A conserved ion channel function of STING mediates noncanonical autophagy and cell death

1The cyclic GMP-AMP (cGAMP) synthase (cGAS)/stimulator of interferon genes (STING) innate immunity pathway is critical in host defense against microbial infection (Ishikawa and Barber,; Ishikawa et al,; Sun et al,; Zhong et al, ). Besides infection-triggered inflammation, the cGAS/STING pathway also promotes sterile inflammation in many other conditions involving cytosolic exposure of self-DNA, such as cell damage, senescence, diseases, and normal aging (Decout et al,; Motwani et al, ). Upon recognition of abnormal cytosolic DNA, cGAS produces a second messenger cGAMP, which binds to and activates STING on the endoplasmic reticulum (ER) (Sun et al,; Wu et al, ). The cGAMP-bound STING undergoes robust trafficking from the ER to the Golgi region and finally forms perinuclear vesicle clusters where TBK1 is recruited to activate both IRF3 and NF-κB, transcription factors that upregulate the expression of type I interferons (IFN-I) and inflammatory cytokines (Tan et al, ).

2STING is also known to restrict microbial infection through noncanonical autophagy, which is independent of autophagy proteins upstream of the ATG5/ATG12/ATG16 complex (Gui et al,; Liu et al,; Moretti et al, ). An extensively studied form of noncanonical autophagy is known as the conjugation of LC3/ATG8 to single membranes or CASM (Durgan and Florey, ). CASM induction is mediated through the direct recruitment of ATG16 by the vacuolar-type ATPase (V-ATPase), a proton pump that senses organelle deacidification upon challenges from weak base chemicals, proton ionophores, or lysosomal membrane damage (Cross et al,; Tan and Finkel,; Wang et al,; Xu et al, ). A recent study from the Youle group identified V-ATPase as a downstream effector of STING for noncanonical LC3 lipidation on single membrane STING vesicles (Fischer et al, ). However, it was still unclear how STING activates V-ATPase.

3Here, we report that STING-induced noncanonical autophagy depends on an unexpected ion channel function of STING. Ligand-bound STING forms an electron-sparse pore in its transmembrane domain that causes proton efflux from post-Golgi STING vesicles and subsequent vesicle deacidification. This is required for STING-dependent LC3 lipidation and cell death, both of which can be abolished by C53, a small molecule compound that blocks the transmembrane channel of STING. This new molecular function of STING resembles other CASM-inducing stresses that deacidify endolysosomes or phagosomes.

4Our recent structural study of apo-STING and cGAMP/STING complex provided new mechanistic insights into STING autoinhibition and activation (Hussain et al,; Liu et al, ). Notably, upon cGAMP binding, the ligand binding domain (LBD) of STING contracted, whereas the transmembrane domain (TMD) underwent dilation. After a thorough analysis of the conformational changes in the TMD from the apo and cGAMP-bound structures of human STING, a ligand-induced transmembrane pore with a radius of ~1.71 Å at its narrowest part was revealed (Fig. 1A,B). In contrast, the pore radius was ~0.86 Å at its bottleneck in the apo-structure (Fig. 1A). The ligand-induced pore space was electron-sparse (Liu et al, ), suggesting that it is not readily accessible by membrane lipids. Thus, we hypothesized that the TMD of STING might function as a potential ion channel.

5We have previously reported the compound C53 as an unusual STING agonist (Lu et al, ). C53 binds precisely to the lower part of STING-TMD and appears to fully block the channel (Fig. 1C). Thus, we tested C53 as a potential inhibitor of STING’s ion channel function and examined its impact on STING-mediated TBK1 activation as well as LC3 lipidation. Remarkably, the presence of C53 fully abolished STING-dependent LC3 lipidation without affecting TBK1 activation (Fig. EV1A), suggesting that STING-mediated noncanonical autophagy is likely triggered through an ion channel function of STING. Although C53 was later found to disrupt STING trafficking (see below), this initial experiment directed our study to STING-mediated LC3 lipidation instead of TBK1 signaling.

6To explore how the potential STING channel could trigger LC3 lipidation, we first investigated where STING might induce LC3 lipidation in cells. Multiple subcellular localizations have been previously described as active sites for the STING complex, including ER-Golgi-intermediate compartments (ERGIC) (Dobbs et al,; Gui et al, ), the Golgi body (Fang et al,; Kong et al,; Mukai et al, ), and post-Golgi vesicles (Fang et al,; Fischer et al,; Saitoh et al, ). We first characterized time-dependent STING trafficking upon ligand stimulation. U2OS cells are widely used for immunofluorescence studies, but no STING was expressed in this cell line. We established a monoclonal U2OS cell line stably expressing low levels of human STING. This cell line showed robust STING trafficking comparable to endogenous STING in human BJ and HT1080 cells, but is a better model to study STING trafficking as it has consistent STING trafficking throughout the cell population, allowing for careful dissection of the trafficking steps upon ligand stimulation.

7Upon cGAMP delivery into U2OS cells, STING showed dramatic subcellular trafficking and strongly colocalized with the cis- and trans-Golgi markers, GM130 and Golgin97, respectively, about 30 min after cGAMP exposure (Fig. 2A–F). Remarkably, the occupation of the Golgi membrane by STING caused striking morphological changes of the Golgi complex, as shown by more extensive overlap between GM130 and Golgin97 (Fig. EV1B–D) as well as swelled Golgi stacks at 30 min (Figs. 2A,D and EV1B,E). After this, a large number of STING vesicles formed out of the Golgi complex (Fig. 2A–F), and the Golgi markers returned to resting morphology (Fig. EV1B–E). Despite the extensive clustering of Golgi-derived STING vesicles that appeared to partially overlap with Golgi markers, these STING vesicles are apparently different membranes outside of the Golgi stacks (Fig. 2A,D).

8STING trafficking caused robust relocalization of the trans-Golgi network (TGN) marker TGN46 (also known as TGN38 or Trans-Golgi Network Protein 2, TGOLN2). While TGN46 showed extensive overlap with Golgin97 in resting conditions, the arrival of STING to the Golgi 30 min after cGAMP binding was accompanied by the budding of TGN46, but not Golgin97, from the trans-Golgi stack (Fig. 2G–I). The TGN46 vesicles and Golgi-derived STING vesicles appeared to be distinct with only partial overlap (Figs. 2J and EV1F).

9STING trafficking through the Golgi complex was also observed when hSTING was activated by diABZI (Fig. EV1G,H), a non-nucleotide-based synthetic STING agonist (Ramanjulu et al, ). Similarly, endogenous STING in HT1080 and BJ cells were both found to traffic through the Golgi upon cGAMP treatment (Fig. EV2A,B). Mouse STING (mSTING) stably expressed in U2OS also trafficked through the Golgi upon cGAMP delivery (Fig. EV2C–H). Thus, in different cell lines, ligand binding activates robust STING trafficking through the Golgi complex to post-Golgi vesicles (Fig. 2K).

10Since STING showed robust trafficking through the Golgi stacks, we asked whether STING stimulated LC3 lipidation directly on the Golgi membrane or post-Golgi vesicles. The cGAMP-stimulated endogenous LC3 puncta had little overlap with the Golgi (Figs. 3A–C and EV3A–C). The LC3 puncta did not appear 30 min after cGAMP exposure when most STING accumulated at the Golgi (Figs. 3A and EV3A,C), and their subsequent occurrence correlated with STING trafficking to post-Golgi vesicles (Fig. 2A–F). The morphology of the triggered LC3 puncta resembled post-Golgi clusters of STING vesicles (Figs. 2A,D, 3A, and EV3A,C). Indeed, three-channel imaging showed that these LC3 puncta largely colocalized with STING but not Golgin97 (Fig. EV3D,E). Thus, these data strongly support that STING stimulates LC3 lipidation primarily on post-Golgi vesicles.

11These observations are consistent with recent findings that cGAMP stimulates LC3 lipidation on single membrane STING vesicles (Fischer et al, ). To further explore the identity of these vesicles, we co-stained LC3 with different organelle markers at different time points after cGAMP treatment. This revealed a significant colocalization between the triggered LC3 puncta with the early endosome marker RAB5, but not Golgin97 (Figs. 3D,E and EV3C). Of note, the total punctate intensities of RAB5 substantially increased after STING left the Golgi (Fig. 3D), suggesting that post-Golgi STING vesicles likely developed endosome-like properties and recruited RAB5. Consistently, a sharp increase in STING colocalization with RAB5 was also observed in its post-Golgi stage (Fig. 3F–H). Further imaging revealed that cGAMP-induced LC3 puncta were typically found on STING vesicles positive for RAB5 (Fig. EV3F). Interestingly, STING and LC3 puncta also appeared to overlap with the recycling endosome marker RAB11 (Fig. 3I–K), consistent with a recent study that STING trafficking to lysosomes for degradation occurred through RAB11-positive vesicles (Kuchitsu et al, ). STING also developed increased colocalization with the late endosome/lysosome markers CD63 and LAMP1, but to a much lower extent compared with RAB5 (Fig. EV3G–L), reflecting partial late endosome properties of STING vesicles or STING delivery to the lysosome for degradation. Together, these results suggest a gradual acquisition of endolysosomal properties by post-Golgi STING vesicles where most LC3 puncta are detected.

12Our data suggest that STING triggered LC3 lipidation mostly on post-Golgi, endosome-like vesicles. Given that C53, a chemical that binds to the transmembrane pore of STING, selectively blocks STING-mediated LC3 lipidation without affecting TBK1 activation or subsequent STING phosphorylation (Fig. EV1A), we hypothesized that a potential ion channel function of STING triggers LC3 lipidation on post-Golgi STING vesicles. STING-mediated LC3 lipidation is known to depend on the V-ATPase proton pump that directly recruits ATG16L1 (Fischer et al, ). In other contexts, the same V-ATPase-ATG16L1 axis senses the deacidification of peripheral organelles such as phagosomes and endolysosomes and triggers direct LC3 lipidation onto these compartments, known as CASM (Cross et al,; Durgan and Florey,; Wang et al, ). Since STING also activates CASM through V-ATPase, we examined whether STING activation could raise the pH of its post-Golgi trafficking vesicles.

13Lyso-pHluorin is a pH sensor of endolysosomes, the fluorescence of which is quenched at acidic pH and activated upon pH neutralization or endolysosomal deacidification (Rost et al, ), as validated in Fig. EV4A as well as in our recent study of lysosomal repair (Tan and Finkel, ). We tested if lyso-pHluorin could serve as a pH sensor of STING vesicles which showed endolysosomal properties (Fig. 3). Remarkably, STING activation triggered the formation of lyso-pHluorin puncta in both U2OS cells reconstituted with human STING (hSTING) and BJ cells expressing endogenous hSTING (Figs. 4A,B and EV4B). Similar lyso-pHluorin puncta were observed when mouse STING was expressed in U2OS cells and activated by DMXAA, a chemical agonist selectively activating mouse STING (Conlon et al,; Gao et al,; Prantner et al, ) (Fig. 4C,D). The activation of neither human nor mouse STING caused any puncta formation of EGFP-galectin3 (Fig. EV4C), a sensor of endolysosomal membrane damage (Maejima et al,; Skowyra et al, ), suggesting that the membrane integrity of STING vesicles was not compromised and that the observed pH neutralization was likely due to the altered activity of potential ion channels. The vesicle deacidification activity of STING appeared to be highly conserved, given that lyso-pHluorin puncta were equally detected when Xenopus tropicalis STING (XtSTING) was activated by cGAMP in U2OS cells (Fig. 4E,F). We consistently noticed the development of brighter, STING-dependent lyso-pHluorin puncta over time. This was confirmed using time-lapse live-cell imaging that showed substantial lyso-pHluorin puncta 60–120 min after the activation of mSTING (Fig. 4G,H), consistent with STING trafficking to endosome-like vesicles at this time (Fig. 3). Thus, vesicle deacidification by STING might be sensed by the V-ATPase proton pump for direct LC3 lipidation on these vesicles. Consistent with vesicle deacidification being upstream of LC3 lipidation, STING-induced lyso-pHluorin puncta formation was normal in ATG5- and ATG7-knockout cells (Fig. EV4D).

14To further characterize whether the observed lyso-pHluorin puncta represent STING vesicles or other nearby, pre-existing endolysosomes, we developed a protocol to stain for STING in the same cells after the induction of lyso-pHluorin puncta by cGAMP (Fig. EV4E). This approach revealed extensive colocalization of cGAMP-induced lyso-pHluorin puncta with STING (Fig. 4I,J) as well as RAB5 (Fig. 4K). It is surprising that Lyso-pHluorin, a pH sensor of late endosome/lysosomes (Rost et al,; Tan and Finkel, ), detected STING-dependent vesicle deacidification on RAB5-positive STING vesicles (Fig. 4K). To further examine such deacidification events, we investigated the relative localization of STING-induced Lyso-pHluorin puncta and different organelle markers in live cells. Consistent with STING colocalization with RAB5 and RAB11 (Fig. 3F–K), cGAMP-induced Lyso-pHluorin puncta also overlapped with both RAB5 and RAB11, but with apparently less association with various late endosome/lysosome markers such as RAB7 and CD63 (Fig. EV4F, quantified in Fig. 4L). These results suggest that STING-induced Lyso-pHluorin puncta are localized to STING vesicles different from late endosome/lysosomes. We added an mCherry tag to the C-terminal end of the Lyso-pHluorin probe so that we could track both the localization and pH-dependent (de)quenching of the probe. This probe again confirmed that STING-induced Lyso-pHluorin puncta represented a fraction of the probe with a weaker mCherry signal different from the pre-existing bright mCherry puncta on late endosome/lysosomes (Fig. EV4G). Together, these data indicate that post-Golgi endosome-like vesicles are deacidified by STING, consistent with V-ATPase-ATG16L1-meidated LC3 lipidation.

15The vesicle deacidification activity of STING further suggests that the transmembrane pore of STING might function as an ion channel allowing protons and/or other ions to pass through the membrane, which is consistent with our structural analysis (Fig. 1). To test this hypothesis, we conducted an in vitro ion flux assay (Su et al, ) using purified STING reconstituted onto proteoliposomes (Fig. 5A). In this assay, proton influx was driven by a strong gradient for the efflux of potassium across the membrane. The proton-sensitive dye 9-Amino-6-chloro-2-methoxyacridine (ACMA) was used to monitor the pH change inside the vesicles. A potassium ionophore valinomycin was added to initiate the flux. Surprisingly, in this system, STING showed constitutive activity in proton flux, which was not further enhanced by the addition of diABZI, a membrane-permeable STING agonist expected to activate STING from the luminal side of the liposomes (Fig. 5B). Nevertheless, the basal proton flux activity was strongly inhibited by C53 (Fig. 5B) which binds to and blocks the transmembrane pore of STING (Fig. 1C). Taken together, our structure analysis, cellular imaging, and in vitro ion flux assay suggest that STING functions as a proton channel to deacidify at least Golgi-derived vesicles, consistent with cGAMP-induced STING trafficking to perinuclear acidic compartments.

16STING-mediated ion flux is likely selective for protons, unique ions with an extremely small size. First, the heavily positively charged cytosolic side of this channel might not allow the efflux of cation ions such as Na+ and K+ (Appendix Fig. S1A). Second, the relatively small size of the pore might preclude the passage of anion ions like hydrated Cl- with an estimated radius of 3.32 Å. Finally, mutations of various amino acid residues in the center of the transmembrane pore in human STING (Appendix Fig. S1B,C) failed to identify any mutants completely dead in vesicle deacidification (Appendix Fig. S1D,E), among which those triggered weaker or no lyso-pHluorin puncta turned out to be unstable (Appendix Fig. S1F). One mutant L54E induced brighter lyso-pHluorin puncta than wild type and any other mutants of STING (Appendix Fig. S1C–E), suggesting increased proton efflux by this mutant. As the L54E mutation is located near the narrowest neck of the channel (Appendix Fig. S1G), the increased proton efflux is likely mediated by an interaction between protons and E54 that allows more efficient proton passage through the channel.

17The data strongly argue for an ion channel function of STING that deacidifies its post-Golgi trafficking vesicles, thus allowing for the activation of the V-ATPase-ATG16L1 axis for LC3 lipidation. Since we were not able to find a STING mutant with a selective loss-of-function in vesicle deacidification (Appendix Fig. S1C–E), we next focused on the STING channel blocker C53 in characterizing the role of ion flux in vesicle deacidification and LC3 lipidation.

18For C53 to be a valid blocker of the ion channel function of STING, we first need to rule out its potential impact on STING trafficking from the ER to post-Golgi vesicles. We have recently shown that although C53 binds to the transmembrane pore of human STING, it also triggers STING oligomerization and TBK1 activation (Lu et al, ). Surprisingly, C53 alone induced large numbers of STING puncta with little colocalization with the Golgi (Appendix Fig. S2A). The addition of C53 together with either cGAMP or diABZI prevented STING trafficking to the Golgi, causing the accumulation of both endogenous and ectopic STING in punctate structures, similar to what was stimulated by C53 alone (Fig. 6A–C; Appendix Fig. S2A–D). Thus, the presence of C53 redirects STING to an abnormal route that blocks STING trafficking at a pre-Golgi stage.

19To assess the impact of C53 without blocking STING trafficking, we tested C53 addition after STING arrived at the Golgi. In U2OS Flag-STING monoclonal cells, most STING accumulated at the Golgi 30 min after cGAMP treatment (Figs. 2 and 6C,D). Adding C53 at this time point did not block STING trafficking to post-Golgi vesicles (Fig. 6C,D; Appendix Fig. S2C,E). However, C53 still almost completely blocked STING-dependent vesicle deacidification (Fig. 6E,F) as well as LC3 lipidation, without affecting TBK1 activation (Fig. 6G–J). Compared with cGAMP, diABZI triggered slower STING trafficking, with most STING found at the Golgi 60 min after stimulation (Fig. EV1G,H). Adding C53 at this time still fully blocked diABZI-induced LC3 lipidation (Appendix Fig. S2F).

20We further tested adding C53 after STING fully arrived at post-Golgi vesicles. Specifically, C53 was added 2 h after cGAMP or diABZI when vesicle deacidification was already detected. Under these conditions, STING-mediated LC3 lipidation was not reversed by C53 as it was already saturated by 2 h (Appendix Fig. S2G). Remarkably, C53 still dramatically suppressed STING-mediated vesicle deacidification, and even fully cleared diABZI-induced lyso-pHluorin puncta (Figs. 7A–C and EV5A). Thus, the fraction of Lyso-pHluorin probe on endosome-like, post-Golgi STING vesicles (Fig. EV4G) can only be detected when these vesicles are deacidified by STING. Taken together, when C53 was used appropriately, it selectively suppresses STING-mediated vesicle deacidification and LC3 lipidation, without affecting STING trafficking or TBK1 signaling.

21Overactivation of STING is known to cause cell death through an interferon-independent function of STING (Tang et al,; Wu et al,; Wu et al, ). A C-terminal “RHLR” motif of STING required for its trafficking from the ER is critical for STING-dependent cell death (Gui et al,; Wu et al, ). We recapitulated STING-mediated cell death in U2OS cells stably expressing hSTING and examined whether it required the ion channel function of STING (Fig. 7D). Cells were briefly treated with cGAMP/digitonin buffer for 10 min and then changed back to original media. Extensive cell death was observed 24 h after cGAMP treatment (Fig. 7D,E). Strikingly, C53 addition 2 h after cGAMP treatment fully blocked cGAMP-induced cell death (Fig. 7D,E). It is of note that C53 alone, which is sufficient to activate TBK1 signaling for interferon production, did not cause any cell death in these tests (Fig. 7E). The L54E mutant of STING, which showed increased vesicle deacidification activity (Appendix Fig. S1C–E), consistently triggered more cell death than wild-type STING expressed in similar levels compared with the mutant (Fig. 7F,G). Thus, STING-mediated cell death under these conditions is dependent on its ion channel function but independent of TBK1 signaling.

22STING-dependent cell death appears to involve multiple mechanisms including ferroptosis (Tang et al,; Wu et al,; Wu et al, ). Two ferroptosis inhibitors both partially suppressed, but failed to fully block, STING-dependent cell death (Fig. EV5B), suggesting the presence of additional cell death mechanisms. Since both STING-dependent LC3 lipidation and cell death relied on the STING channel, we asked if LC3 lipidation was required for STING-dependent cell death. Overexpression of a bacterial effector SopF (Fischer et al,; Xu et al, ) strongly suppressed STING-mediated LC3 lipidation (Fig. 7H,I), but did not cause any defects in cell death (Fig. 7J). Similarly, the deletion of ATG7 also blocked LC3 lipidation but accelerated STING-dependent cell death (Fig. EV5C–E). Thus, STING-dependent LC3 lipidation appears to antagonize STING-mediated cell death.

23In summary, we discovered an unexpected ion channel function of STING that allows the efflux of protons from post-Golgi STING vesicles, leading to vesicle deacidification, noncanonical autophagy, and cell death, all independent of STING-induced TBK1/interferon signaling (Fig. 7K). STING-mediated vesicle deacidification can be captured by lyso-pHluorin, a genetically encoded lysosomal pH sensor that is also targeted to post-Golgi STING vesicles. Although the channel blocker C53 disrupts STING trafficking, its addition after STING arrives at the Golgi or post-Golgi vesicles can selectively block the ion channel function of STING. The new function of STING appears to be highly conserved as STING from humans, mice, and frogs all stimulated vesicle deacidification when activated (Fig. 4A–F). Such changes in vesicle pH likely activate the V-ATPase-ATG16L1 axis for LC3 lipidation onto STING vesicles (Fischer et al, ), resembling other CASM inducers (Cross et al,; Tan and Finkel,; Wang et al,; Xu et al, ).

24How STING triggers noncanonical autophagy and cell death are fundamental questions in innate immunity. In this study, we found that activated STING forms a transmembrane channel that mediates both noncanonical autophagy and cell death. The channel of STING allows the efflux of protons, leading to the deacidification of post-Golgi STING vesicles. STING-mediated vesicle deacidification mimics many other cellular stresses such as lysosomal membrane damage or exposure to proton ionophores, instigating the same downstream effectors for the induction of noncanonical autophagy or CASM (Cross et al,; Durgan and Florey,; Tan and Finkel,; Wang et al,; Xu et al, ).

25Our study suggests that the proton channel function of STING is highly conserved. A recent study identified a bacterial cGAMP receptor as a potential cGAMP-activated Cl- channel (preprint: Tak et al, ). The surface charges on the cytosolic side of the pore suggest that STING might also allow for the flux of anions like Cl- in addition to H+. Alternatively, anion binding to the cytosolic side of the pore might allow for more efficient H+ leakage. In addition, the size of the STING channel might change when it is transported to acidic post-Golgi vesicles, which in turn would affect the channel selectivity in ion flux. Thus, future studies are needed to thoroughly test the ion selectivity of the STING channel.

26It has been controversial regarding the signaling compartments of STING. The extensive clustering of Golgi-derived STING vesicles near the microtubule-organization center (MTOC) has rendered it difficult to differentiate post-Golgi STING vesicles from the Golgi stacks which are themselves considered as part of the MTOC (Fig. EV5F, right). This situation is further complicated by the budding of certain TGN markers such as TGN46 during STING trafficking as well as by often overly simplified cartoon models of how key organelles involved in STING trafficking are organized in the cell (Fig. EV5F, left). Our in-depth analysis of ligand-stimulated STING trafficking in different cell lines strongly suggests the active sites of STING-mediated LC3 lipidation as post-Golgi vesicles. A fraction of lyso-pHluorin probe is localized to endosome-like STING vesicles separated from late endosome/lysosomes and is dequenched by STING-mediated vesicle deacidification. Although it seems likely that STING may neutralize the pH of the Golgi complex with a relatively low level of LC3 lipidation activity, most endogenous LC3 puncta were observed outside of the Golgi. Surprisingly, the presence of C53 blocked STING trafficking at a pre-Golgi stage, leading to STING accumulation in punctate structures that were still capable of activating TBK1 but not LC3 lipidation. The problem can be solved by adding C53 after STING traffics to the Golgi or post-Golgi vesicles, allowing selective inhibition of the ion channel functions of STING by C53.

27During the preparation of this manuscript, a similar study by Liu et al was published (Liu et al, ). The two studies both uncovered STING-mediated deacidification of perinuclear vesicles but through different approaches and pH sensors. Both studies observed STING-mediated proton flux in vitro and achieved a block of STING’s ion channel function by C53 that tightly locks the transmembrane pore. Of note, purified STING was found to be constitutively active in proton flux, suggesting that purified STING might be partially activated when incorporated onto membranes where the autoinhibition through head-to-head binding is absent (Liu et al, ). Both studies also observed a block of STING-dependent vesicle deacidification and cell death by C53, while our study established a more strict protocol in assessing the impact of C53 as a STING channel blocker without disrupting its trafficking, in addition to identifying post-Golgi vesicles as the major sites of action for the STING channel. It is of note that it is unclear how STING-dependent LC3 lipidation contributes to noncanonical autophagy and viral suppression.

28While the cGAS/STING pathway is critical for innate immunity, abnormal activation of this pathway has been increasingly implicated in various human diseases including Parkinson’s and Alzheimer’s diseases, as well as normal aging (Gulen et al,; Sliter et al,; Udeochu et al, ). The ion channel function of STING described here opens an avenue for new therapeutic strategies to target this pathway. Selective manipulation of the inflammation signaling or the ion channel activity, depending on the contexts of specific diseases, might be considered for optimal therapeutic outcomes.

29BJ, U2OS, and 293T cells were from ATCC and were authenticated through short tandem repeat (STR) profiling. All relevant authentication data are publicly available from ATCC. These cell lines differ in their growth rates and morphologies. All cell lines in this study were free of contaminations from other cell lines or mycoplasma. Cells were maintained with mycoplasma reagent and potential contaminations are regularly monitored through polymerase chain reaction (PCR) detection. All cells were cultured at 37 °C with 5% CO2 in Dulbecco’s modified Eagle’s medium (DMEM) supplemented with 8% fetal bovine serum (FBS) and penicillin/streptomycin. Cyclic [G(2’,5’)pA(3’,5’)p] or cGAMP (#CT-CGMAP, ChemieTek) was dissolved in water and stored at −20 °C. The other two STING agonists diABZI (#S8796, Selleck) and C53 (#37354, Cayman) were dissolved in DMSO and stored in aliquots at −80 °C. Monensin (#16488, Cayman) was dissolved in ethanol and stored at −20 °C. cGAMP was delivered to cells using a mild digitonin buffer (50 mM HEPES pH 7.0, 100 mM KCl, 3 mM MgCl2, 0.1 mM DTT, 85 mM sucrose, 0.2% BSA, 1 mM ATP, 8 μg/mL digitonin) (Woodward et al, ). Cells were treated with the buffer containing 1 μM cGAMP for 5-10 min and then changed back to the original culture media. All other chemicals were directly added to the cell culture media.

30Antibodies for LAMP1 (sc-20011, IF 1:200); LAMP2 (sc-18822, IF 1:200); Golgin97 (sc-59820, IF 1:500); GAPDH (sc-365062, WB 1:5000); Tubulin (sc-5286, WB 1:3000) were from Santa Cruz Biotechnology. The GM130 antibody (610822, IF 1:1000) was from BD Biosciences. Flag (M2, IF 1:1000; F7425, WB 1:3000) antibody was from Sigma. LAMP1 rabbit mAb (#9091, IF 1:200); GM130 rabbit mAb (#12480, IF 1:300); Rab5 rabbit mAb (#3547, IF 1:200), Phospho-STING S366 (#50907, WB 1:1000); Phospho-TBK1 Ser172 rabbit mAb (#5483, WB/IF 1:1000) were from Cell Signaling. The following antibodies were from Proteintech: STING (19851-1-AP, WB 1:1000, IF 1:1000); LC3 (14600-1-AP, IF 1:1000, WB 1:1000, reacts with LC3A, LC3B, and LC3C); TGN46 (13573-1-AP, IF 1:1000); Golgin97 (12640-1-AP, IF 1:1000). Alexa-488/594- and Pacific Blue-conjugated secondary antibodies were obtained from ThermoFisher Scientific.

31We used a lentiviral approach for stable protein expression. The DNA sequences of interest were PCR amplified and inserted into a lentiviral vector pCDH-CMV-MCS. When two genes are fused, a GSGSGS linker was used. Small epitope tags were added directly into PCR primers. To generate point mutations, two fragments of the target cDNA were amplified by PCR with their overlapping ends carrying the intended mutations. The two fragments were then fused together into the pCDH vector through infusion reactions. All new plasmids are verified by DNA sequencing.

32All experiments in this study are based on stable lines. No transient transfection of DNA was used in any experiment to avoid DNA-induced activation of the cGAS/STING pathway. We used a lentiviral approach for stable protein expression. Lentiviruses were packaged in 293T cells by transfecting the pCDH vector carrying the gene of interest together with the packaging vectors pMD2.G and pSPAX2 using Lipofectamine 2000 (ThermoFisher). The culture media containing the viruses were collected 48 h after transfection, and were immediately used to infect target cells for stable protein expression. When needed the infected cells were selected by puromycin treatment. The lowest virus titer was used to achieve the desired infection rate. To generate knockout cell lines, CRISPR–Cas9 guide sequences were tested to identify the guide that effectively diminished the target protein in the CRISPR pool. The lentiCRISPR v2 (Addgene, 52961) carrying the selected guide sequence was used for lentivirus packaging. The infection rate of U2OS cells by CRISPR viruses was remarkably high that no cell death was observed after puromycin treatment of infected pools when all uninfected control cells died. ATG7 CRISPR targeting sequence TATACAGTGTTCCAATAGCT was used when generating the ATG7-KO pools. ATG5-KO and ATG7-KO U2OS single clones were previously described (Tan and Finkel, ).

33Cells were seeded on glass coverslips (Electron Microscopy Sciences, #7223001) in 24-well plates. After indicated cell stimulations, cells were fixed with 4% Paraformaldehyde (PFA) in phosphate-buffered saline (PBS) for 30 min at room temperature. Cells were then permeabilized with 0.1% Triton X-100 for 2 min, and blocked for 30 min with 1× fluorescent blocking buffer (ThermoFisher Scientific, #37565). The same buffer was further used for the dilution of primary and secondary antibodies in the following cell staining steps. Cells were incubated with primary antibodies at 4 °C overnight or at room temperature for 2 h. Unbound primary antibodies were washed away with PBS, and cells were further stained fluorescently labeled secondary antibodies. Cells were then washed more than three times to remove any unbound secondary antibodies. When 4′,6-diamidino-2-phenylindole (DAPI) staining was needed, cells were stained with DAPI for 3 min and then washed with PBS. The coverslips were mounted on slides using VECTASHIELD Mounting Medium (Vector Laboratories, #H-1700). Fluorescence images were collected using a Leica SP8 LIGHTNING confocal system with a built-in software Leica Application Suite X 3.5.5.19976. Different positive and negative controls were included to rule out nonspecific staining and any cross-talks between channels. Live-cell imaging was done using the same confocal system with an Okolabstage-top incubator. All images in the same panel were from the same experiment, followed with the same cell staining, image collection settings, and image processing.

34Cells with indicated treatments were briefly washed with cold PBS and then lysed with a lysis buffer containing 50 mM Tris-HCl, pH 7.5, 150 mM NaCl, 0.5% Triton X-100, 2 mM NaF, and a protease inhibitor cocktail. The lysates were briefly sonicated to fully dissolve all membranes, followed by centrifugation at 15,000×g for 10 min. The supernatants were collected and heated at 95 °C for 5 min after mixing with equal volumes of 2× SDS loading buffer (0.1 M Tris-HCl, pH 6.8, 4% SDS, 20% glycerol, 2% 2-mercaptoethanol, 0.01% bromphenol blue). The protein samples were kept at −80 °C or directly moved to immunoblotting analysis. Protein electrophoresis was done using 4-20% precast polyacrylamide gel (Bio-Rad, #4561096), followed by protein transfer to 0.45-µm Nitrocellulose membranes using the Trans-Blot Turbo system. The membranes were then blocked with StartingBlock Blocking Buffer (ThermoFisher, #37542) for 30 min and then incubated sequentially with primary and secondary antibodies. After washing, the target proteins were detected Immobilon Forte Western HRP Substrate (Sigma-Aldrich, # WBLUF0500) in a ChemiDoc MP Imaging System.

35Cells stably expressing Lyso-pHluorin at 50–70% confluency were treated with vehicle or indicated STING agonists to trigger STING trafficking and vesicle deacidification. For cGAMP treatment, the media from the cell culture dish were moved to a new tube and kept at 37 °C. Cells were treated with the digitonin/cGAMP buffer for 10 min, and the buffer was then replaced with the original warm culture media. Other chemicals were directly added to and mixed well in culture media. Cells were subsequently monitored for changes in their lyso-pHluorin signals every 30 min and live-cell images were collected at desired time points.

36The human STING protein was expressed and purified following established protocols (Shang et al,; Zhang et al, ). Briefly, The coding sequences of the human STING gene were inserted into a modified pEZT-BM vector, generating a STING fusion protein with a C-terminal Tsi3 tag (Lu et al, ). The plasmid was transfected into HEK293F cells using PEI with a mass ratio of 1:3 (plasmid: PEI) and the expression was enhanced by 3 mM sodium butyrate 12 h later. After 48 h of culturing, cells were collected and re-suspended into buffer A (20 mM Tris pH 7.5, 150 mM NaCl, 1 mM AEBSF and protease inhibitor cocktail). Then, cells were lysed by French press, and the membranes were collected using ultracentrifugation at 100,000×g. The membrane proteins were extracted using 1% DDM/CHS (5:1) and the insoluble part was removed by ultracentrifugation at 100,000×g. The supernatant was loaded onto the Tse3-conjugated resin pre-equilibrated with buffer B (20 mM Tris pH 8.0, 150 mM NaCl, 1 mM CaCl2, 0.03% DDM and 0.003% CHS). The Tsi3 tag was removed using on-column digestion by adding 3C-protease overnight. The STING protein was further polished by applying to size exclusion chromatography (SEC) in buffer C (20 mM HEPES pH 7.5, 150 mM NaCl, 0.03% DDM, 0.003% CHS, and 1 mM TCEP).The peak was pooled, concentrated at 8 mg/ml and stored at −80 °C for flux assay.

37Lipids of PC (20 mg/ml), PE (20 mg/ml) and PG (20 mg/ml) were mixed in chloroform at a 2:1:1 ratio and dried using a nitrogen stream, and trace mount of chloroform was removed using a vacuum chamber. Dried lipids were suspended in internal buffer C (50 mM HEPES pH 7.4, 450 mM KCl and 5 mM DTT) generating the liposome with a concentration of 10 mg/ml. The unilamellar liposomes was formed by freeze–thawing cycle followed by sonication. An aliquot of liposomes were mixed with 0.7% DDM for 2 h at 25 °C. Human STING protein was then added to the liposome/DDM mixture (ratio 1:100) and incubated for 1 h at 25 °C. DDM was completely removed by adding Bio-beads SM2 (Bio-Rad) every 4 h at 4 °C for three times. Generated proteoliposome vesicles were collected and used for ion channel flux assays. Proteoliposomes (12 μl) were mixed with 160 μl external flux assay buffer D (5 mM HEPES pH 7.4, 450 mM NaCl, 5 mM DTT and 2 μM ACMA) in a 96-well fluorescence assay plate. AMCA fluorescence intensity was measured over time (λEx = 410 nm, λEm = 490 nm). The 0.45 μM valinomycin was added to initiate the flux and fluorescence data were collected at 20-second intervals for 15 min, and then 1 μM CCCP was used to collapse the proton gradient. Data were processed according to established protocol (Su et al, ).

38Cells were treated to activate STING as described above. When obvious cell death was observed, cells were imaged by phase contrast microscopy followed by fixation and permeabilization. Cells were then stained with DAPI and imaged to collect the nuclear images for cell number quantification by a custom code. Dead or dying cells with irregular nuclear shapes were excluded automatically by the software. The total live cells were quantified in each condition and compared with the control cells without STING-dependent cell death.

39For the quantification of fluorescence images, the outlines of randomly selected cells were manually annotated in each image by an investigator blinded to the allocation. The fraction of protein A’s intensity on protein B was quantified by comparing the sum of signal intensities of protein A overlapping with protein B to the total intensity of protein A in a given cell. When quantifying the fraction of protein A positive for protein B puncta, a higher threshold was applied to minimize diffuse protein B signals. Thresholds for both proteins were determined by the target protein’s signal intensity percentile within a single cell, adjusted by a small constant. All cells from the same experiments were applied using the same threshold settings. Manual checking ensured accurate threshold applications. Quantitative data were exported to Excel and visualized using Prism to display mean and SEM values.

40A Leica SP8 LIGHTNING confocal system and its built-in software Leica Application Suite X 3.5.5.19976 were used to collect confocal images as well as for living cell imaging. Images were processed in Adobe Photoshop 20.0.4. Schematic illustration figures were made with Biorender and Microsoft Office PowerPoint Professional Plus 2016. Protein structures were visualized in the PyMOL, Version 2.4.0. Graphs were generated in GraphPad Prism 9.0.0. Images were quantified using custom-coded algorithms tailored according to specific experiments to incorporate quality control measures, minimize errors, and ensure accuracy and reliability.

41All experiments were independently reproduced at least three times unless otherwise indicated. Strict standards were applied to screen for robust and unbiased results. No statistical methods were used to predetermine sample size. The investigators were blinded to allocation during data collection and image quantification. Data were presented as mean ± SEM. Statistical significance was determined by unpaired, two-tailed t tests.