Cellular host defense responses to pathogen invasion principally involves the detection of pathogen associated molecular patterns (PAMPs) such as viral nucleic acid or bacterial cell wall components including lipopolysaccharide or flagellar proteins that results in the induction of anti-pathogen genes[3
]. For example, viral RNA can be detected by membrane bound Toll-like receptors (TLR’s) present in the endoplasmic reticulum (ER) and/or endosomes (e.g. TLR 3 and 7/8) or by TLR-independent intracellular DExD/H box RNA helicases referred to as retinoic acid inducible gene 1 (RIG-I) or melanoma differentiation associated antigen 5 (MDA5, also referred to as IFIH1 and helicard)[3
]. Pathogen DNA can be recognized by TLR 9 present in plasmacytoid dendritic cells, although it is now apparent that important TLR-independent pathways also exist to recognize DNA in alternate tissue, the mechanisms of action of which remain to be determined [3
]. These events culminate in the activation of downstream signaling events, leading to the transcription of NF-κB and IRF3/7-dependent genes, including type I IFN.
To further determine the mechanisms of innate immune signaling, we employed an expression screening system in which approximately 5,500 human and 9,000 murine full length cDNA’s were individually transfected into 293T cells harboring a luciferase gene under control of the IFNβ promoter (IFNβ-Luc). Five genes whose overexpression lead to the significant induction of IFNβ-Luc was found to be IPS-1 (also referred to as VISA/CARDIF/MAVS) (Supplemental Fig. 1)[11
]. However, we also isolated a previously uncharacterized molecule (gi:38093659/NP_938023/2610307O08RIK) which we referred to as STING (for STimulator of INterferon Genes) that harbored 5 predicted TM motifs (in humans) and existed as a 379 amino acid protein in human cells and 378 amino acids in murine cells ( and Supplemental Fig 1). A putative signal cleavage motif was found to exist at position 1–36 and a leucine rich region was apparent between amino acids 21–139 (). The predicted molecular weight of human STING (hSTING) was 42,192 Da, which approximately corresponded to its observed molecular weight in human 293 cells following immunoblot analysis using a rabbit antiserum raised to a STING peptide (). RNAi studies confirmed that the observed 42kDa band was indeed STING (). STING was found to be ubiquitously expressed in a variety of tissues as determined by northern analysis and was found to predominantly reside in the ER region of the cell as determined by confocal microscopy and fractionation analysis ().
STING is an ER protein
Overexpression of STING in 293T cells was subsequently confirmed to robustly induce the expression of the IFN promoter (IFNβ-Luc) up to 400-fold, but not a control TK promoter driving luciferase (pRL-TK), interferon regulatory factor 3 (IRF3) responsive promoters (PRD-III-I-Luc) up to 1000-fold, an NF-κB responsive promoter (NF-κB-Luc) 12-fold and interferon-inducible promoters (interferon sensitive response element- ISRE-Luc) up to 800-fold (). STING did not activate control promoters driving luciferase reporters such as those derived from the Rb, p53 or E2F genes (Supplemental Fig. 2). Increased dimerization of IRF3 was also observed in STING expressing 293T cells confirming that STING regulates the induction of type I IFN at upstream of IRF3 activation, dimerization and translocation (Supplemental Fig. 2)[15
]. Endogenous IFNβ mRNA and IFNβ protein was induced significantly in MEFs transiently transfected with STING (). DNA microarray analysis of STING expression in 293T cells further emphasized STING’s ability to induce primary innate immune response genes (). Accordingly, MEFs cells expressing STING or IPS-1 were significantly resistant to VSV infection (). Subsequent analysis indicated that STING function was ablated in the absence of the IκB kinase family member TBK-1, confirming that STING’s activity involved activation of IRF3 and was indeed upstream of this kinase ()[15
]. Finally, we observed that STING did not exert robust activity in the absence of FADD, which has also been shown to be important for efficient innate immune signaling processes (Supplemental Fig. 2)[16
]. We also established that while STING, HA tagged at the carboxyl region, retained activity, this activity was lost when STING was tagged at the amino terminus or carboxyl terminus with GFP (data not shown). Subsequent analysis indicated that the amino terminus region of STING containing the 5 putative TM regions (amino acids 1–230) or just the carboxyl region of STING (amino acids 173–379) did not exhibit significant ability, alone, to induce the IFNβ-Luc promoter (). Thus, full length intact STING is required for efficient function. However, we further observed that the carboxyl region of STING exerted a dominant-negative inhibitory effect and could impede the ability of full-length STING to stimulate IFNβ-Luc (). Collectively, this data indicates that expression of STING activates the innate immune response including type I IFN leading to the induction of an antiviral state.
STING facilitates IFN induction
To further analyze STING function, we utilized an RNAi approach to ablate STING in a number of cell-types. Our data indicated that knockdown of STING in HEK 293 cells, modestly reduced the ability of the negative-stranded rhabdovirus VSV-GFP to induce IFNβ, presumably since these viruses are only weak activators of IFNβ (Supplemental Fig. 3). However, such cells were rendered extremely susceptible to virus infection and replication. To confirm a requirement for STING in the regulation of type I IFN induction and in host defense, we generated STING deficient (STING
−/−) mice by targeted homologous recombination in ES cells (Supplemental Fig 4). STING
−/− animals were born at the Mendelian ratio and developed and bred normally. Accordingly, MEFs from wild type and STING
−/− animals were infected with VSV-GFP at varying MOI’s (0.01–1) for up to 36 hours post infection. This study confirmed that more progeny virus was produced in MEFs lacking STING compared to controls (2 logs; 24–36 hours) (). This data was verified using VSV expressing a luciferase reporter gene and VSV-ΔM, which exhibits a defect in the viral matrix protein normally responsible for inhibiting cellular mRNA export from the nucleus[17
](Supplemental Fig. 4 and Fig. 3e). Reconstitution of STING to STING −/− MEF’s rescued the susceptibility to VSV infection (Supplemental. ). Similar analysis also indicated that loss of STING also reduced the ability of SeV to induce IFNβ (). In contrast, we did not observe a strong requirement for STING to mediate the ability of transfected poly I:C to induce IFNβ induction, which is largely governed by the intracellular RIG-I homologue MDA5[18
] (Supplemental Fig. 5). We also observed that the positive-stranded encephalomyocarditis virus (EMCV), a member of the picornavirus family, did not effectively induce IFNβ nor replicate differently regardless of the presence of STING (Supplemental Fig. 5). Thus, we conclude that STING may play a more predominant role in facilitating RIG-I mediated innate signaling rather than MDA5. Interestingly, we did notice a significant defect (>5-fold) in the ability of transfected B-form DNA (poly dA-dT) to induce IFNβ in MEFs lacking STING compared to controls (). More strikingly, the non CpG containing interferon stimulatory DNA (ISD) was completely unable to induce IFNβ in STING −/− MEFs, as was the DNA virus herpes simplex virus 1 and bacteria Listeria monocytogenes (). TLR9 is considered to govern CpG DNA-mediated induction of IFNβ, but is not active in MEFs[19
]. Thus, it is plausible that STING may function in TLR9-independent, DNA-mediated induction of type I IFN. This effect was similarly observed in murine STING lacking bone marrow derived macrophages (BMDM) or bone marrow derived dendritic cells (GM-DC) cultured using granulocyte-monocyte colony stimulating factor (GM-CSF) [Supplemental Fig. 6]. However, no significant difference was observed in the ability of exogenous poly I:C or lipopolysaccharide (LPS) to induce IFNβ, when comparing STING −/− BMDM or GM-DC’s to controls, events which depend on TLR 3 and 4 respectively. While, loss of STING rendered MEFs highly susceptible to VSV-GFP, less susceptibility was observed following VSV-GFP infection of STING −/−GM-DC’s or BMDM’s, indicating that STING may be more important in facilitating negative-stranded virus-mediated innate signaling in fibroblasts. Collectively, our data would indicate that loss of STING leads to a defect in RIG-I mediated type I IFN induction but does not affect the TLR pathway. In addition, we report that STING functions in the pathway utilized by intracellular B-form DNA to induce IFNβ.
Loss of STING affects host defense
STING associates with the translocon
To further examine the mechanisms of STINGs function in innate signaling, we attempted to determine if STING interacted with RIG-I and/or MDA5, which are putative innate immune signaling receptors for negative or positive stranded viral RNA respectively[3
]. Co-immunoprecipitation experiments in 293T cells principally indicated that FLAG-tagged RIG-I but not MDA5 could associate with HA-tagged STING in co-transfection experiments (). The binding of RIG-I to STING was augmented upon infection of 293T cells with SeV (). We confirmed in normal human umbilical vein endothelial cells (HUVECs) that endogenous RIG-I could associate directly, or indirectly as a complex, with endogenous STING (). This data would be in agreement with previous data indicating that STING seems to preferentially modulate the RIG-I, rather than the MDA5 regulated pathway ( and Supplemental Fig. 5). We also determined that the CARD domains of RIG-I (amino acids 1-284) were preferentially able to associate with HA-STING in transfected 293T cells (). RIG-I was subsequently shown to colocalize with STING in the co-transfected 293T cells (). It was similarly observed that the carboxyl region of STING could inhibit the function of RIG-I, again indicating that this region of STING can exhibit a dominant-inhibitory effect (). We also noted association of the RIG-I downstream adaptor IPS-1 with STING although similar to the situation with RIG-I, it is not yet clear whether IPS-1 directly interacts with STING or exists as a complex with RIG-I/STING (Supplemental Fig 7). Accordingly, we observed that STING was able to induce the induction of an IFNβ driven luciferase construct in MEFs lacking RIG-I or IPS-1, likely confirming that that STING functions downstream of these latter molecules (Supplemental Fig 8). The ability of ΔRIG-I or IPS-1 to induce IFNβ appeared diminished in STING −/−
MEFs (). However, there was also a marked reduction in the induction of IFNβ by all transfected plasmids (including vector alone) in the absence of STING compared to control MEFs, since endogenous DNA innate signaling pathways are likely defective (). This data indicates that STING may be an important downstream adaptor molecule that facilitates RIG-I and perhaps IPS-1 function.
To shed further insight into the molecular mechanisms of STING action, we screened an IFN-induced, human fibroblast yeast two-hybrid cDNA library using STING (amino acids 173–379) as a bait and repeatedly isolated Ssr2/TRAPβ, a member of the TRAP complex comprising four subunits (α-Δ) that facilitates translocation of proteins into the ER following translation ()[1
]. The TRAP complex is known to associate with the translocon, comprising three subunits, Sec61 α, β and γ. Given this information, we confirmed that TRAPβ can indeed associate with endogenous STING in HEK 293 cells following co-immunoprecipitation experiments (). Using this approach, we confirmed that TRAPβ also co-immunoprecipitated with endogenous Sec61β (). We next verified that STING could also associate not only with TRAPβ but also Sec61β likely as a complex (). STING was also observed to colocalize with TRAPβ in the ER region of the cell (). Indeed, loss of TRAPβ or SEC61β reduced STING’s, ability to induce an IFNβ promoter driving luciferase (; Supplemental Fig. 9). Taken together, this data would indicate that the STING may be involved in translocon function and that the translocon may be able to influence the induction of type I IFN.
Thus, STING is predominantly an ER resident protein that may link RIG-I and DNA-mediated intracellular innate signaling to the translocon. Speculatively, RIG-I may detect translating viral RNA’s at the intersection of ribosome/ER translocon association and/or with mitochondrial associated ER membranes (MAM’s) and require STING to exert effective function[2
]. Alternatively, STING may participate in mediating ER stress response pathways an event that remains to be verified. While it is not clear how signaling from the translocon to IRF-3/NF-kB occurs, it has recently been established that translocon physically associates with the exocyst - the octomeric Sec6-Sec8 complex that also associates with the ER and tethers secretory vesicles to membranes, and facilitates protein synthesis and secretion[21
]. Recently, the exocyst complex was found to recruit and activate TBK1 and play a role in type I IFN-β induction[23
]. Our preliminary analysis indicates that STING also co-immunprecipitates with TBK1 and that RNAi ablation of Sec5 also rendered cells defective in the STING function (). Thus, STING may facilitate the detection of intracellular viral RNA species as well as B-form DNA indicating convergence of these intracellular PAMP recognition pathways.