|
CARMA3 Is a Host Factor Regulating the Balance of Inflammatory and Antiviral Responses against Viral Infection
Abstract
Graphical Abstract Highlights d Deficiency of CARMA3 results in the host resistance to RNA viral infection d CARMA3 positively regulates RIG-I/MAVS-mediated NF-kB activation d CARMA3 negatively regulates RIG-I/MAVS-mediated TBK1/ IRF3 activation d CARMA3 negatively suppresses MAVS oligomerization in mitochondran SUMMARY Host response to RNA virus infection is sensed by RNA sensors such as RIG-I, which induces MAVSmediated NF-kB and IRF3 activation to promote inflammatory and antiviral responses, respectively. Here, we have found that CARMA3, a scaffold protein previously shown to mediate NF-kB activation induced by GPCR and EGFR, positively regulates MAVS-induced NF-kB activation. However, our data suggest that CARMA3 sequesters MAVS from forming high-molecular-weight aggregates, thereby suppressing TBK1/IRF3 activation. Interestingly, following NF-kB activation upon virus infection, CARMA3 is targeted for proteasome-dependent degradation, which releases MAVS to activate IRF3. When challenged with vesicular stomatitis virus or influenza A virus, CARMA3-deficient mice showed reduced disease symptoms compared to those of wild-type mice as a result of less inflammation and a stronger ability to clear infected virus. Altogether, our results reveal the role of CARMA3 in regulating the balance of host antiviral and pro-inflammatory responses against RNA virus infection.
In Brief Jiang et al. reveal that CARMA3, a gene located in a host genomic locus that contributes to the host's susceptibility to RNA respiratory virus infection, is a key molecule that controls the balance of proinflammatory and antiviral responses, through positively regulating NF-kB activation but negatively regulating IRF3 activation.
The innate immune system is the first line of host defense against infection, which is essential for initial detection and recognition of pathogens, activation of acute anti-microbial responses, and subsequent activation of adaptive immunity. This system utilizes pattern recognition receptors such as Toll-like receptors (TLRs) on the cell surface and cytosolic retinoic acid-inducible gene 1 (RIG-I)-like receptors (RLRs) to detect the invading pathogen (Baum and García-Sastre, 2011; Janeway, 2013; Jiang et al., 2011a) . The RLR family of proteins is crucial for detecting viral RNA in cytosol. It is composed of RIG-I, melanoma differentiation-associated protein 5 (MDA5), and laboratory of genetics and physiology 2 (LGP2). RIG-I senses 5 0 -triphosphate RNA as well as short (<2-kb) double-stranded RNA (dsRNA) and is essential for innate immunity to many single-stranded RNA (ssRNA) viruses, including influenza A virus (IAV), Sendai virus (SeV), respiratory syncytial virus (RSV), vesicular stomatitis virus (VSV), etc. In contrast, MDA5 recognizes longer dsRNA (>2 kb) and protects the host from infection of encephalomyocarditis virus (EMCV), Theiler's virus, mengovirus, murine norovirus, and murine hepatitis virus (Kato et al., 2006; McCartney et al., 2008; Roth-Cross et al., 2008) .
Without stimulation, RIG-I is in the closed conformation with the N-terminal CARD domains bound to the central helicase domain. Upon binding of the CTD to viral RNA, RIG-I undergoes conformational changes, oligomerization, and exposure of the CARD domains to recruit a signaling adaptor called mitochondrial antiviral-signaling protein (MAVS). MAVS contains an N-terminal CARD domain, a proline-rich region, and a transmembrane domain (TMD) at the C terminus. The CARD domain is important for its interaction with upstream RLRs (Goubau et al., 2014; Kawai et al., 2005; Meylan et al., 2005; Seth et al., 2005; Xu et al., 2005) . The proline-rich region is required for the recruitment of multiple E3 ligases, such as tumor necrosis factor (TNF) receptor (TNFR)-associated factors (TRAFs). The TMD domain is key for MAVS localization at the mitochondrial outer membrane. Upon activation, MAVS forms a functional prion-like structure at mitochondria and works as a platform to form a MAVS signalosome that further activates IKKa/IKKb/NEMO signaling and TBK1/ IKKε/NEMO signaling (Liu et al., 2013) .
Activation of IKKa/IKKb/NEMO triggers activation of transcription factor necrosis factor kB (NF-kB) and, thus, induction of proinflammatory cytokines (Liu and Gu, 2011) . These cytokines are important to induce inflammatory responses and to restrict viral replication and spread (Dienz et al., 2012) . Elevated levels of proinflammatory cytokines are closely correlated with severity of clinical diseases, including airway inflammation and acute lung injury during influenza infection in children (Chiaretti et al., 2013) . On the other hand, activation of the TBK1/IKKε/NEMO complex leads to activation of IRF3 and production type I interferons (IFNs), including IFNa and IFNb (Fitzgerald et al., 2003; Sankar et al., 2006; Zhong et al., 2008) . Type I IFNs are potent inducers for the expression of hundreds of IFN-stimulated genes (ISGs) in paracrine and autocrine fashions, which induce a cellular antiviral state on the target cells (Au et al., 1995; Schafer et al., 1998; Sun et al., 2010) . Recent mechanistic studies have found that, upon activation, MAVS forms a prion-like fibril at mitochondria and acts as an active platform for the recruitment of E3 ligases, including TRAF6, TRAF2, and TRAF5, through distinct TRAF-binding domains (Hou et al., 2011; Liu et al., 2013; Xu et al., 2014) . These E3 ligases are essential for signaling downstream of MAVS, regulating the polyubiquitination chains, and recruiting downstream IKKa/IKKb/NEMO complex and TBK1/IKKε/NEMO complex (Liu et al., 2013) .
It has been an important question why different individuals display highly variable systemic symptoms across the infected populations during outbreaks of seasonal virus infection. To study how genetic polymorphisms contribute to this variation, Ferris et al. (2013) crossed different incipient lines of mice, which exhibit a broad range of susceptibility to IAV infection, and identified three novel quantitative trait loci (QTLs) that may contribute to the susceptibility for IAV infection. One of these QTLs, Hrl4, contains 13 genes (Ferris et al., 2013) . Among these genes, most of them do not have a clear link to the antivirus response except for the CARD10 gene (Ferris et al., 2013) , which encodes a scaffold protein named CARMA3 (Jiang and Lin, 2012) .
CARMA3 contains multiple protein-protein interaction domains, including an N-terminal CARD domain, a coiled-coil domain, and a C-terminal MAGUK domain (Gaide et al., 2001; Jiang and Lin, 2012) . CARMA3 is expressed only in non-hematopoietic cells, while CARMA1, a related protein, is expressed only in hematopoietic cells. The CARMA proteins share similar structure and functions, albeit with distinct tissue distribution. Upon activation, CARMA proteins form a complex with B cell lymphoma 10 (BCL10) and caspase-like protein mucosa-associated lymphoid tissue lymphoma translocation protein 1 (MALT1), and the CARMA-BCL10-MALT1 (CBM) complex functions to activate the downstream IKK complex, leading to activation of NF-kB (Jiang and Lin, 2012) . Previous studies have shown that CARMA3 is crucial in mediating G protein-coupled receptor (GPCR)-and epidermal growth factor receptor (EGFR)-, but not TLR-or TNFR-, induced NF-kB activation (Grabiner et al., 2007; Jiang et al., 2011b; Klemm et al., 2007; McAllister-Lucas et al., 2007) . However, it is unknown whether CARMA3 also is involved in regulating the host responses to viral infection.
It is known that virus infection induces robust NF-kB activation in host cells to trigger expression of pro-inflammatory cytokines, which help to inhibit virus replication and spread in the host. Since CARMA3 is located in the genomic locus that contributes to host's susceptibility to viral infection and is involved in NF-kB signaling, we investigated its role in host antiviral response. Our data suggest that CARMA3 contributes to inflammatory and antiviral responses via regulating RIG-I/MAVS-induced TBK1/ IRF3 and NF-kB activation. We have found that CARMA3 deficiency resulted in the defect in VSV-and RNA-induced NF-kB activation and production of pro-inflammatory cytokines, but, surprisingly, enhanced TBK1/IRF3 activation and production of type I IFN, thereby displaying a reduced viral load in VSV-infected cells and tissues. Mechanistic studies showed that CARMA3 inhibited IRF3 activation through blocking the formation of MAVS aggregation. Together, these results reveal that CARMA3 is a key molecule that regulates the balance between RNA virus infection-induced inflammatory and antiviral innate immune response.
Recent genetic studies indicate that the CARD10 gene is located in the genomic locus that may contribute to the host's susceptibility to IAV infection (Ferris et al., 2013) . To explore the biological significance of CARMA3 in host antiviral response, we challenged wild-type (WT) and CARMA3À/À (knockout [KO]) mice with IAV strain PR8, a strain that has been highly adapted in mice and causes disease symptoms and mortality in mice. IAV infection caused a significant body weight loss of WT mice, but not that of CARMA3 KO mice ( Figure 1A ). Viral yield was much higher in lungs of WT mice than in those of CARMA3 KO mice at 2 days post-infection (dpi) ( Figure 1B) . Similarly, lung injury caused by IAV infection was greatly attenuated in CARMA3 KO mice ( Figure S1A ), suggesting that CARMA3 plays a negative role in antiviral response against IAV infection. Consistently, we found that CARMA3 KO mice produced more type I IFN IFNb in lungs compared to WT mice ( Figure 1C ), but expressed less pro-inflammatory cytokines IL-6, IL-1a, and IL-1b following IAV infection (Figures 1D and S1B-S1D), suggesting that CARMA3 also plays a positive role in inflammation in response to influenza virus infection.
VSV is another negative sense, ssRNA virus, and its infection induces a flu-like symptom like IAV does. To examine whether CARMA3 also contributes to the host responses to VSV infection, we challenged WT and CARMA3 KO mice intranasally with VSV. Since VSV is a neurotropic virus and brain is one of primary targeting tissues of VSV, we examined the viral load in brain of these mice. Interestingly, we found that viral loads were significantly higher in the olfactory bulbs (OBs) of the brain in WT mice than those in CARMA3 KO mice ( Figures 1E and 1F ), although smaller differences were found in spleen and lung of these mice (Figures S1E and S1F). Consistent with the low viral load in CARMA3 KO mice, we found that CARMA3 KO mice produced significantly higher levels of local IFNb in OBs ( Figure 1G ), whereas the expressions of IFNg and G-CSF in the brain of WT and CARMA3 KO mice were similar (Figures S1G and S1H). Similar to the response to IAV infection, we found that production of pro-inflammatory cytokines, such as IL-6, IL-1a, IL-1b, and TNF-a, in the sera of CARMA3 KO mice was significantly reduced compared to WT mice 2 days following VSV infection (Figures S1I-S1M). Together, these results further support the above observations that CARMA3 plays a negative role in regulating antiviral responses in the host, but it plays a positive role in regulating the expression of pro-inflammatory cytokines in response to viral infection.
Since CARMA3 is only expressed in non-hematopoietic cells, the observed effect of CARMA3 deficiency on viral infection in vivo might be compromised by the contribution of hematopoietic cells. To reveal the molecular mechanism by which CARMA3 affects inflammatory and antiviral responses to virus infection, we prepared primary WT and CARMA3 KO mouse embryonic fibroblast (MEF) cells and stimulated these cells with VSV. Consistent with the in vivo data, we found that VSV infection in WT MEF cells induced significantly higher levels of IL-6 mRNA and protein than those in CARMA3 KO MEF cells (Figures 2A and 2B ). Since IL-6 is a well-known target of NF-kB, we The 6-to 8-week-old CARMA3 WT and KO mice (n R 5 per group) were intranasally inoculated with 600 PFUs of IAV per mouse. The weight loss was monitored daily for 14 days and plotted. (B-D) Mice were infected as in (A). At 2 dpi, mouse lungs were harvested. (B) Viral loads were tittered and plotted. (C) At 4 dpi, lungs were harvested and cytokine production in mRNA level was measured by real-time qPCR. (D) At 4 dpi, lung fluid was harvested and cytokine production was measured by ELISA essay. (E-G) The 6-to 8-week-old CARMA3 WT and KO mice (n R 5 per group) were intranasally inoculated with 10 7 PFUs of VSV-GFP per mouse. (E) At 2 dpi, olfactory bulbs (OBs) were harvested and imaged. (F) Viral loads in OBs were tittered and plotted in log scale. (G) At 1 dpi, OBs were harvested and cytokine production in mRNA level was measured by qPCR. The p values were generated by Student's t test.
examined the NF-kB activation and found that NF-kB activation was partially defective in CARMA3 KO MEF cells following VSV stimulation ( Figure 2C ).
It has been shown that dsRNA virus infection can induce NF-kB through both TLR signaling and RIG-I/MAVS signaling. Given the fact that CARMA3 is not required for TLR-induced NF-kB activation (Grabiner et al., 2007) , we hypothesized that CARMA3 only mediates RIG-I/ MAVS-induced NF-kB, which can explain why VSV-induced NF-kB was only partial defective in CARMA3 KO cells. To test this hypothesis, we stimulated WT and CARMA3 KO MEF cells with RIG-I-specific ligands poly(I:C) or 5 0 triphosphate dsRNA (5 0 ppp-dsRNA). Consistently, we found that NF-kB activation and production of IL-6 were severely defective in CARMA3 KO MEF cells in response to these stimuli ( Figures 2D-2F , S2A, and S2B). Furthermore, we immortalized CARMA3 WT and KO MEF cells and reconstituted the KO cells with HA-tagged CARMA3 to its endogenous level ( Figure S2C ). This reconstitution of CARMA3 rescued the phenotypes ( Figures 2G-2J ). To find out if this is also the case in other cell types, we used BEAS-2B cells, a primary and immortalized human lung epithelial cell line, with knockdown of CARMA3 expression via small hairpin RNA (shRNA) ( Figure S2D ), and consistent data were observed (Figures 2K and 2L) .
Type I IFNs such as IFNa and IFNb are important antiviral cytokines that are induced by RIG-I/MAVS-mediated activation of TBK1-IRF3 signaling. We found that VSV infection induced significantly higher levels of IFNb mRNA and protein in CARMA3 KO MEF cells than in WT MEF cells ( Figures 3A and 3B ). Consistently, IRF3 phosphorylation at S396 and TBK1 phosphorylation were induced at higher levels in CARMA3 KO MEF cells at 6 and 8 hr following VSV infection ( Figure 3C ). To determine if this observation is specific for viral infection, we stimulated these cells with RIG-I-specific ligands, 5 0 triphosphate dsRNA (5 0 ppp-dsRNA), or poly(I:C). Consistently, production of IFN and IRF3 activation were enhanced in CARMA3 KO MEF cells (Figures 3D-3F and S3A-S3E). Furthermore, the virus load was reduced in CARMA3 KO MEF cells compared to that in WT MEF cells ( Figure 3G ).
To further confirm the above data, we reconstituted CARMA3 KO MEF cells with either a CARMA3 expression plasmid or a control vector. We found that the phenotypes observed in CARMA3 KO MEFs were reverted in CARMA3-reconstituted KO MEF cells ( Figures 3H-3K) . Consistently, knockdown of CARMA3 in human lung epithelial cells (BEAS-2B) rendered stronger TBK1/IRF3 phosphorylation and the higher expression level of type I IFN following VSV infection ( Figures 3L, 3M , and S3F). Furthermore, we isolated primary lung cells from CARMA3 WT and KO mice, and higher IFNb was detected in the CARMA3 KO lung cells compared to that in WT cells ( Figure S3G ), although we were not able to detect any significant difference in IL-6 production, which likely was due to the high basal level of IL-6 production in these cells ( Figure S3H ).
Previous studies have shown that BCL10 and MALT1 bind to CARMA3 to form the CBM complex upon EGFR-and GPCRinduced NF-kB activation (Grabiner et al., 2007; RNA was isolated and cytokine production in mRNA level was measured by real-time qPCR, normalized to the internal control GAPDH. (B and E) Cytokine production in culture supernatants was measured by ELISA assay. (C and F) Nuclear fractions were isolated and electrophoretic mobility shift assay (EMSA) analysis was performed to check NF-kB activation. Oct-1 served as an internal control. (G-J) E1A-immortalized CARMA KO MEF cells were generated and reconstituted with HA-tagged CARMA3 or vector as indicated. These cells were infected with VSV at MOI = 3 (G and H) or transfected with poly(I:C) (I and J) for the indicated time. (G and I) Cytokine production in culture supernatants was measured by ELISA assay. (H and J) Nuclear fractions were isolated and EMSA analysis was performed to check NF-kB activation. Oct-1 served as an internal control. (K and L) BEAS-2B cells with stable knockdown by shRNA against CARMA3 or control were established and infected with VSV at MOI = 3 for the indicated time. (K) Nuclear fractions were isolated and EMSA analysis was performed to check NF-kB activation. Oct-1 serves as an internal control. (L) RNA was isolated and cytokine production in mRNA level was measured by real-time qPCR, normalized to the internal control GAPDH. et al., 2007a). Next we wanted to determine if BCL10 and MALT1 play a similar function as CARMA3 in response to virus infection. Similar to CARMA3 deficiency, BCL10 deficiency led to reduced expression of IL-6 but enhanced expression of IFNb and a higher level of IRF3 phosphorylated upon VSV infection or poly(I:C) treatment in primary MEF cells ( Figures 4A-4E) . Consistently, we observed reduced viral load in OBs of BCL10 KO mice compared to WT mice ( Figure 4F ). These results indicate that BCL10 plays a similar role as CARMA3 in response to virus infection.
However, MALT1 deficiency did not result in defective IL-6 production ( Figure S4A ). Although IFNb expression was slightly increased in MALT1 KO MEF cells ( Figure S4B ), IFNa expression was similar to that in WT MEF cells ( Figures S4C and S4D) . Consistently, viral infection-induced NF-kB activation was not significantly changed in MALT1 KO MEF cells ( Figure S4E ). Although IRF3 phosphorylation was slightly enhanced in MALT1 KO MEF cells ( Figure S4F ), it was not as significant as what we observed in CARMA3 KO or BCL10 KO MEF cells. Together, these data suggest that MALT1 does not play a dominant role in regulating viral infection-induced NF-kB and IRF3 activation.
CARMA3 Regulates RIG-I/MAVS-Mediated IKK/NF-kB Activation and TBK1/IRF3 Activation in an Independent Manner Interestingly, we found that VSV infection-induced NF-kB activation could be detected as early as the first hour post-infection, whereas IRF3 phosphorylation was not detectable until 4 hr post-infection ( Figures 2C and 3C ). Since VSV infectioninduced NF-kB activation was partially defective but IRF3 activation was enhanced in CARMA3 KO MEF cells, we decided to examine whether NF-kB-targeted genes expressed at early time points post-infection might inhibit IRF3 activation. To test this possibility, WT MEF cells were pre-treated with NF-kB nuclear translocation inhibitor (NF-kBi) for 1 hr before VSV infection. Although NF-kBi pretreatment partially inhibited NF-kB activation ( Figure 5A ), it did not enhance but instead partially inhibited IRF3 phosphorylation ( Figure 5B ), indicating that it is not NF-kB-inducing genes suppressing IRF3 phosphorylation. To further support this conclusion, we transfected IkBa super-repressor (SR) mutant into immortalized MEF cells, and we found that, although expression of IkBa SR mutant could suppress VSV-induced NF-kB activation, it could not enhance IRF3 phosphorylation (Figures 5C and 5D) . Finally, to determine whether newly synthesized proteins regulate IRF3 activation, primary MEF cells were pretreated with protein synthesis inhibitor cycloheximide (CHX) before poly(I:C) transfection. Pretreatment with CHX did not alter IRF3 activation ( Figure 5E ). Therefore, the impaired NF-kB activation at early hours postinfection is not the cause for the enhanced IRF3 activation in CARMA3 KO MEF cells, indicating that CARMA3 regulates NF-kB and IRF3 activation through an independent event.
CARMA3 and BCL10 Regulate RIG-I/MAVS Signaling by Binding to MAVS To determine the molecular mechanism by which CARMA3 and BCL10 regulate viral infection-induced NF-kB and IRF3 activation, CARMA3 or BCL10 was knocked down by shRNA in HEK293T cells. Overexpression of MAVS induced robust expression of NF-kB-dependent and IRF3-dependent luciferase reporters in control cells, whereas defective NF-kB activation and enhanced IRF3 activation were observed in CARMA3 or BCL10 knockdown cells (Figures 5F, 5I , and S5A). However, when TBK1 or IKKε was overexpressed in these cells, knockdown of CARMA3 or BCL10 did not alter activation of NF-kB or IRF3 ( Figures 5G, 5H , 5J, 5K, S5B, and S5C), indicating that CARMA3 and BCL10 function downstream of MAVS but upstream of TBK1 and IKKε.
Since CARMA3 is a scaffold protein with multiple protein-protein interaction domains, we examined whether CARMA3 is physically associated with MAVS. When overexpressed in HEK293T cells, CARMA3 bound to MAVS but only weakly to RIG-I ( Figure 6A ), whereas BCL10 interacted with both MAVS and RIG-I ( Figure S6A ). In contrast, MALT1 could not bind to MAVS when overexpressed in HEK293T cells ( Figure S6B ). Furthermore, endogenous MAVS was capable of interacting with overexpressed CARMA3 or BCL10 in HEK293T cells (Figure 6B ). More interestingly, inducible interactions, but not constitutive interactions, were observed between endogenous MAVS and BCL10 or between endogenous MAVS and CARMA3 in CARMA3-reconstituted KO MEF cells (Figures 6C and 6D) . The dynamic interaction between MAVS and CARMA3 could be detected in both primary BCL10 Het and KO MEF cells ( Figure 6E ), whereas the MAVS-BCL10 interaction could only be detected in CARMA3 KO cells reconstituted with CARMA3, but not with its vector control ( Figure 6F ). These data suggested that BCL10 might be recruited to the MAVS-containing complex via CARMA3, but not vice versa. To address whether CARMA3 binds to MAVS through CARD-CARD interaction, we co-transfected HEK293T cells with MAVS and CARMA3 WT or truncate mutants ( Figures S6C and S6D ). We found that the C-terminal GUK domain, but not the CARD domain, of CARMA3 appeared to be critical for the binding to MAVS.
It has been shown that infection with SeV, another negative, single-stranded virus, induces the formation of large MAVS aggregates, which is functionally important in mediating IRF3 activation (He et al., 2015; Hou et al., 2011; Moresco et al., 2011; Xu et al., 2014) . Since CARMA3 binds to MAVS, we hypothesized that CARMA3 may block MAVS oligomerization. When CARMA3 was overexpressed together with RIG-I in HEK293T cells, CARMA3 was detected in the mitochondrial fraction in a dosedependent manner ( Figure 7A ). In contrast, only a very small fraction of RIG-I was isolated together with mitochondria ( Figure 7A ). This suggests that CARMA3 can be recruited to mitochondria and bind to MAVS. Importantly, overexpression of CARMA3 was able to disrupt the formation of endogenous MAVS aggregates in semi-denaturing detergent agarose gel electrophoresis (SDD-AGE) gel, which is 2% SDS resistant ( Figure 7A ). This is also the case for BCL10 ( Figures S7A and S7B ). To further confirm that CARMA3 blocks MAVS from forming aggregates, we infected primary MEF cells with VSV and isolated crude mitochondria. We observed significantly stronger formation of endogenous MAVS aggregates in CARMA3 KO MEF cells at 6 hr post-infection ( Figure 7B) , and, consistently, this stronger aggregation of MAVS in CARMA3 KO cells could be reverted when CARMA3 was reconstituted back into the CARMA3 KO MEF cells ( Figure 7C ). Together, these results indicate that CARMA3 plays a negative role in the assembly of MAVS aggregates upon virus infection.
Due to the lack of anti-CARMA3 antibody suitable for detecting endogenous CARMA3 protein by immunoblotting, we generated stable cells that express HA-tagged CARMA3 in CARMA3 KO MEF cells. Interestingly, we found that HA-tagged CARMA3 protein was gradually reduced upon VSV infection (Figures 7D-7G) . However, this reduction in CARMA3 protein was rescued by pretreatment of cells with proteasome inhibitor MG132 ( Figures 7E and 7F ). This indicated that, following VSV infection, CARMA3 is targeted for proteasome degradation. Therefore, we performed immunoprecipitaion with anti-HA agarose, and we found that a high amount of K48-ubiquitinated CARMA3 was observed in the presence of MG132 at 4 hr post-VSV infection ( Figure 7F ). However, in contrast to CARMA3, the protein level of BCL10 was not significantly altered following VSV infection ( Figure 7G ). Together, these results suggest that CARMA3 is targeted to proteasome-mediated degradation following RNA virus infection.
In this study, we reveal that CARMA3 is a regulator of RIG-I/ MAVS signaling. We have found that CARMA3 regulates RIG-I/ MAVS-mediated NF-kB and TBK1/IRF3 activation in a twophase mechanism ( Figure S7C ). Upon RIG-I activation, MAVS is first activated at early time points following viral infection, which can activate IKKa/IKKb/NEMO in a CARMA3/BCL10dependent manner. However, in this early phase of post-infection, MAVS is sequestered by the CARMA3/BCL10 complex via interaction, and, therefore, it cannot form the high-molecular weight aggregates, which is required for downstream activation of TBK1/IRF3 signaling. With time, CARMA3 is targeted to a proteasome-mediated degradation, which releases MAVS that forms the functional aggregates and activate TBK1/IRF3 signaling. Therefore, CARMA3 functions as a host factor to regulate the balance of the RIG-I/MAVS-induced two downstream signaling pathways for antiviral innate immune response and the pro-inflammatory response. On one hand, CARMA3 plays a positive role in MAVS-induced NF-kB activation, leading to the induction of pro-inflammatory cytokines, but, on the other hand, it plays a negative role in MAVS-induced TBK1/IRF3 activation and production of antiviral cytokines, type I IFNs (Figure S7C) . However, the precise mechanism as to how CARMA3 is targeted for ubiquitination-based proteasome degradation requires further investigation. In addition, it will be interesting to see if CARMA3 plays a general effect in response to other RNA viruses as well as DNA viruses that are typically not recognized by cytoplasmic RLRs. Different individuals display highly variable systemic symptoms in response to viral infection. However, host factors contributing to this variability are largely unknown. Recent studies suggest that the CARD10 gene, which encodes CARMA3, is located in a genomic locus that contributes to the host's susceptibility to RNA virus, such as IAV infection (Ferris et al., 2013) . Our data, showing the distinct regulation of CARMA3 in the pro-inflammatory response and the antiviral response, suggest that CARMA3 may be a gene contributing to the host's susceptibility to viral infection. Therefore, it will be important to find out whether CARMA3 protein expression levels or particular CARMA3 polymorphisms exist in different human populations, which may explain the variable susceptibility among different populations in response to virus infection.
CARMA3 is distinguished from many other mediators downstream of MAVS identified so far, which are either positive or negative mediators for both NF-kB activation and IRF3 activation. These mediators include NEMO, TRAF6, A20, and CYLD (Friedman et al., 2008; Liu et al., 2013; Maelfait et al., 2012; Parvatiyar et al., 2010; Saitoh et al., 2005; Zhao et al., 2007) . In contrast, CARMA3 deficiency led to reduced NF-kB activation but an increase in IRF3 activation. CARD9 is structurally related to CARMA3 and contains a CARD domain and a coiled-coil domain, but no MAGUK domain. CARD9 is primarily expressed in myeloid cells. In bone marrow-derived dendritic cells (BMDCs), CARD9 deficiency resulted in defects in NF-kB activation and production of pro-inflammatory cytokines, including IL-6 and IL-1b, in response to 5 0 ppp dsRNA treatment; however, it did not alter the production of type I IFN (Poeck et al., 2010) . In addition to CARD9, CARMA1 is expressed in myeloid cells. It will be interesting to find out how CARMA1, the counterpart of CARMA3 in hematopoietic cells, functions in mediating RIG-I/MAVS signaling in myeloid cells. In contrast to CARMA1, CARMA3, or CARD9, BCL10 is universally expressed. BCL10 functions similarly to CARD9 in BMDCs in response to 5 0 ppp dsRNA treatment, whereas, in primary MEF cells, it functions similarly to CARMA3 upon VSV infection or poly(I:C) treatment. This suggests that BCL10 may regulate RIG-I/MAVS signaling in a cell-type-specific manner.
During RNA virus infection, NF-kB activation can be induced through TLR-dependent and TLR-independent signaling pathways. It has been demonstrated that CARMA3 is not required for TLR-induced NF-kB activation (Jiang et al., 2011b; Pan and Lin, 2013) . In this study, we found CARMA3 deficiency induced only partial defects in NF-kB activation, suggesting that CARMA3 may play an important role in TLR-independent NF-kB activation, which is induced by RIG-I/MAVS signaling. TLR-dependent NF-kB activation may play dominant roles in the induction of local IL-6 production in non-hematopoietic cells following RNA viral infection, which may explain why we do not observe a significant effect of CARMA3 deficiency in local IL-6 mRNA production in OBs following VSV infection.
Virus infection-induced NF-kB activation promotes expression of pro-inflammatory cytokines and chemokine, which are crucial to trigger inflammatory responses in the host. During viral infection, it is important to keep the inflammatory response under control to avoid severe tissue damage. Overly robust NF-kB activation or inflammation can lead to severe health consequences. For example, Ebola virus, a negative sense, singlestranded virus, is a highly virulent pathogen that causes a frequently lethal hemorrhagic fever syndrome, which is well correlated with the high inflammatory response that is induced in the host (Rasmussen et al., 2014) . Therefore, it is important to regulate inflammatory response to a safe level to prevent severe tissue damage in such patients during viral infection. CARMA3 deficiency leads to a reduced NF-kB activation and production of pro-inflammatory cytokines. Challenging WT mice with IAV induced a strong production of pro-inflammatory cytokines, including IL-6, and severe inflammation, lung injury, and significant weight loss, which were attenuated in CARMA3 KO mice, indicating that CARMA3 is a key molecule to regulating inflammatory responses in the host and potentially a therapeutic target.
Previous studies indicated that NF-kB activation contributes to the production of type I IFN in response to viral infection (Basagoudanavar et al., 2011; Wang et al., 2010 Wang et al., , 2007b . In this study, instead of observing a compromised production of type I IFN, we found an even higher production of type I IFN in CARMA3 KO MEF cells. Therefore, the contribution of the enhanced IRF3 activation in the absence of CARMA3 not only counteracted the defect of NF-kB but also resulted in greater overall production of IFNb. This indicates the importance of CARMA3 for inhibiting the expression of type I IFN and antiviral responses, suggesting that IRF3 activation plays a more dominant role in the expression of type I IFN than NF-kB activation in response to RNA virus infection.
In this study, we found that CARMA3 functions downstream of MAVS and upstream of TBK1 or IKKε. As a scaffold protein, CARMA3 interacts with MAVS and prevents the formation of high-molecular weight MAVS aggregates, thereby blocking IRF3 activation. It is interesting that CARMA3 is gradually turned over following VSV infection, which may serve as a mechanism to turn off NF-kB activation, and, meanwhile, it releases MAVS to form functional aggregates to induce IRF3 activation. Further studies are necessary to determine the molecular mechanism as to how CARMA3 is targeted for K48 ubiquitination and degradation, and targeting CARMA3 for degradation may help to reduce inflammation while enhancing the antiviral response. Therefore, understanding this mechanism may provide molecular insights for designing therapeutic agents for suppressing viral infectioninduced inflammation.
Antibodies against Flag (sc-807), IkBa (sc-371), IKKa (sc-7218), HA (sc-7392), His (sc-803), BCL10 (sc-5611), Myc (sc-40), IRF3 (sc-9028), NEMO (sc-8330), b-actin (sc-8432), and a-tubulin (sc8035) were purchased from Santa Cruz Biotechnology. Antibodies against p-IkBa (9246), p-IKKa/b (2681), Caspase 3 (9664), p-IRF3 (4947), p-TBK1 (5483), and TBK1 (3013) were from Cell Signaling Technology. The rabbit polyclonal antibody against Carma3 was homemade against the peptide VRGRILQEQARLVWVEC, matching to the C terminus of human and mouse CARMA3. The anti-MAVS antibody was kindly provided by Dr. Zhijian J. Chen (UT Southwestern). Oligonucleotide probes for NF-kB (E3291) and OCT-1 (E3241) were purchased from Promega. TRIzol (15596-026) and Superscript III First-Strand Synthesis system (18080051) were obtained from Invitrogen and DNase I kits (10104159001) were purchased from Roche Applied Science. Two times SYBR Green PCR Master Mix was purchased from Applied Biosystems. Influenza A/PR/8/34(H1N1) was purchased from Charles River Laboratories. NF-kBi and CHX were purchased from Sigma.
Cells were grown in DMEM (HEK293T, BHK, and MEF) containing 10% fetal bovine serum (FBS) at 37 C and 5% CO 2 except for MEF cells, which were incubated at 37 C and 8.5% CO 2 . BEAS-2B cells were purchased from ATCC, cultured with LHC-8 medium from Thermo Fisher in pre-coated culture dishes with BSA (Fisher Scientific), Fibronectin, and Collagen I bovine (Thermo Fisher).
Expression Plasmids FLAG-tagged human CARMA3 has been described previously (Sun and Lin, 2008) . HA-tagged mouse CARMA3 was cloned into pcDNA3 vector. RIG-I, MAVS, TBK1, IKKε, and IkBa SR plasmids were purchased from Addgene. FLAG-or Myc-tagged BCL10 has been described previously ; shRNA against CARMA3 and BCL10 were purchased from Sigma.
The real-time qPCR and ELISA assay were performed as recommended by the manufacturer's protocol.
The luciferase assay was performed as described previously (Blonska et al., 2005) . Briefly, 293T cells were seeded in triplicates in 12-well plates and transfected with 60 ng NF-kB-dependent luciferase (firefly) reporter plasmid and 6 ng EF1a promoter-dependent Renilla luciferase reporter together with 1.5 mg expression vectors for Carma3, TMEM43 WT or mutants, or vector controls. The transfected cells were cultured in DMEM containing 10% FBS for 16 hr. The cells were harvested and lysed. Luciferase activities in the cell lysates were measured by dual-luciferase kit (Promega).
These experiments were performed as described previously (Blonska et al., 2005) . Briefly, 1-5 3 10 6 cells were seeded, starved for 16 hr, and stimulated with various stimulators for the appropriate time. The cells were then lysed in a buffer containing 50 mM HEPES (pH 7.4), 250 mM NaCl, 1% Nonidet P-40, 1 mM EDTA, 1 mM Na 3 VO 4 , 1 mM NaF, 1 mM PMSF, and a protease inhibitor mixture (Roche Diagnostics). The cell lysates were subjected to SDS-PAGE and western blot or immunoprecipitated with appropriate specific antibodies. The immunoprecipitates were washed with lysis buffer four times eluted with 23 SDS loading buffer. The samples were boiled and separated on 10% SDS-PAGE and transferred to nitrocellulose membranes. Immunoblots were incubated with specific primary antibodies followed by horseradish peroxidase-conjugated secondary antibodies, and they were developed by the enhanced chemiluminescence method according to the manufacturer's protocol (Pierce).
These experiments were performed as described previously (Blonska et al., 2005) . Briefly, 1-5 3 10 6 cells were seeded, starved for 16 hr, and stimulated with various stimulators for the appropriate time, and nuclear extracts were prepared. Nuclear extracts (5-10 mg) were incubated with 32 P-labeled probes at room temperature for 15 min. The samples were separated on a native Tris-Borate-EDTA polyacrylamide gel, which was dried at 80 C for 1 hr, and exposed to X-ray film.
Lentivirus Infection for shRNA Knockdown Lentiviruses were packaged as described previously (Jiang et al., 2011b) . Briefly, HEK293T were co-transfected with shRNA and packaging vectors encoding VSV-G and DVPR using the calcium precipitation method. At 48 hr post-transfection, the supernatant was collected and applied to infect target cells. The infected cells were selected by puromycin for 3-4 days before the conduction of experiments.
These experiments were performed as described previously (Hou et al., 2011) . Briefly, cells were washed with ice-cold PBS twice and harvested by spinning at 600 3 g for 5 min. The cells were swollen in hypotonic buffer (10 mM Tris-Cl [pH 7.5], 10 mM KCl, 0.5 mM EGTA, 1.5 mM MgCl 2 , 1 mM PMSF, and 13 EDTA-free protease cocktail) for 2.5 min at 4 C. The swollen cells were homogenized for 40 strokes at 4 C and spun at 600 3 g for 5 min. The supernatant was transferred to a new tube and spun once more. The resulting supernatant was spun to pellet the mitochondria. The mitochondria pellet was resuspended in mitochondria-resuspending buffer (MRB) (20 mM HEPES-KOH [pH 7.4], 0.5 mM EGTA, 1 mM PMSF, and 13 EDTA-free protease cocktail) containing 0.8 M sucrose and pelleted by spinning at 7,000 3 g for 10 min. The pellet was resuspended and spun as above. The resulting pellet was resuspended, spun at 10,000 3 g for 10 min, and lysed in MRB buffer containing 2% CHAPs on ice for 15 min. The mitochondria lysates were mixed with 23 loading dye (13 TBE, 10% glycerol, 4% SDS, and 0.005% bromophenol blue) containing no reduction reagents.
SDD-AGE was performed as described previously with minor modifications (Halfmann and Lindquist, 2008) . Briefly, 1.5% agarose gel containing 0.1% SDS was prepared in 0.253 TAE buffer. The mitochondria lysates in 13 loading dye were loaded into the wells. After electrophoresis in the running buffer (0.253 TAE and 0.1% SDS) for 7 hr with a constant voltage of 30 V at room temperature, the gel was rinsed with water and transferred to Immobilon membrane (Millipore) by butterfly transfer in 13 TBS (50 mM Tris.Cl [pH 7.5] and 150 mM NaCl). The membrane was then blotted with specific antibody as immunoblotting.
VSV and Influenza Virus Mice Model BALB/c mice (6-8 weeks old, n R 5 per group) were intranasally inoculated with 20 ml H1N1 influenza virus (PFUs = 600 per mice) in PBS under isofluorane sedation. Survival and body weight changes were recorded daily for 14 dpi. Animals that showed signs of severe disease and weight loss >25% of their initial body weight were considered moribund and were humanely sacrificed. Lungs were harvested at 4 dpi. All animal experiments and procedures were conducted under the protocol and were approved by the Institutional Animal Care and Use Committee (IACUC) at the University of Texas MD Anderson Cancer Center. For the VSV mouse model, 6-to 8-week-old BALB/c mice (n R 5 per group) were intranasally inoculated with 20 ml VSV (plaque-forming units [PFUs] = 1 3 10 7 per mouse) in PBS under isofluorane sedation. Mouse serum tissues (lung, brain, and spleen) were harvested 2 days after infection.
The VSV plaque assay protocol was previously described. Briefly, cells were infected with VSV (MOI = 3). At 1 hr post-infection, the cells were washed twice with serum-free DMEM and replenished with complete medium. The supernatant was collected at different time points, diluted with serum-free DMEM, and used to infect BHK cells. At 1 hr post-infection, medium from the BHK cells was removed and replaced with complete medium containing 0.5% methylcellulose (Sigma-Aldrich) for 24 to 48 hr. BHK cells were fixed in fixation solution and stained with crystal violet. Plaques were counted and titers were calculated as PFU per milliliter. Triplicate experiments were performed, and the averages of the virus titers were calculated.
MDCKs were passed in high-glucose DMEM (10% FBS and 1% Pen-Strep) in a T175 (a 1:12 pass dilution every other day) at 37 C. Cells (2 3 10 4 ) were seeded in 96-well plates at the night before assay with 150 ml DMEM (FBS and Pen-Strep). The morning of the next day, influenza samples were thawed on ice and then homogenized, transfering a 600-ml sample into the first tube of the dilution series and serially diluting 184 ml across the tubes. Once samples were diluted across tubes, they were kept on ice and 100 ml sample was added to empty wells. Samples were incubated at 37 C for 1 hr and aspirated off. Then 150 ml DMEM with typsin (1 mg/ml) was added. Samples were incubated at 37 C for 3 days. After the 3 days, the media were aspirated off and 200 ml Crystal violet working stock (40 ml 1% Crystal Violet, 80 ml Methanol, and 300 ml H 2 O) was added. The working stock sat for 1 hr. Crystal violet was removed and samples were rinsed with water.
The statistical analysis was performed by Student's t test.
|
