|Home | About | Journals | Submit | Contact Us | Français|
The intracellular microbial nucleic acid sensors, TLR3 and STING, recognize pathogen molecules and signal to activate the interferon pathway. The TIR-domain containing protein TRIF is the sole adaptor of TLR3. Here we report an essential role for TRIF in STING signaling: various activators of STING could not induce genes in the absence of TRIF. TRIF and STING interacted directly, through their carboxyl terminal domains, to promote STING dimerization, intermembrane translocation and signaling. Herpes simplex virus (HSV), which triggers the STING signaling pathway and is controlled by it, replicated more efficiently in the absence of TRIF and HSV-infected TRIF−/− mice displayed pronounced pathology. Our results indicate that defective STING signaling may be responsible for the observed genetic association between TRIF mutations and Herpes Simplex Encephalitis in patients.
The adaptor TRIF is required for TLR signaling. Wang et al report that TRIF is also required for STING signaling. TRIF interacts directly with STING to promote its dimerization and membrane translocation. TRIF mutations, that cause defective STING signaling, correlate with enhanced HSV-1 replication and pathogenesis in mouse and man.
A variety of pattern recognition receptors (PRR) initiate cellular inflammatory response to external stresses. They recognize not only microbial pathogen-associated molecular patterns (PAMP) but also cellular damage-associated molecular patterns (DAMP). PRRs include Toll-like receptors (TLR), Nod-like receptors (NLR), RIG-I-like receptors (RLR) and various DNA-recognizing receptors, such as cGAS, DAI, IFI16, DDX41 and STING (Janeway and Medzhitov, 2002; Pandey et al., 2015). All PRRs use specific adaptor proteins to recruit various protein kinases, ubiquitin ligases and transcription factors to the signaling complexes; the activated transcriptions factors translocate to the nucleus and induce transcriptions of genes encoding inflammatory cytokines and type I interferon (IFN). STING (MITA, ERIS, MPYS) is the nodal point for receiving signals from a variety of upstream stimulators that include c-di-GMP or c-di-AMP produced by intracellular bacteria and cGAMP synthesized by the mammalian enzyme, cGAS, in response to cytoplasmic DNA (Wu and Chen, 2014). Other DNA receptors, such as IFI16 and DDX41, signal through STING as well (Cai et al., 2014; Paludan and Bowie, 2013). STING provides protection against many bacterial and viral infections (Archer et al., 2014; Dey et al., 2015; Hartlova et al., 2015; Ishikawa et al., 2009); for example, STING signaling protects cells and mice from pathogenesis caused by Listeria monocytogenes and herpes simplex virus 1 (HSV-1). The protective effects are mediated by cytokines, such as type I interferon (IFN), that are induced by STING signaling. In contrast, we have reported a pro-microbial role of STING signaling in cells and mice infected with the protozoan parasite, T. gondii (Majumdar et al., 2015). STING-signaling is also triggered in response to cellular DNA damage (Hartlova et al., 2015) and STING-mediated inflammation has recently been connected to both cancer development and antitumor immunity. STING−/− mice are highly resistant to DMBA-induced skin carcinogenesis (Ahn et al., 2015; Ahn et al., 2014; Barber, 2015). On the other hand, STING promotes tissue repair and protects against colorectal tumorigenesis (Ahn et al., 2015; Zhu et al., 2014). Moreover, in both ionizing radiation-mediated tumor regression (Deng et al., 2014) and T cell-mediated tumor growth control (Woo et al., 2014), STING plays an essential protective role. Therefore, a better understanding of the mechanism of STING signaling is relevant for combating both infectious diseases and cancer.
STING is a transmembrane protein in the ER with a long cytoplasmic domain which binds the activating ligands and signals downstream to induce IFN and other cytokine genes (Barber, 2015; Ran et al., 2014). Structural studies suggest that different activators of STING induce similar, but distinct, conformational changes to the protein; hence, it is likely that the complexity of downstream signals produced by activated STING is dependent on the nature of the activator (Gao et al., 2013; Huang et al., 2012; Ouyang et al., 2012). The ligands promote ubiquitination of STING and its dimerization which are necessary for its ability to signal (Jeremiah et al., 2014; Liu et al., 2014; Tsuchida et al., 2010; Wang et al., 2014; Zhang et al., 2012); consequently a STING mutant, that is constitutively dimeric, signals without any ligand stimulation (Jeremiah et al., 2014; Liu et al., 2014). Activated STING translocates from the ER, in a sequential manner, to other membranes; from the ER, it moves to the ER-Golgi intermediate compartment (ERGIC), where its cytoplasmic carboxyl-terminal domain recruits the signaling proteins. Eventually, it translocates, through the Golgi, to punctated perinuclear microvesicles, where signaling may be turned off (Dobbs et al., 2015).
TRIF is a Toll/IL-1R (TIR) domain containing adapter protein. It is essential for gene induction by TLR3 which signals from the endosomal membrane (Yamamoto et al., 2002); it is also required for the endosomal branch of TLR4 signaling (Fitzgerald et al., 2003). Its interaction with the two TLRs is mediated by its TIR domain (O’Neill and Bowie, 2007). Another domain of TRIF, the RHIM domain, mediates its interaction with RIP1 and RIP3 causing cellular apoptosis (Meylan et al., 2004). TRIF has been linked to herpes simplex encephalitis (HSE) in both mouse and man (Menasria et al., 2013; Sancho-Shimizu et al., 2011). In children with HSE, both autosomal recessive and autosomal dominant mutations in TRIF have been causally connected to the disease. In most of these patients, TRIF protein was completely absent or dysfunctional; however one family carried a specific mutation in TRIF, which did not abrogate its function in TLR signaling indicating that TRIF may be involved in other innate immune signaling pathways activated by HSV1 infection (Sancho-Shimizu et al., 2011).
We report here that TRIF was essential for optimal signaling by STING; it promoted STING dimerization by direct interaction. Consequently, the absence of TRIF eliminated both anti-microbial and pro-microbial effects of STING signaling in cells and mice.
We previously reported that the STING pathway is activated in cells in response to infection by T. gondii and it promotes parasite growth (Majumdar et al., 2015). To identify the components of the signaling pathways activated by parasite infection, we measured the levels of parasite SAG1 protein expression in different knock-out MEFs. Surprisingly, SAG1 expression was low in TRIF−/− MEF. When TRIF expression was restored in these cells to the Wt level (Fig S1A), SAG1 expression was greatly enhanced (Fig S1B) indicating that TRIF was indeed required for efficient parasitic gene expression leading us to hypothesize that TRIF participates in STING signaling. To test our hypothesis, we measured the levels of gene induction, in the paired TRIF−/− and TRIF+ (TRIF restored) MEFs, in response to two direct activators of STING: c-di-GMP and cGAMP. Transfection of c-di-GMP to TRIF+ cells caused strong induction of IFN-β mRNA (Fig 1A) and Ifit2, IFN-α, IL6 and TNF-α mRNAs (Fig S1C); in contrast, none of these mRNAs was induced appreciably in TRIF−/− cells; the same observations were true for induction by cGAMP (Fig 1A and Fig S1D). Similarly, when we transfected ISD and poly (dA:dT), which indirectly activate STING by activating cGAS, there was no induction of IFN-β mRNA in TRIF−/− cells (Fig S1E). Absence of TRIF did not affect the level of STING expression (Fig S1F) nor did it cause a loss of cell viability in response to treatments with various stimulants (Fig S1G). Moreover, the defect of STING-signaling in TRIF−/− cells was not due to a kinetic delay of induction; labile IFN-β mRNA was induced rapidly but transiently in WT cells, whereas in TRIF−/− cells there was hardly any induction even at later time points (Fig S1H).
As expected, TRIF−/− MEFs secreted little TNF-α and IFN-β upon c-di-GMP stimulation (Fig 1B). Similarly, a STING induced cytoplasmic protein, Ifit2, was not induced in c-di-GMP-treated TRIF−/− MEF, but ectopic expression of TRIF in the same cells restored its induction (Fig S1I, upper panel). However, induction of the same protein by a different inducer, IFN-β, was equally strong in Wt and TRIF−/− MEFs (Fig S1I, lower panel), indicating that Ifit2 gene induction did not require TRIF per se. Another TRIF-null mouse strain, Lps2, has an inactivating point mutation in the TRIF gene (Hoebe et al., 2003); Lps2 MEF was equally unresponsive to STING signaling by c-di-GMP (Fig 1C), confirming the need for TRIF in STING signaling. To extend our observation to myeloid cells, gene induction by the STING pathway was measured in mouse primary bone marrow-derived macrophages; Ifit2 was not induced in TRIF−/− cells (Fig 1D). Finally, to test the need of TRIF for STING signaling in human cells, the TRIF gene was disrupted in two human cell lines, HeLa-M and HT1080, using the CRISPR-Cas9 approach; TRIF expression was abolished in both lines without affecting STING expression (Fig S1J). STING signaling failed to induce IFIT1 in HeLa-M cells in the absence of TRIF (Fig 1E). Similarly, there was no induction of IFN-β mRNA in TRIF−/− HT1080 cells (Fig 1F); when TRIF expression was restored in the same cells, there was strong induction of IFN-β mRNA in response to cGAMP. There was a basal level of IFN-β mRNA expression in TRIF-restored cells, which was probably due to TRIF over-expression because ectopic expression of TRIF in WT MEF caused a low basal level of Ifit2 expression without any ligand stimulation (Fig S1K).
To assess the generality of the need of TRIF for gene induction by STING signaling, gene expression profiling was done using microarray analysis of RNAs obtained from untreated and c-di-GMP-treated TRIF+ and TRIF−/− cells. Induction of all genes by the STING pathway required TRIF (Table S1). The levels of the top fifteen induced mRNAs in the two untreated and treated cell lines are shown in Fig 1G. Whereas all of them were strongly induced in TRIF+ cells, none was induced in TRIF−/− cells. It was curious to note that for many of these mRNAs, the level of expression in vehicle-treated cells was higher in TRIF+ cells indicating that TRIF was required for the basal level of signaling by STING as well. The above results demonstrated the universal need of TRIF for gene induction by STING signaling, irrespective of the cell type, the inducer and the induced gene.
IRF3 is a transcription factor, that is activated by STING signaling upon its phosphorylation by TBK1, which itself is also activated by phosphorylation(Ma et al., 2012). Phosphorylation of both IRF3 and TBK1, in response to c-di-GMP treatment, was severely impaired in TRIF−/− cells; phosphorylation of Akt, another kinase involved in IRF3 activation (Joung et al., 2011), was also impaired, but to a lesser degree (Fig 2A). Activation of the NF-κB pathway by STING signaling was monitored by IκB phosphorylation and degradation; 30 min after stimulation, IκB-α was strongly phosphorylated and partially degraded in WT, but not TRIF−/−, cells (Fig 2B). In contrast, STAT6 activation, by Sendai virus (SeV) infection, was not affected in the absence of TRIF (Fig S2A). The need of TRIF for STING signaling was pathway-specific: induction of IFN-β mRNA by R848, a ligand of TLR7, or SeV, an activator of RIG-I, did not require TRIF (Fig S2B). Moreover, gene induction by the STING-TRIF pathway did not require the presence of TLR3, for which TRIF is the exclusive adaptor (Fig S2C).
Because IRF3 and TBK1 form a complex with STING upon its activation(Wu and Chen, 2014), the three proteins co-immunoprecipitated from extracts of stimulated Wt cells (Fig 2C). In contrast, little TBK1 and IRF3 co-precipitated with STING from extracts of TRIF−/− cells demonstrating that the presence of TRIF was required for the signaling complex formation. Moreover, in Wt cells, TRIF co-precipitated with STING indicating that it might be a component of the same complex. Putative interaction between STING and TRIF was further investigated using several complementary approaches. Ectopically expressed TRIF and STING co-precipitated when TRIF was affinity-purified (Fig S2D). The two endogenous proteins were associated with each other as well, even in unstimulated cells (Fig S2E). To ascertain whether their interaction was direct, His-tagged TRIF, expressed in HEK293 cells, was affinity purified to apparent homogeneity (Fig 2D) and HA-tagged STING was immuno-affinity purified under high salt conditions to remove all associated proteins. Agarose bead-bound STING was incubated with purified TRIF in the presence of large excess of BSA; TRIF specifically bound to STING-containing beads (Fig 2E) indicating that the two proteins could interact directly without the assistance of a linker protein. STING is an integral membrane protein of the endoplasmic reticulum (ER) and because we observed association of TRIF with STING, even in un-stimulated cells (Fig S2E), we wondered whether TRIF, a cytoplasmic protein, was ER-associated as well. To test this idea, we purified ER from MEF cells; it contained the ER-marker, calnexin, but was devoid of the cytoplasmic markers, α-tubulin, and caspase-3 (data not shown). As expected, the purified ER fraction was rich in STING; but it also contained TRIF, thus confirming constitutive association of TRIF and STING (Fig 2F).
We further characterized the STING-TRIF interaction by mapping their interaction domains using co-precipitation assays. The TIR domain of TRIF mediates its interaction with TLRs; however, this region of the protein was not needed for STING interaction; a mutant of mouse TRIF, lacking the TIR domain (ΔTIR), bound to STING strongly, as did the mutant, ΔN, lacking residues 1 to 384 (Fig 3A). However, another mutant, ΔC, lacking the C-terminal region (542–732), failed to interact with STING. Similar results were obtained with human TRIF: ΔTIR and ΔN mutants bound to STING but ΔC did not (Fig 3B). A natural splice variant of human TRIF (Han et al., 2010), TRIS, also interacted with STING. The above results indicated that the C-terminal region of TRIF mediated the STING interaction. Indeed, TRIF mutant C, containing only residues 542–712 of TRIF, interacted strongly with STING (Fig 3B). Another domain, called the RHIM domain, is located near the C-terminus of TRIF and specific mutations in this domain perturb its interaction with RIP-1, a signaling protein (Meylan et al., 2004). However, the RHIM mutant of TRIF interacted with STING as strongly as the Wt protein (Fig S3A) indicating that TRIF-STING interaction was not mediated by RIP-1. We investigated the functionality of different TRIF mutants in STING signaling by expressing them in TRIF−/− MEF and measuring cytokine mRNA induction in response to c-di-GMP stimulation (Fig 3C). Full length (FL) and ΔTIR TRIF mediated the induction of both IFN-α and IL-6 mRNAs and ΔC, which did not interact with STING, was functionally inactive, as expected. Interestingly, ΔN, which could interact with STING (Fig 3A), could not support gene induction (Fig 3C) indicating that the N-terminal region of TRIF was required for its functioning in the STING pathway; similar results were obtained when Ifit2 induction was monitored by western blot (Fig S3B). The reciprocal domain mapping of STING revealed that its C-terminal region, comprising of residues 220 to 378, mediated its interaction with TRIF. The complete cytoplasmic domain of STING (110–378) or its deletion mutant (220–378) interacted with TRIF (Fig 4A); but even a small deletion from the C-terminus disrupted the interaction (Fig S3C, S3D). Two residues S365 and L373, located near the C-terminus of STING, are required for its interaction with IRF-3 (Tanaka and Chen, 2012); however, these residues were not required for TRIF interaction (Fig 4B). The above results indicated that the TRIF-STING interaction is mediated directly by the C-terminal domains (CTD) of both proteins. Indeed, we observed strong interaction between two proteins containing only TRIF CTD and STING CTD respectively (Fig 4C).
Activators of STING facilitate its dimerization and intracellular membrane translocation, processes that are required for its ability to signal. Because dimeric STING partially resists dissociation in SDS-PAGE, it can be easily identified by its slower mobility in western blots (Sun et al., 2009; Tsuchida et al., 2010). c-di-GMP stimulation of Wt MEF caused STING dimerization but no dimer was detected in similarly treated TRIF−/− MEF (Fig 5A). In a different assay for STING dimerization, we expressed together HA tagged STING and myc-tagged STING and assayed their co-immunoprecipitation from extracts of ligand-stimulated cells. Myc-STING was co-immunoprecipitated with HA-STING only in Wt cells, not in TRIF−/− cells (Fig 5B). Both monomeric and dimeric Myc-STINGs were observed in Wt cells, indicating partial dissociation of the dimer in SDS-PAGE; in contrast neither form of Myc-STING immunoprecipitated with HA-STING in TRIF−/− cells. Overexpression of TRIF in Wt MEFs caused STING dimerization even without ligand stimulation (Fig 5C) demonstrating that TRIF can promote STING dimerization even in the absence of an exogenous ligand of STING. We have shown above that the ΔN mutant of TRIF cannot support gene induction although it binds STING (Fig 3C); the defect of this mutant was traced to its inability to promote STING dimerization (Fig 5D).
Signaling by STING is preceded by its translocation from the ER membrane to another membrane compartment, called ERGIC, followed by its clustering (Dobbs et al., 2015). We monitored this process in Wt and TRIF−/− cells, after c-di-GMP stimulation, using two STING visualization methods: either GFP tagged STING was visualized directly (Fig 5E) or HA tagged STING was visualized by staining with HA antibody (Fig S4A). Using either method, we observed distinct clustering of STING upon ligand stimulation of Wt MEF, but not TRIF−/− MEF. Cells expressing diffused STING or clustered STING were counted to quantitate STING mobilization and it was apparent that in TRIF−/− MEF, STING clustering did not happen even after ligand stimulation (Fig 5F, S4A). These results demonstrated that STING dimerization and membrane translocation, steps essential for its ability to signal, required its interaction with TRIF. To strengthen the above conclusion, we tested the properties of a STING mutant, V155M, which is constitutively dimeric (Jeremiah et al., 2014; Liu et al., 2014). As expected, dimeric STING was detected in TRIF−/− MEF expressing ectopic mutant STING, but not Wt STING (Fig 5G). In TRIF−/− MEF, the mutant was located in the signaling compartment, as evidenced by its clustering (Fig S4B). The dimeric mutant STING was fully functional even in the absence of TRIF; it could activate TBK1 and IRF3 and induce Ifit2 gene expression in TRIF−/− cells (Fig 5H). These results established that the requirement of TRIF in STING signaling was at the stage of its dimerization and subsequent translocation to the signaling compartment.
STING signaling is important for its role in innate immune defense of the host against DNA viruses, such as HSV1. HSV1 infection activates STING signaling causing induction of IFN and other cytokines which, in turn, attenuate virus replication. Once we established the need of TRIF in STING signaling, we wanted to examine the role of TRIF in HSV1 infection. In Wt or TRIF+ MEFs, HSV1 infection induced Ifit2 synthesis (Fig 6A); however, because this induction was mediated by the STING pathway, there was no Ifit2 induction in TBK1−/− or STING−/− MEFs as well as TRIF−/− MEFs. In the latter cells, cytokine induction by virus infection was strongly impaired too (Fig 6B). STING-mediated innate immunity was inoperative in TRIF−/− cells; consequently HSV1 replicated better in those cells, as evidenced by higher levels of the viral protein ICP0 (Fig S5A) indicating that TRIF was required for STING-mediated attenuation of HSV1 replication. To further document TRIF-mediated inhibition of virus replication, we measured viral DNA levels in the infected cells. There was no difference in virus adsorption by Wt and TRIF−/− cells as measured by cell-associated viral DNA right after virus adsorption (0 h). Viral DNA replication is a late event, and it was observed only after 12 h, but not at 4 h, post infection; much more viral DNA accumulated in the infected TRIF−/− cells as compared to Wt cells (Fig 6C).
TRIF deficiency has been associated with HSE in children. Sancho-Simizer et al (Sancho-Shimizu et al., 2011) reported a specific TRIF mutation, P625L, in a boy suffering from HSE. However, it was not clear how this particular mutation affected TRIF function because this TRIF mutant was active in the TLR signaling pathway that use TRIF as an adaptor. In light of our observation of the requirement of TRIF in STING signaling, known to be activated by HSV1, we wondered whether TRIF P625L was unable to support STING signaling. To test this idea, we expressed Wt and mutant TRIFs to similar levels in TRIF−/− cells (Fig S5B). When we compared the properties of these cells, it was apparent that the mutant TRIF could not support STING signaling; c-di-GMP treatment did not cause gene induction in cells expressing the mutant TRIF (Fig 6D). The signaling defect was due to a failure to promote STING dimerization (Fig 6E) and STING mobilization (Fig S5C). The functional defect of TRIF P625L was traced to its inability to bind to STING strongly (Fig 6F). Consequently, HSV1 replicated more efficiently in cells expressing the mutant TRIF compared to those expressing the Wt TRIF, as evidenced by the difference in the levels of viral DNA 16 h after infection (Fig 6G). The difference was not due to different susceptibility of the two cell populations to HSV1 infection because similar low amounts of input viral DNA were present in all cells early after infection (Fig S5D). These results indicate that more efficient HSV1replication and resultant pathogenesis in the patient, expressing P625L TRIF, could be due to defective STING signaling. To directly test the role of TRIF in regulating HSV-1 pathogenesis, Wt and TRIF−/− mice were infected with HSV-1 and disease progression was followed for 9 days (Fig 7). Whereas no Wt mouse succumbed, 6 out of 9 TRIF−/− mice died during this period (Fig 7A); the rates of weight loss were similar but the Wt mice recovered more rapidly (Fig 7B) and late after infection, there was a much higher titer of virus in the brains of TRIF−/− mice (Fig 7C). These results demonstrated that in a mouse model, HSV-1, which is known to activate STING signaling, was highly pathogenic in the absence of TRIF.
In contrast to its antiviral role, STING signaling, which activates IRF-3, is required for robust replication of T. gondii (Majumdar et al., 2015) ; consequently the parasite could not replicate well in TRIF−/− MEF (Fig S1B). To determine whether TRIF was needed for parasite replication in other, more physiologically relevant, cell types as well, we measured the efficacy of T. gondii replication and the resultant pathogenesis in TRIF−/− mice (Fig S6). As reported before (Majumdar et al., 2015), all Wt mice succumbed to the infection (Fig S6A) after losing substantial body weight (Fig S6B) within 1 week after infection whereas IRF3−/− mice were quite resistant. The response of TRIF−/− mice was similar to that of IRF3−/− mice: there was little weight loss (Fig S6B) and fewer deaths (Fig S6A) even late after infection. These results established the need of TRIF in triggering STING signaling by parasite infection vivo.
We have uncovered a function for TRIF in mediating innate immune response to PAMPs and DAMPs. Our results demonstrated universal need of TRIF for STING signaling; in a variety of cell types, STING ligands failed to induce genes in the absence of TRIF (Fig 1) because the STING signaling complex did not assemble (Fig 2A, 2B, 2C). In TRIF−/− cells, the defect was traced back to impaired dimerization and membrane translocation of STING (Fig 5), processes that are required for initiating signaling. STING dimerization-promoting effects of TRIF were mediated by direct interaction between the two proteins which was ligand-independent (Fig 2C, 2D and 2E). Even in unstimulated cells, endogenous TRIF co-precipitated with endogenous STING (Fig S2E) and because STING is an ER protein, a portion of TRIF was ER-bound as well (Fig 2F). Although TRIF and a cyclic dinucleotide ligand were both needed for efficient STING dimerization and signaling, overexpression of TRIF could promote STING dimerization even in the absence of exogenous ligands (Fig 5C) causing gene induction (Fig S1K). It is safe to conclude from these observations that TRIF’s role in STING signaling is to promote dimerization and/or stabilize STING dimers. Accordingly, the constitutively dimeric V155M mutant of STING did not require TRIF to signal (Fig 5H).
Our results support a simple model for the role of TRIF in STING signaling. In this model, dimerization and conformational change of STING are both required for its ability to signal; hence, in the absence of TRIF, there is too little STING dimer to signal even in the presence of the ligand. Binding of TRIF shifts the monomer-dimer equilibrium of STING to the dimeric side and binding of the ligand promotes the conformational change. The literature suggests that any process that shifts the equilibrium of STING to the dimeric form, promotes signaling (Jeremiah et al., 2014; Sun et al., 2009). STING can form dimers without any ligands; for example, cellular overexpression of STING can promote its dimerization through mass action. Physiologically, its ligands, such as cGAMP, need the dimeric interface to bind strongly and activate signaling by changing the protein conformation (Gao et al., 2013). Indeed, STING activity can be observed in a cell-free in vitro system containing STING at a high concentration (Tanaka et al., 2013). Our findings indicate that TRIF promotes STING dimerization, possibly by stabilizing the dimeric form of STING, and it is required for effective signaling by STING at the physiologically relevant cellular concentrations of both proteins. Residues at the very C-terminus of STING were required for TRIF binding (Fig S3) as well as for its ability to signal (Tanaka et al., 2013). This region was observed to have a flexible structure in the STING crystals produced in the absence of TRIF (Gao et al., 2013); TRIF might be stabilizing the structure of this region by binding to it. Moreover, TRIF itself can form dimers (Han et al., 2010; Tatematsu et al., 2010) and that property of TRIF may also contribute to its ability to stabilize STING dimers. TRIF is known for its role as an adaptor for TLR3 and TLR4 signaling and in that role, it interaction with the TLRs requires the TIR domains of both partners (Fitzgerald et al., 2003; Yamamoto et al., 2002). Ligand-induced conformational change of the TLR3 cytoplasmic domain exposes its TIR domain and allows TRIF binding and the recruitment of signaling proteins (Tatematsu et al., 2010). In contrast, TRIF’s function in STING signaling is quite distinct; as we report here, TRIF does not require the TIR domain for STING interaction; instead, the interaction between STING and TRIF was mediated by the carboxyl terminal domains of the two proteins (Fig 4C). Previously identified motifs in these regions are the RHIM domain of TRIF and the IRF-3 interacting residues of STING (Meylan et al., 2004; Tanaka and Chen, 2012); however, these motifs were not needed for TRIF-STING interaction. It is not clear which one of the two proteins in the TRIF-STING complex recruits the signaling proteins including the kinases and the transcription factors. It appears that STING itself can provide the docking sites for all of them because the STING V155M mutant could signal in the absence of TRIF. It is curious to note, however, that the TRIF ΔN mutant failed to support gene induction by Wt STING although it could bind to STING (Fig 3) and the N-terminal region of TRIF, missing from this mutant, contains known binding sites for signaling proteins, including TBK1 and several TRAFs; it masks the TIR domain, thus preventing signaling in unstimulated cells (Tatematsu et al., 2010). It remains to be seen whether any of the above properties of the TRIF N-terminal region is relevant for STING dimerization and signaling.
We previously reported (Majumdar et al., 2015) the surprising observation that induction of the IFN-stimulated genes is required for the replication of T. gondii. These genes are induced in the parasite-infected cells through the action of IRF-3. By systematic analyses, we demonstrated that the cGAS-STING pathway, and not the TLR or RLR pathways, is used by the parasite to activate IRF-3; consequently the parasite grows poorly in STING−/− cells. Here, we show that the same was true for TRIF−/− MEF (Fig S1B), supporting the need of TRIF for STING signaling. We reported the need of IRF-3 for T. gondii replication in mice (Majumdar et al., 2015). When, using the same approach, we tested TRIF−/− mice, it was apparent that in the absence of TRIF, the parasite was not lethal to the animals (Fig S6). Our in vitro results suggest that cell-intrinsic defect of parasite replication, rather than an immunological anti-parasitic response, was responsible for the resistance of TRIF−/− mice.
In contrast to its pro-microbial role in supporting the replication of T. gondii, the STING pathway is more commonly used by the host to dampen the growth of viruses such as HSV-1. HSV1 triggers STING signaling with the help of the DNA-binding protein IFI16 and cGAS (Johnson et al., 2014; Orzalli et al., 2015). In the absence of STING, the virus replicates more efficiently demonstrating a protective antiviral role of the STING pathway; consequently, HSV-1 infection is lethal in STING−/− mice (Ishikawa et al., 2009; Parker et al., 2015). Consistent with our conclusion that TRIF was required for STING signaling, HSV-1 failed to induce cytokines and replicated more efficiently in TRIF−/− cells. Moreover, when injected intra-peritoneally (Zhou et al., 2014), HSV-1 was more pathogenic in TRIF−/− mice (Fig 7). Using a different infection protocol, it has been shown that HSV-1 replicates better in TRIF−/− mice and causes higher lethality (Menasria et al., 2013); but this observation was interpreted as an evidence for the involvement of TLR3 signaling, which uses TRIF, in host defense against HSV-1. In the light of the observations reported here, it is highly likely that the higher susceptibility to the TRIF−/− mice is due to the lack of activation of STING signaling by HSV-1. A connection between TRIF deficiency and herpes simplex infection has also been observed in humans. Children with autosomal dominant or autosomal recessive TRIF mutations suffer from herpes simplex encephalitis (Sancho-Shimizu et al., 2011). Although the absence of functional TRIF, resulting from missense mutations, can cause impairment of TLR3 signaling, our results suggest that the patients suffer from impaired STING signaling as well and the defect in STING signaling could be the primary cause of enhanced virus replication and pathogenesis. Strong support for the above assertion was provided by the nature of the TRIF mutation in one case in which HSE was associated with a specific TRIF mutation, P625L. However, no biochemical basis for the disease could be ascertained because this mutant TRIF was competent in supporting TLR3 signaling (Supplementary Figs 7A and 8F in Sancho-Shimizu et al.). In contrast, we found that TRIF P625L could not support STING signaling because of its impaired binding to STING (Fig 6F). These results provide further support for the physiological need of TRIF for STING signaling and a reasonable explanation for the etiology of HSE in this particular patient. Overall, our results confirm a protective role of TRIF against HSE and indicate that, in this context, its role in STING signaling may be the critical function.
HeLa-M, HT1080 and HEK293T cells were transfected with lipofectamine 2000 following manufacturer’s instructions. For primary BMDMs, c-di-GMP (8 μg/ml) was transfected using lipofectamine 2000. 15 min later, medium was removed and fresh medium was added. For other cells, cells were transfected with lipofectamine 2000 and c-di-GMP (8 μg/ml) or cGAMP (4 μg/ml) for 30 min following manufacturer’s instructions except as indicated in the paper. R848 was directly added into culture medium to the final concentration of 1 μg/ml.
HSV-1 (KOS) propagation and infection were performed as previously described (Blaho et al., 2005). Virus titers were determined by TCID50 assay. Male C57/Bl6 mice were injected intra-peritoneally with 5 × 107 pfu of HSV-1 (Zhou et al., 2014). Sendai virus (Cantell strain) was used.
For IP, cells were lysed in lysis buffer containing 20 mM Hepes (pH 7.4), 50 mM NaCl, 1.5 mM MgCl2, 2 mM dithiothreitol (DTT), 2 mM EGTA, 10 mM NaF, 12.5 mM β-glycerophosphate, 1 mM Na3VO4, 5 mM Na-pyrophosphate, 0.2% (v/v) Triton X-100, and protease inhibitors (Roche Applied Science). Cellular lysates were pre-cleared with mouse immunoglobulin G (IgG) agarose or rabbit immunoglobulin G (IgG) agarose (Sigma), dependent on the species of antibody used in IP, for 1 hour, incubated overnight with IP antibodies and then with protein A/G beads for 1 hour with rotation. After incubation, beads were washed with lysis buffer and boiled with 4× Laemmli sample buffer. For pull-down, cells were lysed in lysis buffer supplemented with 5 mM imidazole. Cellular lysate was mixed with Ni-NTA agarose (Thermo Fisher) for at least 4 hours with rotation and Ni-NTA agarose was washed with lysis buffer supplemented with 20 mM imidazole. Samples were boiled in 4× Laemmli sample buffer. For Western-blot, cellular lysates were boiled in 1× Laemmli sample buffer and fractionated in SDS-PAGE gel.
For ELISA, culture supernatants from indicated cells were collected at 5 hours post c-di-GMP treatment, and cell debris was pelleted by centrifugation. Clarified supernatant were measured by ELISA kits following manufacturers’ instructions.
For microarray analysis, experiments were performed in duplicates. RNA hybridization to chips was performed by the Lerner Research Institute Genomics Core at the Cleveland Clinic. Messenger RNA expression microarray was analyzed via Illumina Mouse Ref-8 V2 beadchip and GenomeStudio software V2010.2 (Illumina, Inc.). The average signal values for two samples in each gene group was calculated and shown. Genes with “detection p value”≤0.005 were picked as significant signals.
Immunofluorescence staining (IF) was performed as previous described (Wang et al., 2013). Microscopy analysis was performed in the Lerner Research Institute Image Core at the Cleveland Clinic (more details in supplemental experimental procedure).
TRIF and STING were purified from 293T cells, washed by high salt buffer, re-folded at 4 °C overnight, and then used in in vitro interaction assay (more details in supplemental experimental procedure). To analyze STING-TRIF interaction in vitro, 200 ng of TRIF in combination with 2 mg BSA was added into 100 μl binding buffer (20 mM Tris-HCl, 150 mM NaCl, pH 8.0) and incubated with STING for 4 hours with rotation at 4 °C. Beads were washed with buffer containing 20 mM Tris-HCl, 150 mM NaCl and 0.1% Triton X-100, pH 8.0 for 3 times.
We thank Glen Barber, Kate Fitzgerald, Jae Jung, Karen Mossman, Nan Yan and Eain Murphy for important reagents and advice. This work was supported in part by the National Institutes of Health grants CA062220 (GCS), CA068782 (GCS), AI109569 (SB) and S10 OD010381 (SB).
AUTHOR CONTRIBUTIONS:X.W., S.C., S.B. and G.C.S. designed the study. X.W. conducted the majority of experiments. T.M., E.O. and S.B. conducted the T. gondii associated experiments. P.K. conducted CRISPR/CAS9 approaches. Y. Z. conducted the HSV1 mouse experiments. X.W., S.C. S.B. and G.C.S. wrote the manuscript.
GEO accession number for the microarray analysis is GSE78205.
Publisher's Disclaimer: This is a PDF file of an unedited manuscript that has been accepted for publication. As a service to our customers we are providing this early version of the manuscript. The manuscript will undergo copyediting, typesetting, and review of the resulting proof before it is published in its final citable form. Please note that during the production process errors may be discovered which could affect the content, and all legal disclaimers that apply to the journal pertain.