TREX1 – Apex predator of cytosolic DNA metabolism

Abstract

The cytosolic Three prime Repair EXonuclease 1 (TREX1) is a powerful DNA-degrading enzyme required for clearing cytosolic DNA to prevent aberrant inflammation and autoimmunity. In the absence of TREX1 activity, cytosolic DNA pattern recognition receptors of the innate immune system are constitutively activated by undegraded TREX1 substrates. This triggers a chronic inflammatory response in humans expressing mutant TREX1 alleles, eliciting a spectrum of rare autoimmune diseases dependent on the nature of the mutation. The precise origins of cytosolic DNA targeted by TREX1 continue to emerge, but DNA emerging from the nucleus or taken up by the cell could represent potential sources. In this Review, we explore the biochemical and immunological data supporting the role of TREX1 in suppressing cytosolic DNA sensing, and discuss the possibility that TREX1 may contribute to maintenance of genome integrity.

1. Introduction

Infectious organisms utilize nucleic acid polymers to encode genetic information. DNA and RNA therefore represent excellent pathogen-associated molecular patterns (PAMPs) for innate immune recognition of microbial threats. Pattern recognition receptors (PRRs) have evolved to detect pathogen-derived nucleic acids within both the endosomal and cytosolic compartments. PRR ligation initiates signaling cascades leading to production of inflammatory signaling molecules, such as the type I interferons (IFN-I), which stimulate innate and adaptive immune responses against the perceived microbial threat (reviewed in [1]). DNA PRRs are generally not sequence-specific and provide broad detection coverage regardless of pathogen origin. This lack of specificity presents a potential danger, however. The innate immune system cannot distinguish self- from pathogen-derived DNA, so cells carefully manage and compartmentalize their own genetic material to prevent self-DNA sensing and aberrant immune activation. By sequestering DNA within the nucleus and mitochondria, cells establish the cytosol as a DNA-free space in which non-specific DNA-sensing molecules can be safely deployed for pathogen detection.

The Three-prime Repair EXonuclease 1 (TREX1) is required to maintain the cytosol as a DNA-free space. TREX1 is a potent 3′→5′ cytosolic exonuclease which rapidly degrades both double-stranded and single-stranded DNA (ds- and ssDNA) [2–5]. TREX1 was initially thought to participate in DNA repair, based on sequence and structural homology with bacterial DNA repair enzymes. However, TREX1 ablation in mice demonstrated an aggressive, immune-activating phenotype and not a mutator phenotype [6]. TREX1 mutations are now known to cause a spectrum of human autoimmune diseases associated with constitutive low-level IFN-I signaling in the absence of a microbial threat [7,8]. Murine models of TREX1 deficiency have demonstrated that the cyclic GMP-AMP Synthase - Stimulator of INterferon Genes (cGAS-STING) pathway of cytosolic DNA sensing is required for this inflammatory signaling [9,10], supporting a model in which TREX1 degrades cytosolic DNA to prevent inappropriate DNA PRR activation. TREX1 DNA substrates may arise from within the cell itself or from exogenous sources, but critically, must be cleared to maintain normal DNA compartmentalization and suppress persistent stimulation of cytoplasmic DNA sensors, inflammation, and autoimmunity.

The precise nature of TREX1 substrates continues to emerge, as do the specific molecular pathways of DNA sensing and inflammatory signaling underlying TREX1-mediated autoimmune pathogenesis. In this review, we summarize current knowledge regarding the activity, regulation, and function of the TREX1 enzyme. We explore the accumulated mouse and human immunological data demonstrating a role for TREX1 in suppression of cytosolic DNA sensing and inflammatory signaling. Additionally, we discuss current concepts regarding sources of TREX1 cytosolic DNA substrates, and the cell populations participating in DNA sensing to initiate inflammation. Finally, we examine data from other nucleic acid-processing enzymes which have been linked to autoimmunity, including SAMHD1 and RNASEH2 [11,12]. Despite acting on unique substrates in distinct nucleic acid metabolism processes, mutations in these genes cause forms of autoimmunity similar to TREX1-deficiency. We discuss the possibility that genome instability may represent a common source of DNA sensing in TREX1 dysfunction and other disorders.

2. TREX1 activity, structure, localization, and regulation

The TREX1 gene is a single open reading frame located on chromosome 3p21.31 encoding a 314 amino acid polypeptide with a molecular weight of ~33 kDa [13]. TREX1 homologs are found in metazoans, are well-conserved, and are expressed in most tissues [5,13,14], likely indicating an important house-keeping function in most cells. The gene encodes a magnesium-dependent, non-processive DNA exonuclease which degrades both double- and ssDNA from the 3′-terminus, cleaving the phosphodiester bond and releasing single nucleoside monophosphates [2–5,13]. The TREX1 N-terminal 242 amino acids contain all of the necessary structural elements for full catalytic activity, and the C-terminal region 72 amino acids are required for localization to the perinuclear space or endoplasmic reticulum (ER), where it is oriented with its active sites exposed to the cytosol (Fig. 1A) [15–22]. TREX1 is an obligate homodimer, with residues from one protomer reaching across the dimer interface to contribute to catalysis in the opposing protomer. This structure is unique among exonucleases and is mediated by an extensive hydrogen-bonding network between protomers, including backbone contacts between β-strands, side-chain pairings, and hydrophobic packing of antiparallel helices (Fig. 1B). The strength of this interaction creates a kinetically stable dimer that does not dissociate and re-equilibrate at measurable rates [23–25].

Fig. 1. TREX1 localization and structure.

(A) Microscopy images demonstrating full-length protein (AA 1–314) and C-terminal region (AA 243–314) localizing to perinuclear region/ER, and diffuse localization of TREX1 C-terminal truncation mutant (AA 1–286). Images borrowed from published data [18]. (B) The dimer interface and active site of TREX1 with DNA bound (yellow and orange). Magnesium ions are denoted as magenta spheres. Protomers are differentiated by color. Interactions along the dimer interface are represented by dashed lines. Catalysis in the highlighted TREX1 active site (right) is mediated by several key residues in the green protomer, including D18 and D200, as well as residues from the blue protomer, including R62 and R114 (not shown). Image was borrowed from published data [25]. (C) A space-filling model of substrate binding in the TREX1 active-site. The 3′-terminal nucleotide (orange spheres) of the 4-mer oligonucleotide substrate is tightly constrained inside the nucleotide binding pocket, such that bulky DNA damage adducts would be sterically hindered. TREX1 atoms are represented as cyan spheres. Graphic was generated in PyMol using Protein Data Bank structure ‘2OA8’.

The TREX1 sequence contains the DEDD motif typical of many DNA exonucleases, and exhibits both sequence and structural homology with the E. coli 3′→5′ exonuclease domain of DNA Pol I and the DnaQ subunit of Pol III [4,5]. Because these bacterial enzymes play a critical role in DNA proofreading and repair, it was initially hypothesized that TREX1 was a component of the mammalian DNA repair machinery. However, TREX1 exhibits very little degradative activity against DNA containing 3′ obstructive lesions that result from DNA damage, indicating that its primary substrate is undamaged DNA 3′-termini [26,27]. This notion is further supported by the crystal structures of TREX1 in complex with substrate or product, which reveal a nucleotide binding pocket within the active site too constrictive to allow binding of bulky 3′-modifications such as phosphate, phosphoglycolate, or tyrosyl residues (Fig. 1C) [15,20]. TREX1 has been reported to translocate to the nucleus following genotoxic stress [22,28] and to interact with the PARP1 complex [29], suggesting that it could contribute to DNA repair. However, TREX1 ablation in mice does not increase mutation rates or oncogenesis [6] and TREX1 does not co-localize with major DNA Damage Response (DDR) components γH2AX, pATR, or p53BP following genotoxic stress [28]. In addition, TREX1-deficient cells repair UV-induced dsDNA breaks and cyclobutane pyrimidine dimers at the same rate as TREX1 wild type (WT) cells [21]. Therefore, TREX1 does not appear to participate directly in DNA damage repair.

TREX1 exhibits complex and multi-layered regulation, suggesting a fluctuating level of exonuclease activity dependent upon the physiological status of the cell. TREX1 transcription is induced by numerous DNA-damaging agents, including UV light, polycyclic aromatic hydrocarbons, and H₂O₂ [28] as well as high doses of gamma radiation [30] and the cytokine IFN-γ [31]. TREX1 induction following DNA damage might indicate that TREX1 acts on DNA which “leaks” from the nucleus to the cytosol following DNA damage, a concept that will be revisited later in the review. Upregulation of TREX1 in response to genotoxic stress or to IFN-γ requires c-Jun, a member of the AP-1 transcription factor complex, and the TREX1 promoter contains an AP-1 binding site [28]. In addition, TREX1 is induced by IFN-I signaling and is therefore considered an interferon-stimulated gene (ISG), suggesting a potential role for TREX1 in the anti-viral response [14,32]. The TREX1 gene contains two splice sites, allowing for three potential splice variants [13]. At least one TREX1 splice variant results in truncation of key catalytic residues, revealing a potential pathway for post-transcriptional regulation. In addition, microRNA-mediated RNA interference has been shown to modulate TREX1 message levels in endothelial cells [33]. Finally, our lab has demonstrated that the TREX1 C-terminus is post-translationally modified by ubiquitination, a modification which promotes shuttling to and degradation in the autophagosome [18]. Additional study of how and when TREX1 activity is modulated by the cell might contribute to our understanding of its normal physiological role.

3. Immunological consequences of TREX1 dysfunction

TREX1 mutations induce a chronic state of IFN-I signaling and inflammation, presumably due to the accumulation or persistence of undegraded DNA in the cytosol and constitutive activation of DNA PRRs. A multitude of redundant PRRs detect DNA in the cytoplasm of eukaryotic cells (reviewed in [34]). The key sensor of cytosolic DNA resulting from TREX1 dysfunction appears to cGAS, a sequence-independent obligate dsDNA sensor localized predominantly within the cytoplasm [35]. DNA binding induces cGAS dimerization and production of cyclic GMP-AMP (cGAMP), a small molecule ligand of the ER-localized signaling adaptor protein STING [36–39]. Binding of cGAMP to STING leads to STING dimerization, translocation to the Golgi apparatus, and recruitment of both IKK and TANK-binding kinase 1 (TBK1) [40,41]. IKK phosphorylates IκB, leading to its degradation, NF-κB activation, and the expression of pro-inflammatory cytokines such as TNF-α and IL-1 [42]. Binding of TBK1 to STING induces TBK1 auto-phosphorylation and subsequent phosphorylation of IRF3 [43]. IRF3 dimerizes, translocates to the nucleus, and stimulates production of type I interferons (IFN-I), secreted signaling molecules which induce inflammation and mediate anti-viral immunity. IFNs bind to the IFN-alpha receptor (IFNAR), expressed on all cells, inducing the expression of hundreds of ISGs as a coordinated anti-viral defense program. TREX1-deficient mice exhibit constitutive IFN-I signaling and ISG expression, and ablation of cGAS [9,44,45], STING [10,46,47], IRF3 [48], or IFNAR [47,48] completely abrogates this chronic inflammatory state. Inhibition of TBK1 in TREX1-deficient mice mediates similar effects [49]. Thus, TREX1-dependent inflammation likely arises from the inappropriate sensing of undegraded TREX1 substrates in the cytosol by cGAS, leading to chronic STING activation and constitutive IFN-I signaling (Fig. 2).

Fig. 2. TREX1 dysfunction triggers the cGAS-STING cytosolic DNA sensing pathway.

Model of how TREX1 catalytic inactivity initiates IFN-I-dependent inflammation. TREX1 dysfunction allows cytosolic accrual of undegraded DNA, cGAS-dependent sensing, cGAMP synthesis, and activation of STING. STING recruitment of TBK1 and TBK1-dependent IRF3 phosphorylation triggers transcription of IFN-I genes, leading to chronic inflammation and autoimmunity.

Inflammation arising from TREX1 mutations causes a spectrum of autoimmune diseases in humans and mice, and the nature of the specific mutation determines the resulting disease phenotype. Certain TREX1 alleles give rise to Aicardi-Goutières Syndrome (AGS), a severe neurodegenerative disorder typically presenting in infancy [7,8,50]. AGS TREX1 mutations include single amino acid variants and frameshift mutations leading to early translational STOP codons and presumably dysfunctional TREX1 polypeptides. Most disease-causing alleles exhibit recessive genetics, but some dominant alleles have been identified. AGS patients demonstrate sterile IFN-I production, dramatic expansion of white blood cells, neural inflammation and calcification, skin lesions, multi-organ inflammation, and significant mortality early in life [51–54]. Similarly, TREX1 ablation in mice leads to an aggressive lupus-like autoimmune phenotype, as indicated by robust ISG induction, extensive myocarditis, nuclear autoantibody production, and markedly reduced survival [9,10,44,46,55,56]. TREX1^−/− mice do not exhibit a neuroinflammatory phenotype, but the severity and rapid course of disease in these animals mirrors the aggressiveness of AGS. In contrast, humans expressing catalytically inactive TREX1 D18N develop less severe autosomal dominant Familial Chilblain Lupus (FCL) [8,57,58]. FCL patients exhibit temperature-sensitive chilblain lesions on the hands and feet as well as systemic IFN-I signaling, autoantibody production, and multi-organ inflammation. Neurological involvement is occasionally observed, leading to speculation that FCL may represent a milder form of AGS [50]. Our laboratory generated a mouse model of FCL by replacing the TREX1 WT allele with the catalytically inactive TREX1 D18N allele. This mouse model exhibits a less aggressive lupus-like phenotype relative to TREX1^−/− mice, paralleling the human TREX1 disease spectrum [47,59,60]. Finally, frameshift mutations in TREX1 that eliminate the C-terminal region from the protein cause mislocalization throughout the cytosol and autosomal dominant Retinal Vasculopathy with Cerebral Leukodystrophy (RVCL) [16,19,61]. RVCL most often presents in middle age as progressive vision loss and neural degeneration, likely through destruction of small retinal and neural blood vessels [19,62]. In contrast with AGS and FCL, systemic IFN-I signaling is not observed [63]. Curiously, mice expressing RVCL-causative TREX1 C-terminal mutants develop elevated levels of autoantibodies, but not specifically against nuclear antigens. Furthermore, systemic ISG expression is not observed, but certain purified cell populations may exhibit ISG upregulation in vitro, potentially indicating a more localized IFN-I response [64,65]. This suggests a disease mechanism distinct from AGS and FCL.

Different TREX1 mutations cause a spectrum of human diseases, indicating that total loss of TREX1 protein, fractional or complete loss of catalytic activity, and enzyme mislocalization differentially impact disease phenotype. This implies that TREX1 possesses additional function(s) beyond cytosolic DNA degradation, and that the C-terminal region must be properly localized in the perinuclear space or ER to fulfill at least one of these functions. Supporting this concept are studies demonstrating TREX1 participation in the oligosaccharyltransferase (OST) complex [64,66]. OST is an ER-localized membrane protein complex responsible for transferring short sugar moieties from lipid-linked oligosaccharides (LLOs) to nascent proteins. TREX1 interacts with this complex through its hydrophobic C-terminal region, and appears to mediate proper OST function through a DNA catalysis-independent mechanism. Loss of the TREX1-OST interaction disrupts normal OST function, leading to an increase in non-LLO-linked free glycans in TREX1^−/− cells. Glycan mismanagement has been previously linked to inflammation and lupus-like autoimmunity [67,68], and indeed, free glycans induced by TREX1 dysfunction have been shown to directly induce IFN-I signaling through a STING-TBK1-dependent mechanism [69]. Thus, the human and mouse disease spectrum resulting from different TREX1 mutations likely indicates that multiple molecular sources of inflammation contribute to disease (Fig. 3). Furthermore, the severe phenotypes resulting from full TREX1 ablation likely reflect the combined effects of failed DNA degradation and OST dysregulation.

Fig. 3. Heterogeneity in TREX1-linked human diseases likely relates to distinct sources of inflammation.

Cartoon models depicting active sources of inflammation caused by TREX1 mutations. Full ablation of TREX1 or highly deleterious mutations that result in no polypeptide being produced (upper right) lead to both failed cytosolic DNA degradation and OST dysregulation, and are associated with the severe AGS phenotype. Point mutation D18N (lower left) renders TREX1 catalytically inactive and leads to failed cytosolic DNA degradation, but preserves OST interactions, consistent with the more mild FCL phenotype observed in humans expressing the D18N allele. C-terminal mutations (lower right) induce TREX1 mislocalization throughout the cytosol but preserve catalytic activity, likely allowing clearance of cytosolic DNA but negatively impacting OST regulation. C-terminal mutations cause RVCL in humans, but a direct link between RVCL and free glycan mismanagement has yet to be definitively established.

Systemic TREX1 dysfunction in multiple cell populations could drive inflammation and trigger autoimmunity. The inflammatory myocarditis phenotype of the TREX1^−/− mouse initially suggested cardiomyocytes as a key cell population sensing and responding to undegraded cytosolic DNA [6,46]. Consistent with this model, ISG induction was observed in TREX1^−/− hearts early during development, suggesting active DNA sensing in cardiac tissue. However, cardiomyocytes do not appear to possess a functional cGAS-STING signaling axis [10], suggesting that DNA sensing and IFN-I production might originate within the immune cell infiltrates that develop in TREX1^−/− mice. One model may be that TREX1-deficient cardiomyocytes release undegraded cytoplasmic DNA into the microenvironment, perhaps within exosomes, which could act as an attractant and source of cGAS-stimulatory DNA for circulating immune cells [70]. Potentially consistent with this model, a recent study found that cardiomyocytes release DNA following ischemic injury, triggering an IRF3-dependent influx of IFN-I-producing phagocytic cells into cardiac tissue [71]. Multiple studies have reported an immune cell origin of inflammatory signaling in TREX1^−/− mice, and evidence supports both monocytes [10] and dendritic cells [56] as sources of DNA sensing and IFN-I production. In addition to innate immune populations, our laboratory has recently demonstrated that T cells respond to TREX1 catalytic inactivity by producing IFN-I [47]. Using the TREX1 D18N model of FCL, we have observed T cell IFNA expression and TBK1/IRF3 phosphorylation in vivo, and demonstrate that TREX1 D18N T cells produce IFN-I protein when activated. This would suggest that TREX1-dysfunctional T cells can supply an autocrine inflammatory signal promoting autoimmunity. In addition, distinct cell types may be involved in the inflammatory responses generated by self-DNA sensing versus DNA-independent free glycan accumulation. Studies focused on isolating the variables of failed DNA catalysis (TREX1 D18N mice) vs. OST dysregulation (TREX1 C-terminus mutant mice) may contribute to separating the influence of these competing inflammatory pathways, leading to a better understanding of autoimmune pathogenesis in TREX1-linked human diseases.

4. TREX1 DNA substrates

The molecular targets of TREX1 degradation are an active area of research, and several potential DNA sources have been proposed. This may suggest that TREX1 acts on DNA arising from multiple independent physiological processes. One model posits that TREX1 acts in the SET complex to degrade genomic dsDNA during granzyme A-mediated cell death [72]. Granzyme A is released by cytolytic immune cells, initiating a death program in cells targeted for deletion. This process involves nuclear translocation of the SET complex, allowing the NM23-H1 endonuclease to cleave phosphodiester bonds in genomic DNA. TREX1 binds to the free 3′-hydroxyl termini of NM23-H1-generated nicks and fully dismantles the genome, killing the target. However, granzyme A-mediated cell killing is primarily linked to virally-infected cells and cancer cells, raising questions about the mechanism of constitutive inflammation detected from TREX1 dysfunction in humans and mice. Recently, our laboratory has proposed that NM23-H1 and TREX1 could act in concert to degrade the genomic DNA of developing erythroblasts [60]. Over ~200 billion enucleated erythrocytes generated in humans each day [73], so TREX1 dysfunction in this context could lead to the persistence of undegraded erythroblast DNA, dysregulated erythropoiesis, and constitutive self-DNA sensing.

Mounting evidence suggests that genomic DNA may move from the nucleus to the cytosol, perhaps resulting from DNA damage, repair processes, DNA replication, or turnover of endogenous retroelements. TREX1 likely acts as a general disposal mechanism for these DNA species to limit self-sensing. Two models have been proposed in which TREX1 degrades nucleus-derived ssDNA to prevent inflammation, but they differ on how ssDNA is generated. In one model, ssDNA fragments are thought to be generated in the nucleus during S phase following failed processing of aberrant replication intermediates [22,74,75]. TREX1^−/− murine embryonic fibroblasts (MEFs) and primary fibroblasts from AGS patients exhibit cytosolic ssDNA accrual, and BrdU-pulse experiments suggest that this DNA is released from the nucleus during S phase. Cytosolic ssDNA accumulation in TREX1-deficient cells is reportedly correlated with cell cycle defects, p53 activation and p21 induction, reduced CHK2 expression, and DNA comet formation, consistent with a state of elevated DNA damage and chronic DDR activation [21,22,76,77].The potential relationship between TREX1 inactivity and DNA damage will be discussed later in the review. Another model proposes a source of accumulating cytosolic ssDNA derived from retroviral replication. ssDNA retroelements were found to be more abundant in TREX1^−/− versusWT hearts [48] and in TREX1-deficient cells [76,78,79], and overexpression of WT TREX1 but not catalytically-inactive TREX1 led to reduction in retrotransposition events in vitro [48]. Related to this idea, TREX1 also appears to exhibit degradation activity against HIV retroviral cDNA during infection. It has been proposed that TREX1 degradation of aborted, non-productive HIV cDNA contributes to HIV avoidance of detection by cGAS [80,81].

TREX1 dysfunction causes cGAS-STING activation, and cGAS requires dsDNA binding for activation [36,82–85]. Furthermore, TREX1^WT/D18N heterodimers that could form in FCL patients exhibit no functional activity against dsDNA, but retain significant ssDNA degradative activity [123]. Inflammation arising from TREX1 dysfunction therefore likely reflects the sensing of dsDNA. Cytosolic ssDNA originating from the nucleus during DNA replication, reportedly ~60–65 nucleotides in length [22] would need to exhibit double-stranded structure to activate cGAS. HIV cDNAs have been shown to form hairpin structures permitting cGAS-DNA binding and activation [86], potentially suggesting that a similar property may extend to retroelement cDNAs, but attempts to relate retroelement activation of cGAS-STING to TREX1 dysfunction have generated conflicting results. The reverse transcriptase (RT) inhibitor AZT had no apparent effect on TREX1^−/− mouse survival [55], but a marked improvement in survival was reported using a combination of Truvada and Viramune targeting a broader range of RT enzymes [87]. A more comprehensive study using broad-spectrum RT inhibitor cocktails found no change in survival, tissue inflammation scores, or IFN-I signaling between treated and untreated TREX1^−/− animals [88]. Finally, the RT inhibitors 3TC and d4T appear to cause a reduction in inflammatory signaling within TREX1-mutant neuronal cell lines in vitro [79]. TREX1 substrates continue to emerge, and further research will likely reveal that TREX1 acts on many DNA species to maintain a DNA-free cytosol.

5. Nucleic acid metabolism, genome instability, and inflammation

Five genetic loci involved in nucleic acid-metabolism or detection are causally linked to AGS. In addition to TREX1, mutations in any of the three RNASEH2 genes [12], SAMHD1 [11], ADAR1 [89], and IFIH1 [90,91] are causative for AGS. Mutations in these genes can be partitioned into those involving cytosolic RNA sensing and those involving DNA sensing. ADAR1 is an RNA-editing enzyme which may target and destabilize endogenous dsRNA species, preventing chronic activation of the RIG-I/MDA5 RNA sensing pathway (reviewed in [92]). Likewise, IFIH1 encodes MDA5 [93], and gain-of-function mutations in IFIH1 cause AGS. Thus, one cluster of AGS mutations likely induce aberrant cytosolic RNA sensing. TREX1, RNaseH2, and SAMHD1 mutations impinge on the cGAS-STING pathway of cytosolic DNA sensing. RNaseH2 excises misincorporated ribonucleotides from genomic DNA. These ribonucleotide lesions increase susceptibility to single- and double-stranded breaks (DSBs) and loss of RNaseH2 activity induces massive DNA damage, impaired cell cycle progression, and perinatal lethality in mice dependent on cGAS-STING [94–96]. SAMHD1 is a triphosphohydrolase which regulates dNTP pools to maintain DNA replication fidelity [97,98]. SAMHD1 also likely contributes to dsDNA break repair by homologous recombination [99], and might promote resolution of stalled replication forks by MRE11 exonuclease activity [100]. SAMHD1 deficiency leads to chronic DNA damage, DDR activation, cellular senescence, and cell cycle defects [101], an increased mutational burden in cultured fibroblasts [102], and cGAS-STING-dependent IFN-I signaling [100,103]. Thus, AGS-causing mutations occur in multiple genes responsible for genome maintenance, suggesting a model in which failure to maintain genome integrity can lead to cGAS-STING activation and autoimmunity.

Ample experimental evidence supports the concept that genome instability can lead to the appearance of DNA in the cytosol and the activation of cGAS-STING. RNaseH2-deficient MEFs exhibit increased formation of micronuclei, extra-nuclear structures containing damaged chromatin that bud from the nucleus following missegregation of genetic material during mitosis [104,105]. Micronuclei are initially compartmentalized by a nuclear envelope but frequently rupture, allowing cGAS access to the damaged chromatin [104,106,107]. In addition, DNA damage caused by genotoxic agents, depletion of DNA repair factors, and the normal aging process are all associated with “leakage” of naked ss- and dsDNA fragments directly into the cytosol, which induces a cGAS-STING-dependent IFN-I response [30,108–113]. These observations suggest that cGAS-STING might participate in the detection of genotoxic stress in addition to PAMPs. Aging cells and cells exposed to genotoxic insults risk oncogenic transformation. cGAS-STING-dependent cytosolic DNA sensing and IFN-I production in these cells could act to promote immune recognition and clearance, as well as encourage senescence [114–116].

TREX1 does not appear to directly participate in DNA repair. It is therefore curious that TREX1^−/− cells exhibit chronic DNA damage similar to RNaseH2 and SAMHD1-deficient cells [21,22,76,77]. It may be that the failure to degrade cytosolic DNA by TREX1 can cause genomic instability by triggering a “feedback” response in the nucleus. One model suggests that undegraded TREX1 ssDNA substrates in the cytosol can diffuse back into the nucleus, where they non-productively associate with DNA repair factors RPA and RAD51 [21]. Depletion of free RPA/RAD51 may chronically sensitize cells to DNA damage and genome instability. However, the structural and biochemical properties of TREX1 indicate that dsDNA is a key substrate, and genotoxic stress can lead to the appearance of double-stranded genomic DNA in the cytosol. The fate of this dsDNA after it leaves the nucleus could involve TREX1-mediated degradation. Consistent with a model of extranuclear dsDNA disposal, ablation of lysosomal endonuclease DNaseII in fibroblasts is associated with cytosolic accumulation of nucleus-derived dsDNA fragments [117]. Furthermore, these cells exhibit chronic DNA damage similar to AGS-mutant cells. DNaseII-deficient cells and mice also exhibit cGAS-STING-dependent IFN-I signaling [117–119]. This implies that 1) dsDNA is regularly exported from and degraded outside the nucleus, and 2) dysfunction in this process may lead to both cytosolic pattern recognition and DDR activation. A precise mechanism describing how failed cytosolic dsDNA degradation leads to DNA damage remains to be elucidated, but a similar model to that proposed for ssDNA TREX1 substrates could apply. dsDNA fragments diffusing back into the nucleus would represent excellent substrates for DNA-PK, a nuclear kinase which mediates recruitment of repair factors to dsDNA breaks, and which is readily activated by fragments as small as 14 nucleotides in vitro [120–122]. The localization of TREX1 at the nuclear border, its high affinity for dsDNA, and the phenotypic similarities between TREX1 and DNaseII deficiency could suggest that both participate in disposal of leaked nuclear dsDNA to prevent genome instability and cytosolic DNA sensing.

6. Concluding remarks

TREX1 is required for efficient cytosolic DNA disposal to effectively prevent inappropriate cGAS-STING-mediated DNA sensing and immune activation. The consequences for human health resulting from dysfunction in this process include severe and debilitating autoimmune diseases. Many aspects of TREX1 biology remain to be fully elucidated. Additional work examining the precise molecular targets of TREX1 degradation is needed. Furthermore, additional clarification of the TREX1-deficient cell populations driving pro-autoimmune inflammatory signaling in vivo could lead to better, more targeted therapies for human diseases caused by TREX1 mutation. Answers to these questions should provide a better understanding of how autoimmunity arises from failed nucleic acid processing, and how DNA waste products are normally managed to maintain cellular homeostasis and genome stability.

Abbreviations:

AGS

Aicardi-Goutieres syndrome

cGAMP

cyclic GMP-AMP

cGAS

cyclic GMP-AMP synthase

DDR

DNA damage response

DSBs

double-stranded breaks

dsDNA

double-stranded DNA

ER

endoplasmic reticulum

FCL

Familial Chilblain Lupus

IFN-I

Type I interferon

IFNAR

interferon-α/β receptor

ISG

interferon-stimulatedgene

LLOs

lipid-linked oligosaccharides

MEFs

murine embryonic fibroblasts

OST complex

oligosaccharyltransferase complex

PAMP

pathogen-associated molecular pattern

PRR

pattern recognition receptor

RVCL

retinal vasculopathy with cerebral leukodystrophy

RT

reverse transcriptase

ssDNA

single-stranded DNA

STING

stimulator of interferon genes

TBK1

TANK-binding kinase 1

TREX1

three-prime repair exonuclease 1

WT

wild type