Cyclic Dinucleotides at the Forefront of Innate Immunity

Abstract

Cyclic dinucleotides (CDNs) have emerged as ubiquitous signaling molecules in all domains of life. In eukaryotes, CDN signaling systems are evolutionarily ancient and have developed to sense and respond to pathogen infection. On the other hand, dysregulation of these pathways has been implicated in the pathogenesis of autoimmune diseases. Thus, cyclic dinucleotides have garnered major interest over recent years for their ability to elicit potent immune responses in the eukaryotic host. Similarly, ancestral CDN-based signaling systems also appear to confer immunological protection against infection in prokaryotes. Therefore, a better understanding of the host processes regulated by CDNs will be of tremendous value in many areas of research. Here, we aim to review the latest discoveries and recent trends in cyclic dinucleotide research with a particular focus on the molecular mechanisms by which these small molecules mediate innate immunity.

Introduction

Immunity is the physiological process whereby organisms detect and defend themselves against disease by infectious agents. Within this working definition, the immune system has classically been divided into two branches: the innate arm and the adaptive arm. In contrast to adaptive or acquired immune defenses, which are often delayed and involve genetic rearrangement events, innate immunity is considered the first and immediate line of defense against invading pathogens [1]. This is accomplished by a suite of invariant, germline encoded receptors that recognize general, conserved, molecular features of infectious agents that are rarely or never found in the healthy host and, hence, does not confer any lasting, immunological memory [1]. Upon sensing the infectious insult, receptors of the immune system must then relay signals to effector proteins, culminating in pathogen elimination. In the context of innate immunity, this is often mediated by the employment of second messengers—small molecules that orchestrate key shifts in cellular function in response to altered environmental conditions. An emerging paradigm in prokaryotic and eukaryotic innate immune systems is the use of cyclic dinucleotide (CDN) second messengers [2–6]. As their name implies, CDNs are heterocyclic compounds synthesized by cyclases or nucleotidyltransferases (NTases) through the cyclization of two ribonucleoside triphosphate moieties via two ‘canonical’ (3’,5’) and/or ‘non-canonical’ (2’,5’) phosphodiester linkages. CDN signaling is resolved by the tightly regulated action of phosphodiesterase enzymes (PDEs), which catalyze the metal dependent hydrolysis of one or both phosphodiester bonds, yielding a linear (pNpN) dinucleotide intermediate or two nucleoside monophosphates, respectively.

While a number of CDN-based signaling systems have been identified to date (reviewed in detail by Krasteva and Sondermann [7]), innate immune sensing of pathogen infection is mediated by a conserved family of NTases including the dinucleotide cyclase in Vibrio cholerae (DncV) and its metazoan homologue cyclic GMP-AMP Synthase (cGAS) [4, 6, 8, 9]. This family of proteins was hence given the name cGAS/DncV-like nucleotidyltransferases (CD-NTases) [9]. Upon activation, CD-NTases catalyze the synthesis of CDN second messengers, including cyclic GMP-AMP (cGAMP), which function as immune ‘alarmones’ produced in the context of pathogen infection. Notably, prokaryotic 3’3’-CDNs (c[N(3′,5′)pN(3′,5′)p]) contain two ‘canonical’ (3’,5’) phosphodiester linkages; however, as this signaling system emerged in eukaryotes, it eventually transitioned to the use of 2’3’-cGAMP (c[G(2′,5′)pA(3′,5′)p]) containing mixed phosphodiester bonds in higher metazoans including humans [7, 10–12]. While the biochemical underpinnings for this transition still remain unclear, it is intriguing to hypothesize that this may afford some degree of immune discrimination between exogenous and endogenous sources of CDNs in higher eukaryotes. Despite this dissimilarity, cGAMP-based signaling systems constitute an ancient immune signaling pathway with evolutionary origins in prokaryotes [6]. This review will highlight the most recent discoveries and emerging themes in CDN-based innate immune signaling with an emphasis on CDN effectors, the regulation of CDN levels, and the strategies used by viruses to subvert this important signaling axis.

A Broad Family of CD-NTases in Bacteria Synthesizes Diverse Nucleotide Second Messengers

cGAS and DncV are the founding members of a large class of nucleotidyltransferases in both prokaryotes and eukaryotes. Whereas the evolutionary origins of cGAS in metazoans can be traced as far back as the sea anemone, Nemostella vectensis, CD-NTases have recently been found in more than ten percent of bacterial genomes spanning nearly every bacterial phylum including human commensal and pathogenic microorganisms [6, 9, 13, 14]. Although these enzymes are divergent at the amino acid level, they display common structural features including a DNA polymerase β-like nucleotidyltransferase fold [9]. Unlike other CDN cyclase families, CD-NTases display extreme diversity in their capacity to synthesize cyclic dinucleotides; moreover, cyclic trinucleotide products have also been observed including cyclic AMP-AMP-GMP produced by a CD-NTase from Enterobacter cloacae [9]. In addition to utilizing purines, bacterial CD-NTases can synthesize a number of pyrimidine containing 3’3’-CDNs, including: c-di-UMP and cyclic UMP-CMP as well as pyrimidine-purine hybrids (cyclic UMP-AMP and cyclic UMP-GMP) [9]. Although synthetic pyrimidine-based CDNs have been described, pyrimidine containing nucleotide second messengers have not been observed anywhere else in nature [15, 16]. Remarkably, the substrate specificity of these enzymes can be reprogrammed by single amino acid substitutions within the active site lid [9]. Such plasticity in CD-NTases may have conferred a fitness advantage, for example in the context of evolutionary arms races between bacteria and the phages that infect them.

CD-NTases and Innate Anti-Phage Immunity

In bacteria, CD-NTases are found within the genomic context of a cyclic oligonucleotide-based anti-phage signaling system (CBASS) comprised of a core CD-NTase-effector unit and two accessory proteins with domains commonly found in eukaryotic ubiquitination machinery including an E1/E2-containing enzyme and a JAB de-ubiquitinase [6]. CBASS systems confer marked resistance against a wide array of DNA-phages in a manner dependent on both CD-NTase and effector enzyme activity [6]. Interestingly, the accessory proteins appear to be conditionally essential during infection with select phages, raising the intriguing possibility that these proteins may counteract viral immune evasion strategies [6]. Similar to metazoans, CD-NTase enzymatic activity appears to be triggered by bacteriophage infection as 3’3’-cGAMP is undetectable in uninfected cells [4, 6]. The ligand and/or activation mechanism for bacterial CD-NTases remains to be fully elucidated, but, because bacteria do not compartmentalize their DNA, it is unlikely to be viral nucleic acids as is the case for metazoan CD-NTases [4]. Intriguingly, some CD-NTases appear to become activated following recognition of specific, bacteriophage peptide motifs by Hop1, Rev7, and Mad2 (HORMA) domain containing receptors [17]. Although several, non-redundant 3’3’-cGAMP-selective PDEs have been identified in V. cholerae, the roles of these enzymes, if any, in CBASS signaling are unclear [18]. One possibility is that PDEs may function as a rheostat to control CBASS activation; alternatively, because canonical CBASS signaling ultimately results in cell death, as discussed below, resolution of CDN signaling may not be necessary.

A number of CBASS effector proteins have been identified, but the best characterized effector to date is a CDN-activated patatin-like phospholipase originally identified in V. cholerae and hence given the name CapV (Figure 1) [6, 19]. Upon activation by 3’3’-CDNs, CapV-like phospholipases hydrolyze phospholipids in the bacterial inner cell membrane resulting in loss of membrane integrity and cell lysis [6, 19]. Thus, the CD-NTase-phospholipase signaling axis constitutes a ‘cellular suicide’ program intended to protect the bacterial community through abortive phage infection [6]. In lieu of a phospholipase, a number of alternative effectors have been identified in CBASS systems including HNH-type endonucleases, a membrane spanning protein, a domain of unknown function, as well as a Toll Interleukin Receptor (TIR) domain, which also appear to mediate innate anti-phage immunity [6, 17, 20]. Interestingly, in a few cases, the CBASS systems contain a hybrid effector containing an N-terminal TIR domain fused to a C-terminal domain resembling the metazoan cGAMP effector Stimulator of Interferon Genes or STING (see below) [6, 21–23]. Although it is unclear if and how the TIR-STING effector mediates anti-phage immunity, it is possible that metazoan cGAS-cGAMP-STING signaling systems originate from these ancestral bacterial CBASS systems and have retained their role in innate antiviral defense along the way.

Figure 1: Overview of CBASS Signaling.

Following bacteriophage infection, activation of CD-NTases results in the production of diverse cyclic oligonucleotide species (cNNs or cNNNs, where N is any NMP) including 3’3’-CDNs. Binding of 3’3’-CDNs to a latent phospholipase results in its activation and the subsequent hydrolysis of the bacterial inner cell membrane. Dissolution of the inner membrane results in bacteriolysis, abortive infection, and ultimately protection of the larger bacterial community. Alternative CBASS effectors include endonucleases, transmembrane effectors, Toll Interleukin Receptor (TIR)-like effectors, domains of unknown function (DUFs), among others, although these are less well characterized.

cGAS-STING-Mediated Innate Immune Responses

As CD-NTase-based innate immune systems made their way into eukaryotes, they increasingly became specialized to detect foreign nucleic acid species as is the case for cGAS, a double-stranded (ds)DNA sensor, and its homologue 2’−5’ oligo-adenylate synthase (OAS), a dsRNA sensor [4, 9]. Upon engaging foreign or mislocalized self dsDNA species, cGAS oligomerizes and forms phase-separated droplet-like structures that function as ‘microreactors’ for the production of 2’3’-cGAMP [24–27]. This ultimately activates the ER-resident, transmembrane effector STING (also known as TMEM173, MITA, ERIS, and MPYS) [4, 5, 21, 22]. Structurally, STING is comprised of an N-terminal four-pass transmembrane domain followed by a cytosolic CDN binding domain (CBD) and a loosely structured C-terminal tail (CTT) [21]. In addition to binding 2’3’-cGAMP, STING can also recognize 3’3’-CDNs of bacterial origin including 3’3’-cGAMP, 3’3’-c-di-AMP, and 3’3’-c-di-GMP, albeit with reduced affinities; pyrimidine-containing CDNs appear to be much weaker inducers of STING-mediated responses [7, 9, 15, 23].

Unlike in prokaryotes where CDNs promote anti-phage immunity at the post-translational level, in metazoans CDN-mediated activation of STING fosters innate anti-microbial immunity through both post-translational and transcriptional responses. In the inactive state, STING is held in an auto-inhibited state by its CTTs [28]. Upon engaging CDNs, the CTTs of STING are released thereby exposing a polymerization interface that triggers the formation of STING homo-oligomers, which are stabilized further by intermolecular disulfide bonds [28, 29]. Activated STING eventually traffics to a perinuclear compartment where exposure of the CTTs subsequently facilitates the recruitment, activation, and oligomerization of the kinase, Tank-binding kinase 1 (TBK1) [30, 31]. Finally, the phosphorylation of STING on its CTT at a conserved motif by TBK1 recruits the transcription factor, Interferon Regulatory Factor 3 (IRF3), resulting in its subsequent phosphorylation and activation [30, 31]. IRF3 then dimerizes and translocates to the nucleus where it induces the expression of Type I Interferons and other innate immune gene programs in order to restrict viral replication and spread (Figure 2).

Figure 2: Schematic of the cGAS-cGAMP-STING Signaling Axis.

Accumulation of pathogenic or mislocalized self dsDNA species in the host cell cytosol activates cGAS, resulting in the formation of cGAS-DNA liquid-phase condensates and the production of 2’3’-cGAMP. Binding of 2’3’-cGAMP or 3’3’-CDNs from bacterial sources to STING results in its activation culminating in STING oligomerization and translocation to the perinuclear Golgi. There STING recruits the kinase TBK1 to the transcription factor IRF3 via its C-terminal tails ultimately resulting in IRF3 phosphorylation, dimerization, and nuclear translocation. In the nucleus IRF3 induces Type I Interferon antiviral gene programs to foster an antiviral state. In addition, STING can traffic to the ER-Golgi Intermediate Compartment (ERGIC) where it promotes the formation of autophagosomes to remove microbial pathogens from the cytosol through autophagy.

As mentioned above, cGAS-STING signaling systems in eukaryotes can be found as far back as early metazoans including the sea anemone, N. vectensis [14]. Interestingly, nvSTING was shown to lack the CTT domain required for signaling by TBK1, hinting at the possibility of interferon-independent functions of the CDN-STING axis [14, 32, 33]. It is now apparent that this evolutionarily ancient function of STING is the transcription-independent induction of autophagy, a process that allows the cell to remove cytosolic pathogens including viruses and bacteria [33]. Indeed, STING-induced autophagy is sufficient to promote the clearance of cytosolic DNA and herpes-simplex virus-1 (HSV-1); interestingly, full-length STING displays a much stronger anti-viral activity, suggesting that interferons may cooperate with autophagy to restrict viral infection (Figure 2) [33].

In addition to inducing autophagy, STING is a functionally plastic immune-signaling scaffold, the downstream signaling specificity of which can be modified through discrete signaling modules linearly organized along its CTT [34]. Interestingly, the STING-CTT domains found in higher vertebrates only contain modules for recruiting TBK1 and IRF3 and, thus, inducing Type I Interferons [34]. On the other hand, the CTT of STING from ray-finned fish has undergone an expansion to include modules to recruit and activate TNF Receptor Associated Factor 6 (TRAF6) in order to promote signaling by Nuclear Factor Kappa-light-chain-enhancer of activated B cells (NF-κB), a transcription factor that induces the expression of pro-inflammatory cytokines [34]. It has been postulated that acquisition of the TRAF6-module may have provided a strong selective advantage in ray-finned fish, perhaps against a pathogen that is not well controlled by Type I Interferons [34]. It remains to be understood why higher vertebrates including humans lack this signaling module.

2’3’-cGAMP Transporters and Phosphodiesterases

One advantage of second messenger signaling systems is that they allow for the amplification of input signals. In higher eukaryotes, this is of particular significance as cyclic dinucleotides can potentially function as ‘alarmones’ or ‘immunotransmitters’ to signal the threat of infection to neighboring cells. Indeed, 2’3’-cGAMP can be transferred to neighboring ‘bystander’ cells through gap-junctions as well as by being packaged into virions [35–37]. In addition to direct transfer, 2’3’-cGAMP can also be secreted into the extracellular milieu where it can be sensed by patrolling immune cells [38, 39]. This was recently shown to be important for restricting tumor growth in vivo through the activation of tumor-associated natural killer (NK) and dendritic cells [38–39]. While the mechanism of 2’3’-cGAMP export is still unclear, the reduced folate carrier, also known as SLC19A1, moonlights as an importer of CDNs in human immune cells [40, 41]. SLC19A1 appears to indiscriminately transport all CDNs tested, and this may be of importance for the field of tumor immunotherapy where non-hydrolyzable CDNs are currently being developed and tested [40, 41]. Interestingly, SLC19A1 does not appear to be the sole CDN transporter as SLC19A1-deficient cells can still respond to exogenous CDNs, albeit to a lesser extent [40, 41]. Follow-up studies will be necessary to identify the other CDN transporter(s) and to elucidate the role (if any) of cGAMP transport in the control of viral infection.

CDN export also provides a unifying mechanism for resolution of 2’3’-cGAMP signaling in higher metazoans [39]. Indeed, the catalytic domain of the only known mammalian 2’3’-cGAMP PDE, ectonucleotide pyrophosphatase phosphodiesterase 1 or ENPP1, is found in the extracellular space either as a transmembrane protein or as a soluble secreted factor [42, 43]. Thus, 2’3’-cGAMP secretion may function to promote infiltrating immune cell activation as well as to return 2’3’-cGAMP levels to pre-induction homeostasis through ENPP-1-mediated degradation (Figure 3) [38, 39, 42, 43]. While ENPP1 appears to be the major cGAMP hydrolase in mammals, a new class of 2’3’-cGAMP-selective phosphodiesterases was recently identified in some lower metazoans. These enzymes were initially identified in mammalian poxviruses and were hence given the name poxvirus immune nucleases or poxins [44]. Subsequent bioinformatic analysis then identified poxin homologues with conserved 2’3’-cGAMP hydrolase activity in the genomes of insects, particularly moths and butterflies [44]. Interestingly, poxin homologues can also be found in the alphabaculoviruses that infect these insect hosts, raising the intriguing hypothesis that poxins may have been domesticated by these viruses and eventually transferred to pathogenic, mammalian viruses in order to antagonize the cGAS-STING pathway [44].

Figure 3: cGAMP Transport and Homeostasis.

Following activation by cytosolic dsDNA species, cGAS catalyzes the in-situ formation of 2’3’-cGAMP resulting in STING and Type I Interferon activation within the infected or transformed cell. Transfer of 2’3’-cGAMP to neighboring bystander cells via gap junctions propagates STING-mediated signaling across the infected or damaged tissue. 2’3’-cGAMP levels can return to their pre-induction state through secretion via an unidentified transporter followed by ENPP-1 mediated hydrolysis. Secreted or exogenous CDNs can also be taken up by immune cells via a folate/organic phosphate (OP⁻) antiporter, SLC19A1, to further amplify the innate immune response.

Viral Evasion of cGAMP-STING Signaling

While cGAMP is critical for controlling virus infection, many viruses appear to have developed countermeasures including degradation mechanisms to evade this signaling system, highlighting the evolutionary arms race that exists between viruses and their hosts [45–54]. One such strategy, commonly employed by herpesviruses, is to limit cGAMP production by inhibiting cGAS activation [47–50]. This can be accomplished through the use of tegument proteins that directly bind to cGAS and trigger its dissociation from viral DNA as is the case for Kaposi-sarcoma herpesvirus and human cytomegalovirus [47–49]. Interestingly, HSV-1 has developed a sophisticated strategy for the species-specific inhibition of cGAS-signaling involving the tegument protein, UL37 [50]. Mechanistically, UL37 was found to possess deamidase activity, which allows it to inactivate cGAS through the deamidation of a single asparagine residue on its activation loop [50]. Interestingly, UL37 can deamidate mouse and human cGAS isoforms but not isoforms from non-human primates, again underscoring how evolutionary arms races may specify host tropism [50]. Thus, direct targeting of cGAS appears to be a common strategy employed by herpesviruses to evade host DNA-sensing pathways. While we aimed to highlight some of the recent publications and common themes, this is certainly not an exhaustive list of the strategies used by DNA viruses to evade cGAMP-STING signaling, some of which involve directly antagonizing STING (for a more detailed review on this topic please see Ni et. al.) [51–54]. It is possible that the mechanisms for bystander cell activation discussed above may have developed out of a necessity to counteract these viral immune evasion strategies in infected cells. It will be interesting to see whether bacteriophages may employ similar mechanisms for evading CBASS-mediated immunity, perhaps through use of poxin homologues.

Concluding Remarks

In summary, CD-NTase signaling networks constitute an evolutionarily ancient form of innate anti-viral defense originating as prokaryotic anti-phage systems. It is possible that as bacteria fused with archaea they brought along these anti-viral defense mechanisms ultimately giving rise to the cGAS-STING pathway in metazoans as we know it today. Although pathogens have developed strategies to evade these signaling systems, sensing of foreign and damaged- or misplaced-self DNA species by the enzyme cGAS has proven to be essential for the control of pathogen infection and tumorigenesis, respectively, both in vitro and in vivo. Although much is now known about the mechanisms of cGAS and STING activation, as cGAS indiscriminately binds DNA, a more thorough understanding of self versus non-self discrimination by cGAS will be necessary. Several recent studies have begun to shed light on this process, which will likely be an active area of research over the coming years [55–59]. Indeed, cyclic dinucleotides are a rapidly expanding class of signaling molecules in all domains of life with many exciting biological roles to be uncovered ahead.