Main

The innate immune system uses pattern-recognition receptors to detect microbial signatures (microbe-associated molecular patterns) or cellular damage (damage-associated molecular patterns)1,2. Given that nucleic acids are central to the replication and propagation of pathogens, the recognition of aberrant RNA and DNA serves as a fundamental mechanism of host defense3. The activation of cytosolic nucleic acid receptors initiates the production of type I interferons and other cytokines and leads to the activation of adaptive immunity and restriction of infection4,5,6.

Several DNA sensors, including Sox2 (ref. 7), TLR9 (ref. 8), AIM2 (ref. 9), DAI10, IFI16 (ref. 11) and DDX41 (ref. 12), have been reported to monitor pathogenic DNA that induces the secretion of type I interferons and proinflammatory cytokines. The cyclic GMP-AMP (cGAMP) synthase cGAS has been defined as a key sensor of cytosolic DNA13,14,15. After binding DNA, cGAS catalyzes the synthesis of cGAMP, which in turn associates with and activates the adaptor STING and then elicits innate immune responses1,16,17. However, it is still unknown how the activity of cGAS is regulated during host defense.

Post-translational modifications of proteins, such as phosphorylation, acetylation, ubiquitination and glycosylation, have important roles in regulating the activity of the target proteins during immune responses18,19. Glutamylation adds glutamate side chains onto the γ-carboxyl groups of glutamic acid residues in the sequence of target proteins20,21. Glutamylation is catalyzed by glutamylases, such as TTLL ('tubulin tyrosine ligase–like') enzymes22,23. Glutamylation is a reversible modification akin to phosphorylation, whereby glutamates can be removed by a family of cytosolic carboxypeptidases (CCPs)21. Polyglutamylation of tubulins, the well-known targets of glutamylation, modulates the interaction between microtubules and their partners, which regulates microtubule-related processes such as ciliary motility and neurite outgrowth21,24.Glutamylation of the spindle-checkpoint protein Mad2 has a pivotal role in the development of megakaryocytes25. Here we identified differential glutamylation of cGAS that regulated its activity in antiviral defense.

Results

Susceptibility of CCP5- or CCP6-deficient mice to DNA viruses

Mice deficient in the gene encoding the cytosolic carboxypeptidase CCP6 (Agbl4; called 'Ccp6' here) display underdevelopment of megakaryocytes and abnormal thrombocytosis25. We therefore sought to explore whether glutamylation was involved in the regulation of host defense against viral infection. In addition to mouse strains deficient in CCP1 (Agtpbp1; called 'Ccp1' here) and Ccp6 studied before25, we generated mouse strains deficient in the genes encoding CCP2 (Agbl2; called 'Ccp2' here), CCP3 (Agbl3; called 'Ccp3' here), CCP4 (Agbl1; called 'Ccp4' here) or CCP5 (Agbl5; called 'Ccp5' here) via genome-editing approached based on clustered regularly interspaced short palindromic repeats (CRISPR) and the endonuclease Cas9 (Supplementary Fig. 1a) and further verified deletion of these genes in bone marrow–derived macrophages (BMDMs) (Fig. 1a). We infected all six mutant mouse strains with DNA viruses such as herpes simplex virus (HSV) and vaccinia virus (VACV). We noted that mice deficient in Ccp5 (Ccp5−/−) or Ccp6 (Ccp6−/−) were more vulnerable to HSV and VACA infection than were the other mutant mice (Fig. 1b and Supplementary Fig. 1b). Moreover, Ccp5−/− or Ccp6−/− mice had lower concentrations of type I interferons in serum than did their wild type littermates, after infection with HSV (Fig. 1c,d). In contrast, mice deficient in Ccp1, Ccp2, Ccp3 or Ccp4 had concentrations of type I interferons in serum similar to those of their wild-type littermates after infection with HSV (Fig. 1e,f). Consistent with that, after infection with HSV, viral titers were much higher in Ccp5- or Ccp6-deficient mouse brains than in their wild-type counterparts (Fig. 1g). At the same time, after infection with HSV, viral titers in mice deficient in Ccp1, Ccp2, Ccp3 or Ccp4 were similar to those of their wild-type counterparts (Supplementary Fig. 1c). Similar results were obtained by infection with VACV (Supplementary Fig. 1d–f).

Figure 1: Mice deficient in CCP5 or CCP6 are susceptible to infection with DNA viruses.
figure 1

(a) Immunoblot analysis of CCPs (left margin) and β-actin (loading control throughout) in BMDMs from wild-type mice (+/+) and mice deficient in Ccp1, Ccp2, Ccp3, Ccp4, Ccp5 or Ccp6 (−/−). (b) Survival of wild-type (WT) and mutant mice as in a (n = 10 per strain) after intravenous injection of HSV (1 × 106 plaque-forming units (PFU) per mouse). (c,d) Enzyme-linked immunosorbent assay (ELISA) of IFN-α (c) and IFN-β (d) in serum obtained from wild-type, Ccp5−/− and Ccp6−/− mice (n = 9 per strain) at various times (horizontal axes) after intravenous injection of HSV (1 × 107 PFU per mouse). (e,f) ELISA of IFN-α (e) and IFN-β (f) in serum obtained from wild-type, Ccp1−/−, Ccp2−/−, Ccp3−/− and Ccp4−/− mice (n = 7 per strain) after intravenous injection of HSV (1 × 107 PFU per mouse). (g) Viral titers in homogenates of brains from wild-type, Ccp5−/− and Ccp6−/− mice (n = 7 per strain) after intravenous injection of HSV (1 × 106 PFU per mouse). (hj) RT-PCR analysis of Ifnb mRNA in peritoneal macrophages (MΦ) (h), peripheral CD11chi DCs (i) and lung fibroblasts (j) obtained from wild-type, Ccp5−/− and Ccp6−/− mice (n = 5 per cell type) after intravenous injection of HSV (1 × 107 PFU per mouse); results are presented relative to those of wild-type cells immediately after infection with HSV (0 h). *P < 0.05, **P < 0.01 and ***P < 0.001 (Student's t-test). Data are representative of at least three independent experiments (mean and s.d. in cj).

We then assessed interferon-encoding mRNA in Ccp5- or Ccp6-deficient cells of the innate immune system following viral infection. As expected, after infection with HSV, Ccp5- or Ccp6-deficient cells had much lower expression of Ifnb mRNA (encoding interferon-β (IFN-β)) than did their wild-type counterparts after infection with HSV (Fig. 1h–j). In contrast, cells of the innate immune system from mice deficient in Ccp1, Ccp2, Ccp3 or Ccp4 had expression of Ifnb mRNA similar to that of their wild-type counterparts after infection with HSV (Supplementary Fig. 1g–i). Similar observations were obtained with VACV-infected Ccp5- or Ccp6-deficient cells (data not shown). These results indicated involvement of CCP5 and CCP6 in regulating immune responses to DNA viruses.

Requirement for CCP5 and CCP6 in Ifnb induction

We next sought to assess the innate immune responses of BMDMs to viral challenge by analyzing the activation of the transcription factor IRF3, which undergoes homodimerization after it is phosphorylated26. We observed that Ccp5- or Ccp6-deficient BMDMs failed to trigger activation (homodimerization) of IRF3 or production of IFN-β following infection with a DNA virus (Fig. 2a). However, after challenge with an RNA virus, Ccp5- or Ccp6-deficient BMDMs were able to induce IRF3 activation and IFN-β secretion (Fig. 2a). Consistent with that, transfection of herring testis DNA or Escherichia coli DNA did not induce activation of IRF3 in or secretion of IFN-β from Ccp5- or Ccp6-deficient BMDMs (Supplementary Fig. 2a). As expected, the synthetic RNA duplex poly(I:C) triggered IRF3 activation and IFN-β production in Ccp5−/− or Ccp6−/− BMDMs (Supplementary Fig. 2a). These data confirmed the proposal that CCP5 and CCP6 were specifically involved in the regulation of immune responses to DNA viruses but not those to RNA viruses.

Figure 2: CCP5 and CCP6 are required for the activation of IRF3 and induction of Ifnb.
figure 2

(a) ELISA of IFN-β (top) and immunoblot analysis of IRF3 dimerization (below) in BMDMs obtained from wild-type, Ccp5−/− and Ccp6−/− mice at 18 h (top) or 8 h (below) after mock infection (far left) or infection with HSV, VACV, vesicular stomatitis virus (VSV) or Sendai virus (SeV) (bottom), each at a multiplicity of infection (MOI) of 1. (b) Immunoblot analysis of IRF3 and CCP5 in Ccp5+/+ or Ccp5−/− BMDMs transfected with empty vector (EV) or vector encoding wild-type CCP5 (CCP5-wt) or CCP5-mut (grid above lanes), followed by no viral incubation (HSV −) or incubation for 8 h with HSV (MOI = 1) (HSV +). (c) RT-PCR analysis of Ifnb mRNA in wild-type, Ccp5−/− and Ccp6−/− BMDMs transfected with empty vector or vector encoding wild-type CCP5, CCP5-mut, wild-type CCP6 (CCP6-wt) or CCP6-mut (horizontal axis) and mock infected or infected for 8 h with HSV (MOI = 1) (key); results are presented relative to those of empty vector–transfected, mock-infected wild-type cells. (d) Microscopy of Ccp5+/+ and Ccp5−/− BMDMs transfected as in b and mock infected (bottom row) or infected for 24 h with green fluorescent protein–tagged HSV (HSV-GFP) (MOI = 1). Scale bars, 400 μm. (e) Frequency of GFP+ cells among wild-type, Ccp5−/− and Ccp6−/− BMDMs transfected as in c (horizontal axis) and infected as in d (key). (f,g) Immunoblot analysis of IRF3, CCP5 and CCP6 (f) and RT-PCR analysis of Ifnb mRNA (g) in wild-type BMDMs transfected for 24 h with empty vector or plasmid for the overexpression of wild-type CCP5, CCP5-mut, wild-type CCP6 or CCP6-mut (above lanes (f) or below plot (g)) and incubated for 8 h with HSV (MOI = 1) (HSV +) or not (HSV −); results in g are presented relative to those of empty vector–transfected, uninfected cells. (hj) Immunoblot analysis of IRF3, CCP5 and CCP6 (h,i) and RT-PCR analysis of Ifnb mRNA (j) in wild-type, Ccp5−/− and Ccp6−/− BMDMs left untreated (−) or treated (+) for 6 h with 10 μM CoCl2 (h,j) or 2 μM phenanthroline (Phen) (i,j) and incubated for 8 h with HSV (MOI = 1) or not; results in j are presented relative to those of uninfected wild-type cells. (k) Frequency of GFP+ cells among wild-type, Ccp5−/− and Ccp6−/− BMDMs treated with CoCl2 or phenanthroline as in h and infected for 24 h with HSV-GFP (MOI = 1) or not. *P < 0.05, **P < 0.01 and ***P < 0.001 (Student's t-test). Data are representative of at least three independent experiments (mean and s.d. in a,c,e,g,j,k).

To determine whether the enzymatic activity of CCP5 and CCP6 was required for the activation of IRF3, we generated enzymatically inactive mutants of CCP5 (CCP5-mut; with the substitutions H252S and E255Q)21 and CCP6 (CCP6-mut; with the substitutions H230S and E233Q)25. We then restored expression of CCP5 in Ccp5−/− BMDMs via wild-type CCP5 or CCP5-mut. We found that restoration with CCP5-mut resulted in no activation of IRF3 after infection with HSV, whereas wild-type CCP5 triggered IRF3 activation (Fig. 2b). In parallel, restoration of the expression of CCP6 in Ccp6−/− BMDMs via CCP6-mut resulted in no activation of IRF3 after infection with HSV (Supplementary Fig. 2b). Consequently, similar restoration via CCP5-mut or CCP6-mut did not induce Ifnb expression after challenge with HSV, whereas similar restoration with wild-type CCP5 or wild-type CCP6 did initiate elevated expression of Ifnb after challenge with HSV (Fig. 2c). Restoration via CCP5-mut and CCP6-mut did not eradicate HSV infection (Fig. 2d,e and Supplementary Fig. 2c). To further define the direct effect of CCP5 and CCP6 on IRF3 activation and interferon production, we overexpressed CCP5 or CCP6 in wild-type BMDMs. As expected, overexpression of CCP5 or CCP6 augmented IRF3 activation and IFN-β expression after challenge with HSV (Fig. 2f,g), and promoted restriction of HSV propagation (Supplementary Fig. 2d). These results indicated that the deglutamylation hydrolysis activity of CCP5 and CCP6 was required for the induction of IRF3 activation and IFN-β production during infection with a DNA virus.

CoCl2 is an agonist for carboxypeptidases27, and phenanthroline serves as an inhibitor of carboxypeptidases21. We noted that after infection with HSV, treatment with CoCl2 substantially enhanced IRF3 activation in wild-type BMDMs, whereas such IRF3 activation was abrogated in Ccp5−/− or Ccp6−/− cells infected with HSV and treated with CoCl2 (Fig. 2h). Conversely, treatment with phenanthroline hindered the activation of IRF3 in wild-type BMDMs (Fig. 2i). Consequently, after challenge with HSV, treatment with CoCl2 increased Ifnb expression and antiviral activity in wild-type BMDMs but not in Ccp5−/− or Ccp6−/− BMDMs (Fig. 2j,k and Supplementary Fig. 2e,f). In contrast, treatment with phenanthroline decreased Ifnb expression and antiviral activity in wild-type BMDMs but not in Ccp5−/− or Ccp6−/− BMDMs (Fig. 2j,k and Supplementary Fig. 2e,f). In addition, treatment with CoCl2 or phenanthroline did not affect Ifnb expression after infection of wild-type BMDMs or even Ccp5−/− or Ccp6−/− BMDMs with vesicular stomatits virus (Supplementary Fig. 2g). This confirmed the proposal that CCP5- or CCP6-mediated deglutamylation participated in the regulation of antiviral responses to DNA viruses.

cGAS as a substrate for CCP5 and CCP6

We next sought to identify candidate substrates for deglutamylation during antiviral responses to DNA viruses. The antibody GT335 specifically recognizes the branch points of glutamate side chains and detects all glutamylated forms of target proteins21. After immunoblot analysis, two bands of around 58 kilodaltons (kDa) and 50 kDa appeared in the lanes for lysates of BMDMs and dendritic cells (DCs) from CCP5-deficient mice (Fig. 3a). The upper band (58 kDa) was undetectable and the lower band (50 kDa) had diminished intensity in the corresponding lanes for lysates of BMDMs and DCs from their wild-type littermates (Fig. 3a). Thus, these two bands might have represented proteins that could be potential substrates for CCP5. We next generated wild-type CCP5 and CCP5-mut and immobilized them on affinity resin for neutral or basic proteins, for passage through BMDM lysates by affinity chromatography. The eluted fractions were resolved by SDS-PAGE, followed by silver staining (Fig. 3b). These two bands were present in the gel analyzing CCP5-mut and were cut for mass spectrometry. The lower, 50-kDa band was identified as α-tubulin (data not shown), a reported CCP5 substrate21. The upper, 58-kDa band was identified as cGAS (Supplementary Fig. 3a), a previously unknown candidate substrate for CCP5.

Figure 3: cGAS is a substrate for CCP5 and CCP6.
figure 3

(a) Immunoblot analysis of glutamylation (assessed with GT335; top) and of CCP5 (below) in Ccp5+/+ and Ccp5−/− BMDMs and DCs. (b) Affinity chromatography of glutathione S-transferase (GST) alone or GST-tagged wild-type CCP5 or CCP5-mut immobilized on affinity resins, assessed by the addition of BMDM lysates, resolution of the eluted fractions by SDS-PAGE and silver staining (presented here), followed by liquid chromatography–linear trap quadrupole tandem mass spectrometry of the 58-kDa and 50-kDa bands (Supplementary Fig. 3a). Left margin, size in kDa. (c) GST-precipitation assay of GST or GST-tagged wild-type CCP5 or CCP5-mut incubated for 4 h with BMDM lysates, assessed by immunoblot analysis of cGAS, GST-tagged CCP5 or GST among the precipitates (GST ppt) and of cGAS in lysates without precipitation (below). (d) Confocal microscopy of wild-type BMDMs stained with antibody to (anti-) β-tubulin, anti-cGAS and anti-CCP5. Area outlined at top right is enlarged 6× in row below. Scale bar, 10 μm. (e) Immunoassay of wild-type BMDMs left untreated (−) or treated (+) for 6 h with 10 μM CoCl2 or 2 μM phenanthroline (above lanes), assessed by immunoprecipitation (IP) with immunoglobulin G (IgG), as a control, or with anti-CCP5 and immunoblot analysis of immunoprecipitates (top two blots) and total lysates (below) with anti-cGAS or anti-CCP5. (f) GST-precipitation assay of GST or GST-tagged wild-type CCP6 or CCP6-mut incubated for 4 h with BMDM lysates, assessed by immunoblot analysis of cGAS, GST-tagged CCP6 or GST among the precipitates (GST ppt) and of cGAS in lysates without precipitation (below). (g) Immunoassay of wild-type BMDMs treated as in e, assessed by immunoprecipitation with IgG or anti-cGAS and immunoblot analysis of immunoprecipitates with GT335 or anti-cGAS and of total lysates with anti-cGAS. (h) Immunoassay of Ccp5−/− and Ccp6−/− BMDMs by immunoprecipitation with IgG or anti-cGAS and immunoblot analysis of immunoprecipitates with GT335 or anti-cGAS and of total lysates with anti-CCP5, anti-CCP6 or anti-cGAS. (i) Immunoassay of wild-type BMDMs left untransfected (−) or transfected (+) for 24 h with vector encoding Flag-tagged CCP5 or CCP6, assessed by immunoprecipitation with IgG or anti-cGAS and immunoblot analysis of immunoprecipitates with GT335 or anti-cGAS and of total lysates with anti-Flag or anti-cGAS. (j) Immunoassay of wild-type BMDMs left untransfected (−) or transfected (+) for 24 h with vector encoding Flag-tagged wild-type cGAS (cGAS-wt), cGAS(E272A), cGAS(E302A) or cGAS(E272A,E302A), assessed by immunoprecipitation with IgG or anti-Flag and immunoblot analysis with polyE, GT335 or anti-Flag and of total lysates with anti-Flag. Data are representative of at least three independent experiments.

CCP5-mut was able to precipitate cGAS from BMDM lysates (Fig. 3c). Colocalization of cGAS with CCP5 in the cytoplasm of BMDMs was visualized by confocal microscopy (Fig. 3d). Moreover, the colocalization of cGAS with CCP5 did not overlay with microtubular structures (Fig. 3d). Staining of cGAS displayed punctate structures (Fig. 3d), as previously observed15. Moreover, the CCP inhibitor phenanthroline augmented the association between CCP5 and cGAS, whereas treatment with CoCl2 abrogated this association (Fig. 3e). Notably, deletion of CCP5's enzymatic domain (residues 160–424) abolished the interaction between CCP5 and cGAS (Supplementary Fig. 3b), which suggested that the interaction of CCP5 with cGAS required the enzymatic activity of CCP5. As noted for CCP5, the interaction of CCP6 with cGAS was also verified (Fig. 3f and Supplementary Fig. 3c,d). The interaction of CCP6 with cGAS was also dependent on its enzymatic activity (Supplementary Fig. 3e). Collectively, these results indicated that the association of cGAS with CCP5 or CCP6 required the enzymatic activity of CCP5 or CCP6, which suggested that cGAS might be a potential substrate for CCP5 and CCP6.

Notably, recombinant CCP5-mut or CCP6-mut did not precipitate together with recombinant cGAS (data not shown), whereas CCP5-mut or CCP6-mut bound to endogenous cGAS (Fig. 3c,f). These results suggested that CCP5 or CCP6 might need to bind to glutamylated cGAS for their hydrolysis. Indeed, cGAS was glutamylated in BMDMs (Fig. 3g). Moreover, treatment with the CCP inhibitor phenanthroline substantially enhanced the glutamylation of cGAS, whereas the CCP agonist CoCl2 rendered the glutamylation of cGAS undetectable (Fig. 3g). More notably, more glutamylation of cGAS was observed in Ccp5−/− and Ccp6−/− BMDMs than in wild-type cells (Fig. 3h). In contrast, overexpression of CCP5 or CCP6 decreased the glutamylation of cGAS (Fig. 3i). Additionally, CCP1, CCP2, CCP3 and CCP4 failed to precipitate together with cGAS in BMDM lysates (data not shown). Finally, deficiency in Ccp1, Ccp2, Ccp3 or Ccp4 did not affect the glutamylation of cGAS (data not shown). Therefore, in these conditions, cGAS was a substrate for both CCP5 and CCP6 but not for other CCPs.

Deglutamylation of cGAS by CCP5 and CCP6

Glutamylation is the ATP-dependent addition of glutamates catalyzed by glutamylases of the TTLL family, either singly or sequentially20. As noted above, the antibody GT335 can recognize branch-point glutamates, while the polyglutamylation-specific antibody polyE is able to detect three or more consecutive branched glutamates21. On the basis of amino acid mapping and structural prediction of cGAS, we identified two putative glutamylation site on mouse cGAS (Supplementary Fig. 3f–h). We then substituted alanine for the glutamic acid at positions 272 (E272A) and 302 (E302A), separately or together, and transfected vector encoding these mutants into BMDMs. We found that cGAS(E272A) had branched glutamates that were detected only by GT335, whereas cGAS(E302A) had branched glutamates that were detected by both GT335 and polyE (Fig. 3j). Moreover, the double mutant cGAS(E272A,E302A) lacked branched glutamates (Fig. 3j). These results suggested that cGAS had three or more glutamates added at Glu272 and two or fewer glutamates at Glu302.

We next investigated how cGAS was deglutamylated by CCP5 and CCP6. We cotransfected BMDMs with vector encoding cGAS(E272A) or cGAS(E302A), together with vector encoding CCP5 or CCP6, and assessed the cells by immunoblot analysis. We observed that overexpression of CCP5 abolished the GT335 signals for BMDMs transfected to express cGAS(E272A) but not in those transfected to express cGAS(E302A) (Supplementary Fig. 4a). In contrast, overexpression of CCP6 eliminated the GT335 signals for BMDMs transfected to express cGAS(E302A) but not those transfected to express cGAS(E272A) (Supplementary Fig. 4b). Moreover, Ccp5-deficient BMDMs overexpressing cGAS(E272A) displayed stronger GT335 signals than those in similarly transfected wild-type cells, whereas overexpression of cGAS(E302A) had no such effect (Supplementary Fig. 4c). In addition, overexpression of cGAS(E302A) in Ccp6-deficient cells resulted in stronger GT335 and polyE signals than those of similarly transfected wild-type cells, while overexpression of cGAS(E272A) did not influence GT335 signals of immunoprecipitated cGAS (Supplementary Fig. 4d). As expected, CCP5-mut associated with cGAS(E272A) but not with cGAS(E302A) in lysates of wild-type BMDMs transfected to express these (Supplementary Fig. 4e). In contrast, CCP6-mut interacted with cGAS(E302A) but not with cGAS(E272A) in lysates of wild-type BMDMs transfected to express these (Supplementary Fig. 4f). Therefore, CCP5 was responsible for hydrolyzing the glutamate chain of cGAS at E302, and CCP6 removed the glutamate chain of cGAS at Glu272.

Differential glutamylation of cGAS by TTLL4 and TTLL6

To further determine how the glutamylases catalyzed the glutamylation of cGAS, we screened nine glutamylases for their interactions with cGAS. We found that only TTLL4 and TTLL6 associated with cGAS (Fig. 4a). Furthermore, TTLL4 and TTLL6 showed high constitutive expression in cells of the mouse innate immune system, such as BMDMs, DCs and fibroblasts (Supplementary Fig. 4g). The interaction of cGAS with TTLL4 and TTLL6 was verified by a precipitation assay (Fig. 4b,c). cGAS localized together with TTLL4 and TTLL6 in the cytoplasm of BMDMs, and this colocalization did not overlay with microtubular structures (Fig. 4d,e). These observations suggested that TTLL4- and TTLL6-mediated modifications of cGAS were not restricted to microtubular structures. It has been reported that TTLL4 is a monoglutamylase and that TTLL6 polyglutamates tubulins23. After incubation with TTLL4 in an in vitro glutamylation system, GT335 (glutamate) signals appeared (by immunoblot analysis) only for samples containing wild-type cGAS or cGAS(E272A), not those containing cGAS(E302A) (Fig. 4f), which suggested that cGAS was monoglutamylated by TTLL4 at Glu302. In contrast, following incubation with TTLL6, both GT335 and polyE signals were present (by immunoblot analysis) for samples containing wild-type cGAS or cGAS(E302A) but not those containing cGAS(E272A) (Fig. 4g), which indicated that cGAS was polyglutamylated by TTLL6 at Glu272.

Figure 4: cGAS is glutamylated differentially by TTLL4 and TTLL6.
figure 4

(a) GST-precipitation assay of GST or GST-tagged members of the TTLL family (above lanes) incubated for 4 h with BMDM lysates, assessed by immunoblot analysis of cGAS, GST-tagged TTLL glutamylases or GST among the precipitates (GST ppt) and of cGAS in lysates without precipitation (below). (b) GST-precipitation assay of GST or increasing amounts (wedge) of GST-cGAS incubated for 4 h with BMDM lysates, assessed by immunoblot analysis of TTLL4, TTLL6, GST-tagged cGAS or GST among the precipitates (GST ppt) and of TTLL4 and TTLL6 in lysates without precipitation (below). (c) Immunoassay of wild-type BMDMs subjected to immunoprecipitation with IgG or anti-cGAS, assessed by immunoblot analysis of immunoprecipitates (top three blots) or total lysates (below) with anti-TTLL4, anti-TTLL6 or anti-cGAS. (d,e) Confocal microscopy of wild-type BMDMs stained with anti-β-tubulin, anti-cGAS and anti-TTLL4 (d) or anti-TTLL6 (e). Area outlined at top right is enlarged 5× in row below. Scale bars, 10 μm. (f,g) GST-precipitation assay of Flag-tagged TTLL4 (f) or TTLL6 (g) overexpressed for 24 h in mouse embryonic fibroblasts (+) or not (−), followed by immunoprecipitation with anti-Flag and elution with Flag peptides, then incubation of immunoprecipitates for 1 h with GST or GST-tagged wild-type cGAS, cGAS(E272A) or cGAS(E302A) in glutamylation buffer, assessed by immunoblot analysis of the precipitates with polyE, GT335 or anti-GST (GST ppt) and of total lysates (below) with anti-TTL4 (f) or anti-TTL6 (g). (h,i) Immunoassay of Ttll4+/+ and Ttll4−/− BMDMs (h) or Ttll6+/+ and Ttll6−/− BMDMs (i) by immunoprecipitation with IgG or anti-cGAS and immunoblot analysis of the immunoprecipitates with polyE, GT335 or anti-cGAS (top three blots) and of total lysates (below) with anti-cGAS and anti-TTL4 (h) or anti-TTL6 (i). (j,k) Immunoassay of Ttll4+/+ and Ttll4−/− BMDMs (j) or Ttll6+/+ and Ttll6−/− BMDMs (k) left untransfected (−) or transfected (+) for 24 h to express Flag-tagged wild-type cGAS, cGAS(E272A) or cGAS(E302A), assessed by immunoprecipitation with IgG or anti-Flag and immunoblot analysis of the immunoprecipitates with polyE, GT335 or anti-Flag (top three blots) and of total lysates (below) with anti-Flag and anti-TTL4 (j) or anti-TTL6 (k). Data are representative of at least three independent experiments.

Through domain mapping, we found that TTLL4 interacted with a fragment consisting of residues 280–320 of cGAS, whereas TTLL6 associated with a fragment consisting of residues 240–280 of cGAS (Supplementary Fig. 4h). Additionally, TTLL4 and TTLL6 associated with wild-type cGAS or either cGAS mutant (Supplementary Fig. 4i,j), which suggested that the interaction of cGAS with TTLL4 and TTLL6 was independent of the glutamylation of cGAS. We next generated Ttll4- orTtll6-deficient mice via CRISPR-Cas9 technology (Supplementary Fig. 4k). We noted that the GT335 signals for immunoprecipitated cGAS were substantially diminished in Ttll4−/− BMDMs, whereas polyE signals were unchanged, relative to those in wild-type cells, as detected by immunoblot analysis (Fig. 4h). In contrast, polyE signals for immunoprecipitated cGAS were undetectable and GT335 signals were substantially diminished in Ttll6−/− BMDMs, relative to those in wild-type cells, as detected by immunoblot analysis (Fig. 4i). Finally, restoration of Ttll4−/− BMDMs with cGAS(E272A) did not result in GT335 signals, whereas Ttll4−/− BMDMs restored with cGAS(E302A) displayed GT335 signals, as detected by immunoblot analysis (Fig. 4j). In parallel, restoration of Ttll6−/− BMDMs with cGAS(E272A) resulted in GT335 signals of cGAS but a lack of polyE signals, whereas restoration of Ttll6−/− BMDMs with cGAS(E302A) impaired both GT335 signals and polyE signals of cGAS, as detected by immunoblot analysis (Fig. 4k). cGAS was therefore glutamylated differentially by TTLL4 and TTLL6.

cGAS polyglutamylation suppresses its DNA-binding ability

After binding DNA, cGAS undergoes a conformational change to catalyze cGAMP synthesis28,29. To investigate whether glutamylation of cGAS affected its DNA-binding ability, we performed a DNA-precipitation assay using cell lysates pretreated with the CCP agonist CoCl2 or the inhibitor phenanthroline. After treatment with CoCl2, more cGAS associated with double-stranded DNA (dsDNA) (Fig. 5a). In contrast, after treatment with phenanthroline, less cGAS was precipitated by dsDNA (Fig. 5a). To further analyze the DNA-binding ability of cGAS in vivo, we transfected dsDNA into BMDMs, followed by incubation of the cells with CoCl2 or phenanthroline. Consistent with the results reported above, in the presence of phenanthroline, cGAS associated with less dsDNA in BMDMs, whereas it bound much more dsDNA in the presence of CoCl2 (Fig. 5b). These results suggested that glutamylation of cGAS impeded its DNA-binding ability.

Figure 5: TTLL6-mediated polyglutamylation of cGAS suppresses its DNA-binding ability.
figure 5

(a) DNA-precipitation assay of untreated lysates alone (far left) or BMDM lysates pretreated (+) or not (−) with 10 μM CoCl2 or 2 μM phenanthroline (above lanes) and incubated with resin-conjugated 200–base pair dsDNA, assessed by immunoblot analysis of the precipitates (DNA ppt) and total lysates (below) with anti-cGAS. (b) Chromatin-immunoprecipitation assay (with anti-cGAS) of BMDMs treated with vehicle, CoCl2 or phenanthroline (horizontal axis) and transfected with 200–base pair dsDNA. (c) DNA-precipitation assay of lysates of BMDMs transfected to express Flag-tagged wild-type cGAS, cGAS(E272A) or cGAS(E302A) and incubated with resin-conjugated 200–base pair dsDNA (+) or not (−), assessed by immunoblot analysis of the precipitates (DNA ppt) and total lysates (below) with anti-Flag. (d) DNA-precipitation assay of Ccp5−/− and Ccp6−/− BMDM lysates incubated with resin-conjugated 200–base pair dsDNA (+) or not (−), assessed by immunoblot analysis of the precipitates (DNA ppt) with anti-cGAS and of total lysates (below) with anti-cGAS, anti-CCP5 or anti-CCP6. (e) Confocal microscopy of Ccp6+/+ and Ccp6−/− BMDMs transfected with fluorescein isothiocyanate (FITC)-conjugated 50–base pair dsDNA (left), and colocalization of cGAS with the labeled DNA at left (Pearson's correlation coefficient; right). Area outlined (top right image of each pair at left) is enlarged 5× in row below. Scale bars (left), 10 μm. (f) Chromatin-immunoprecipitation assay of Ccp6+/+ and Ccp6−/− BMDMs transfected as in c and further transfected for 4 h with 200–base pair dsDNA. (g) DNA-precipitation assay of Ttll4−/− and Ttll6−/− BMDM lysates incubated with resin-conjugated 200–base pair dsDNA, assessed by immunoblot analysis of the precipitates (DNA ppt) with anti-cGAS and of total lysates (below) with anti-cGAS, anti-TTLL4 or anti-TTLL6. (h) Chromatin-immunoprecipitation assay of Ttll6+/+ and Ttll6−/− BMDMs transfected as in c and further transfected for 4 h with 200–base pair dsDNA. (i) DNA-precipitation assay of TTLL6 incubated for 1 h with GST, GST-tagged wild-type cGAS, cGAS(E272A) or cGAS(E302A) in glutamylation buffer, assessed by immunoblot analysis of the precipitates (DNA ppt) with anti-cGAS and of total lysates (below) with anti-GST or anti-TTLL6. (j) Bead-based proximity assay of 200–base pair dsDNA (50 ng/μl) incubated with various concentrations (horizontal axis) of GST or wild-type or mutant cGAS not glutamylated or glutamylated in vitro by TTLL6 (+TTLL6) (left) or cGAS as at left (50 ng/μl) mixed with various concentrations (horizontal axis) of biotin-labeled 200–base pair dsDNA (Biotin-200-mer) (right); results are presented in units generated by the assay system. (k) RT-PCR analysis of Ifnb mRNA in Ttll6+/+ and Ttll6−/− BMDMs mock infected or incubated for 8 h with HSV (MOI = 1); results are presented relative to those of mock-infected wild-type cells. *P < 0.01 and **P < 0.001 (Student's t-test). Data are representative of at least three independent experiments (mean and s.d. in b,e,f,h,j,k).

We next sought to determine which glutamic acid's glutamylation affected the DNA-binding ability of cGAS. We transfected dsDNA, along with vector encoding wild-type cGAS or either of the two cGAS mutants, into BMDMs for precipitation assays. More cGAS(E272A) than wild-type cGAS bound dsDNA, whereas cGAS(E302A) and wild-type cGAS bound dsDNA to a similar degree (Fig. 5c). These data suggested that glutamylation of Glu272 hindered the binding of DNA to cGAS. Additionally, cGAS did not bind dsDNA in Ccp6−/− BMDMs (Fig. 5d,e). However, similar amounts of cGAS bound dsDNA in Ccp5−/− and wild-type BMDMs (Fig. 5d and Supplementary Fig. 5a). Accordingly, cGAS(E302A) overexpressed in CCP6−/− BMDMs failed to bind dsDNA (Fig. 5f), whereas cGAS(E272A) overexpressed in CCP5−/− BMDMs bound the same amount of dsDNA as in CCP5+/+ cells (Supplementary Fig. 5b). The polyglutamylation of cGAS at Glu272, therefore, suppressed its DNA-binding ability.

Given that TTLL6 catalyzed the polyglutamylation of cGAS at Glu272, we used Ttll6−/− cells to further investigate the dsDNA-binding ability mediated by the polyglutamylation of cGAS. We noted that more cGAS was able to bind dsDNA in Ttll6−/− BMDMs than in wild-type cells, whereas similar amounts of cGAS were precipitated by dsDNA in Ttll4−/− BMDMs and wild-type cells (Fig. 5g). Consistent with that, overexpression of cGAS(E302A) in Ttll6−/− BMDMs enhanced its DNA-binding ability, as did overexpression of cGAS(E272A), whereas overexpression of cGAS(E302A) in Ttll6+/+ BMDMs did not affect its DNA-binding affinity (Fig. 5h). However, Ttll4 deficiency did not affect the DNA-binding ability of cGAS (Supplementary Fig. 5c). Finally, cGAS(E272A) pre-incubated with TTLL6 in vitro was precipitated by dsDNA as efficiently as recombinant wild-type cGAS was, but wild-type cGAS and cGAS(E302A) lost their DNA-binding ability after incubation with TTLL6 in vitro (Fig. 5i,j). In contrast, incubation with TTLL4 did not alter the DNA-binding ability of cGAS(E272A) or cGAS(E302A) (Supplementary Fig. 5d). These results indicated that the TTLL6-mediated polyglutamylation of cGAS at Glu272 abolished its DNA-binding ability.

As a consequence of the results reported above, Ttll6−/− BMDMs exhibited higher expression of Ifnb mRNA than that of Ttll6+/+ cells after infection with HSV (Fig. 5k). Moreover, Ttll6 deficiency substantially suppressed the propagation of HSV (Supplementary Fig. 5e). We then generated BMDMs deficient in the gene encoding cGAS (Mb21d1; called 'Cgas' here) via a CRISPR-Cas9 approach, as reported7, and restored cGAS expression via wild-type cGAS or cGAS(E272A) in these Cgas−/− cells. We found that restoration with cGAS(E272A) accelerated interferon expression relative to its expression after restoration with wild-type cGAS and consequently suppressed viral amplification (Supplementary Fig. 5f,g). Therefore, TTLL6-mediated polyglutamylation of cGAS at Glu272 inhibited its DNA-binding ability, which led to suppression of its activity against viral infection.

Blockade of cGAS synthase activity via monoglutamylation

We further explored how the synthase activity of cGAS was regulated by its glutamylation. We observed, by thin-layer chromatography (TLC), that treatment with CoCl2 substantially enhanced the formation of cGAMP catalyzed by cGAS (Fig. 6a). Conversely, treatment with phenanthroline abrogated the synthesis of cGAMP (Fig. 6a), which suggested that glutamylation of cGAS participated in the negative regulation of its enzymatic activity. cGAS(E302A) promoted much more formation of cGAMP than did wild-type cGAS (Fig. 6b), which suggested that monoglutamylation of cGAS at Glu302 inhibited its synthase activity.

Figure 6: TTLL4-mediated monoglutamylation of cGAS blocks its synthase activity.
figure 6

(a) TLC assay of cGAS immunoprecipitated (with IgG or anti-cGAS) from wild-type BMDMs pretreated with 10 μM CoCl2 or 2 μM phenanthroline (+) or not (−), in the presence of 50–base pair dsDNA, [α32p]ATP, GTP and MgCl2 (top left); dashed lines indicate origin and solvent fronts (right margin). Right, ratio of cGAMP to cGAS in the samples at left. (b) TLC assay of BMDMs transfected for 24 h with Flag-tagged wild-type cGAS or cGAS(E302A), assessed by immunoprecipitation with anti-Flag (left), and ratio of cGAMP to cGAS in the samples at left (right). (c,d) TLC assay (as in a) of Ccp5+/+ and Ccp5−/− BMDMs (c) or Ttll4+/+ and Ttll4−/− BMDMs (d). Right, ratio of cGAMP to cGAS in the samples at left. (e) GST-precipitation assay and TLC assay as in a of TTLL4 incubated for 1 h with various combinations (above lanes) of dsDNA and GST or GST-tagged wild-type cGAS or cGAS(E302A) in glutamylation buffer. Right, ratio of cGAMP to cGAS in the samples at left. (f) RT-PCR analysis of Ifnb mRNA in Ttll4+/+ and Ttll4−/− BMDMs mock infected or incubated for 8 h with HSV (MOI = 1); results are presented relative to those of mock-infected wild-type cells. (g) Microscopy (left) of Ttll4+/+ and Ttll4−/− BMDMs mock infected or incubated for 24 h with HSV-GFP (MOI = 1), and frequency of GFP+ cells among those BMDMs (right). Scale bars (left), 400 μm. (h) RT-PCR analysis of Ifnb mRNA in Cgas+/+ and Cgas−/− BMDMs transfected with empty vector or vector encoding wild-type cGAS or cGAS(E302A) and mock infected or incubated for 8 h with HSV (MOI = 1); results are presented relative to those of empty vector–transfected, mock-infected wild-type cells. (i) Microscopy (left) of Cgas+/+ and Cgas−/− BMDMs transfected as in h and mock infected or incubated for 24 h with HSV-GFP (MOI = 1) , and frequency of GFP+ cells among those BMDMs (right). Scale bars (left), 400 μm. Bottom left (ae), immunoassay of the samples above, subjected to immunoprecipitation as above, assessed by immunoblot analysis of immunoprecipitates (IP) or total lysates (below) with anti-cGAS (a,c,d), anti-Flag (b) anti-CCP5 (c), anti-TTL4 (d,e) and/or anti-GST (e). *P < 0.05, **P < 0.01 and ***P < 0.001 (Student's t-test). Data are representative of at least three independent experiments (mean and s.d.).

We observed that CCP5 removed the monoglutamylation of cGAS at Glu302 (Supplementary Fig. 4c). As expected, Ccp5 deficiency disrupted the synthase activity of cGAS (Fig. 6c). We also found that the monoglutamylation of cGAS at Glu302 was catalyzed by TTLL4 (Fig. 4f). Therefore, Ttll4 deficiency resulted in significantly augmented cGAMP synthesis (Fig. 6d). In addition, cGAS(E302A) still displayed cGAMP-synthetic activity after incubation with TTLL4 in vitro, whereas wild-type cGAS lost its cGAMP-synthetic ability after incubation with TTLL4 (Fig. 6e). These data indicated that monoglutamylation of cGAS at Glu302 catalyzed by TTLL4 impeded the synthase activity of cGAS for cGAMP synthesis.

As expected, Ttll4−/− BMDMs displayed higher expression of Ifnb mRNA than that of Ttll4+/+ cells after infection with HSV (Fig. 6f). Consequently, Ttll4 deficiency substantially suppressed the propagation of HSV (Fig. 6g). We then restored cGAS expression via wild-type cGAS or cGAS(E302A) in Cgas-deficient BMDMs. We found that restoration with cGAS(E302A) promoted Ifnb expression and consequently restricted viral propagation (Fig. 6h,i). Thus, TTLL4-mediated monoglutamylation of cGAS at Glu302 blocked its synthase activity, which resulted in suppression of its antiviral activity.

Modulation of antiviral immunity via cGAS glutamylation states

To further confirm the proposal of a role for glutamylation and deglutamylation in innate immune responses to DNA viruses, we generated Ccp5−/−Ccp6−/− double-deficient mice by crossing Ccp5−/− mice with Ccp6−/− mice and found that the Ccp5−/−Ccp6−/− mice lacked expression of CCP5 and CCP6 (Fig. 7a). As expected, Ccp5−/−Ccp6−/− BMDMs showed augmented glutamylation of cGAS relative to that of Ccp5−/− or Ccp6−/− cells (Fig. 7b,c). Consequently, Ccp5−/−Ccp6−/− mice had a blockade in the production of IFN-β (Fig. 7d) and were highly susceptible to infection with HSV (Fig. 7e), whereas Ccp5−/− or Ccp6−/− mice produced some IFN-β (Fig. 7d) but had compromised resistance to infection with HSV (Fig. 7e). These data indicated that CCP5- and CCP6-mediated deglutamylation of cGAS enhanced the production of type I interferons and eradication of DNA viruses.

Figure 7: Glutamylation and deglutamylation of cGAS tightly modulates antiviral immunity.
figure 7

(a) Immunoblot analysis of CCP5 and CCP6 in BMDMs from mice deficient in Ccp5 and/or Ccp6 (above lanes; n = 7 mice per strain in ae). (b) Immunoprecipitation of cGAS from BMDMs from mice deficient in Ccp5 and/or Ccp6 (above lanes), assessed by immunoblot analysis with GT335 or anti-cGAS. (c) Microscopy of BMDMs deficient in Ccp5 and/or Ccp6 (left margin) stained with anti-cGAS and GT335 (left), and colocalization of cGAS and GT335 signals in those cells (right). Scale bars (left), 10 μm. (d) ELISA of IFN-β in serum from mice deficient in Ccp5 and/or Ccp6 (key), at 8 h after intravenous injection of vehicle (Mock) or HSV (1 × 107 PFU per mouse). (e) Viral titers in homogenates of brains from mice deficient in Ccp5 and/or Ccp6 (key), analyzed immediately (0 d) or 2 d after intravenous injection of HSV (1 × 106 PFU per mouse). (f) Immunoblot analysis of TTLL4 and TTLL6 in BMDMs from mice deficient in Ttll4 and/or Ttll6 (above lanes; n = 7 mice per strain in fj). (g) Immunoprecipitation of cGAS from BMDMs deficient in Ttll4 and/or Ttll6 (above lanes), assessed by immunoblot analysis with GT335 or anti-cGAS. (h) Microscopy of BMDMs from mice deficient in Ttll4 and/or Ttll6, stained with anti-cGAS and GT335 in those cells (left), and colocalization of cGAS and GT335 signals (right). Scale bars (left), 10 μm. (i) ELISA of IFN-β in serum from mice deficient in Ttll4 and/or Ttll6 (key) at 8 h after infection as in d. (j) Viral titers in homogenates of brains deficient in Ttll4 and/or Ttll6, analyzed immediately (at 0 d) or 2 d after infection as in d. *P < 0.05 and **P < 0.01 (Student's t-test). Data are representative of at least three independent experiments (mean and s.d. in ce,hj).

We also established Ttll4−/−Ttll6−/− double-deficient mice by crossing Ttll4−/− mice with Ttll6−/− mice and found that Ttll4−/−Ttll6−/− mice lacked expression of TTLL4 and TTLL6 (Fig. 7f). We observed that Ttll4−/−Ttll6−/− BMDMs lacked glutamylation of cGAS, whereas Ttll4−/− or Ttll6−/− cells exhibited some glutamylation of cGAS (Fig. 7g,h). Moreover, Ttll4−/−Ttll6−/−mice produced greater amounts of IFN-β than did Ttll4−/− or Ttll6−/− mice (Fig. 7i) and were much more resistant to infection with HSV than were Ttll4−/− or Ttll6−/− mice (Fig. 7j). Thus, TTLL4- and TTLL6-mediated glutamylation of cGAS suppressed the production of type I interferons and clearance of DNA viruses. Glutamylation and deglutamylation of cGAS, therefore, tightly regulated the immune response during infection with DNA viruses.

Discussion

cGAS is a member of the nucleotidyltransferase family that catalyzes the synthesis of cGAMP from GTP and ATP in the presence of dsDNA15,30. In the resting state, the catalytic regions of cGAS are embedded inside cGAS, which hinders the catalytic generation of cGAMP from ATP and GTP16,31,32. Once bound to DNA, cGAS undergoes a conformational change that renders the catalytic pocket of cGAS accessible for cGAMP synthesis28,29,33. Collectively, such studies suggest that engagement of DNA is a prerequisite for activation of the nucleotidyltransferase activity of cGAS. Here we found that the activity of cGAS was tightly regulated by glutamylation and deglutamylation, which had a critical role in the modulation of innate immune responses.

Post-translational modifications act as regulatory signals for control of the stability, localization, activity and function of proteins20,25,34,35. Glutamylation is a modification that adds monoglutamic or polyglutamic acid residues to target proteins in an ATP-dependent manner20,21,35. Glutamylation is catalyzed by the TTLL family of glutamylases22. Gutamylation was originally identified on tubulins as a dominant tubulin modification in the adult brain of mammals36. Tubulins and nucleosome-assembly proteins are well-known substrates for polyglutamylation21,35. In this study, we found that TTLL4 and TTLL6 glutamylated the previously unknown substrate cGAS differentially. TTLL6-mediated polyglutamylation of cGAS at Glu272 impeded its DNA-binding ability, whereas TTLL4-mediated monoglutamylation of cGAS at Glu302 blocked its synthase activity. Of note, TTLL glutamylases differ in their 'preferences' for their substrates, as well as for the catalysis of glutamate chains, side-chain initiation or elongation23. Thus, TTLL glutamylases direct their effects to different target substrates.

Glutamylation is a reversible modification, and deglutamylation is hydrolyzed by CCPs21,34,37. Members of the CCP family have specific expression profiles in different tissues and are involved in a variety of physiological functions21,38. Members of this family have 'preferred' enzymatic specificities for hydrolyzing deglutamylation. Among these members, CCP1, CCP4 and CCP6 remove the shortening of penultimate polyglutamate chains of α-tubulin, while CCP5 specifically hydrolyzes the branching site glutamate21. Our results indicated non-redundant roles for CCP5 and CCP6 in the regulation of cGAS activity. Of note, we observed that the amount of glutamylated cGAS diminished rapidly after infection with HSV. Meanwhile, the amount of TTLL4 and TTLL6 protein decreased substantially after infection with HSV, whereas the amount of CCP5 and CCP6 remained unchanged during the process of infection. Moreover, after infection with HSV, cGAS associated with larger amounts of CCPs but smaller amounts of TTLL glutamylases. These observations suggested that the activity of cGAS was tightly regulated by its dynamic glutamylation and deglutamylation modifications during infection with HSV.

cGAS detects cytosolic DNA in a sequence-independent manner, which elicits the cGAS-STING pathway to prime innate immune responses to various DNA viruses, retroviruses and even bacterial pathogens39,40,41. Manipulating the glutamylation state of cGAS could thus potentially either initiate the cGAS-STING pathway to eradicate DNA viruses or shut down this pathway to prevent excessive immune responses. Together our findings have shown that the glutamylation and deglutamylation of cGAS regulated its activity and, as a consequence, controlled the response to DNA viruses. Therefore, manipulation of the glutamylation state of cGAS might be of clinical importance.

Methods

Antibodies and reagents.

Antibodies used were as follows: anti-CCP1 (LM-1A7), anti-CCP2 (S-13), anti-CCP3 (S-15), anti-CCP4 (T-17), anti-CCP5 (N-18), anti-CCP6 (N-14), anti-TTLL4 (S-14), anti-cGAS (N-17), anti-IRF3 (D-3), anti-β-tubulin, anti-TBK1 (M-375), anti-STING (M-12) and anti-Myc (9E10) (all from Santa Cruz Biotechnology); polyclonal anti-TTLL4 (PAB22002) and polyclonal anti-TTLL6 (H00284076-K) (both from Abnova); antibody to the polyglutamylation modification (GT335) and antibody to the polyglutamate chain (polyE) (IN105) (both from Adipogen); and polyclonal anti-GST (G7781), anti-Flag (M1), anti-His (6AT18) and anti-β-actin (SP124) (all from Sigma-Aldrich). The secondary antibody donkey anti–rabbit IgG conjugated to Alexa Fluor 488 (A11008), Alexa Fluor 594 (A11012) or Alexa Fluor 405 (A31556) and secondary antibody donkey anti–mouse IgG conjugated to Alexa Fluor 488 (A11029) or Alexa Fluor 594 (A11032) were purchased from Molecular Probes. Paraformaldehyde (PFA), phenanthroline, CoCl2 and DAPI (4,6-diamidino-2-phenylindole) were from Sigma-Aldrich.

Cells and culture.

BMDMs were generated as described below. Bone marrow cells aspirated from mouse femurs were cultured for 7 d in RPMI-1640 medium containing 10% FBS and 50 ng/ml macrophage colony-stimulating factor. For macrophage transfection, BMDMs (1 × 106) were resuspended in 100 μl Nucleofector Solution buffer (Lonza) containing 5 μg DNA or other substrates, followed by transfection using the Nucleofector Program Y-001 on Amaxa nucleofector II device (Lonza). Cells were recovered for 6 h in RPMI-1640 medium containing 4 mM L-glutamine, 1.5 g/l sodium bicarbonate and 10% heat-inactivated FBS, followed sorting of viable cells by flow cytometry. To exclude the possible effects of transfected DNAs on subsequent experiments, plasmids transfected cells were further cultured for 48 h to allow stable exogenous gene expression for further examination as described30. Cells tested negative for mycoplasma contamination.

Animals and viruses.

Mouse experiments complied with ethical regulations and were approved by the Institutional Animal Care and Use Committees at the Institute of Biophysics, Chinese Academy of Sciences. Ccp1- or Ccp6-deficient mice were described previously25. Ccp2-, Ccp3-, Ccp4-, Ccp5-, Ttll4- or Ttll6-deficient mouse strains were generated using CRISPR-Cas9 approaches as described42. Vesicular stomatits virus and VACV virus strains were gifts from J. Bennink (National Institute of Allergy and Infectious Diseases). The viral strain HSV-1f was a gift from H. Peng (Institute of Biophysics, Chinese Academy of Sciences). For preparation of virus, viruses were incubated with Vero cells, followed by supernatant collection 48 h later. Supernatants were ultra-centrifuged at 25, 000g for 2 h. Pellets were resuspended in RPMI-1640 medium. For virus infection in vivo, mice were given intravenous injection of virus or were inoculated intranasally with virus. Viruses were injected intravenously into mice for determination of viral load in brain.

Plasmid construction and protein purification.

cDNA was cloned from a bone marrow cDNA library7. cDNA encoding CCP5, CCP6, TTLL4 or TTLL6 was subcloned to pFlag-CMV2 (Sigma-Aldrich) or pcDNA4/TO/myc-His B (Invitrogen) for expression in mammalian cells. Protein mutants were generated by site-directed mutagenesis method as previously described25. cDNAs encoding CCPs and TTLL glutamylases were subcloned into a modified pGEX-6P-1 vector (GE Healthcare) with a carboxy-terminal six-histidine tag. Plasmids were transformed into Escherichia coli strain BL21 (DE3). DE3 clones were cultured (absorbance at 600 nm, 0.6), followed by induction with 0.2 mM IPTG at 16 °C for 24 h. Cells were collected and lysed through the use of an ultrasonic cell disruptor followed by successive purification through Ni-NTA resins and GST resins.

Immunoprecipitation assay.

Cells were lysed in lysis buffer containing 150 mM NaCl, 50 mM Tris-Cl, 1% TritonX-100 and protease inhibitor cocktail, pH 7.4. Lysates were centrifugated on 15,000g for 15 min at 4 °C, and supernatants were incubated for 6 h at 4 °C with the appropriate antibodies (identified above), followed by immunoprecipitation with 20 μl protein A/G–conjugated agarose (Santa Cruz Biotechnology). Precipitates were completely washed with lysis buffer and separated by SDS-PAGE, followed by transfer to nitrocellulose membranes and immunoblot analysis.

Immunofluorescence

Immunostaining was performed as described previously43. Cells were plated on 0.01% poly-L-lysine treated coverslips and fixed with 4% PFA for 10 min, followed by permeabilization with 0.5% Triton X-100 for 20 min at room temperature. Primary antibodies (identified above) were added for 2 h, followed by further staining with Alexa Fluor 488– or Alexa Fluor 594–conjugated secondary antibodies (identified above). Cells were visualized by confocal microscopy (Olympus FV1000).

AlphaScreen.

Purified cGAS variants and biotinylated DNA were incubated at various concentrations according to the manufacturer's instructions (AlphaScreen; PerkinElmer)7.

DNA-precipitation assay DNA-precipitation assays were performed as previously described44.

dsDNA was synthesized with an NH2 modification at the 5′ end and the two complementary sequences were annealed, followed by coupling to CNBr-activated Sepharose 4B resins (GE Healthcare). For the DNA-precipitation assay, recombinant proteins or cell lysates were incubated with DNA-linked resins in the presence of Protease Inhibitor Cocktail Set III (Calbiochem), followed by washing with buffer containing 150 mM KCl, and immunoblot analysis (as described above) of precipitates.

Chromatin immunoprecipitation.

BMDMs overexpressing Flag-tagged cGAS variants were transfected with dsDNA, followed by crosslinking with 1% formaldehyde at 37 °C for 10 min. Lysates were incubated with anti-Flag for 6 h after preclearance with salmon sperm DNA–protein A agarose. Flag-tagged cGAS was then immunoprecipitated by salmon sperm DNA–protein A agarose. Immunoprecipitates were washed sequentially with low-salt wash buffer, high-salt wash buffer, LiCl wash buffer, and TE buffer, followed by eluting with elution buffer (1% SDS and 0.1 M NaHCO3). The cGAS-DNA complex was reversed by heating at 65 °C for 4 h. DNA was further purified by phenol-chloroform extraction and precipitated with ethanol, followed by RT-PCR analysis.

RT-PCR analysis.

Total RNA was extracted from cells using Trizol reagent and cDNA was reverse-transcribed using Superscript II (Invitrogen). RT-PCR was performed usinga StarScript II Two-step RT-PCR Kit (Genestar) with the following primers: Ifna primers, sense, 5′-ACTCATAACCTCAGGAACAAG-3′, and anti-sense, 5′-CTTTGATGTGAAGATGTTC AG-3′; Ifnb primers, sense, 5′-AGTACAACAGCTACGCCTGG-3′, and anti-sense, 5′-GAGTCC GCCTCTGATGCTTA-3′; Ccp1: sense, 5′-GGGGTCGAAGAGCGAGTTT-3′, and anti-sense, 5′-GAATGGAGTGAGTCTGCACCA-3′; Ccp2: sense, 5′-ATGAATGTCCTGCTTGAGATG G-3′, and anti-sense, 5′-CAAACGCGCTGATGAGTGC-3′; Ccp3: sense, 5′-AGCTGAAGATG CTTACAAAGAGC-3′, and anti-sense, 5′-CGCACAGTCAACTCGTATTCAT-3′; Ccp4: sense, 5′-CCAGCAGTGCCTATACCTTCC-3′, anti-sense, 5′-TGCTCAGATCAGTTTCCAAGT C-3′; Ccp5: sense, 5′-CTGCTCATTCTCGTCTTCAGG-3′, and anti-sense, 5′-ATCGAGTCCT AATGCAAGGGA-3′; and Ccp6: sense, 5′-AGGCAGGCAATGATACAGGAA-3′, and anti-sense, 5′-GGTTACCACTTTCAAAGCAAGCA-3′.

In vitro glutamylation assay.

Flag-tagged TTLL4 or TTLL6 was overexpressed in mouse embryonic fibroblasts, followed by immunoprecipitation with anti-Flag. Immunoprecipitates were eluted with Flag peptides, followed by incubation with recombinant cGAS in glutamylation buffer containing 50 mM Tris-HCl, 5 mM MgCl2, 2.5 mM dithiothreitol, 10 mM sodium glutamate and 2 mM ATP, pH 7.0. Reactions were incubated at 30 °C for 1 h.

Statistical analysis.

Student's t-test was used for statistical analysis by using Microsoft Excel45.