DNA-PK interacts with cyclic dinucleotides and inhibits type I interferon responses

The presence of double-stranded DNA (dsDNA) in the cytosol is sensed as a danger signal and promotes the activation of inflammatory responses (Decout et al., 2021). A major pathway involved in the detection of cytosolic dsDNA relies on the cyclic GMP-AMP (cGAMP) synthase (cGAS) (Sun et al., 2013) that catalyzes the production of the 2′3′-cGAMP cyclic dinucleotide (CDN), which in turn binds and activates the stimulator of IFN genes (STING) adaptor protein (Sun et al., 2013). The interaction of 2′3′-cGAMP with STING drives the assembly of a signalosome where the Tank-binding kinase 1 catalyzes phosphorylation-dependent activation of transcription factors, such as the IFN regulatory factor 3 (IRF3), to drive the expression of inflammatory cytokines and type I IFNs (Sun et al., 2009; Zhong et al., 2008). Activation of the cGAS–STING signaling axis is essential in response to pathogen infection, cellular stress, or tissue damage, which are contexts where immune stimulatory dsDNAs are exposed in the cytosol. However, chronic activation of this signaling pathway is reported in several human pathologies, such as autoinflammatory diseases or cancer, where it is responsible for chronic low-grade disease-promoting inflammation, leading to immune dysfunction and tissue damage (Decout et al., 2021).

To mitigate the deleterious effects of unwanted activation of the cGAS–STING pathway, several regulatory layers have been described to directly control cGAS and/or STING activation (Hertzog and Rehwinkel, 2020; Vila et al., 2022). In contrast, few mammalian direct regulators of the 2′3′-cGAMP molecule have been documented (Hou et al., 2023; Li et al., 2014; Mardjuki et al., 2024, Preprint), all of which act on extracellular 2′3′-cGAMP. This is striking since 2′3′-cGAMP is an essential component of the cGAS–STING axis, which is notably involved in signal amplification and propagation (Jutte et al., 2021), and its intracellular negative regulation could be essential for signal termination. Additionally, despite several CDNs of bacterial origin being active in mammalian cells (McWhirter et al., 2009; Woodward et al., 2010), no cellular mechanism ensuring their regulation has been described to date.

Prior to the identification of cGAS, the DNA-PK holoenzyme was proposed to operate as a cytosolic dsDNA sensor (Ferguson et al., 2012). DNA-PK is comprised of three subunits, with the KU70 and KU80 scaffold proteins forming heterodimers that ensure the recruitment of the DNA-PK catalytic subunit (DNA-PKcs) to nuclear double-strand breaks and their repair by nonhomologous end joining (NHEJ) (Yue et al., 2020). Several studies support a role of DNA-PKcs in innate immune activation in the presence of cytosolic dsDNA, both through direct detection of cytosolic dsDNAs (Burleigh et al., 2020; Morales et al., 2017; Scutts et al., 2018; Taffoni et al., 2023) or as a co-activation of cGAS (Hristova et al., 2024; Taffoni et al., 2023) following viral infections or genotoxic stress. Yet, absence of DNA-PKcs was also shown to lead to a potentiation of type I IFN responses (Sun et al., 2020), suggesting dichotomous functions of DNA-PKcs in the regulation of cGAS–STING-associated inflammatory responses and highlighting a lack of mechanistic insight into the relationship between those two pathways.

Here, we show that DNA-PKcs acts as a direct regulator of intracellular levels of the 2′3′-cGAMP CDN. We also uncover that the action of DNA-PKcs on 2′3′-cGAMP takes place downstream of its role in the primary detection of cytosolic dsDNA and serves as a signal terminator downstream of 2′3′-cGAMP production. We demonstrate that DNA-PKcs is also capable of regulating the bioactivity of the bacterial 3′3′-cGAMP CDN and of specific synthetic STING agonists. This finding, therefore, bears important implications for our understanding of the regulation of the cGAS–STING pathway regulation but also for the development and use of STING agonists in immunosuppressed contexts.

Since the absence of DNA-PKcs was shown to promote type I IFN responses, suggesting an inhibitory function downstream of cGAS activation (Sun et al., 2020), we hypothesized that it may directly modulate 2′3′-cGAMP bioactivity. We thus tested whether DNA-PKcs interacts with 2′3′-cGAMP.

We performed in vitro–binding assays, using a FLAG-tagged WT-DNA-PKcs (FLAG-DNA-PKcs; Fig. 1 A), which was FLAG-purified in conditions where the KU scaffold proteins are not detectable (Fig. 1 B). We found a significant enrichment of 2′3′-cGAMP with DNA-PKcs in these conditions (Fig. 1 C). Similarly, when endogenous DNA-PKcs was immunoprecipitated (Fig. S1 A), the presence of DNA-PKcs in immunoprecipitates (Fig. S1 B) correlated with the detection of 2′3′-cGAMP following in vitro–binding assays (Fig. S1 C). To further exclude the contribution of DNA-PKcs cofactors in the binding to 2′3′-cGAMP, we performed similar experiments using recombinant DNA-PKcs (Fig. 1 D and Fig. S1 D). We found a significant enrichment of 2′3′-cGAMP when recombinant DNA-PKcs was immunoprecipitated (Fig. 1 E). Conversely, we conducted experiments where FLAG-immunoprecipitated FLAG-DNA-PKcs was peptide-eluted under native conditions prior to incubation with streptavidin bead–immobilized biotinylated 2′3′-cGAMP or biotin (Fig. 1 F). Western blot (WB) analyses showed enrichment of DNA-PKcs with biotinylated 2′3′-cGAMP (Fig. 1 G). Immunofluorescence analyses also showed co-localization between DNA-PKcs and 2′3′-cGAMP (Fig. 1 H and I). Thermal shift assays (TSAs) were then conducted using either whole-cell lysates (Fig. 1, J and K; and Fig. S1, E and F) or FLAG-purified FLAG-DNA-PKcs (Fig. 1, L and M). Incubation with 2′3′-cGAMP led to stabilization of DNA-PKcs both in whole-cell extracts (Fig. 1 K and Table 1) and immunoprecipitates (Fig. 1 M). TSA performed using the NU7441 DNA-PKcs competitive inhibitor (Leahy et al., 2004) showed selective stabilization of DNA-PKcs but not of STING (Fig. S1, E and F).

Taken together, these data show that DNA-PKcs physically interacts with 2′3′-cGAMP.

We next aimed to identify the domain of DNA-PKcs interacting with 2′3′-cGAMP. We first performed molecular modelling and docking analyses to identify putative binding regions. We used the resolved crystal of human DNA-PKcs (RCBS: 5LUQ) for docking experiments using 2′3′-cGAMP. This predicted that 2′3′-cGAMP can dock into DNA-PKcs’ catalytic pocket (Fig. 1 N and Fig. S1, G and H), adopting a stable conformation (energy −28940.5 Kcal/mol; Table 2), notably through strong H-bonds between DNA-PKcs residues Asp3723, Arg3741, and His1069 and Asp3744, Arg3746, and His3748 with the phosphodiester bonds within the 2′3′-cGAMP molecule (Fig. S1, I and J). To control for specificity, we used other DNA repair–related enzymes belonging to the phosphatidylinositide 3-kinase–related kinase family, to which DNA-PKcs belongs, namely ataxia telangiectasia mutated (ATM; 8OXQ) and ataxia telangiectasia and Rad3-related protein (ATR; 5YZ0), which possess analogous catalytic sites competent for ATP binding and hydrolysis. The available structures of ATM and ATR were superposed with that of DNA-PKcs, and comparative analysis of their catalytic cleft predicted that 2′3′-cGAMP is unlikely to interact with these catalytic domains (Fig. S1, K and L). Altogether, these in silico data suggest selective interaction of 2′3′-cGAMP with DNA-PKcs in its catalytic pocket.

We reasoned that if 2′3′-cGAMP interacts with DNA-PKcs’ catalytic pocket, it could be expected to reduce DNA-PKcs’ ability to hydrolyze ATP. We thus used an in vitro assay to measure the kinase activity of DNA-PKcs in the presence of increasing doses of 2′3′-cGAMP. The NU7441 competitive DNA-PKcs inhibitor was used as a positive control. Increasing 2′3′-cGAMP concentration led to inhibition of DNA-PKcs catalytic activity in a dose-dependent manner (Fig. 1 O), supporting binding of 2′3′-cGAMP in the catalytic cleft of DNA-PKcs. We next used recombinant DNA-PKcs for in vitro 2′3′-cGAMP–binding assays, in which NU7441 was used as a competitor. NU7441 displaced the interaction of 2′3′-cGAMP with DNA-PKcs (Fig. 1 P and Fig. S1, M and N), suggesting that the interaction of 2′3′-cGAMP with DNA-PKcs occurs in its catalytic cleft.

To confirm that the interaction of 2′3′-cGAMP with DNA-PKcs occurs in the catalytic domain, we generated DNA-PKcs mutants where the kinase domain is deleted (FLAG-DNA-PKcs-ΔK) or corresponds to the kinase domain alone (FLAG-kinase). TSAs were conducted on full-length and truncation mutants of DNA-PKcs, showing that 2′3′-cGAMP stabilized both FLAG-DNA-PKcs and FLAG-kinase, but not FLAG-DNA-PKcs-ΔK (Fig. 1 Q). Binding assays were then performed using full-length and DNA-PKcs truncations. Full-length FLAG-DNA-PKcs and FLAG-kinase both interacted with 2′3′-cGAMP, while FLAG-DNA-PKcs-ΔK was not retrieved in the pulldown assay (Fig. 1 R).

Altogether, our data show that DNA-PKcs interacts directly with 2′3′-cGAMP through its catalytic domain.

Interaction of 2′3′-cGAMP with DNA-PKcs may affect 2′3′-cGAMP biological activity. We thus assessed the impact of DNA-PKcs on 2′3′-cGAMP–associated signaling.

To this aim, we treated T98G cells with 2′3′-cGAMP in the presence or absence of NU7441 and analyzed STING activation. Of note, T98G cells do not express detectable levels of cGAS and are therefore a good model to study the direct effect of DNA-PKcs on 2′3′-cGAMP, excluding confounding effects associated with the role of DNA-PKcs as a co-activator of cGAS (Taffoni et al., 2023). Expectedly, 2′3′-cGAMP treatment alone led to STING degradation, STING phosphorylation (pSTING), and IRF3 phosphorylation (pIRF3) (Fig. 2 A), accompanied by upregulation of IFNB1 expression (Fig. 2 B). Attesting to the production of bioactive IFNs, increased expression of the ISG15, IFIT1, and CCL5 IFN-stimulated genes (ISGs) was measured following 2′3′-cGAMP treatment (Fig. 2 B). Combining 2′3′-cGAMP treatment with NU7441-mediated DNA-PKcs inhibition led to increased pSTING and pIRF3 levels (Fig. 2 A, compare lanes 2 and 4) and enhanced the expression of IFNB1, ISG15, IFIT1, and CCL5 (Fig. 2 B) when compared with treatment with 2′3′-cGAMP alone. The induction of type I IFN responses is also observed upon measurement of produced IFNβ (Fig. 2 C), CXCL10, and CCL5 (Fig. S2 A). Similar results were obtained with a panel of DNA-PKcs inhibitors (Fig. S2 B), confirming that DNA-PKcs inhibition is associated with boosted 2′3′-cGAMP signaling. This is in contrast to what is observed when T98G cells stably expressing cGAS (T98G^cGas) are challenged with dsDNAs, which results in decreased IFNB1, CXCL10, and IFIT2 expression (Fig. S2 C).

To confirm a direct role of DNA-PKcs in inhibition of 2′3′-cGAMP signaling, we next used siRNAs targeting DNA-PKcs or KU70 (Fig. S2 D). Treatment with 2′3′-cGAMP following knockdown of DNA-PKcs, but not of KU70, increased expression of STING-associated transcripts such as IFNB1, ISG15, CXCL10, and IFIT2 (Fig. 2 D), phenocopying the effect of NU7441 treatment. Similarly, we found that in DNA-PKc knockout cells (Fig. S2 E), 2′3′-cGAMP–associated type I IFN signaling was boosted (Fig. 2 E). Thus, DNA-PKcs inhibition or ablation enhances 2′3′-cGAMP–associated type I IFN responses.

We next sought to verify if DNA-PKcs–mediated type I IFN signaling relies on the canonical STING-dependent pathway. To this aim, we used THP-1 human myeloid cell lines knocked out for cGAS (THP-1^cGAS−/−) or STING (THP-1^STING−/−) and isogenic control cells (THP-1^CTRL). Cells were mock treated or treated with 2′3′-cGAMP and NU7441 individually or in combination. Gene expression and WB analyses showed that combining 2′3′-cGAMP treatment with DNA-PKcs inhibition boosted the activation of STING-dependent signaling in THP-1^CTRL and THP-1^cGAS−/−, resulting in increased phosphorylation of IRF3 and STING (Fig. S2, F–H) and increased expression of IFNB1, CXCL10, ISG15, and IFIT2 (Fig. 2 F). However, in STING-deficient THP-1 cells, the 2′3′-cGAMP-associated response was abrogated, regardless of NU7441 treatment (Fig. 2 F; and Fig. S2, F–H). This pattern is also reflected at the cytokine production level (Fig. 2 G and Fig. S2 I). To confirm the observed increase in type I IFN responses upon combining 2′3′-cGAMP stimulation and DNA-PKcs inhibition, we performed similar experiments in THP-1 cells expressing a DNA-PKc–targeting guide RNA (THP-1^gDNA-PKcs) (Fig. 2 H and Fig. S2 J). Similar results were obtained in T98G cells knocked out for STING (T98G^STING−/−) (Fig. S2, K and L). In T98G cells knocked out for the IFN-α/β receptor (IFNAR), combining DNA-PKcs inhibition with 2′3′-cGAMP boosted hallmarks of STING activation (Fig. S2 M). Thus, DNA-PKcs–mediated inhibition of 2′3′-cGAMP signaling is independent of cGAS, requires the presence of STING, but is not a consequence of IFNAR-dependent signaling.

We next asked how DNA-PKc–mediated 2′3′-cGAMP inhibition impacts dsDNA-associated signaling. To this aim, T98G^cGAS were transfected or not with dsDNA for up to 48 h, in the presence or absence of NU7441. As expected, gene expression analyses showed that DNA-PKcs inhibition impaired dsDNA-associated type I IFN responses at early time points (up to 6 h; Fig. 2 I). However, this phenotype was reverted at later time points (from 16 to 48 h; Fig. 2 I). Proteome profiler analyses were conducted at 24 h after stimulation with dsDNA in the presence or absence of NU7441, showing that DNA-PKcs inhibition enhanced the production of numerous cytokines and chemokines (Fig. 2 J). Similar experiments were performed in THP-1 cells, where we found that NU7441 treatment, in combination with dsDNA transfection, led to decreased type I IFN responses at early time points (3–6 h; Fig. S2 N) and boosted type I IFN responses at later time points (16–24 h; Fig. S2 N). The net inflammatory output resulted in DNA-PKcs inhibition, enhancing the production of cytokines and chemokines (Fig. S2 O).

Finally, human STING haplotypes deficient for optimal type I IFN response induction have been reported (Aybar-Torres et al., 2024; Simchoni et al., 2025; Yi et al., 2013). We questioned whether DNA-PKcs inhibition may alter their response threshold to 2′3′-cGAMP stimulation. STING knockout THP-1 cells were engineered to express human STING haplotypes (STING-H232, STING-AQ, and STING-HAQ) prior to DNA-PKcs inhibition and 2′3′-cGAMP stimulation. We found that DNA-PKcs inhibition led to increased 2′3′-cGAMP activation of type I IFN responses in cells expressing STING-HAQ and STING-AQ, but not STING-H232 (Fig. S2 P). Interestingly, as observed in Fig. S2 K, DNA-PKcs inhibition led to decreased dsDNA-induced stimulation of type I IFN responses (Fig. S2 Q).

Together, our data establish that DNA-PKcs inhibits 2′3′-cGAMP signaling, thereby contributing to signal termination following activation of the cGAS–STING axis.

Mammalian regulators of 2′3′-cGAMP have been shown to operate selectively on 2′3′-cGAMP (Hou et al., 2023; Li et al., 2014; Mardjuki et al., 2024). We thus questioned the spectrum of DNA-PKcs activity on CDNs.

We assessed the capacity of DNA-PKcs to inhibit CDNs of bacterial origin: 3′3′-cGAMP (Zhang et al., 2013), bis-(3′-5′)-cyclic dimeric adenosine monophosphate (c-di-AMP), and bis-(3′-5′)-cyclic dimeric guanosine monophosphate (c-di-GMP) (Burdette et al., 2011; Jin et al., 2011). We assessed the impact of these CDNs on DNA-PKcs catalytic activity in vitro. We found that 3′3′-cGAMP, but not c-di-AMP nor c-di-GMP, inhibited DNA-PKcs (Fig. 3 A and Fig. S3 A). Docking analyses supported that 3′3′-cGAMP, but not c-di-AMP or c-di-GMP, can interact with DNA-PKcs in its catalytic cleft, adopting a stable conformation (energy −29,962.4 Kcal/mol; Table 2). We also observed that recombinant DNA-PKcs interacts with 3′3′-cGAMP but not with c-di-AMP (Fig. 3 B).

To test whether the interaction between DNA-PKcs and 3′3′-cGAMP translates into signaling inhibition, we treated THP-1 and T98G cells with 3′3′-cGAMP alone or in combination with NU7441. Combining 3′3′-cGAMP treatment with DNA-PKcs inhibition enhanced pSTING (Fig. 3 C and Fig. S3 B) and increased expression of STING activity-associated transcripts such as IFNB1, ISG15, IFIT1, and CCL5 in both cell types (Fig. 3 D and Fig. S3 C) when compared with cells treated only with 3′3′-cGAMP. We also measured increased production of IFNβ, CXCL10, and CCL5 when 3′3′-cGAMP and NU7441 treatments were combined in THP-1 cells (Fig. 3 E). To confirm that this effect is directly mediated by DNA-PKcs, we performed 3′3′-cGAMP treatment in THP-1^CTRL and THP-1^gDNA-PKcs. We observed that type I IFN responses were potentiated in THP-1^gDNA-PKcs as compared with control cells (Fig. 3 F). Similar experiments were next conducted on human primary macrophages (Fig. S3, D and E). Gene expression analyses showed that combining NU7441 with 3′3′-cGAMP treatment boosted IFNB, ISG15, IFIT1, and CCL5 expression (Fig. 3 G and Fig. S3 F) when compared with treatment with 3′3′-cGAMP alone. Thus, the negative regulatory role of DNA-PKcs on 3′3′-cGAMP bioactivity can also be measured in primary human cells.

Combined, our data show that DNA-PKcs selectively interacts with 3′3′-cGAMP to inhibit STING-dependent signaling in immune and nonimmune cells but does not interfere with other tested bacterial CDNs.

STING is an important biomedical drug target for which a plethora of small molecule agonists are in clinical trials (Hines et al., 2023). We questioned whether DNA-PKcs may interact with these compounds and impact their potency.

We first performed DNA-PKcs activity assays in the presence of increasing doses of two STING agonists, E7766 and ADU-S100. E7766, but not ADU-S100, inhibited DNA-PKcs activity (Fig. 3 H), suggesting interaction of E7766 with DNA-PKcs catalytic domain. Functional assessment of the impact of DNA-PKcs on the activity of these compounds was tested by treating cells with E7766 or ADU-S100 in combination with NU7441 or not. DNA-PKcs inhibition boosted type I IFN responses when combined with E7766 (Fig. 3 I and Fig. S3 G) but not with ADU-S100 (Fig. 3 J and Fig. S3 H). These data were supported by in silico docking analyses that showed that E7766 presents a strong capacity to interact with DNA-PKcs, as compared with ADU-S100 (Table 2).

Additional experiments were conducted using the diABZI STING agonist, showing that combination with NU7441 potentiated type I IFN responses (Fig. 3 K and Fig. S3 I). Similar to what was observed for 2′3′-cGAMP and 3′3′-cGAMP, we found that treatment of THP-1^gDNA-PKcs with E7766 and diABZI led to enhancement of type I IFN responses when compared with control cells (Fig. S3, J and K).

These data, thus, show that DNA-PKcs inhibition selectively alters the potency of STING agonists.

We next sought to assess whether the inhibitory effect of DNA-PKcs on STING agonists is recapitulated in vivo. Conflicting data were reported concerning the capacity of DNA-PKcs to act as a cytosolic dsDNA sensor in murine cells (Ferguson et al., 2012; Burleigh et al., 2020), warranting assessment of the inhibitory role of DNA-PKcs on 2′3′-cGAMP in murine models. We thus evaluated the impact of combining NU7441 with diABZI or E7766 on STING-dependent signaling in the RAW264.7 myeloid murine cell line. Similar to what we observed in human cell lines, the combination of DNA-PKcs inhibition with STING agonists boosted the expression of Ifnβ, as well as of the Isg15, Ifit1, Ccl5, and Cxcl10 ISGs (Fig. 4 A and Fig. S4 A). We next isolated primary splenocytes from mice and assessed the impact of combining 3′3′-cGAMP or diABZI with NU7441 (Fig. 4 B). We found that combining NU7441 with 3′3′-cGAMP boosted inflammatory gene expression (Fig. 4 C) as well as cytokine production (Fig. 4, D and E). Similarly, co-treatment with NU7441 and diABZI boosted expression and production of inflammatory cytokines (Fig. 4, F and G). Thus, DNA-PKcs inhibits CDN-associated signaling in murine cells.

Sting-deficient (Sting^−/−) and Ifnar-deficient (Ifnar^−/−) mice and WT littermates were subjected or not to i.p. injections of NU7441 prior to injection of diABZI, followed by analysis of hallmarks of STING activation in peritoneal fluids, peritoneal macrophages, peritoneum, and spleens (Fig. 4 I). Gene expression analyses showed that combining DNA-PKcs inhibition with a STING agonist boosted the expression of transcripts associated with STING signaling in the peritoneum (Fig. 4 J), peritoneal macrophages (Fig. 4 K), and the spleen (Fig. 4 L), accompanied by enhanced Cxcl10 production in the peritoneal fluid and in the spleen (Fig. 4, M and N). IFN response genes were not increased in any of the tested conditions in Sting-deficient mice (Fig. 4, J–N), supporting that DNA-PKcs–dependent regulation of type I IFN responses relies on Sting. Absence of upregulation in Ifnar-deficient mice supports that in WT mice (Fig. 4, J–N), bioactive IFNs are produced to stimulate the expression of tested ISGs.

Altogether, these data demonstrate DNA-PKcs inhibits agonist-dependent STING signaling in vivo.

We next asked whether the negative role of DNA-PKcs on STING agonists translates into a biologically relevant functional response. We therefore assessed the impact of DNA-PKcs inhibition on the establishment of an antiviral state. We used two DNA viruses, namely the Herpes simplex virus 1 (HSV-1) and monkeypox virus (MPXV), and one RNA virus, the vesicular stomatitis virus (VSV), all of which are sensitive to IFN responses.

We treated T98G cells with 3′3′-cGAMP, diABZI, and E7766 STING agonists, individually or in combination with NU7441, prior to infection with the 3 aforementioned viruses and measurement of the percentage of infected cells (Fig. 5 A). As expected, treatment with STING agonists alone was sufficient to decrease infection by HSV-1 (Fig. 5, B and C), MPXV (Fig. 5 D and Fig. S5 A), and VSV (Fig. 5 E and Fig. S5 B). Combining STING agonists with NU7441 further decreased viral infection (Fig. 5, B–E; and Fig, S5, A and B). Dependency on DNA-PKcs for this phenotype was confirmed using DNA-PKcs–deficient cells (Fig. S5 C). Of note, treatment with the NU7441 alone increased infection, which likely reflects a direct antiviral role of DNA-PKcs.

We next assessed the impact of DNA-PKcs antagonism on the STING agonist–induced antiviral status in human primary cells. Human primary macrophages (Fig. S5 D) were treated with NU7441, in combination or not with STING agonists, prior to infection with VSV. As observed in T98G cells, combining DNA-PKcs inhibition with STING agonism increased the inhibition of infection with VSV (Fig. 5, F and G; and Fig. S5, E and F). Thus, combining STING activation and DNA-PKcs inhibition promotes the establishment of a higher antiviral state, supporting the role of DNA-PKcs in the negative regulation of STING agonists.

Finally, we investigated the contribution of DNA-PKcs in the resolution of virus-induced inflammatory responses. To this aim, mouse embryonic fibroblasts were infected with HSV-1 in the presence of NU7441 (“pre-treatment”) or treated with NU7441 6 h after infection (“post-treatment”), prior to gene expression analyses (Fig. 5 H). Gene expression analyses showed that while DNA-PKcs inhibition during HSV-1 infection led to decreased inflammatory responses (Fig. 5 I), treatment with NU7441 6 h after infection led to increased inflammatory gene expression (Fig. 5 I). This suggests that during early steps of infection (pretreatment), DNA-PKcs is antiviral, while at later steps, it plays a proviral role. These data further support a role of DNA-PKcs in the resolution of 2′3′-cGAMP signaling.

Together, these data show that DNA-PKcs is a regulator of STING-dependent antiviral responses in cell lines and primary human cells, likely through the control of 2′3′-cGAMP. This supports that in addition to previously reported roles of DNA-PKcs in activating type I IFN responses alongside the cGAS–STING pathway, DNA-PKcs buffers STING overactivation through the trapping of 2′3′-cGAMP.

We describe a conserved function of DNA-PKcs in the regulation of 2′3′-cGAMP and 3′3′-cGAMP CDN-associated inflammatory responses. We show that DNA-PKcs binds to those CDNs selectively, decreasing their availability for interaction with STING and, thereby, buffering their activity. This is complementary to other mechanisms described to control CDN activity through the regulation of extracellular 2′3′-cGAMP levels (Hou et al., 2023; Li et al., 2014; Mardjuki et al., 2024). That DNA-PKcs also ensures termination of 3′3′-cGAMP–associated inflammatory responses appears as a distinctive feature and may have evolved as a defense mechanism toward 3′3′-cGAMP–producing pathogenic bacteria (Davies et al., 2012). Since DNA-PKcs is highly conserved in the tree of life (Lees-Miller et al., 2021), one may question whether this mechanism operates as a general mechanism to prevent overactivation of inflammatory responses in response to pathogen infection.

Conflicting reports exist in the literature, opposing a pro- and anti-inflammatory role of DNA-PKcs in nucleic acid immunity (Burleigh et al., 2020; Hristova et al., 2024; Taffoni et al., 2023; Sun et al., 2020). Our work reconciles these prior studies, showing that DNA-PK is indeed involved as a detector of incoming DNA viruses, which potentiates the initial activation of type I IFN responses (Burleigh et al., 2020; Ferguson et al., 2012; Taffoni et al., 2023), but also inhibits CDN to ensure signal termination. Combined, the current evidence therefore suggests that DNA-PK has a dual role in modulating cGAS–STING signaling, regulating the amplitude and kinetics of dsDNA responses, as well as homeostatic inflammation, ultimately decreasing overall signaling output. Early upon exposure to dsDNA species in the cytosol, DNA-PK senses incoming DNA in parallel to the cGAS–STING pathway (Burleigh et al., 2020; Ferguson et al., 2012; Taffoni et al., 2023) and catalyzes activating phosphorylation of cGAS (Taffoni et al., 2023), while excess of produced 2′3′-cGAMP is trapped by DNA-PKcs to prevent overactivation of STING. This fits with the multifunctional roles of DNA-PK in the regulation of inflammatory responses (Wu et al., 2024).

We also demonstrate that 2′3′-cGAMP and 3′3′-cGAMP block DNA-PKcs catalytic activity, which may have an impact on its ability to repair double-strand breaks. However, previous studies have shown that the levels of DNA-PKcs do not correlate with DNA repair functions (Meek et al., 2008) and that other pathways can take over NHEJ-mediated repair when DNA-PKcs is nonfunctional (Blackford and Jackson, 2017). Disentangling the relative contribution of DNA-PK repair function and regulation of inflammatory responses is therefore a difficult task. It is, however, possible to speculate that CDN-mediated DNA-PKcs inhibition may have a distinct impact on specific repair functions, such as in micronuclei damage (Harding et al., 2017) or immunoglobulin class switching (Bjorkman et al., 2015), which are contexts where DNA-PK inhibition would impact immune function. This further suggests potential cell type–specific roles of DNA-PKcs signaling functions. In addition, owing to the abundance of DNA-PKcs, the small percentage of the protein sufficient for repair function and functional redundancy of repair pathways, it has been previously speculated that DNA-PKcs role beyond DNA repair may be its most prominent function (Taffoni et al., 2021; Wu et al., 2024). How CDN-mediated inhibition of DNA-PKcs impacts those other DNA-PKcs functions would require further investigations.

The ability of DNA-PKcs to control CDN-associated inflammatory responses is relevant for drugs designed to activate STING in immunosuppressed contexts such as the tumor microenvironment. Existing STING agonists have distinct features. Among tested compounds in the present study, ADU-S100, E7766, and diABZi all present the ability to interact with the STING CDN-binding domain. ADU-S100 is a synthetic CDN STING agonist, structurally similar to c-diAMP (Corrales et al., 2015), and both compounds present a poor capacity to interact with DNA-PKcs. In contrast, E7766 and diABZI that are macrocycle-bridged and non-nucleotide-based STING agonists (Huang et al., 2022) were found to have their bioactivity regulated by DNA-PKcs, suggestive of interaction with DNA-PKcs. Precise assessment of the contact points of STING agonists with the catalytic cleft of DNA-PKcs is warranted to fully evaluate what explains these distinctive features. Furthermore, whether the capacity of these compounds to interact with DNA-PKcs correlates with DNA repair deficiency-associated side-effects remains to be assessed. Thus, the status of DNA-PKcs in target tissues may be important to evaluate in order to predict the capacity of STING agonists to induce beneficial inflammation. For instance, in the context of cancer immunotherapy where STING agonists are used to boost inflammatory responses (Hines et al., 2023), the presence of DNA-PKcs anomalies, such as frequent copy number variations (Yang et al., 2020), must be assessed. It is therefore important to take into account the status of DNA-PKcs in patients enrolled in STING agonist–based clinical trials for better stratification.

Further work will be required to demonstrate the impact of DNA-PKcs inhibition in the context of cancer. The risks associated with DNA-PK–mediated DNA repair impairment should also be assessed in order to evaluate the full impact on anti-tumoral responses. In addition, DNA-PKcs has been reported to be involved in B cells Immunoglobulin class switching (Bjorkman et al., 2015). Although alternative routes to NHEJ may be used to ensure IgG class switching, the impact on B cell development and functionality must be assessed formally. Finally, we identify that CDNs bind into the catalytic pocket of DNA-PKcs. Point mutations may allow the identification of precise contact points and allow for rationalized drug design of STING agonists that do not impede DNA-PKcs catalytic activity.

T98G cells were a gift from Caroline Goujon (Institut de Recherche en Infectiologie de Montpellier, Montpellier, France). Control THP-1, cGAS-deficient THP-1 (THP-1^cGas−/−), and STING-deficient THP-1 (THP-1^Sting−/−) were a gift of S.R. Paludan (Aarhus University, Aarhus, Denmark). Vero (African green monkey), HEK293T (Homo sapiens), and THP-1 (H. sapiens) cells were obtained from the American Type Culture Collection (ATCC).

T98G^STING−/−, T98G^IRF3−/−, and their control counterparts were previously generated using the CRISPR-Cas9 technology (Taffoni et al., 2023). Parental and genetically engineered T98G, HEK293T, and Vero cells were maintained in DMEM supplemented with 10 % FBS (Eurobio), 1% L-glutamine (Lonza), and 1% penicillin/streptomycin (Lonza). THP-1 cells (Cat#TIB-202, ATCC) were cultured in Roswell Park Memorial Institute (RPMI, Lonza) supplemented with 20% FBS, 1% penicillin/streptomycin, and 1% L-glutamine. All cell lines were maintained at 37°C under 5% CO₂.

Buffy coats were obtained from the Etablissement Français du Sang (EFS). Human primary monocytes were isolated using the StraightFrom Buffy Coat CD14 Microbead Kit (Miltenyi) and differentiated into macrophages for 5 days using 10 µg/ml GM-CSF.

Buffy coats from healthy donors were obtained from the EFS. CD14⁺ monocyte isolations from three healthy donors were performed using the EasySep Human CD14 Positive Selection Kit II (#17858; STEMCELL Technologies) and differentiated straight toward monocyte-derived macrophages (MDMs) upon 5–7 days incubation with 500 U/ml M-CSF in IMDM supplemented with 10% FBS, 100 U/ml penicillin, 100 μg/ml streptomycin, 1 mM HEPES (H3537-100 Ml; Sigma-Aldrich), 1 mM sodium pyruvate (S8636-100 Ml; Sigma-Aldrich), and 1× MEM Non-Essential Amino Acids (M7145-100 Ml; Sigma-Aldrich). Expression of the following macrophage markers were routinely analyzed by flow cytometry for each donor: CD14, CD68, and HLA-DR.

THP-1 cells stably expressing human STING haplotypes were generated using THP1-Dual KO-STING as detailed in (Guha et al., 2025, Preprint). In brief, cells were transduced with retroviral vectors expressing N-terminal HA-tagged HsSTING haplotypes/mutants (HsSTING-R232H, HsSTING-AQ, and HsSTING-HAQ) and selected with 2 μg/ml puromycin.

WT mice and mice deficient for STING (Sting^−/−) and IFNAR (Ifnar^−/−) were bred and housed under pathogen-free conditions at CNRS animal facility (ZEFI UMR5535, Montpellier, France). Mouse strains are C57/BL6J, and backcrosses to reference strains are performed every five generations. They were maintained in a 12-h light-dark cycle with food and water ad libitum, following European and local legislation. Age-matched, 8- to 14- wk-old mice were used for experiments. All animal experiments complied with the French government animal experiment regulations and ARRIVE guidelines. The protocols were submitted to the “Ethics Committee for Animal Experimentation of CNRS Campus Montpellier” under number G3417216 and approved by the French minister under APAFIS #41092.

The following drugs were used: NU7441 (Bio-Techne/Tocris, #3712, CID: 11327430); DMSO (Sigma-Aldrich, D2650, CID: 679); 2′3′-cGAMP (InvivoGen, tlrl-nacga23-02, CID: 137120248); 3′3′-cGAMP (InvivoGen, tlrl-nacga, CAS number: 849214-04-6); c-di-AMP (InvivoGen, tlrl-nacda, CAS number: 2734909-87-4/54447-84-6 [free acid]); c-di-GMP (InvivoGen, tlrl-nacdg, CAS number: 2222132-40-1/61093-23-0 [free acid]); diABZI (InvivoGen, tlrl-diabzi-2, CID: 137701219, CAS number: 2138299-34-8); E7766 (MedChemExpress, HY-111999A); ADU-S100 ammonium salt (MedChemExpress, HY-12885B); fluorinated 3′3′-cGAMP (InvivoGen, tlrl-nacgaf-05; CAS number: not available).

To generate the T98G^IFNAR−/− knockout and control cell lines, lentiviral particles were produced by co-transfection of 2 × 10⁶ 293T cells with 5 μg of LentiCRISPRv2GFP plasmid (#82416; Addgene) expressing the gRNA targeting the gene of interest or nontargeting control gRNA, 5 μg of psPAX2, and 1 μg of pMD2.G, using the standard calcium phosphate transfection protocol.

T98G cells were transduced with lentiviral particles, and 72 h after transduction, GFP-positive cells were sorted and pooled in a 6-well plate using a BD FACS melody. Cells were next amplified, and IFNAR levels were controlled by WB.

HEK293T^{DNA-PKcs−/−} clones and HEK293T^CTRL cells were generated by jetPRIME (Polyplus)-mediated transfection of the LentiCRISPRv2GFP plasmid expressing the gRNA targeting the gene of interest or nontargeting control gRNA. 2 days later, GFP⁺ single-cell clones were selected on a BD FACSAria. Absence of DNA-PK expression was verified by WB.

The following primers were used:

CTRL: 5′-CAC​CGA​GCA​CGT​AAT​GTC​CGT​GGA​T-3′, 5′-AAA​CAT​CCA​CGG​ACA​TTA​CGT​GCT​C-3′;

IFNAR: 5′-CACCGAGCTGACACTCACCTTCCCC-3′, 5′-AAACGGGGAAGGTGAGTGTCAGCTC-3′;

PRKDC: 5′-CACCGGCAGGAGACCTTGTCCGCTG-3′, 5′-AAACCAGCGGACAAGGTCTCCTGCC-3′.

DNA-PK–knockout THP-1 cells were generated using the CRISPR/Cas9 ribonucleoprotein (RNP) gene-editing system according to the manufacturer’s instructions (Synthego) and have been described by (Rabinowitz et al., 2025). Briefly, purified recombinant Cas9 protein was complexed with target-specific single-guide RNA (sgRNA) to form Cas9–sgRNA RNP complexes by incubation for 10 min at room temperature (RT) immediately prior to delivery. Cas9 RNPs were introduced into THP-1 cells by nucleofection using the Lonza 4D-Nucleofector System, optimized for THP-1 cells with the SG Cell Line 4D-Nucleofector X Kit S (Lonza; Catalog #V4XC-3032). Transient delivery of Cas9 RNPs enables efficient genome editing while minimizing prolonged Cas9 expression and reducing off-target effects. To generate stable knockout cell lines, edited cells were single-cell cloned by limiting dilution. Individual clones were expanded and validated at both the genomic and protein levels to confirm biallelic gene disruption.

pcDNA3.1-nFlag-DNA-PK was cloned from synthesized gene fragments (eBlocks, IDT) based on consensus sequence ENST00000314191.7 modified for codon optimization. The gRNA-binding site targeted for knockout was modified to prevent targeting by CRISPR/Cas9 constructs. An N-terminal 3X Flag sequence, including a (GGGGS)₃ linker, was added to the synthesized template. The pcDNA3.1 backbone was digested using BamHI and EcoRI, dephosphorylated using Quick CIP (NEB), and purified on agarose gel. eBlocks were amplified with KAPA HIFI HotStart Ready Mix (Roche) and purified on agarose gel. Gibson cloning was performed in two steps, with amplified and purified intermediate constructs, using NEBuilder HiFi DNA Assembly Master Mix and NEB Stable Competent Escherichia coli grown at 30°C. Resulting plasmid preparations were verified by whole plasmid sequencing.

Silencing of DNA-PKcs was achieved in T98G cells using siRNAs and INTERFERin (Polyplus) following the manufacturer’s instructions. siRNAs (Dharmacon; Horizon Discovery) were used:

siCTRL: 5′-CGU​ACG​CGG​AAU​ACU​UCG​AUU-3′;

siDNA-PKcs: 5′-GAU​CGC​ACC​UUA​CUC​UGU​UUU-3′;

siKU70: 5′-ACA​AGC​AGU​GGA​CCU​GAC​UU-3′.

Cells were plated in 6-well plate or 100-mm dishes. 18 h later, cells were pretreated with 2 μM NU7441 (#3712; Bio-Techne/Tocris) for 1 h in DMEM or RPMI prior to transfection with 2′3′-cGAMP (10 µg/ml, InvivoGen) using Lipofectamine 2000 (Thermo Fischer Scientific), following the manufacturer’s instructions. 1, 3, or 6 h after transfection, cells were harvested and stored at −80°C prior to 2′3′-cGAMP, protein, or RNA extraction.

For other drug treatments, following 1-h preincubation with Nu7441, cells were treated with 2′3′-GAM(PS)2 (Rp/Sp) (InvivoGen) or fluorinated 3′3′-cGAMP (InvivoGen) at 10 µg/ml, with diABZI compound 3 (InvivoGen) at 10 µM, with E7766 (MedChemExpress) at 1 µM or ADU-S100 (MedChemExpress) at 50 µM. 6 h later, cells were harvested and stored at −80°C prior to protein or RNA extraction.

Mice were pretreated i.p. with either NU4474 (8 µg in 200 μl) or DMSO (4% in 200 μl) 1 h before being challenged with the diABZI (1 µg in 200 µl of PBS). After 6 h, the mice were euthanized.

Peritoneal lavage (PL) was performed 6 h after the last challenge by rinsing the peritoneal cavity with 5 ml of cold PBS via a cannula inserted in the peritoneum.

The acellular fraction of the lavage (supernatant) was collected after centrifugation and was used to assess Cxcl10 by ELISA. The cellular fraction containing peritoneal macrophages was used for the quantification of mRNA expression by RT-qPCR. The peritoneum and spleen were collected for gene expression analysis and Cxcl10 production measurement.

Total RNA was isolated with TRIzol (Invitrogen). RNA concentration was measured with a NanoDrop spectrophotometer (ND-1000, NanoDrop Technologies) prior to treatment with TURBO DNase (Ambion) and cDNA synthesis from 1 to 2 μg RNA using SuperScript IV (Invitrogen) with Oligo(dT) and quantification with a LightCycler 480 cycler (Roche) using SYBR Green Master Mix (Takara) and appropriate primers. Relative quantities of the transcript were calculated using the ΔΔCt method, using the GAPDH for normalization, except for THP1 analyses, where a normalization with hypoxanthine phosphoribosyltransferase 1 (HPRT1) and actin beta (ACTB) was performed. For normalization in mouse cell lines, Gapdh or heat shock protein 90 (Hsp90) were used. The RT-qPCR primers used are listed below:

GAPDH: For: 5′-CTG​GCG​TCT​TCA​CCA​CCA​TGG-3′, Rev: 5′-CAT​CAC​GCC​ACA​GTT​TCC​CGG-3′;

ACTB: For: 5′-GGA​CTT​CGA​GCA​AGA​GAT​GG-3′, Rev: 5′-AGC​ACT​GTG​TTG​GCG​TAC​AG-3′;

HPRT1: For: 5′-TGA​CAC​TGG​CAA​AAC​AAT​GCA-3′, Rev: 5′-GGT​CCT​TTT​CAC​CAG​CAA​GCT-3′;

IFNB1: For: 5′-GAA​TGG​GAG​GCT​TGA​ATA​CTG​CCT-3′, Rev: 5′-TAG​CAA​AGA​TGT​TCT​GGAGCATCTC-3′;

ISG15: For: 5′-GAT​CAC​CCA​GAA​GAT​CGG​CG-3′, Rev: 5′-GTT​CGT​CGC​ATT​TGT​CCA​CC-3′’

CXCL10: For: 5′-GAA​AGC​AGT​TAG​CAA​GGA​AAG​GTG-3′, Rev: 5′-ATGTAGGGAAGTGATGGGAGAGG-3′;

IFIT1: For: 5′-GCC​TTG​CTG​AAG​TGT​GGA​GGA​A-3′, Rev: 5′-ATC​CAG​GCG​ATA​GGC​AGA​GAT​C-3′;

IFIT2: For: 5′-GGA​GCA​GAT​TCT​GAG​GCT​TTG​C-3′, Rev: 5′-GGA​TGA​GGC​TTC​CAG​ACT​CCA​A-3′;

CCL5: For: 5′-CCT​GCT​GCT​TTG​CCT​ACA​TTG​C-3′, Rev: 5′-ACA​CAC​TTG​GCG​GTT​CTT​TCG​G-3′;

Gapdh: For: 5′-TTC​ACC​ACC​ATG​GAG​AAG​GC-3′, Rev: 5′-GGC​ATC​GAC​TGT​GGT​CAT​GA-3′;

Hsp90: For: 5′-GTC​CGC​CGT​GTG​TTC​ATC​AT-3′, Rev: 5′-GCA​CTT​CTT​GAC​GAT​GTT​CTT​GC-3′;

Ifnb: For: 5′-CTG​CGT​TCC​TGC​TGT​GCT​TCT​CCA-3′, Rev: 5′-TTC​TCC​GTC​ATC​TCC​ATA​GGG​ATC-3′;

Cxcl10: For: 5′-ATG​ACG​GGC​CAG​TGA​GAA​TG-3′, Rev: 5′-TCA​ACA​CGT​GGG​CAG​GAT​AG-3′;

Ccl5: For: 5′-CAG​CAA​GTG​CTC​CAA​TCT​TGC-3′, Rev: 5′-CCA​CTT​CTT​CTC​TGG​GTT​GGC-3′;

Ifit1: For: 5′-TTT​CCG​TAG​GAA​ACA​TCG​CGT-3′, Rev: 5′-TGT​TGC​TTG​TAG​CAG​AGC​CC-3′;

Isg15: For: 5′-GTG​CTC​CAG​GAC​GGT​CTT​AC-3′, Rev: 5′-CTC​GCT​GCA​GTT​CTG​TAC​CA-3′;

Il6: For: 5′-GAC​TTC​CAT​CCA​GTT​GCC​TTC​T-3′, Rev: 5′-TTC​TCT​CCG​GAC​TTG​TGA​AGT-3′;

Tnfα: For: 5′-CTG​TAG​CCC​ACG​TCG​TAG​C-3′, Rev: 5′-TTG​AGA​TCC​ATG​CCG​TTG-3′.

IFNβ, CCL5, and CXCL10 concentrations in the PL or spleen were measured by ELISA (R&D System, Minneapolis) according to the manufacturers’ instructions. Data were acquired on the FLUOstar Omega equipment (BMG-LABTECH).

Cells were lysed in five packed cell volumes of TENTG-150 (20 mM Tris-HCl [pH 7.4], 0.5 mM EDTA, 150 mM NaCl, 10 mM KCl, 0.5% Triton X-100, 1.5 mM MgCl2, and 10% glycerol, supplemented with 10 mM β-mercaptoethanol, 0.5 mM PMSF, and phosphatase inhibitor [Sigma-Aldrich]) for 30 min at 4°C. Lysates were centrifuged for 30 min at 14,000 g, and supernatants were collected for WB. Protein quantification was performed using Bradford assay (Bio-Rad).

Protein samples were prepared in Laemmli buffer and heated at 95°C for 5 min prior to resolution by sodium dodecyl sulfate-polyacrylamide gel electrophoresis (SDS-PAGE) using 4–15% Mini-PROTEAN TGX Precast Protein Gels (Bio-Rad). Proteins were transferred onto nitrocellulose membranes (Bio-Rad Trans Blot Turbo). Proteins were visualized on membranes using Ponceau S solution (Sigma-Aldrich) prior to 30 min blocking with PBS containing 0.1% Tween (PBS-T) supplemented with 5% milk. Membranes were subsequently incubated with primary antibodies in 5% milk/PBS-T (1:1,000 dilution, except when indicated) for 2 h at RT. Primary antibodies used include anti-DNA-PKcs (A300-517, Bethyl, 1:500), anti-cGAS (15102) Cell Signaling Technology), anti-pSTING Ser366 (19781; Cell Signaling Technology), anti-STING (13647, Cell Signaling Technology), anti-pIRF3 Ser386 (ab76493; Abcam), anti-IRF3 (11904; Cell Signaling Technology), anti-HSP90 (4877; Cell Signaling Technology), anti-GAPDH (60004-1-Ig; Proteintech Europe, 1:5,000), and αTUBULIN (66031-1-Ig; Proteintech Europe, 1:10,000). Membranes were incubated with horseradish peroxidase (HRP)-coupled secondary antibodies (Cell Signaling Technology) at 1:2,000 dilution for 1 h at RT. Immunoreactivity was detected by chemiluminescence (SuperSignal West Pico or Femto, Thermo Fisher Scientific). Images were acquired on a ChemiDoc (Bio-Rad) or Amersham Imager 680 (GE Healthcare).

For immunoprecipitation of endogenous DNA-PKcs, T98G cells were lysed in five packed cell volume of the lysis buffer 20 mM Tris-HCl (pH 7.4), 150 mM NaCl, 10 mM KCl, 0.5% Triton X-100, 1.5 mM MgCl₂, and 10% glycerol, supplemented with 10 mM β-mercaptoethanol and 0.5 mM PMSF, for 30 min at 4°C on a wheel. Lysates were centrifuged for 30 min at 14,000 g, and supernatants were collected for immunoprecipitation. Endogenous immunoprecipitation was performed using DNA-PKcs–targeting antibody (A300-517A; Bethyl) or Rabbit IgG (Santa Cruz) as a negative control. After an overnight incubation at 4°C on a wheel, immunoprecipitation was performed using Protein G Sepharose Fast Flow beads (Sigma-Aldrich). After three washes in lyses buffer, part of the bound material was either eluted in Laemmli buffer for WB analyses or used to assess DNA-PKcs:2′3′-cGAMP interaction.

FLAG-DNA-PKcs immunoprecipitation was performed by expressing FLAG-DNA-PKcs in 293T^{DNA-PKcs−/−} for 48 h prior to whole-cell extract preparation as above and immunoprecipitation using anti-FLAG dynabeads. Immunoprecipitates were analyzed by WB or used to assess DNA-PKcs:2′3′-cGAMP interaction.

To assess DNA-PKcs:2′3′-cGAMP interaction, bound material was incubated with an excess of 2′3′-cGAMP diluted in lyses buffer, for 30 min on ice. After three washes with lyses buffer, bound material was subjected to proteinase K treatment (15 min at RT) to denature DNA-PKcs and allow the release of potential bound 2′3′-cGAMP. Proteinase K were inactivated, incubating samples 10 min at 95°C, prior to measurement of 2′3′-cGAMP level by ELISA.

For the immunoprecipitation of recombinant DNA-PKcs, 2 µg of anti-DNA-PKcs antibody or the corresponding IgG control was incubated with 70 ng of recombinant DNA-PKcs protein (Promega) in 500 μl of lysis buffer (20 mM Tris-HCl [pH 7.4], 150 mM NaCl, 10 mM KCl, 0.5% Triton X-100, 1.5 mM MgCl2, and 10% glycerol, supplemented with 10 mM β-mercaptoethanol and 0.5 mM PMSF). The mixture was incubated overnight at 4°C on a rotating wheel. The next day, the samples were centrifuged at 12,000 rpm for 10 min, and the supernatant was collected. 30 µl of Sepharose beads (5 μl per immunoprecipitation point, for a total of six points) were added to the supernatant and incubated at 4°C for 45 min on a rotating wheel. Following incubation, the beads were centrifuged at 2,500 rpm and washed three times with the lysis buffer (without β-mercaptoethanol and PMSF). After washing, the immunoprecipitated complexes were incubated with the tested CDNs, including 2′3′-cGAMP, 3′3′-cGAMP, and c-di-AMP, with or without NU7441. The incubation was performed with a CDN or nu7441 concentration 10 times higher than that of DNA-PKcs for 30 min on ice. After the incubation, the beads were washed, and the samples were digested with proteinase K before measuring the amount of CDNs bound to DNA-PKcs by ELISA. 2′3′-cGAMP, 3′3′-cGAMP, and c-di-AMP ELISAs were performed according to the manufacturer’s protocol using the Cayman Chemical 2′3′-cGAMP ELISA Kit (CAY501700), 3′3′-cGAMP ELISA Kit (CAY502130), and c-di-AMP ELISA Kit (CAY501960), respectively.

The ability of DNA-PK to hydrolyze ATP was measured by using the DNA-PK Kinase Enzyme System coupled to the ADP-Glo Assay (Promega), following manufacturer’s instructions. Briefly, 30 units of human recombinant DNA-PK were incubated for 60 min at RT with 1 µM ATP, 1× activator, and 0.2 µg/μl substrate in presence of increasing doses of 2′3′-cGAMP (InvivoGen), 3′3′-cGAMP (InvivoGen), c-di-GMP (InvivoGen), c-di-AMP (InvivoGen), ADU-S100 (MedChemExpress), or E7766 (MedChemExpress). Incubation with 4 μM NU7441 was included as a positive control. After 1 h, the reactions were incubated with 25 μl of ADP-Glo reagent for 40 min at RT, followed by incubation with 50 μl of Kinase Detection reagent for 30 min at RT. Luminescence was measured using the FLUOstar Omega microplate reader (BMG Labtech).

For TSA from immunoprecipitation of purified proteins, HEK-293T cells were transfected with plasmids encoding WT or mutant of DNA-PKcs protein. After 48 h, cells were lysed in TENTG-150 buffer. Anti-FLAG immunoprecipitation was performed, and proteins were eluted with FLAG peptide in TSA buffer (20 mM Tris-HCl, pH 7.5, 150 mM NaCl, 5% glycerol, 1 mM DTT, and 0.5 mM PMSF). For TSA from whole-cell lysate, cells were lysed in TSA buffer by five freeze–thaw cycles (liquid nitrogen/37°C). Lysates were centrifuged at 10,000 rpm for 5 min at 4°C to remove debris. For both approaches, eluates or whole-cell lysates were incubated with water or 2′,3′-cGAMP (1,000 µM) on ice for 30 min. Samples were aliquoted (20 μl) into PCR strip tubes and subjected to a thermal gradient from 46.9 to 61.7°C for 3 min. Following the thermal gradient, samples were cooled on ice for 5 min and centrifuged at 14,000 rpm for 20 min at 4°C. The supernatant containing soluble and non-denatured proteins was collected and mixed with Laemmli buffer before loading onto a gel for WB analysis

1 × 10⁵ cells were seeded per well on 96-cell U-bottom plate for cytometer and on Revvity PhenoPlate 96-well for microscopy. Cells were pretreated with Nu7441 (DNA-PK inhibitor) at a concentration of 2 μM for 1 h, followed by incubation of the agonist 3′3′-cGAMP (fluorinated) at concentrations of 0.1 and 1 μg/ml for 2 h. Then, cells were infected with VSV-GFP at multiplicity of infection (MOI) 0.5 and for 1 h. Cell medium was changed, and cells were either left incubated for 24 and 48 h for VSV-GFP. Cells were then fixed with 4% paraformaldehyde (PFA), and plates were acquired either using the NovoCyte apparatus (Agilent) for flow cytometry or the Opera Phenix (Revvity) for microscopy. Flow cytometry data were then analyzed using the NovoExpress software (Agilent), while fluorescence microscopy data were analyzed with Harmony software (Revvity).

T98G cells grown on coverslips were transfected with 1 µg/ml 2′3′-cGAMP–iFluor 488 (#20320; AAT BioQuest) or DMSO (control) using JetPRIME (Sartorius). After 6 h, cells were washed with PBS, fixed with 4% formaldehyde for 20 min, then blocked/permeabilized in PBS supplemented with 0.5% Triton X-100 and 5% BSA for 1 h. Cells were incubated with anti–DNA-PK (#ab32566; Abcam, 1:100) for 1 h, washed, and exposed to Alexa Fluor 594 goat anti-rabbit (#R37117; Invitrogen) for 1 h. Nuclei were stained with DAPI (1 µg/ml, 10 min) and then mounted with VECTASHIELD PLUS (#H-1900; Vector).

Confocal images were acquired on a Zeiss LSM 980 NLO. 10 to 20 images were obtained per condition with identical channel adjustments. Image analysis (CellProfiler) was used to identify 2′3′-cGAMP and DNA-PK foci (8–80 pixels), define cytosolic DNA-PK by inverted masks, and quantify colocalization (Relate Objects).

2 × 10⁵ MEFs were seeded per well in a 6-well plate in a final volume of 2 ml of DMEM. Cells were either infected or not with HSV-1-KOS-64 at MOI 5, then treated with either NU7441 or DMSO as control. In the first, cells were pretreated with NU7441, then infected with HSV. In the second setup, cells were infected with HSV and then at 6 h after infection, treated with NU7441. For this condition, the two 6-well plates were collected only at 16 h after infection for qPCR analysis.

Spleens were harvested from euthanized WT C57BL/6J mice under sterile conditions. Spleens were placed in cold RPMI-1640 medium (10% FBS, 1% penicillin-streptomycin, and 2 mM L-glutamine) and mechanically dissociated by pressing through a 70-µm cell strainer using a syringe plunger. The single-cell suspension was then collected in RPMI-1640 medium, followed by lysing red blood cells with Gibco ACK Lysing Buffer (Thermo Fisher Scientific) according to the manufacturer’s protocol. After lysis, cells were washed with RPMI-1640 and filtered through 40-µm trainer to remove debris. Splenocytes were used immediately to optimize cell viability. Whole splenocytes were seeded at a density of 2 × 10⁶ cells per well in 6-well plates pretreated with 2 μM NU1744 for 1 h at 37°C or DMSO at 0.02% as a control. Subsequently, Cells were stimulated or not with STING agonists (diABZI 10 µM, 2′3′-cGAMP 10 µg/ml, and 3′3′-cGAMP 10 µg/ml). After 6 h at 37°C, cells were centrifuged at 300 g for 5 min at 4°C, and the supernatant was stored at −80°C for a multiplex cytokines assay using the LEGENDplex Multi-Analyte Flow Assay Kit (Cat. No. 740622; BioLegend). The remaining cell pellets were resuspended in TRIzol Reagent (15596018; Invitrogen) for gene expression analysis as described above.

3D coordinates were obtained from the x-ray solved, crystal structure of the human DNA-PKcs with RCSB code: 5LUQ. The full amino acid sequences for ATM, ATR, and mouse DNA-PK were obtained from the GenBank database. The sequence alignment was done using ClustalW. The alignment was repeated using hidden Markov models, yielding similar results as obtained by ClustalW, due to the conservation of several anchoring motifs throughout the alignment. Molecular docking studies were executed using the ZDOCK suite, which relies on a 3D fast Fourier transformation algorithm. ZDOCK uses a scoring function that returns electrostatic, hydrophobic, and desolvation energies and performs a fast pairwise-shape complementarity evaluation. Moreover, it uses the contact propensities of transient complexes to perform an evaluation of a pairwise atomic statistical potential for the docking molecular system.

RDOCK was utilized to refine and evaluate the results obtained by ZDOCK. RDOCK performs a fast energy minimization step on the ZDOCK molecular complex outputs and ranks them according to their recalculated binding free energies. Energy minimizations were used to remove any residual geometrical strain in each molecular system, using the Amber99 force field as it is implemented into the Gromacs suite, version 4.5.5. All Gromacs-related simulations were performed through our previously developed graphical interface. An implicit generalized Born solvation was chosen at this stage to accelerate the energy minimization process. Subsequently, molecular systems were subjected to unrestrained molecular dynamics simulations (MDS) using the Gromacs suite, version 4.5.5. MDS took place in a simple point charge water-solvated, periodic environment. Water molecules were added using the truncated octahedron box extending 7 Å from each atom. Molecular systems were neutralized with counterions as required. For the purposes of this study, all MDS were performed using the NVT ensemble (i.e., constant number of atoms, volume, and temperature throughout the simulation) in a canonical environment at 300 K and a step size equal to 2 femtoseconds for a total of 10-nanoseconds simulation time.

The quality and reliability of produced models in terms of 3D structural conformation were evaluated using the Gromacs package using a residue packing quality function, which depends on the number of buried nonpolar side chain groups and on hydrogen bonding and the PROCHECK suite. Finally, the Molecular Operating Environment suite was used to evaluate the 3D geometry of the models in terms of their Ramachandran plots, omega torsion profiles, phi/psi angles, planarity, C-β torsion angles, and rotamer strain energy profiles.

The HSV-1 KOS-64-GFP strain (HSV-1-GFP) was a gift from S.R. Paludan (Aarhus University, Aarhus, Denmark). HSV-1 was amplified on Vero cells, aliquoted, and frozen at −80°C. Titration was performed on Vero cells by serial dilutions and plaque formation assessment to determine the MOI.

VSV-GFP was a gift from Dr. Sebastien Pfeffer (Institut de Biologie Moléculaire et Cellulaire, Strasbourg, France). VSV-GFP was amplified on BHK21 cells, aliquoted, and frozen at −80°C. Single-round titration of the viral stock was performed on NIH-3T3 cells to determine the MOI.

2 × 10⁴ T98G cells were seeded in 96-well plates in technical triplicates. The next morning, cells were pretreated or not for 1 h with 2 µM of NU7441. Cells were subsequently treated for 6 h with either 10 µM of diABZi, 1 µM of E7766, or 10 µg/ml of fluorinated 3′3′-cGAMP in the presence or not of 2 µM NU7441. The cells were then infected with either HSV-1-GFP (for 1.5 h at MOI 2) or VSV-GFP (for 1 h at MOI 0.1) in DMEM without serum. Media was then replaced with either DMEM-10% FBS (VSV-GFP) or DMEM-1% human serum overnight and replaced with DMEM-10% FBS the next morning (HSV-1-GFP). Cells were fixed with 4% PFA for 20 min either 16 h after infection (for VSV-GFP) or 48 h after infection (for HSV-1-GFP). Cells were permeabilized with 0.2% Triton X-100 for 10 min at RT, and nuclei were stained with DAPI.

The percentage of infected cells (i.e., GFP-positive cells) was quantified using an ImageXpress Pico (Molecular Devices) with a 4× lens. Values were averaged between technical triplicates.

Human MDM (1 × 10⁵ cells) were pretreated with NU7441 (2 µM) for 1 h and then washed with PBS before adding the STING agonists 3′3′-cGAMP (fluorinated) (10 µg/ml) for 3 h. Cells were washed again in PBS, and infections were performed with VSV-GFP (MOI of 0.5) for 1 h. Finally, cells were washed from viral inputs and incubated back into IMDM (Dutscher, Bernolsheim, France) for 16 h before fixation with PFA (2%). Cells were then analyzed for GFP expression by flow cytometry on a NovoCyte (Agilent Technologies).

The MPXV clade 2b strain (MPXV/2022/FR/Medical Center of Institut Pasteur [CMIP]) was isolated from a pustular lesion of a 36-year-old French man who consulted at the CMIP in June 2022. The clinical specimen was inoculated on Vero E6 cells, whose supernatant was harvested after 3 days and tested positive for the presence of MPXV by PCR (Batejat et al., 2022; Hubert et al., 2023). The third cell passage of the virus was used.

Statistical analysis was performed using GraphPad Prism software. To compare data from two conditions, a standard unpaired two-tailed Student’s t test was performed. One-way ANOVA with Tukey’s multiple comparisons test was used to analyze the ATP hydrolyses assay. The number of replicates in each experiment is indicated in the figure legends. All data are expressed as mean ± SEM. Results were considered significant when P < 0.05. ns, nonsignificant. *P < 0.05, **P < 0.01, ***P < 0.001, and ****P < 0.0001.

Fig. S1 supports that DNA-PKcs interacts with 2′3′-cGAMP, specifically at its catalytic site. Fig. S2 shows that DNA-PKcs regulates 2′3′-cGAMP–associated STING signaling. Fig. S3 illustrates that DNA-PKcs modulates the activity of 3′3′-cGAMP. Fig. S4 reveals that DNA-PKcs regulates the activity of STING agonists in a mouse cell line. Fig. S5 indicates that inhibition of DNA-PKcs–associated CDN signaling affects the establishment of an antiviral state.

Correspondence and requests for material should be addressed to Nadine Laguette or Dimitrios Vlachakis. Raw data for immunoblots are provided with the manuscript.

We thank all members of the Molecular Basis of Inflammation laboratory for reading and editing the manuscript. We thank Caroline Bertin, Johanna Marines, Amel Bouzid, Amandine Kremer, Mathilde Saccas, Robin Charpentier, and Tommy Chastel for technical assistance. We thank Soren Paludan for control, STING- and cGAS-deficient THP-1 cell lines, and HSV-1-GFP. We thank Dr. Jin Lei (University of Florida, Gainesville, USA) for the gift of Sting-deficient mice; we thank Dr. Miereia Pellegrin (IRMB, Montpellier, France) for the gift of Ifnar-deficient mice. We thank Dr. Christelle Langevin (INRAE-Université Paris-Saclay, Jouy-en-Josas, France) for the gift of the VSV-GFP molecular clone. We thank the ZEFI (Zone d'Expérimentation et de Formation) and the PCEA (Plateau Central d'Elevage et d'Archivage) animal facilities for animal housing. We thank the CEMIPAI BSL-3 level laboratory for housing virology experiments. We acknowledge the MRI imaging and cytometry facility, a member of the national infrastructure France-BioImaging infrastructure supported by the French National Research Agency (ANR-10-INBS-04, “Investments for the future”). We thank the MAMMA metabolomics platform for support.

This work was co-funded by the European Union (European Research Council, SENTINEL 101087092 to N. Laguette and DELV to K. Majzoub). Views and opinions expressed are, however, those of the author(s) only and do not necessarily reflect those of the European Union or the European Research Council. Neither the European Union nor the granting authority can be held responsible for them. This work was also co-funded by the Agence Nationale de la Recherche (AlphA [N. Laguette] and dezincRNA to [K. Majzoub]); l’Institut National du Cancer (INCa_1884 [N. Bidère] and Inca_19379 [N. Laguette]); La Ligue contre le cancer (AAPARN 2021.LCC/JuF [N. Laguette] and équipe labellisée [N. Bidère]); the Agence Nationale de Recherches sur le SIDA et les Hépatites Virales (ANRS) (ECTZ117448 [N. Laguette] and fellowships to C. Taffoni and JM); the I-SITE Excellence Program of the University of Montpellier, under the Investissements France 2030 (RETTiNA [N. Laguette] and ChoiCe [N. Laguette]); the Fondation ARC (ARCPJA2021060003720 COPALYS [N. Laguette] and ARCPJA2021060003886 [N. Bidère]); La Région Languedoc Roussillon (Prématuration 2021 MODULON 21015964 [N. Laguette]); Fondation de France (fellowship [J. Jardine]); Fondation pour la Recherche Médicale (FRM; SPF202409019744 [Y. Messaoud-Nacer]); and the Centre National de La Recherche Scientifique (Prématuration CNRS [N. Laguette]). The O. Schwartz lab is funded by Institut Pasteur, FRM, ANRS, the Vaccine Research Institute (ANR-10-LABX-77), and Labex IBEID (ANR-10-LABX-62-IBEID); the HERA projects DURABLE (grant 101102733) and LEAPS. J. Postal is supported by DURABLE. This work was also supported by the ANRS (ECTZ192763) and the ANR (ANR-23-CE15-0028 METABODEN) to F.P. Blanchet.

Author contributions: Isabelle K. Vila: conceptualization, formal analysis, investigation, methodology, validation, visualization, and writing—review and editing. Yasmine Messaoud-Nacer: conceptualization, investigation, methodology, and writing—review and editing. Clara Taffoni: formal analysis, investigation, methodology, resources, validation, visualization, and writing—review and editing. Jane Jardine: investigation and writing—review and editing. Roger J. Eloiflin: investigation. Adeline Augereau: investigation. Soumyabrata Guha: investigation and methodology. Moritz Schussler: investigation and writing—review and editing. Pierre Le Hars: data curation, formal analysis, and writing—review and editing. Joe McKellar: investigation, validation, and visualization. Tamara Carvalho: investigation. Jeanne Postal: investigation and methodology. Morgane Chemarin: investigation and writing—review and editing. Joanna Re: investigation, formal analysis, and writing—review and editing. Florence Guivel-Benhassine: investigation. Raphaelle Lopez: investigation. Kilian Trillet: formal analysis and investigation. Jennifer Barrat: conceptualization, investigation, and writing—review and editing. Maximin Serbier: investigation, writing—review and editing. Insaf El Mansouri: conceptualization, formal analysis, investigation, and visualization. Charlotte Luchsinger: conceptualization, formal analysis, and methodology. George P. Chrousos: conceptualization, supervision, and writing—review and editing. Francoise Porrot: investigation and resources. Felipe Diaz-Griffero: conceptualization, formal analysis, methodology, supervision, and writing—original draft. Olivier Schwartz: investigation, resources, and writing—review and editing. Fabien P Blanchet: supervision and writing—review and editing. Karim Majzoub: project administration, resources, supervision, validation, and writing—review and editing. Nicolas Bidere: funding acquisition, supervision, and writing—review and editing. Dimitrios Vlachakis: conceptualization, data curation, formal analysis, investigation, methodology, validation, visualization, and writing—original draft, review, and editing. Nadine Laguette: conceptualization, funding acquisition, methodology, project administration, visualization, and writing—original draft, review, and editing.