The host innate immune system provides a critical first line of defense against invading microorganisms including pathogenic bacteria. Utilizing germ-line encoded pattern recognition receptors (PRRs), mammalian cells detect a wide variety of highly invariant molecular structures (known as PAMPs, for pathogen associated molecular patterns) that are often required by microbes for their survival or pathogenicity1-3. Indeed, many bacterial pathogens employ cyclic-diguanosine monophosphate (c-di-GMP) or cyclic-diadenosine monophosphate (c-di-AMP), two key secondary messengers that play essential roles in regulating metabolism, motility and virulence 4-6. During infection with certain bacterial species, these bacterially derived secondary messengers can also act as PAMPs, triggering a host type I interferon (IFN) innate immune response, characterized by the activation of NF-κB and IRF3 transcription factors7, 8.
The mechanism by which c-di-GMP and c-di-AMP activate the host type I IFN response remains poorly understood. Indeed, intracellular detection of c-di-GMP or c-di-AMP leads to the activation of the type I IFN response in a manner independent of the cytoplasmic PRR RIG-I (also known as DDX58) or its downstream adaptor IPS-1. Moreover, c-di-GMP or c-di-AMP require neither MyD88 nor TRIF adaptors, suggesting that the Toll-like receptor (TLR) family of PRRs also do not play a role in the detection of c-di-GMP or c-di-AMP9. The activation of type I IFNs by c-di-GMP and c-di-AMP was however, shown to require STING (aka MITA, MPYS, ERIS or TMEM173), suggesting that these cyclic dinucleotides are detected via a PRR that signals via STING10, 11. In response to certain viral nucleic acids and B-DNA, STING functions as an adaptor protein and has been demonstrated to facilitate downstream signal transmission to IRF3 and NF-κB12-14. Here, we provide evidence that the recently identified PRR DDX41 is the primary sensor that directly binds to c-di-GMP or c-di-AMP and can trigger the type I IFN host immune response via STING.
RESULTS
DDX41 mediates c-dinucleotide sensing in murine DCs and human monocytes
We stably knocked down DDX41 via short hairpin RNA (shRNA) in the murine splenic dendritic cell line D2SC (Fig. 1a) and examined IFN-β induction in response to c-di-GMP and c-di-AMP. Control cells infected with L. monocytogenes displayed a robust IFN-β response, whereas DDX41-shRNA cells showed a marked reduction in the IFN-β response (Fig. 1b). Consistent with published data10, 11, STING-shRNA cells also showed impaired IFN-β induction in response to L. monocytogenes (Fig. 1b). Cytoplasmic delivery of either c-di-GMP or c-di-AMP via lipofection also yielded strong IFN-β activation which was largely diminished in DDX41-shRNA cells, which paralleled the impairment displayed in STING-shRNA cells (Fig. 1c,d). Induction of IFN-β by the synthetic DNA poly (dA:dT) but not by the RNA ligand poly (I:C) was impaired in DDX41-shRNA cells (Supplementary Fig. 1a,b), demonstrating the specificity of DDX41 for c-di-GMP, c-di-AMP and B-DNA. Type I IFNs mediate the innate immune response via the IFN-α/β receptor, where receptor ligation leads toward the activation of hundreds of interferon stimulated genes (ISGs)15. Both c-di-GMP and c-di-AMP activated the ISG Mx1 in control cells, however, induction of Mx1 was significantly reduced in DDX41-shRNA or STING-shRNA cells (Fig. 1e,f). Similarly, c-di-GMP-mediated activation of other ISGs was also impaired in DDX41-shRNA cells (Supplementary Fig. 1c-f). These results indicate that DDX41 plays a critical role in c-dinucleotide-mediated activation of type I IFN and IFN-mediated signaling. Cytosolic detection of bacterial secondary messengers also leads to the activation of NF-κB, a key transcription factor important for the induction of pro-inflammatory cytokines such as interleukin 6 (IL-6) and TNF16. We found that IL-6 and TNF expression levels were significantly reduced in DDX41-shRNA cells compared to control-shRNA cells that were treated with c-di-GMP and/or c-di-AMP (Fig. 1g,h and Supplementary Fig. 1g). To determine whether DDX41 mediates c-di-GMP/c-di-AMP sensing and type I IFN activation in human cells, we stably knocked down DDX41 via shRNA in the human monocyte cell line THP-1 (Fig. 1i). Consistent with our findings in murine cells, both c-di-GMP and c-di-AMP induced the production of IFN-β in control–shRNA cells, whereas c-di-GMP or c-di-AMP-mediated production of IFN-β was significantly defective in DDX41- or STING-shRNA THP-1 cells (Fig. 1j,k). Induction of ISGs in response to c-di-GMP was also impaired in DDX41-shRNA THP-1 cells (Supplementary Fig. 1h-l). In addition, the DDX41-STING signaling axis was required for the induction of IFN-β in THP-1 cells stimulated with B-DNA (Supplementary Fig. 1m). These results suggest that DDX41 functions in regulating type I IFN and pro-inflammatory gene inductions in response to c-dinucleotides in both murine and human cells.
DDX41-mediates c-dinucleotide sensing in primary cells
The role of DDX41 in facilitating c-di-GMP or c-di-AMP induced activation of type I IFN in primary cells was also examined. BMDCs (mDCs derived from bone marrow with granulocyte-macrophage colony-stimulating factor (GM-CSF)) were prepared and transfected with shRNA targeting DDX41 (Fig. 2a). Control-shRNA BMDCs infected with L. monocytogenes displayed robust production of IFN-β, while IFN-β induction in DDX41-shRNA BMDCs was markedly reduced (Fig. 2b). c-di-GMP and c-di-AMP stimulation also induced the production of IFN-β in control-shRNA BMDCs, while IFN-β production was highly impaired in DDX41-shRNA BMDCs (Fig. 2c,d). Similarly, primary thioglycollate-elicited mouse peritoneal macrophages transfected with siRNA targeting DDX41 (Fig. 2e) displayed reduced activation of IFN-β in response to L. monocytogenes infection or c-di-GMP stimulation compared to control-siRNA transfected cells (Fig. 2f,g). Consistently, induction of Mx1 and IL-6 were also impaired in DDX41-siRNA peritoneal macrophages infected with L. monocytogenes or treated with c-di-GMP (Supplementary Fig. 2a-d).
To determine whether DDX41 plays a role in sensing c-di-GMP in primary human cells, we transfected peripheral blood mononuclear cells (PBMCs) with an siRNA specific for DDX41 (Fig. 2h). While control-siRNA PBMCs infected with L. monocytogenes or treated with c-di-GMP elicited the activation of IFN-β, DDX41-siRNA PBMCs showed defective IFN-β activation in response to L. monocytogenes or c-di-GMP (Fig. 2i,j). Similar reductions in IFN-β activation were displayed in DDX41-siRNA PBMCs obtained from two additional human donors (Supplementary Fig. 2e,f). These results further indicate a critical role for DDX41 in facilitating type I IFN responses induced by c-di-GMP or c-di-AMP in primary immune cells.
DDX41 is a direct sensor of c-di-GMP
DDX41 is known not only to signal via the STING adaptor to activate type I IFN, but to also function as a PRR that directly binds viral DNA and B-DNA17. We therefore investigated whether DDX41 functioned as a direct sensor (PRR) for c-di-GMP or c-di-AMP. Biotin-labeled c-di-GMP or c-di-AMP pulled down ectopically expressed DDX41 from 293T cell lysates (Fig. 3a and Supplementary Fig. 3a). Immunofluorescence microscopy further revealed that DDX41 and c-di-GMP co-localized upon their co-transfection into D2SC cells (Fig. 3b). In addition, we found that GST-purified DDX41 or His-purified DDX41 from E.coli directly bound c-di-GMP (Fig. 3c and Supplementary Fig. 3b). The c-di-GMP interaction with DDX41 was specific as only unlabeled c-di-GMP, c-di-AMP and poly (dA:dT) (B-DNA), but not poly (I:C), could competitively disrupt the c-di-GMP-DDX41 interaction (Fig. 3d). Furthermore, the structurally similar molecules GMP and AMP, but not the c-di-GMP and c-di-AMP precursors, GTP or ATP, also disrupted the DDX41-c-di-GMP complex (Supplementary Fig. 3c).
To determine whether c-di-GMP binding was specific to the DDX41 helicase, we compared c-di-GMP binding with another DEAD box helicase PRR. c-di-GMP bound to DDX41, but showed no interaction with DDX58 (Supplementary Fig. 3d). To define which domain of DDX41 was important for binding c-di-GMP or c-di-AMP, we used a series of deletion mutants of DDX41 (Fig. 4a,b) and evaluated their interactions with biotinylated c-di-GMP or c-di-AMP in 293T cells. c-di-GMP and c-di-AMP failed to interact with DDX41 lacking the central DEAD box domain (Fig. 4c,d). To determine the physiological relevance of this domain in terms of type I IFN induction, we reconstituted DDX41-shRNA THP-1 cells with either full length DDX41 or the DDX41 deletion mutant lacking the DEAD box domain (Fig. 4e). As expected, DDX41-shRNA cells displayed defective IFN-β production in response to c-di-GMP or c-di-AMP compared to control-shRNA cells (Fig. 4f,g). However, DDX41-shRNA cells reconstituted with full length DDX41, but not the DDX41 deletion mutant lacking the DEAD box domain, “rescued” the defective IFN-β production in DDX41-shRNA cells in response to c-di-GMP or c-di-AMP stimulation (Fig. 4f,g). Thus c-dinucleotide mediated induction of IFN-β requires the central DEAD box domain of DDX41.
DDX41-dependent signaling downstream of c-di-GMP and c-di-AMP
DDX41 interacts and co-localizes with the STING adaptor (Supplementary Fig. 4a,b) to facilitate DNA ligand dependent signal transduction17. Introduction of either c-di-GMP or c-di-AMP into D2SC cells led to enhanced DDX41-STING complex formation (Fig. 5a). In DNA dependent signaling pathways, STING further binds to the downstream kinase TBK1 to activate the type I IFN response18, 19. Both c-di-GMP and c-di-AMP when transfected into control D2SC cells activated the formation of a STING-TBK1 complex, however, c-di-GMP and c-di-AMP mediated activation of the STING-TBK1 complex was almost completely abrogated in DDX41-shRNA cells (Fig. 5b). Consequently, c-di-GMP and c-di-AMP mediated activation of TBK1, IRF3 and the downstream type I IFN effector, STAT1 was impaired in DDX41-shRNA cells (Fig. 5c and Supplementary Fig. 5). Activation of NF-κB was also impaired in DDX41-shRNA cells in response to either c-di-GMP or c-di-AMP (Fig. 5c). Together, our findings suggest that DDX41 is a critical PRR for c-di-GMP and c-di-AMP mediated IFN induction, and that its absence generates a defect in downstream STING-dependent signaling.
c-di-GMP signals to the STING adaptor via DDX41
A recent study found that c-di-GMP could bind to the C-terminal domain of STING (STING CTD) and suggested that the STING adaptor could function as an immune sensor of c-di-GMP20. We therefore performed binding assays to determine the affinities of c-di-GMP or c-di-AMP for DDX41 and for STING in parallel. Biotin-labeled c-di-GMP pulled down ectopically expressed DDX41 with greater affinity over ectopically expressed STING from 293T cell lysates (Fig. 6a). Physiologically, binding of c-di-GMP with endogenous DDX41 was also greater than the association between c-di-GMP and endogenous STING (Fig. 6b). Immunofluorescence imaging further revealed greater co-localization between c-di-GMP and DDX41 in comparison to STING (34.13% vs 6.25%) (Fig. 6c). Affinity capillary electrophoresis (ACE) experiments were also performed to examine the binding affinities between c-di-GMP and recombinant DDX41 or recombinant STING CTD. c-di-GMP bound DDX41 with a Kd of ~5.65 μM, whereas c-di-GMP associated with STING CTD with a Kd of ~14.54 μM (Fig. 6d). Consistent with these findings, c-di-GMP bound to purified recombinant DDX41 with stronger affinity than purified recombinant STING CTD in pulldown binding assays (Fig. 6e). We therefore hypothesized that DDX41 is the major primary sensor of c-di-GMP and c-di-AMP, operating upstream of STING, and upon binding these PAMPs, yields enhanced complex formation with STING to facilitate downstream signaling and type I IFN activation (Supplementary Fig. 6). Accordingly, c-di-GMP was significantly impaired in its capacity to associate with ectopically expressed STING in 293T cells transfected with siRNA targeting DDX41 (Fig. 7a). Consistently, co-localization between c-di-GMP and STING was largely reduced in DDX41-shRNA cells, whereas c-di-GMP-DDX41 interactions remained intact in STING-shRNA cells (Fig. 7b,c). Taken together, our findings indicate that DDX41 is the primary PRR for c-di-GMP and c-di-AMP, which signals via STING for type I IFN induction.
DISCUSSION
Many bacterial pathogens including Staphylococcus, Streptococcus, Pseudomonas, Yersiniae, Listeria and Mycobacterial species employ key secondary messengers including c-di-GMP or c-di-AMP that play essential modulatory roles in bacteria4, 7. Although several substrates and effectors have been identified for these cyclic dinucleotide monophosphate species within the bacterial cell, our understanding of these bacteria-specific secondary messengers on the innate immune response within the mammalian host cell is just beginning to emerge. c-di-GMP and c-di-AMP activate the host type I IFN response in a manner dependent on the STING adaptor10, 11 and our findings indicate that the DNA sensor helicase DDX41 functions as a direct PRR for these cyclic dinucleotides in both murine and human cells. Our results showed that unlabeled c-di-GMP or c-di-AMP could disrupt the DDX41-c-di-GMP interaction. We found that GMP and AMP, but not the bulkier GTP or ATP molecules, could also competitively disrupt the DDX41-c-di-GMP complex. Although two molecules of GMP or AMP are structurally similar to c-di-GMP or c-di-AMP, respectively, they are not known to function as PAMPs in mammalian cells9. It will therefore be of further interest to determine how these species play a modulatory role in the type I IFN response. Our competition experiments additionally revealed that B-DNA could disrupt the DDX41-c-di-GMP complex. Indeed DDX41 has been shown to function as a sensor for B-DNA as well17. The mechanism by which DDX41 binds B-DNA, as well as cyclic dinucleotides, as revealed by co-crystallization studies and point mutation analysis will be the subject of future investigation.
Our results additionally indicated that c-di-GMP and c-di-AMP mediated activation of innate signaling and type I IFN induction were similarly defective between cells in which DDX41 or STING had been knocked down , suggesting DDX41 and STING share a common signaling pathway. STING-deficient cells displayed a very modest defect in NF-κB activation in response to c-di-GMP or c-di-AMP. The reason for this phenomenon is not entirely clear10, however it may be possible that there is redundancy or compensation in signaling to NF-κB. Another DNA sensor, IFI16 (also known as p204) was also shown to facilitate some viral DNA triggered signaling via the STING adaptor21, 22. Although a role for IFI16 cannot be ruled out in the c-di-GMP and c-di-AMP signaling pathway, it is however unlikely that IFI16 functions as a primary sensor for these molecules since its’ basal expression is low and is rather induced in a type I IFN-dependent manner. DDX41 expression on the other hand, is greater at the basal state and is not modulated by type I IFNs17.
Our data suggests that DDX41 serves as the PRR for c-di-GMP and c-di-AMP, which upon receptor binding signals to TBK1-IRF3 via the STING adaptor. Lending further credibility as a scaffolding molecule, STING was recently shown to bridge TBK1-IRF3 interactions for optimal signaling23. Nevertheless, consistent with a published report20, we also found that c-di-GMP associated with STING, however, with lower affinity than DDX41. Although the physiological relevance of this interaction requires further investigation, our data shows that c-di-GMP interaction with STING is significantly enhanced in the presence of DDX41 in cells. The solved structure of the C-terminal domain of STING in complex with c-di-GMP revealed that one molecule of c-di-GMP binds one dimer of STING24-28. We propose that c-di-GMP detection and binding to DDX41 promotes enhanced DDX41-STING interactions leading to an increase in binding affinity of STING toward c-di-GMP, ultimately driving downstream signaling events. Thus, STING may function as a secondary receptor or co-factor in the cyclic dinucleotide signaling pathway.
The significance of type I IFN induction in the context of anti-bacterial innate immunity is currently unclear and somewhat controversial, particularly due to conflicting reports on whether type I IFN either functions to support or inhibit bacterial growth29-32. It will therefore be of interest to further study how different bacteria and host cells use secondary messengers and DDX41 as virulence factors and innate immune receptors, respectively, in their battle of infection and immunity33-35. As such, cyclic dinucleotide species and DDX41 represent new targets such that modulation of their interaction during certain bacterial infections can alter the host immune response in a manner to suppress bacterial replication and spread.
METHODS
Methods and any associated references are available in the online version of the paper at http://www.nature.com/natureimmunology/.