD8+ T cells without the cGAS-STING cascade are still viable.

Fig. S9. Quantification of Tcf7 mRNA in CD8+ T cells upon stimulation.

Fig. S10. The effect of various quantities of IFN-β on regulating TCF1 expression in PD-1+CD44+CD8+ T cells from WT and cGAS−/− mice.

Fig. S11. The cGAS-STING cascade prevents the overactivation of CD8+ T cells.

Fig. S12. The rescue of TCF1 expression in cGAS-deficient CD8+ T cells by an Akt inhibitor but not by an mTOR inhibitor.

Fig. S13. The staining of cytosolic dsDNA and dsRNA in CD8+ T cells with or without DNase or RNase treatment.

Fig. S14. The gating strategy of TSCM and TEFF in CD8+ T cells from patients with cancer and healthy volunteers.

Fig. S15. The viability and STING phosphorylation of CD8+ T cells stimulated with a STING agonist.

Fig. S16. Schematic of proposed mechanism for the differentiation of stem cell–like CD8+ T cells mediated by the cGAS-STING pathway in T cell therapy.

Table S1. Studied patients.

Data file S1. Primary data.

REFERENCES AND NOTES

Acknowledgments: We acknowledge Y.-X. Fu, L. Liang, S. Wang, and B. Ge for helpful scientific discussion and manuscript editing. Funding: This study was supported by the National Natural Science Foundation of China (81771682 to L.D., 81772770 to W.D., 81971467 and 81671643 to X.Y., 81702804 to W.L., and 31900649 to M.W.), National Thousand Youth Talents Program (to L.D.), Science and Technology Commission of Shanghai Municipality (16JC1406000 to L.D.), Innovative research team of high-level local universities in Shanghai (to L.D. and W.L.), the National Key Research and Development Program of China (2016YFC1302900 to W.D. and 2016YFC1303400 to X.Y.), Shanghai Municipal Commission of Health and Family Planning [ZY(2018-2020)-FWTX-3006 and 2017ZZ02016 to W.D. and 20194Y0625 to W.L.], and Open grant of Shanghai Key Laboratory of Gynecologic Oncology at Ren Ji Hospital (FKZL-2018-01 to W.L.). Author contributions: W.L. and L.L. performed the experiments and wrote the paper. J.L. and X.W. performed part of experiments. Y.Y., J.J., and C.Y. aid in the experiments of flow cytometry assay. M.W., L.W., X.H., and D.C. aided in some animal experiments. F.L. performed xenograft tumor model. W.L., L.L., and J.L. analyzed the data. B.S. and J.C. helped in the editing of the manuscript. X.Y., W.D., and L.D. conceived and supervised all experiments and the writing of the manuscript. All authors approved the paper. Competing interests: The authors declare that they have no competing interests. Data and materials availability: All data associated with this study are present in the paper or the Supplementary Materials.

Stay Connected to Science Translational Medicine

Navigate This Article