Distinct families of germline-encoded pattern recognition receptors can sense both microbial and endogenous nucleic acids. These DNA and RNA sensors include endosomal TLRs and cytosolic sensors upstream of stimulator of type I IFN genes (STING) and MAVS. The existence of overlapping specificities for both foreign and self nucleic acids suggests that, under optimal conditions, the activity of these receptors is finely tuned to effectively mediate host defense yet constrain pathogenic self-reactivity. This equilibrium becomes disrupted with the loss of either TLR9 or STING. To maintain immune protection, this loss can be counterbalanced by the elevated response of an alternative receptor(s). Unfortunately, this adjustment can lead to an increased risk for the development of systemic autoimmunity, as evidenced by the exacerbated clinical disease manifestations of TLR9-deficient and STING-deficient autoimmune-prone mice. These studies underscore the delicate balance normally maintained by tonic signals that prevent unchecked immune responses to nucleic acids released during infections and cellular duress or death.
Nucleic acids (NAs) are the principal means of information transfer in most organisms. The conveyance of information from DNA (nuclear) to RNA (cytosolic) in eukaryotic cells relies on the precise segregation of NAs into appropriate nuclear, endosomal, and cytosolic compartments. These processes are highly systematized, actively maintained, and closely monitored by intrinsic NA sensors. This strict regulation of endogenous NAs allows abrupt shifts in the quantity and quality of NAs to serve as surrogate indicators of microbial infection that, in turn, initiate host defense mechanisms. However, because these sensors also detect endogenous NAs, inappropriate accumulation of these self-derived molecules can also provoke host responses, in some cases fostering autoimmunity and autoinflammation. Accordingly, the responses elicited by NA sensors must be programmed to optimize host defense, as well as to properly constrain responses to self-NAs. Further, because most microbes can engage multiple NA sensors, regulatory cross-talk likely exists to integrate the aggregate of signals generated by individual sensors. We propose that, under homeostatic conditions, these NA-sensing regulatory networks are finely tuned to the “tonic” receptor engagement levels mediated by endogenous NAs. Accordingly, the loss or inactivation of one sensor system impacts the remaining regulatory network, adjusting the calibration set point and affording heightened sensitivity to exogenous NAs. However, although such compensatory mechanisms may ensure adequate host defense, they may also confer an increased risk for the development of autoimmune responses.
In this article, we briefly review the evidence for NA sensor involvement in autoimmunity and autoinflammation and provide examples of endogenous ligands that are likely to promote these conditions. We also summarize studies that document the connection between loss of the endosomal DNA sensor TLR9, or loss of the cytosolic DNA sensor stimulator of type I IFN genes (STING), and more severe system lupus erythematosus (SLE). Potential molecular mechanisms that might account for these paradoxical observations are discussed.
Endosomal and cytosolic NA sensors contribute to autoimmunity and autoinflammation
The importance of sensing inappropriate NA accumulation emerged with the identification of TLR9 as an endosomal sensor for bacterial DNA (1). Thus TLR9, as well as subsequently described RNA-specific TLRs (TLR3, TLR7, TLR8, and TLR13), clearly plays critical roles in microbial immunity (2). However, autologous DNA and RNA also activate these TLRs, so the aberrant distribution of endogenous NAs can similarly foster immune activity, including the activation of autoreactive B cells, IFN-producing plasmacytoid dendritic cells, neutrophils, and other myeloid-derived APCs (3–5). As a result, endosomal TLRs can play key roles in the initiation and progression of systemic autoimmune diseases. In fact, endosomal TLRs have been implicated in all murine models of spontaneous SLE, because autoimmune-prone mice, deficient in the expression of MyD88, Unc93B1, IRF5, both TLR7 and TLR9, or TLR7 alone, invariably exhibit less severe disease manifestations than do the corresponding gene-sufficient strains (6–14). Moreover, hydroxychloroquine, a drug that blocks endosome acidification and thus TLR activation, is routinely used to treat SLE patients.
The contributions of TLR7 and TLR9 are particularly clear in B cells, where TLR9-deficient autoimmune-prone mice fail to make autoantibodies reactive with dsDNA or nucleosomes, and TLR7-deficient autoimmune-prone mice lack autoantibodies against RNA or RNA-binding autoantigens found in macromolecular complexes, such as splicesomes, nucleosomes, or ribosomes (6, 14). Conversely, elevated expression of TLR7 causes more severe disease in autoimmune-prone strains (15–18), and very high TLR7 copy number yields additional organ-specific autoinflammation (19). TLR8 also was implicated in murine SLE (20), and overexpression of human TLR8 exacerbates joint inflammation in a collagen-induced arthritis model (21). Finally, TLRs were linked to macrophage activation and the ensuing fetal cardiac fibrosis that develops in the offspring of mothers expressing autoantibodies reactive with the RNA-binding protein Ro60 (22), illustrating a role for TLRs in human autoimmune disease. There is also considerable genetic data linking TLRs to SLE. Polymorphisms in IRF5, a transcription factor downstream of both TLR7 and TLR9, were associated with SLE (23). Moreover, individuals whose cells cannot properly degrade extracellular DNA as a result of reduced expression of DNase1 or DNase1L3 are at increased risk for developing SLE (24, 25).
Nuclease deficiencies also implicated cytosolic NA sensors in systemic inflammation. For example, mutations in the cytosolic DNase Trex1, the cytosolic RNase RNaseH2A, or the RNA-editing enzyme ADAR1, are linked to Aicardi-Goutieres syndrome, a debilitating neuroinflammatory condition (26–29), as well as to different forms of SLE (30). In mice, Trex1 deficiency results in a type I IFN–driven systemic inflammation, causing myocarditis (31, 32) and inflammation of skeletal muscle, tongue, skin, and glandular stomach (33). Genetic deletion of murine DNase2a, an endonuclease primarily expressed in phagolysomes or autophagosomes, results in an even stronger IFN response and embryonic lethality, at least in part due to the leakage of DNA from saturated lysosomes into the cytosol (34, 35). However, NA sensors can also activate IFN-independent pathways and, in the absence of the type I IFNR, DNase2a deficiency leads to inflammatory arthritis and SLE-like autoimmunity (36, 37). In patient populations, single-nucleotide polymorphisms in the Dnase2a promoter region that correlate with low DNase2a serum activity are risk factors for rheumatoid arthritis (38). Other defects in lysosome formation, storage, or function are associated with various forms of arthritis (39, 40) and illustrate the need for proper lysosomal NA degradation in the prevention of systemic inflammation.
Numerous cytosolic DNA receptors were identified recently, including cGAS and Ifi204 (41, 42). These sensors, or second messengers derived from these sensors, converge on the ER-associated protein STING to activate downstream pathways leading to the expression of IFN-inducible genes and proinflammatory cytokines (43). Importantly, the systemic inflammation resulting from Trex1 deficiency and the arthritis resulting from DNase2 deficiency depend on STING expression (33, 44). In addition, gain-of-function STING mutations in patient populations were recently linked to SAVI, a clinical syndrome associated with elevated type I IFN, severe vasculopathy, arthritis, pulmonary fibrosis, and, in some cases, SLE-like autoantibody production (45, 46). Thus, NA sensors orchestrate the onset or progression of chronic inflammatory diseases, in many cases driven by autologous NAs. Intriguingly, the same STING mutation, V155M, results in highly variable disease outcomes (45, 46), pointing to critical interactions between STING-dependent pathways and other genetically inherited or environmentally triggered disruptions of immunoregulatory networks.
Endogenous NAs from diverse sources target distinct sensor systems
The origins of autologous NA ligands are diverse and include cell-extrinsic and -intrinsic sources. For instance, the majority of autoantibodies in SLE patients, as well as related systemic autoimmune diseases, such as Sjögren’s syndrome or systemic sclerosis, bind autologous DNA- or RNA-associated protein complexes often found on dying cells or persistent apoptotic blebs (47). This cell-extrinsic apoptotic debris is normally “silent” and rapidly cleared through noninflammatory mechanisms (48, 49). However, when cell debris is not properly removed, it can be endocytosed by autoreactive BCRs and delivered to TLR-containing compartments, leading to activation, autoantibody production, immune complex formation, and amplification of the response by FcγR+ APCs (50). Moreover, failure to clear apoptotic cells may lead to secondary necrosis or other immunogenic forms of cell death (48, 49). The premise that the excessive accumulation of extracellular cell debris is the source of the endogenous TLR ligands in SLE was explored experimentally by the development of mice that express a bovine RNase transgene. Autoimmune-prone RNase-transgenic mice are protected from TLR7-driven disease (19).
Cell-intrinsic sources of autologous NAs include transcribed and reverse-transcribed retro-elements (32), damaged genomic DNA (51), and oxidized mitochondrial DNA (52, 53). These accumulate in the cytosol, independently of receptor-mediated internalization. Endogenous retroelements (EREs) form ∼40% of the mammalian genome and have a long evolutionary history with host cells (54). Several steps in the ERE lifecycle that involve active transcription of their genomes into RNA and then reverse transcription to cDNA occur in the cytosol (54). It is at this interface of ERE replication and innate sensing where EREs are a potent source of endogenous NA damage-associated molecular patterns (32). Cytosolic nucleases, such as Trex1, RNaseH2, and a deoxynucleoside triphosphohydrolase SAMHD1, limit the exposure of NA sensors to ERE load (32). MyD88/TLR7-dependent B cell–dependent Ab responses are also required to curb the reactivation and emergence of infectious endogenous retroviruses (55). However, reactivated EREs share a complex relationship with host cells, because BCR-mediated induction of ERE transcription appears to provide a second signal to cytosolic NA sensors and, thereby, facilitate B cell activation by type II T-independent Ags (56).
Dysregulated catabolism of cellular components, such as lipids, proteins, and self-NAs resulting from lysosomal storage disorders, is another source of endogenous ligands (57). Self-NAs can be sensed in the lysosomal compartment by the endosomal TLRs, and the ensuing loss of lysosomal integrity due to excessive swelling or frustrated/repeated fusion events can further release their components into the cytosol, where cytosolic NA sensors can be engaged (37, 57, 58). Mitochondria are an additional major source of endogenous NA (59). Normal turnover of stressed and damaged mitochondria via autophagy results in the access of mitochondrial-derived NA (i.e., mitochondrial DNA) to endolysosomal TLRs with some regularity, and this can cause inflammation when clearance is perturbed (59). Moreover, mitochondrial instability during cellular stress or cell death can release mitochondrial DNA into cytosolic compartments and activate cytosolic NA-sensing pathways (52, 53, 60). These examples reflect the variety and overlapping sources of DNA and RNA that gain access to both endosomal and cytosolic receptors during chronic infection, autoimmunity, and autoinflammation, raising the possibility that simultaneous activation of multiple pathways may lead to persistent or fatal inflammation.
Negative regulatory function of TLR9
In the presence of excess ligand, NA sensors play a critical role in immune activation. However, NA sensors can also negatively regulate immune responses. For example, despite the inability to make anti-dsDNA autoantibodies, all strains of TLR9-deficient autoimmune mice produce elevated Ab titers against RNA and RNA-associated autoantigens and develop accelerated and more severe clinical disease (6, 14, 61–63). Exactly how TLR9 deficiency promotes disease remains unresolved, but, in general, Tlr9−/− mice appear to be hyperresponsive to TLR7 ligands. In vitro analyses of transduced cell lines and myeloid lineage cells suggest that endosomal sensors compete with one another for association with Unc93B1 (64) and imply that, in the absence of TLR9, more Unc93B1 is available for the RNA sensors, leading to their increased activity. TLR9 and TLR7 differentially traffic to endosomal compartments (64, 65) where they could potentially interact with distinct signaling complex components and trigger TLR-specific functions. However, remarkably little is known about distinguishing components of the TLR9 and TLR7 signaling cascades, how they interface with one another, or why a TLR7 signal would be more pathogenic than a TLR9-elicited response, especially in myeloid cells.
The distinct functional outcomes of TLR9 versus TLR7 activation have been rigorously explored in B cells. Mixed bone marrow chimera studies suggest that TLR9 deficiency in B lineage cells alone is sufficient to drive exacerbated autoimmunity (14). A variety of mechanisms may be involved. These include a unique requirement for TLR9 in the production of autoantibodies involved in the clearance of apoptotic debris (66) or the increased pathogenicity of autoantibodies specific for RNA or RNA-associated autoantigens. Alternatively, BCR/TLR7-stimulated B cells may have a greater capacity to activate autoreactive T cells (14) and/or to differentiate toward the plasma cell lineage (67). It is also possible that TLR9 preferentially contributes to the depletion of autoreactive B cells from the developing bone marrow repertoire (68).
The most direct comparisons between BCR/TLR9 and BCR/TLR7-dependent activation were carried out with B cells derived from autoreactive BCR-transgenic mice. Importantly, BCR/TLR9 coengagement appears to limit the survival of mature autoreactive B cells and, thus, preclude sustained Ab secretion, germinal center formation, and affinity maturation (67, 69). This strategy may allow for the production of low-affinity IgM autoantibodies that facilitate immune clearance of endogenous NA cell debris without the risk for a sustained and focused self-reactive IgG autoantibody production. It is tempting to speculate that either cell-intrinsic or -extrinsic factors capable of compromising this checkpoint might underlie some autoimmune etiology. In the context of ongoing inflammation, these inherently short-lived responses might be sustained and redirected through the receipt of additional survival or differentiation signals that could extend the duration of Ab secretion or afford GC initiation, affinity maturation, and establishment of long-lived plasma cell pools, essentially converting the negative regulatory function of TLR9 to disease-promoting activity.
Negative regulatory function of the cytosolic DNA sensor STING
Recent studies also revealed an unanticipated negative regulatory role for cytosolic DNA sensors. As described above, STING clearly plays a fundamental role in driving type I IFN production when triggered by excess ligand, and genetic variants are associated with human SLE. Nevertheless, STING−/− lpr/lpr mice develop more severe SLE than their STING-sufficient counterparts (70). Unexpectedly, IRF3 (the type I IFN inducing transcription factor downstream of STING) is not required for this apparent STING-mediated immune suppression, because IRF3+ and IRF3−/− lpr/lpr mice developed comparable levels of disease (70). In line with the STING−/− lpr/lpr data, C57BL/6 STING−/− mice injected i.p. with the proinflammatory mineral oil pristane developed a more severe TLR-dependent inflammatory response than littermate STING-sufficient mice (70). This remarkable parallel between TLR9 and STING is summarized in Fig. 1.
This STING-mediated suppression may reflect the capacity of STING to control the phosphorylation of SHP1/2 and downregulate JAK1/STAT1 signaling (71). STING also was reported to directly activate STAT6 (72), and STAT6 can promote the activation of M2 (anti-inflammatory, wound healing) macrophages (73) that could ameliorate disease pathology. STING deficiency was further associated with a reduction in the number of regulatory T cells (Tregs) in STING−/− lpr/lpr secondary lymphoid organs. Apoptotic debris induces the production of IDO, an enzyme that generates tryptophan derivatives that, in turn, promote Treg differentiation (74); therefore, reduced IDO levels correspond to a loss of Treg-mediated tolerance and increased autoimmunity. Remarkably, very little IDO could be found by immunostaining of STING−/− lpr/lpr spleens compared with age-matched lpr/lpr spleens, and loss of IDO correlated with decreased Treg numbers (70). The identification of the signaling molecules and transcription factors downstream of STING that are major players in these suppressive functions remains to be determined. Potential signaling intermediates include NIK and p52, because these can suppress type I IFN responses (75). Alternatively, other IRFs, such as IRF1 or IRF5, may play a cell-specific role in modulating cytokine profiles downstream of STING, similar to their roles in RIG-I–like receptor pathways (76). It will be important to explore the impact of STING deficiency in additional models of systemic autoimmunity and determine whether loss of its homeostatic function results in similar increases in disease severity.
How are NA sensors calibrated?
Endosomal and cytosolic sensors serve overlapping functions in protection against infectious agents. For example, both TLR9- and STING-dependent pathways are activated by malarial parasites (77–79), and multiple endosomal TLRs play a role in murine viral immunity (80, 81). Moreover, TLR9 synergizes with TLR2 in protection against HSV-1 and other viral infections. Analogously, cGAS plays a major role in activating innate immune responses important for protection against RNA and DNA viral infections (82). These interdigitating pathways suggest that such synergy is advantageous, and perhaps necessary, for an appropriate sterilizing immune response in some of these cases. Therefore, if one arm of the innate immune system is compromised, then one might anticipate a need to bolster an alternative innate immune pathway. This adjustment requires an intrinsic calibration mechanism to establish thresholds prior to frank infection. Inasmuch as these receptors also sense endogenous NA ligands, an attractive possibility is that cells titer their capacity for NA reactivity against homeostatic levels of endogenous NAs.
This premise is supported by the documented heightened response of STING−/− myeloid cells to TLR ligands compared with STING-sufficient controls. This hyperresponsiveness corresponded to a reduction in the basal expression level of a number of negative regulators of TLR signaling (e.g., A20, Nlrc3, SOCS1, and SOCS3) (70) (S. Sharma, unpublished observations). Moreover, when STING was overexpressed in the RAW264.7 macrophage line, the cells were hyporesponsive to TLR ligands and expression of negative regulators was increased compared with vector control cells (70). Thus, under homeostatic conditions, the normal turnover of NAs under noninflamed conditions appears to provide a low-level tonic signal to STING or STING-dependent sensors and, thereby, calibrate TLR-dependent responses.
In the context of host defense, this retuning of the regulatory network may enhance TLR-mediated microbial immunity in a setting where the cytosolic DNA sensors are inactive. However, the same adjustments appear to disrupt the balance that limits self-reactivity because STING−/− SLE-prone mice develop more severe disease. It is important to note that loss of one pathway is often sufficient to cause increased susceptibility to pathogens still detected by alternate pattern recognition receptor pathways, although it is not always clear whether increased pathogenicity comes from increased microbial burden or unchecked activation of the complementary pathways.
Whether TLR9 deficiency similarly impacts regulatory networks remains to be determined. However, increased TLR7 activity due to the greater availability of Unc93B1 in the absence of TLR9 essentially leads to the same outcome: greater TLR7-mediated host defense but more severe autoimmunity. Future studies need to address whether distinct sources of endogenous ligands mediate negative versus positive regulatory effects, as well as how ongoing inflammatory responses intersect these pathways. Most importantly, it will be important to understand how these pathways promote disease, as well as how they can be downregulated to most effectively manipulate these pathways therapeutically.
Both endosomal and cytosolic NA sensors detect autologous ligands, and the excessive accumulation of endogenous NAs can promote fatal inflammation. Nonetheless, the normal turnover of endogenous NAs and their capacity to modestly engage NA sensors, even under homeostatic conditions, likely play key roles in adjusting the balance between innate immune components. For example, in the absence of TLR9 or STING, responses initiated by the remaining innate sensors are tuned up, presumably as a means to better cope with potential microbial challenge. Unfortunately, such an adjustment comes with an increased risk for poorly controlled autoimmune responses. Whether tonic signaling of NA sensors has a similar impact in human populations remains unresolved. However, the data from mouse models highlight the need for caution in the design and application of STING and TLR inhibitors for the treatment of systemic autoimmunity and/or autoinflammation, because there is the potential to perturb an equilibrium that facilitates appropriate protective immunity but guards against autoimmune pathology. A better understanding of the integrated network governing NA-sensing pathways should reveal points amenable to intervention in autoimmunity or autoinflammation.
We thank Dr. M. Atianand for advice and helpful suggestions.
This work was supported by Alliance for Lupus Research Grant 260146 (to S.S.), National Institutes of Health Grant AR050256 (to A.M.-R.), and Congressionally Directed Medical Research Programs Grant PR130769 (to M.P.C.).
Abbreviations used in this article:
systemic lupus erythematosus
stimulator of type I IFN genes
regulatory T cell.
The authors have no financial conflicts of interest.