The deubiquitinase CYLD is a specific checkpoint of the STING antiviral signaling pathway

Stimulator of interferon genes (STING) is critical for cytosolic DNA-triggered innate immunity. STING is modified by several types of polyubiquitin chains. Here, we report that the deubiquitinase CYLD sustains STING signaling by stabilizing the STING protein. CYLD deficiency promoted the K48-linked polyubiquitination and degradation of STING, attenuating the induction of IRF3-responsive genes after HSV-1 infection or the transfection of DNA ligands. Additionally, CYLD knockout mice were more susceptible to HSV-1 infection than their wild-type (WT) littermates. Mechanistically, STING translocated from the ER to the Golgi upon HSV-1 stimulation; CYLD partially accumulated with STING and interacted selectively with K48-linked polyubiquitin chains on STING, specifically removing the K48-linked polyubiquitin chains from STING and ultimately boosting the innate antiviral response. Our study reveals that CYLD is a novel checkpoint in the cGAS-STING signaling pathway and sheds new light on the dynamic regulation of STING activity by ubiquitination.


Author summary
STING is critical for mediating the production of type I interferons and other proinflammatory cytokines. The appropriate activation of STING signaling is precisely modulated to maintain immune homeostasis. It is well established that covalent modification of STING by different types of polyubiquitin chains serves to fine-tune STING activity in response to extracellular and intracellular stresses. However, it remains poorly understood how these polyubiquitin chains on STING are dynamically removed in response to different stimuli. In this study, we characterized the deubiquitinase CYLD, which partially accumulates with STING upon HSV-1 infection and interacts selectively with the K48-linked a1111111111 a1111111111 a1111111111 a1111111111 a1111111111 Introduction The innate immune system represents the first line of host defense against invading pathogens and employs germline-encoded pattern-recognition receptors (PRRs) to detect conserved microbial molecules known as pathogen-associated molecular patterns (PAMPs). Upon sensing their corresponding PAMPs, PRRs activate signaling cascades that trigger the expression of downstream genes, which collaboratively restrain microbes and activate adaptive immune responses [1].
The action of STING is extensively regulated by various types of ubiquitin modifications [16]. For example, the E3 ubiquitin ligases RNF5 and TRIM29 catalyze K48-linked polyubiquitination of STING, facilitating its proteasome-mediated degradation [17,18]. TRIM56 and TRIM32 target STING for K63-linked polyubiquitination [19,20] and enhance the downstream signaling. AMFR/INSIG1 catalyze the K27-linked polyubiquitination of STING, which is crucial for recruiting and activating TBK1 [21]. RNF26 promotes K11-linked polyubiquitination of STING at lysine 150, a residue that is also targeted by RNF5 for K48-linked polyubiquitination [22]. Notably, it remains largely unknown how polyubiquitin chains on STING are dynamically removed in response to different stimuli.
CYLD was identified as a tumor suppressor that is mutated in familial cylindromatosis, and it plays important roles in development and tumorigenesis [23][24][25][26]. Biochemically, CYLD is a deubiquitinase that can remove K63-or K48-linked polyubiquitin chains in a reaction catalyzed by its C-terminal USP domain [27][28][29][30][31]. A recent study reported that SDC4 could recruit CYLD to remove polyubiquitin chains from RIG-I, thus inhibiting RIG-I signaling [32]. The potential role of CYLD in the STING signaling pathway remains to be explored.
In this study, we demonstrated that silencing CYLD markedly attenuated the STING-mediated induction of antimicrobial genes. Consistent with this finding, CYLD knockout mice were more susceptible to HSV-1 infection than their WT littermates. Mechanistically, CYLD partially accumulated with STING upon HSV-1 stimulation and interacted selectively with K48-linked polyubiquitin chains on STING. CYLD specifically removed K48-linked polyubiquitin chains from STING and ultimately boosted host innate antiviral responses. This study uncovers an essential function of CYLD in the STING signaling pathway and provides a new perspective for restricting infections caused by DNA pathogens.

CYLD is a new regulator of DNA-triggered antiviral signaling pathway
To explore the potential function of CYLD in STING signaling, we screened two different siR-NAs against murine CYLD (siCYLD#1 and siCYLD#2 for mouse). Knockdown of Cyld significantly inhibited the expression of IRF3-responsive genes (Ifnb, Ifna4, and Cxcl10) induced by either poly(dA:dT) or interferon stimulatory DNA (ISD) (a cytosolic DNA mimic) in mouse embryonic fibroblasts (MEFs) (Fig 1A and 1B). Moreover, silencing of Cyld markedly attenuated antiviral gene expression upon HSV-1 infection, as shown in Fig 1C. Furthermore, we measured the secretion of IFN-β. Consistent with the above results, Cyld knockdown in MEFs resulted in decreased production of IFN-β upon poly(dA:dT) and ISD stimulation (Fig 1D and  1E). In addition, the expression of IRF3-responsive genes triggered by cyclic GMP-AMP (cGAMP), a ligand that specifically activates STING, was markedly decreased in Cyld-silenced MEFs (Fig 1F). We also observed that silencing of Cyld increased the expression of IRF3-responsive genes induced by poly(I:C) (S1A Fig), which is consistent with the results of previous studies [32,33]. The efficiency of siRNA-mediated knockdown of Cyld mRNA is shown in S1B Fig. Collectively, these data suggest that CYLD is a new positive regulator of the DNA-triggered signaling pathway.

The deubiquitylation activity of CYLD is essential for potentiating DNAtriggered signaling
Given that CYLD is a deubiquitinase, we investigated whether its effect was dependent on its deubiquitylation activity. We constructed a mouse CYLD C597S mutant (cysteine mutated to serine at position 597), which lacked deubiquitination activity (S2E Fig) [34]. As expected, ectopic expression of CYLD WT enhanced poly(dA:dT)-or ISD-induced IFN gene expression. In contrast, ectopic expression of the CYLD C597S mutant failed to exert the same effect (Figs 2A, 2B, S2A and S2B).
To rule out potential off-target effects of CYLD siRNAs, we generated two RNA interference (RNAi)-resistant CYLD constructs, namely, rCYLD WT and rCYLD C597S, in which silent mutations were introduced into the sequence targeted by the siRNA without changing the amino acid sequence of the corresponding protein. As shown in Figs 2C, 2D, S2C and S2D, the diminished IFN production in Cyld-silenced cells was restored by introducing rCYLD WT but not rCYLD C597S after stimulation with poly(dA:dT) or ISD. Substantiating this finding, the nuclear translocation of IRF3 was markedly reduced when Cyld was knocked down in MEFs (Figs 2E and S2F). Furthermore, knocking down Cyld facilitated HSV-1-GFP virus infection, as evidenced by stronger GFP-positive signals (Fig 2F and 2G). Next, Cyldsilenced MEFs or control MEFs were challenged with HSV-1, and the titers of HSV-1 were analyzed by a standard plaque assay. Consistent with the above findings, Cyld knockdown resulted in a marked increase in viral titers compared with those of the controls, as shown in S2G Fig. Taken together, these results suggest that the deubiquitylation activity of CYLD is indispensable for regulating the DNA-triggered signaling pathway.

CYLD deficiency impairs the DNA-triggered signaling pathway
We further investigated the role of CYLD in bone marrow-derived macrophages (BMDMs). BMDMs from WT or Cyld -/mice were stimulated separately by poly(dA:dT) or ISD. Compared with that in WT BMDMs, the induction of IFN-β and IFN-α4 mRNAs was markedly attenuated in Cyld -/-BMDMs (Fig 3A and 3C). We also observed that the induction of IFN-β and IFN-α4 mRNAs was increased in Cyld -/-BMDMs compared with that in WT BMDMs stimulated by poly(I:C) (S3A Fig), which is consistent with the data shown in S1A Fig. Moreover, ELISA demonstrated that the production of IFN-β was impaired upon poly(dA:dT) or ISD stimulation (Fig 3B and 3D) in Cyld -/-BMDMs. Furthermore, the gene expression of IFN-β and IFN-α4 was impaired in Cyld -/-BMDMs after challenge with HSV-1 ( Fig 3E). Since IFN-β protects host cells against virus infection, we examined the role of CYLD in restricting HSV-1 infection. Fresh MEFs were pretreated with culture supernatants collected from ISD-stimulated Cyld-deficient BMDMs or WT BMDMs, followed by HSV-1 infection. Fresh MEFs pretreated with culture supernatants from Cyld-deficient BMDMs were more sensitive to HSV-1 infection (S3B Fig).
IRF3 dimerization/phosphorylation and TBK1 phosphorylation are hallmarks of STING signaling pathway activation. Compared to those in WT BMDMs, these processes were markedly attenuated upon HSV-1 challenge in Cyld -/-BMDMs ( Fig 3F). Cyld knockout also drastically inhibited the expression of IRF3-responsive genes (Ifnb and Ifna4) induced by cGAMP ( Fig 3G). Consistent with this finding, IRF3 dimerization/phosphorylation and TBK1 phosphorylation triggered by cGAMP were almost abolished in Cyld -/-BMDMs ( Fig 3H). In addition, the phosphorylation and dimerization of STING were impaired in Cyld -/-BMDMs compared to those in WT BMDMs (S3C Fig). Taken together, these results indicate that CYLD is an important checkpoint for the DNA-triggered STING-dependent signaling pathway.

CYLD associates with STING in a ubiquitination-dependent manner
To identify the potential target of CYLD, we expressed the indicated plasmids in HEK293 cells as shown in S4A Fig, which triggered the expression of the IFN-β-luciferase reporter. We observed that the activation of the IFN-β-luciferase reporter by cGAS and STING was markedly reduced when Cyld was silenced. In contrast, knockdown of Cyld did not affect the activation of the IFN-β-luciferase reporter when TBK1 or IRF3-5D was ectopically expressed. CYLD had a remarkable effect on cGAMP-induced activation of IRF3-responsive genes (Figs 1F, 3G and 3H). Therefore, we wondered whether CYLD could interact directly with STING. Exogenous STING and CYLD were transfected individually or together into HEK293T cells, followed by coimmunoprecipitation (co-IP) assays (Fig 4A and 4B). HA-tagged STING did not interact with Myc-tagged CYLD. It was demonstrated in our previous work that K27-linked polyubiquitin chains on STING can recruit TBK1 and facilitate its translocation to the perinuclear microsomes [21]. We speculated that the polyubiquitin chains of STING may recruit CYLD to STING. To test this possibility, we coexpressed STING and CYLD together with different E3 ligases, namely, RNF5, TRIM32, or TRIM56, in HEK293T cells. RNF5 is known to catalyze K48-linked polyubiquitination of STING [17], while TRIM32 and TRIM56 catalyze the formation of K63-linked polyubiquitin chains on STING [19,20]. Interestingly, the association between CYLD and STING was detected only when the E3 ligase RNF5 was coexpressed. In contrast, the E3 ligase Trim32 or Trim 56 failed to promote the association between STING and CYLD (Fig 4A and 4B). Moreover, a catalytically dead mutant of RNF5 ) or CYLD siRNAs were stimulated with poly(dA:dT) (3 μg per well), ISD (5 μg per well) or HSV-1 (MOI = 1) for 6 h. Then, the induction of Ifnb, Ifna4, and Cxcl10 mRNAs was measured by quantitative PCR. (D and E) MEFs transfected with the indicated siRNA were stimulated with poly(dA:dT) (3 μg per well) or ISD (5 μg per well) for 6 h. IFN-β production was assayed by ELISA. (F) MEFs transfected with negative control (N.C.) or CYLD siRNAs were incubated with or without cGAMP (1 μg per well) for 30 minutes at 37˚C, and the medium was then replaced. After cGAMP delivery for the indicated time periods, the induction of Ifnb and Ifna4 mRNAs was measured by quantitative PCR. Graphs show the mean ± s.d., and the data (A-F) shown are representative of three independent experiments. � P <0.05; �� P <0.01 (two-tailed ttest).
https://doi.org/10.1371/journal.ppat.1007435.g001 (C42S) did not promote the interaction between STING and CYLD (Fig 4C and 4D). Furthermore, as it was reported that RNF5 ubiquitinated STING at K150 [17], we examined whether lysine 150 of STING could influence the association of STING with CYLD. We coexpressed CYLD and RNF5 with STING or STING K150R and found that the STING K150R mutant failed to associate with CYLD even when RNF5 was present (S4D Fig). Additionally, as shown in S4B Fig, we observed that RNF5 could interact with STING but not CYLD. Recently, TRIM29 was reported to catalyze K48-linked polyubiquitination of STING at lysine 370. TRIM29 is also known to be specifically expressed in airway epithelial cells (AECs) and intestinal epithelial cells [18]. Therefore, we examined if TRIM29 has any effect on the association of STING with CYLD. As shown in S4C Fig, unlike RNF5, TRIM29 could not promote CYLD-STING association. To further investigate if STING and CYLD interact directly, we performed the indicated experiments in S4E Fig. We found that ubiquitinated STING, but not unmodified STING, interacted with CYLD in vitro, suggesting that CYLD associates with STING in a ubiquitination-dependent manner. To test whether the catalytically dead CYLD mutant influenced binding with STING, we coexpressed a catalytically dead mutant of human CYLD C601S in the presence of RNF5, as indicated in Fig 4E [ 34]. As expected, we observed that the human CYLD C601S mutant bound more tightly to STING than the WT human CYLD. A similar result was observed for the catalytically dead murine mutant CYLD C597S (S6A Fig). Then, the endogenous association between STING and CYLD was further investigated. Notably, the endogenous interaction between STING and CYLD was substantially enhanced upon HSV-1 infection ( Fig 4F). Consistent with a previous report [17], we also found that RNF5 expression was not changed during HSV-1 infection, while the association of RNF5 with STING was gradually increased. Consistent with this finding, confocal microscopy imaging revealed that the colocalization of STING and CYLD was enhanced in the presence of RNF5 (Figs 4G and S4F). We knew that STING could translocate from the ER to the Golgi upon stimulation with HSV-1, which was also confirmed by our results, as shown in S5A and S5B Fig. Interestingly, we observed that CYLD partially accumulated with STING, which was instrumental for deubiquitinating STING. Taken together, these data establish that CYLD associates with STING in a ubiquitination-dependent manner and colocalizes with STING in the Golgi upon HSV-1 infection.

Fig 2. CYLD is essential for potentiating DNA-triggered signaling. (A and B)
MEFs were transfected with 1 μg of empty vector (EV) or plasmids for the expression of wild-type HA-CYLD or HA-CYLD C597S. Then, the induction of Ifnb and Ifna4 mRNAs was measured by quantitative PCR after poly(dA:dT) (3 μg per well) or ISD (5 μg per well) stimulation for 6 h. (C and D) MEFs were transfected with 2 μl of negative control (N.C.) or CYLD siRNA#1 for 24 h and then transfected with the indicated siRNA-resistant constructs for another 24 h, followed by stimulation with poly(dA:dT) (3 μg per well) or ISD (5 μg per well) for 6 h. Then, the induction of Ifnb and Ifna4 mRNAs was measured by quantitative PCR. (E) MEFs (12-well plate) transfected with negative control (N.C.) or CYLD siRNA#1 were stimulated with poly(dA:dT) (3 μg per well) or ISD (5 μg per well) for 4 h, then stained with an antibody specific for IRF3 and imaged by confocal microscopy. Scale bars represent 50 μm. (F and G) MEFs (12-well plate) were transfected with the negative control (N.C.) or CYLD siRNA#1 and infected with HSV-1-GFP (MOI = 1) 24 h later. GFP expression was visualized 16 h later by fluorescence microscopy (scale bars represent 100 μm) (F) and quantified (G) from five independent experiments. Graphs show the mean ± s.d., and the data (A-G) shown are representative of three independent experiments. n.s., not significant; � P <0.05; �� P <0.01 (two-tailed t-test).
We explored whether CYLD could catalyze the deubiquitination of STING. Flag-tagged STING was cotransfected with CYLD or the catalytically dead mutant CYLD C601S. The cell lysates were denatured and subjected to immunoprecipitation with an anti-Flag antibody, followed by immunoblotting, as indicated in Fig 5B. Notably, STING was markedly deubiquitinated in the presence of CYLD, whereas CYLD C601S had no such effect. Consistent with the previous study, RNF5 catalyzed the formation of K48-linked polyubiquitin chains on STING. CYLD efficiently removed the polyubiquitin chains from STING, whereas CYLD C601S failed to do so ( Fig 5C). Murine CYLD had the same function as human CYLD (S6B and S6C Fig).
We also confirmed that human CYLD and murine CYLD deubiquitinated STING in vitro, but the catalytically dead mutants (human CYLD C601S and murine CYLD C597S) could not per-  5E). These results indicated that CYLD selectively deconjugates the K48-linked ubiquitination of STING. In addition, compared with that in WT BMDMs, endogenous STING polyubiquitination was markedly enhanced in Cyld -/-BMDMs upon HSV-1 stimulation in the presence of the proteasome inhibitor MG132 (Fig 5F). Consistent with this finding, Cyld -/-BMDMs exhibited a marked reduction in the STING protein level upon HSV-1 infection, but this reduction was reversed by MG132 treatment (Fig 5G). Taken together, these data indicate that CYLD prevents STING from undergoing proteasome-mediated degradation by specifically removing K48-linked ubiquitination, thus sustaining STING and promoting the host antiviral response after HSV-1 infection.

CYLD is required for host defense against HSV-1 infection in mice
To address the in vivo function of CYLD during infection with a DNA pathogen, WT and Cyld -/mice were injected intravenously with HSV-1, and their survival rates were monitored. We observed that all Cyld -/mice died within 3 days, whereas 66.7% of WT mice survived until 7 days after HSV-1 infection ( Fig 6A). As expected, HSV-1 virions were more abundant in the brains of Cyld -/mice than in the brains of WT mice (Fig 6B). Notably, Cyld -/mice displayed a more severe defect in the production of serum IFN-β after HSV-1 invasion than WT mice ( Fig 6C). The expression of Ifnb, Ifna4, and Cxcl10 mRNAs was markedly lower in spleen , and IFN-β production was determined by ELISA. (C) WT and Cyld-deficient BMDMs were stimulated with mock or ISD (5 μg per well) for the indicated time periods. The induction of Ifnb and Ifna4 mRNAs was measured by quantitative PCR. (D) WT or Cyld-deficient BMDMs were treated as described in (C), and IFN-β production was determined by ELISA. (E) WT and Cyld-deficient BMDMs were mock infected or infected with HSV-1 (MOI = 5) for the indicated time periods. The induction of Ifnb and Ifna4 mRNAs was measured by quantitative PCR. (F) WT and Cyld-deficient BMDMs were mock infected or infected with HSV-1 (MOI = 5) for the indicated time periods, and cell extracts were analyzed for IRF3 dimerization and for IRF3 phosphorylation and TBK1 phosphorylation by native PAGE and SDS-PAGE, respectively. (G) WT and Cyld-deficient BMDMs were incubated with or without cGAMP (1 μg per well) for 30 minutes at 37˚C, and the medium was then replaced. After cGAMP delivery for the indicated time periods, the induction of Ifnb and Ifna4 mRNAs was measured by quantitative PCR. (H) WT and Cyld-deficient BMDMs were incubated with or without cGAMP (1 μg per well) for 30 minutes at 37˚C, and the medium was then replaced. After cGAMP delivery for the indicated time periods, the cell extracts were analyzed for IRF3 dimerization and for IRF3 phosphorylation and TBK1 phosphorylation by native PAGE and SDS-PAGE, respectively. Graphs show the mean ± s.d., and the data (A-H) shown are representative of three independent experiments. � P <0.05; �� P <0.01 (two-tailed t-test).
https://doi.org/10.1371/journal.ppat.1007435.g003 tissues from Cyld -/mice than in spleen tissues from WT mice (Fig 6D). Consistent with this finding, the expression of IFN-β mRNA was also significantly reduced in liver and lung tissues from Cyld -/mice (Fig 6E and 6F). We also examined the viral titers in the spleens and brains of Cyld -/mice, which were much higher than those in WT mice, as indicated by a standard plaque assay, as shown in Fig 6G and 6H.
It is well established that STING is required to initiate effective IFN production when the host is infected by HSV-1 [35,36]. According to several seminal studies, STING KO mice had a higher mortality rate than WT mice upon HSV-1 infection [35][36][37]. Significant amounts of HSV-1 could be detected in the brains of infected STING KO mice but not in the brains of controls. Mice defective in STING abolished microglia-dependent type I IFN expression upon HSV-1 infection and exhibited increased susceptibility to acute Herpes simplex encephalitis [37]. Additionally, analysis of serum from STING KO mice indicated a profound defect in the production of type I IFNs after HSV-1 infection compared to that of infected control animals. Here, we revealed that CYLD KO mice had similar characteristics after injection of HSV-1. Taken together, our data indicate that CYLD is also required in vivo for the effective production of type I IFNs and is essential for sustaining protection against HSV-1 infection.

Discussion
As the convergence point for monitoring cytosolic aberrant DNAs, STING integrates upstream danger signals to activate TBK1/IRF3, which ultimately induces the production of type I IFNs and proinflammatory cytokines [38]. However, misregulation of STING signaling has been implicated in cancers and autoimmune diseases. Thus, the strength and duration of STING signaling are dynamically modulated in a multilayered and highly ordered manner to maintain immune homeostasis. It is well established that modification of STING by different types of polyubiquitin chains serves to fine-tune STING activity in response to extracellular and intracellular stresses [16]. Similarly, removing polyubiquitin chains from STING acted as an effective feedback mechanism to counterbalance the relevant immune responses. A recent report demonstrated that USP13 could remove K27-linked polyubiquitin chains from STING, thus downregulating STING signaling [39]. It remains unknown how other types of STING polyubiquitination are terminated and how STING activity is sustained during viral infection.
In this study, we characterized the ability of CYLD to specifically deconjugate K48-linked polyubiquitin chains from STING, thus sustaining STING activity and promoting host antiviral responses against DNA viruses. Several lines of evidence substantiate these claims. First, silencing CYLD resulted in a marked decrease in the STING-mediated induction of IRF3-responsive genes, which could be rescued by ectopic expression of WT CYLD but not a catalytically dead CYLD mutant. Second, CYLD enhanced STING signaling via its deubiquitination activity. Third, CYLD knockdown impaired the nuclear translocation of IRF3 and facilitated HSV-1 infection. Fourth, knocking out Cyld in BMDMs drastically attenuated STING signaling stimulated by DNA mimics, cGAMP or HSV-1. Fifth, CYLD associated with STING via the K48-linked polyubiquitin chains on STING, and this association appeared to be enhanced upon HSV-1 infection. Sixth, Cyld -/mice displayed a significant decrease in the production of interferons and cytokines became vulnerable to HSV-1 infection in vivo.
CYLD was originally reported as a tumor suppressor and later characterized as a deubiquitinase [23,26]. CYLD can suppress the activation of NF-κB or JNK signaling by removing K63-linked polyubiquitin chains from the signaling complex [27][28][29]40]. A recent study reported that SDC4 enhances the interaction between CYLD and RIG-I, thus promoting the removal of K63-linked polyubiquitin chains on RIG-I [32]. Importantly, CYLD can also deconjugate K48-linked polyubiquitin chains from some target proteins. For example, CYLD specifically removes K48-linked polyubiquitin chains from TRAF2, thereby preventing the proteasome-mediated degradation of TRAF2 [29]. CYLD can also deconjugate K48-linked polyubiquitin chains on LCK, blocking T cell receptor (TCR) downstream signaling [27]. Our study revealed that CYLD preferentially deconjugates K48-linked polyubiquitin chains on STING, leaving other types of polyubiquitin chains intact. Interestingly, we observed that STING and CYLD colocalized in the Golgi but not in the ER upon HSV-1 stimulation. Therefore, STING translocated from the ER to the Golgi upon HSV-1 stimulation, and CYLD partially accumulated with STING to promote STING deubiquitination. Consequently, CYLD actively prevented the proteasome-mediated degradation of STING and sustained antiviral responses in innate immunity.
Notably, the STING signaling pathway plays a crucial role in dampening cancer development. It is intriguing to consider whether CYLD acts as a tumor suppressor by potentiating STING signaling in cancer. It is also unknown whether human CYLD mutations correlate with the activity of the STING signaling pathway. Insights from these investigations will be instrumental in developing CYLD as a potential target for cancer therapy.

Ethics statement
Animal experiments were carried out in strict accordance with the regulations in the Guide for the Care and Use of Laboratory Animals issued by the Ministry of Science and Technology of the People's Republic of China. The protocol was approved by the Institutional Animal Care and Use Committee of the Shanghai Institute of Biochemistry and Cell Biology, Chinese Academy of Sciences (Permit Number: IBCB0027 Rev2).

Fig 5. CYLD specifically removes K48-linked polyubiquitin chains from STING. (A) K48-linked di-ubiquitin chains (Lys 48-diUb) or
K63-linked di-ubiquitin chains (Lys 63-diUb) were subjected to a Ni-NTA pull-down assay with or without His-CYLD-USP (residues 583-956 aa), followed by immunoblotting with the indicated antibodies. (B) HEK293T cells were transfected with the indicated plasmids. Thirty-six hours after transfection, the cell lysates were subjected to denaturing immunoprecipitation with an anti-Flag antibody and then analyzed by immunoblotting with the indicated antibodies. (C) HEK293T cells were transfected with the indicated plasmids. Thirty-six hours after transfection, the cell lysates were subjected to denaturing immunoprecipitation with an anti-Flag antibody and then analyzed by immunoblotting with the indicated antibodies. (D) HEK293T cells were transfected with Flag-STING, 5×Myc-CYLD, HA-tagged-Ub (WT) and various mutants. Thirty-six hours after transfection, the cell lysates were subjected to denaturing immunoprecipitation with an anti-Flag antibody and then analyzed by immunoblotting with the indicated antibodies. (E) HEK293T cells were transfected with Flag-STING, 5×Myc-CYLD, HA-tagged-Ub (WT) and a K48R mutant of HA-Ub. Thirty-six hours after transfection, the cell lysates were subjected to denaturing immunoprecipitation with an anti-Flag antibody and then analyzed by immunoblotting with the indicated antibodies. (F) WT and Cyld -/-BMDMs were infected with or without HSV-1 (MOI = 10) for the indicated time periods in the presence of MG132 (10 μM). Cell lysates were subjected to denaturing immunoprecipitation with an anti-STING antibody and then analyzed by immunoblotting with the indicated antibodies. (G) WT and Cyld -/-BMDMs were infected with or without HSV-1 for the indicated time periods in the presence or absence of MG132 (10 μM) for 6 h, and the cell lysates were analyzed by immunoblotting with the indicated antibodies.

Mice
Cyld knockout mice (in C57BL6/DBA mixed genetic background) were generated and genotyped as described previously [27]. Cyld +/mice were intercrossed to generate Cyld +/+ and Cyld -/littermates. The mice were maintained under specific pathogen-free (SPF) conditions at the Shanghai Institute of Biochemistry and Cell Biology.

RNA interference
The siRNAs duplexes were synthesized from Gene-Pharma. The sequences of siRNAs are shown as follows: Cyld siRNA #1 for mouse,

Immunoprecipitation assay and immunoblot analysis
For immunoprecipitation assay, cell extracts were prepared by using lysis buffer (50 mM Tris-HCl pH 7.4, 150 mM NaCl, 0.5% Triton X-100, 1 mM EDTA) supplemented with a protease inhibitor cocktail (Roche). The lysate was incubated with appropriate antibodies for 4 hours to overnight at 4˚C before adding protein A/G agarose beads for another 2 hours. The beads were washed three times with the lysis buffer and eluted with SDS-loading buffer for 5 minutes.
For denaturing immunoprecipitation, cells were lysed in 1% SDS buffer (50 mM Tris-HCl pH 7.5, 150 mM NaCl, 1% SDS, 10 mM DTT) and denatured by heating for 30 minutes. The lysates were centrifuged and diluted with Lysis buffer (50mM Tris-HCl pH 7.5, 150 mM NaCl, 1 mM EDTA, 1% Triton X-100) until the concentration of SDS was decreased to 0.1%. The diluted lysates were immunoprecipitated with the indicated antibodies for 4 hours to overnight at 4˚C before adding protein A/G agarose for 2 hours. After extensive wash, the immunoprecipitates were subjected to immunoblot analysis.
For immunoblot analysis, the immunoprecipitates samples were subjected to SDS-PAGE. The separated proteins were then electrically transfected to PVDF membrane (Millipore). Immunoblotting was probed with indicated primary and secondary antibodies. The protein bands were visualized by using a SuperSignal West Pico chemiluminescence ECL kit (Pierce).
For in vitro deubiquitination assays, denature-IP was performed to obtain ubiquitin-modified STING from HEK293T cells cotransfected with Flag-tagged STING and HA-Ubiquitin. The immunoprecipitates were eluted by the Flag peptide (0.3 mg ml -1 , 60 μl). CYLD, CYLD-C601S, CYLD-USP, mCYLD, and mCYLD-C597S protein were obtained by an in vitro transcription and translation kit (Promega). The ubiquitin-modified Flag-STING was incubated with the obtained protein at 37˚C for 2 h followed overnight incubation at 16˚C in the presence of ATP (1 μM). The mixtures were analyzed by immunoblot with the indicated antibodies.

Luciferase reporter assay
Cells were seeded in 12-well plates and transfected with reporter gene plasmids (100 ng) combined with siRNAs and other constructs as indicated. The total amount of DNA was kept constant by supplementing with empty vectors. pTK-Renilla reporter plasmid (2 ng) was added to normalize transfection efficiency. Luciferase activity was analyzed with the Dual Luciferase Reporter Assay System (Promega) [45].

Immunofluorescence and confocal microscopy
Cells seeded onto glass coverslips were fixed for 15 minutes with 4% paraformaldehyde in PBS and permeabilized in 0.25% Triton X-100 in PBS for another 15 minutes, followed by using 5% BSA in PBS for 1 h. Then, cells were stained with indicated primary antibodies followed by incubation with fluorescent-conjugated secondary antibodies. AlexaFluor 488 (Goat anti-Mouse) and Cyanin3 (Goat anti-Rabbit) were used for immunofluorescence in Fig 4G. Alexa-Fluor 647 (Goat anti-Rabbit) and Cyanin3 (Goat anti-Mouse) were used for immunofluorescence in S5 Fig. The nuclei were counterstained with DAPI (Sigma-Aldrich). Slides were mounted with fluorescent mounting medium (Dako). Images were captured using Leica laser scanning confocal under a 64×oil objective.

Measurement of cytokines
Concentrations of cytokines in culture supernatants or mouse serum were measured by Veri-Kine Kit (PBL Assay Science) according to the manufacturer' s instructions.

Native PAGE assay
Native gel electrophoresis for IRF3 dimerization was carried out as described previously [46].

STING dimerization assay
STING dimerization assay was performed as described previously [47]. Briefly, cells were lysed in the lysis buffer (50 mM Tris-HCl pH 7.4, 150 mM NaCl, 0.5% Triton X-100, 1 mM EDTA) with 1 mM NaF and protease cocktail inhibitor. Cell lysates were mixed with sample loading buffer without 2-Mercaptoethanol before loading on the SDS-PAGE electrophoresis. After transferring to the PVDF membrane, 5% milk in TBS with 0.1% Tween 20 (TBST) was used to block the membrane at room temperature for 1 h. STING antibodies were diluted in 5% BSA in TBST, and incubated at 4˚C overnight. The membrane was washed with TBST to remove the non-specific binding. Secondary antibody, in 5% milk TBST, was added to the membrane for 1 h at room temperature, the membrane was washed again and imaged by films.

Histological analysis
Tissues were fixed in 4% paraformaldehyde, embedded in paraffin, cut into sections, and placed on adhesion microscope slides. Sections were subjected to immunohistochemical (IHC) staining according to standard procedures. The anti-HSV-1 antibody (Abcam) was used for staining.
For Ni-NTA pulldown analysis, the purified recombinant His-CYLD-USP protein was incubated with 20 μl Ni-NTA agarose beads (Qiagen) for 2 h at 4˚C followed by extensive washing. The preloaded Ni-NTA agarose beads were incubated with di-Ub (K48-linked) or di-Ub (K63-linked) (Boston Biochem), respectively, for 1 h at 4˚C. Precipitates were extensively washed and subjected to SDS-PAGE followed by immunoblot analysis.

Plaque assay
The supernatants of MEFs cultures and the homogenates of brains and spleens from infected mice (or the serial dilutions) were used to infect monolayers of Vero cells. 2 h later, the supernatants or homogenates were removed and the infected Vero cells were washed with prewarmed PBS twice followed by incubation with a mixed media, which consists of 2×MEM (gibco), 4% FBS, and 2% Low Melting Point Agarose (SIGMA) for another 48 h. Then the cells were fixed with 4% paraformaldehyde for 15 min and stained with 1% crystal violet for 30 min before counting the plaques.

Statistics
Student's t-test was used for statistical analysis of two independent treatments. Mouse survival curves and statistics were analyzed with the Mantel-Cox test. P values of less than 0.05 were considered to be statistically significant.
Supporting information S1 Fig. (Related to Fig 1). CYLD deficiency impairs cGAMP-mediated type I interferon production. (A) MEFs transfected with negative control (N.C.) or CYLD siRNAs were stimulated with poly (I:C) (2 μg per well) for 6 h. Then, the induction of Ifnb, Ifna4, and Cxcl10 mRNAs was measured by quantitative PCR. (B) MEFs were transfected with the indicated siRNA, and the Cyld mRNA was measured by quantitative PCR. Graphs show the mean ± s.d., and the data shown are representative of three independent experiments. � P <0.05; �� P <0.01 (two-tailed t-test). (TIF) S2 Fig. (Related to Fig 2). CYLD is essential for HSV-1 restriction. (A and B) MEFs were transfected with 1 μg of empty vector (EV) or plasmids for the expression of wild-type HA-CYLD or HA-CYLD C597S for 24 h, followed by stimulation with poly(dA:dT) (3 μg per well) or ISD (5 μg per well) for 6 h. Then, the cell lysates were analyzed by immunoblotting with the indicated antibodies. (C and D) MEFs were transfected with 2 μl of negative control (N.C.) or CYLD siRNA#1 for 24 h and then transfected with the indicated siRNA-resistant constructs for another 24 h, followed by stimulation with poly(dA:dT) (3 μg per well) or ISD (5 μg per well) for 6 h. Then, the cell lysates were analyzed by immunoblotting with the indicated antibodies. (E) The amino acid sequence alignment of mouse CYLD and human CYLD. (F) MEFs (12-well plate) transfected with negative control (N.C.) or CYLD siRNA#1 were stimulated with poly(dA:dT) (3 μg per well) or ISD (5 μg per well) for 4 h. Then, cell lysates were analyzed by immunoblotting with the indicated antibodies. (G) MEFs transfected with the nonspecific control (N.C.) or CYLD siRNA#1 were infected with HSV-1 (MOI = 1) for 6 h. The titers of HSV-1 were determined by a standard plaque assay. Graphs show the mean ± s.d., and the data shown are representative of three independent experiments. �� P <0.01 (twotailed t-test). (TIF) S3 Fig. (Related to Fig 3). CYLD deficiency enhances RNA-triggered type I IFN expression. (A) WT and Cyld-deficient BMDMs were stimulated with mock or poly(I:C) (2 μg per well) for the indicated time periods. The induction of Ifnb and Ifna4 mRNAs was measured by quantitative PCR. (B) WT and Cyld -/-BMDMs were transfected with mock or ISD (10 μg per well). Equal volumes of culture supernatants from these treatments were applied to fresh MEFs, followed by HSV-1 (MOI = 10) infection. The proliferation of cells was examined by crystal violet staining. Scale bars represent 200 μm. (C) WT and Cyld-deficient BMDMs were mock infected or infected with HSV-1 (MOI = 5) for the indicated time periods, and the cell extracts were analyzed for STING dimerization and STING phosphorylation. Graphs show the mean ± s.d., and the data shown are representative of three independent experiments. � P <0.05; �� P <0.01 (two-tailed t-test). (TIF) S4 Fig. (Related to Fig 4). Ubiquitinated STING associates with CYLD. (A) The negative control (N.C.) or CYLD siRNA#3 was transfected into HEK293 cells together with IFN-β-luciferase and pTK-Renilla reporter plasmids by using Lipofectamine 2000. Forty-eight hours after transfection, the cells were transfected again with cGAS, STING, TBK1 or IRF3-5D by using Lipofectamine 2000 for 16 h before luciferase assays were performed. (B) HEK293T cells were transfected with the indicated plasmids. Thirty-six hours after transfection, the cell lysates were immunoprecipitated with an anti-Flag antibody and then immunoblotted with the indicated antibodies. (C) HEK293T cells were transfected with the indicated plasmids. Thirty-six hours after transfection, the cell lysates were immunoprecipitated with an anti-HA antibody and then immunoblotted with the indicated antibodies. (D) HEK293T cells were transfected with the indicated plasmids. Thirty-six hours after transfection, the cell lysates were immunoprecipitated with an anti-HA antibody and then immunoblotted with the indicated antibodies. (E) Flag-STING was transfected into HEK293T cells with or without HA-ubiquitin for 36 h, and cell lysates were immunoprecipitated with an anti-Flag antibody. Then, ubiquitinated STING or STING was eluted from the anti-Flag precipitates by a Flag peptide (0.3 mg ml -1 , 60 μl). The eluted ubiquitinated-Flag-STING or Flag-STING was subjected to a Ni-NTA pull-down assay with or without recombinant His-CYLD-USP at 37˚C for 1 h in the presence of ATP (1 μM). The precipitates were analyzed by immunoblotting with the indicated antibody. (F) HeLa cells were transfected with empty vector or Flag-tagged RNF5 for 24 h, and the cell lysates were then immunoblotted with the indicated antibodies. Graphs show the mean ± s.d., and the data shown are representative of three independent experiments. n.s., not significant; �� P <0.01 (two-tailed t-test). (TIF) S5 Fig. (Related to Fig 4). CYLD partially accumulates with STING in the Golgi upon HSV-1 stimulation. (A) MEF cells were stimulated with or without HSV-1 (MOI = 1) for 4 h and then stained with the indicated antibodies before imaging by confocal microscopy. STING (Red); ER (Green); CYLD (Purple); Nucleus (Blue). (B) MEF cells were stimulated with or without HSV-1 (MOI = 1) for 4 h and then stained with the indicated antibodies before imaging by confocal microscopy. STING (Red); Golgi (Green); CYLD (Purple); Nucleus (Blue). Scale bars represent 25 μm. (TIF) S6 Fig. (Related to Fig 5). CYLD deubiquitinates STING in vitro. (A) HEK293T cells were transfected with the indicated plasmids. Thirty-six hours after transfection, the cell lysates were immunoprecipitated with an anti-Myc antibody and then immunoblotted with the indicated antibodies. (B) HEK293T cells were transfected with the indicated plasmids. Thirty-six hours after transfection, the cell lysates were subjected to denaturing immunoprecipitation with an anti-Flag antibody and then analyzed by immunoblotting with the indicated antibodies. (C) HEK293T cells were transfected with the indicated plasmids. Thirty-six hours after transfection, the cell lysates were subjected to denaturing immunoprecipitation with an anti-Flag antibody and then analyzed by immunoblotting with the indicated antibodies. (D) In vitro deubiquitination analysis of ubiquitin-modified STING eluted from the denatured IP (anti-Flag) from HEK293T cells transfected with Flag-STING and HA-ubiquitin with Flag peptide, followed by incubation with in vitro generated CYLD, CYLD-C601S, and CYLD-USP by an in vitro transcription and translation kit. The mixtures were analyzed by immunoblot analysis with the indicated antibodies. (E) In vitro deubiquitination analysis of ubiquitin-modified mSTING eluted from the denatured IP (anti-Flag) from HEK293T cells transfected with Flag-mSTING and HA-ubiquitin with Flag peptide, followed by incubation with mCYLD and mCYLD-C597S, which were generated by an in vitro transcription and translation kit. The mixtures were analyzed by immunoblot analysis with the indicated antibodies. (TIF)