Citrobacter rodentium NleB Protein Inhibits Tumor Necrosis Factor (TNF) Receptor-associated Factor 3 (TRAF3) Ubiquitination to Reduce Host Type I Interferon Production*

Interferon signaling plays important roles in both intestinal homeostasis and in the host response to pathogen infection. The extent to which bacterial pathogens inhibit this host pathway is an understudied area of investigation. We characterized Citrobacter rodentium strains bearing deletions in individual type III secretion system effector genes to determine whether this pathogen inhibits the host type I IFN response and which effector is responsible. The NleB effector limited host IFN-β production by inhibiting Lys63-linked ubiquitination of TNF receptor-associated factor 3 (TRAF3). Inhibition was dependent on the glycosyltransferase activity of NleB. GAPDH, a target of NleB during infection, bound to TRAF3 and was required for maximal TRAF3 ubiquitination. NleB glycosyltransferase activity inhibited GAPDH-TRAF3 binding, resulting in reduced TRAF3 ubiquitination. Collectively, our data reveal important interplay between GAPDH and TRAF3 and suggest a mechanism by which the NleB effector inhibits type I IFN signaling.


Interferon signaling plays important roles in both intestinal homeostasis and in the host response to pathogen infection. The extent to which bacterial pathogens inhibit this host pathway is an understudied area of investigation. We characterized Citrobacter rodentium strains bearing deletions in individual type III secretion system effector genes to determine whether this pathogen inhibits the host type I IFN response and which effector is responsible. The NleB effector limited host IFN-␤ production by inhibiting Lys 63 -linked ubiquitination of TNF receptorassociated factor 3 (TRAF3)
. Inhibition was dependent on the glycosyltransferase activity of NleB. GAPDH, a target of NleB during infection, bound to TRAF3 and was required for maximal TRAF3 ubiquitination. NleB glycosyltransferase activity inhibited GAPDH-TRAF3 binding, resulting in reduced TRAF3 ubiquitination. Collectively, our data reveal important interplay between GAPDH and TRAF3 and suggest a mechanism by which the NleB effector inhibits type I IFN signaling.
Many bacterial pathogens utilize a type III secretion system (T3SS) 2 to inject virulence proteins (effectors) into host cells to subvert various host functions (1). Characterizing the identity and mechanism of bacterial effectors expressed by attaching/ effacing pathogens, which include enterohemorrhagic Escherichia coli (EHEC), enteropathogenic E. coli (EPEC), and Citrobacter rodentium, has been a subject of extensive research in recent years (2). Effector subversion of host pathways regulated by the transcription factor NF-B has been relatively well studied (3), but less attention has been given to the potential inhibition of host interferon signaling, which also functions as an important early mediator of host defense.
Virally infected cells secrete interferon to protect other cells from subsequent infection (4). Type I IFNs (e.g. IFN-␣ and IFN-␤) bind to a common IFN-␣/␤ receptor (5), initiating a signal transduction cascade that ultimately induces the transcription of genes with promoters containing IFN-stimulated response elements (6). Type I IFNs induce antiviral activity via double-stranded RNA-dependent protein kinase and 2Ј,5Ј-oligoadenylate synthetase to promote viral mRNA degradation and inhibit translation, respectively (7).
Type I IFN production is stimulated not only by viral infection but also by bacterial components, including LPS and flagellin (8). LPS binding to Toll-like receptor 4 (TLR4) induces a type I IFN response mediated by the TRIF-related adaptor molecule and TIR-domain-containing adapter-inducing interferon-␤ adaptor proteins (9). TRIF-related adaptor molecule-TIR-domain-containing adapter-inducing interferon-␤ signaling results in the activation of TRAF family member-associated NF-B activator-binding kinase (TBK1) and inducible IB kinase (IKK⑀ (10)). Serine phosphorylation of interferon-regulatory factor 3 (IRF3) is mediated by TBK1/IKK⑀ to induce IRF3 dimerization and nuclear translocation (11). Phosphorylated IRF3 (and IRF7) dimers then associate with AP-1, high-mobility group proteins, and NF-B to form an IFN-␤ enhanceasome (12).
IFNs are important for maintaining intestinal homeostasis and for responding to pathogen infection (13). Accordingly, pathogens have also evolved strategies to interfere with host type I IFN signaling. For example, the anthrax lethal factor inhibits host IFN production by cleaving MAPKK6 (14) and by inhibiting STAT1 and ISGF3 activation (15). The deubiquitinating protease activity of Yersinia YopJ blocks TLR induction of type I IFN at the level of TAK1 activation by inhibiting TRAF3/6 ubiquitination (16). Vaccinia virus protein K7 prevents IRF3 activation by inhibiting the interaction between DEAD box protein 3 and IKK⑀ (17). The Ebola virus VP35 protein blocks TBK1/IKK⑀ interaction with IRF3/7 (18).
A recent study found that IFN-␤ is induced by EPEC infection and that the T3SS effector NleD both inhibits IFN-␤ induction and enhances TNF expression to promote barrier disruption (19). However, whether C. rodentium T3SS effectors may also inhibit host IFN-␤ induction is unclear. Here, we screened C. rodentium strains bearing deletions in individual effector genes to determine the extent to which this pathogen might inhibit the host IFN-␤ response. We found that the T3SS effector NleB limits host IFN-␤ production by inhibiting TRAF3 Lys 63 -linked ubiquitination.

Results
NleB Inhibits the Generation of a Type I IFN Response-To determine whether C. rodentium T3SS effectors inhibit the host type I IFN response, we monitored the survival of a recombinant vesicular stomatitis virus (VSV-G) expressing GFP (20). HeLa cells were first infected with C. rodentium strains possessing (WT) or lacking (⌬escN) a functional T3SS for 2 h or were instead treated with LPS for 30 min as a positive control. HeLa cell supernatants were then removed, filtered, and applied to Vero cells, which are highly responsive to exogenous IFN-␤. After incubation for 12 h with the supernatants, the Vero cells were infected with VSV-G/GFP for 30 h. Cells were then fixed, and viral loads were quantified by measuring GFP intensity.
Transferring supernatants from cells infected with C. rodentium ⌬escN inhibited VSV-G/GFP to levels similar to those induced by LPS treatment. By contrast, transferring supernatants from cells infected with WT C. rodentium had no discernible ability to inhibit VSV-GFP growth (Fig. 1A). These data suggested that a T3SS effector inhibits the production of a host factor involved in virus inhibition. We then infected HeLa cells with C. rodentium strains lacking individual T3SS effector genes and screened the cell supernatants for antiviral activity. Among the 12 individual effector deletion strains screened (Fig.  1B), ⌬nleB inhibited virus replication most significantly (Figs. 1,  A and B), with a minor phenotype also observed for ⌬nleE.
To test the hypothesis that this phenotype was due to the release of type I IFN from the infected HeLa cells, we used an antibody that blocks the IFN-␣/␤ receptor. Pretreating Vero cells with this antibody blocked the ability of HeLa cell supernatants to inhibit virus replication (Fig. 1C). Thus, C. rodentium induces a host IFN-␤ response, and the T3SS effector NleB blocks the transduction of this response.
NleB Inhibits IFN-␤ and RANTES Gene Transcription-We then conducted in vitro experiments to determine the impact of NleB on IFN-␤ signaling pathways. To determine whether NleB inhibits transcription of the IFN-␤ gene, we performed quantitative RT-PCR analysis of IFNB1 transcript abundance after infection with WT, ⌬escN, or ⌬nleB C. rodentium. Although infection with ⌬escN stimulated IFNB1 expression similar to LPS treatment, infection with the WT did not ( Fig. 2A). This phenotype was attributable to NleB, as deleting nleB relieved the inhibition of IFNB1 gene transcription observed with WT C. rodentium. We obtained similar data for the chemokine gene RANTES (CCL5, Fig. 2B).
IRF3 is a transcription factor that binds IFN-␤ and other innate immune gene promoters (21). Because NleB inhibited IFNB1 and RANTES expression, which are both direct targets of IRF3, we tested whether IRF nuclear abundance was altered after bacterial infection. By fractionating cell lysates after bacterial infection, we observed that, although WT C. rodentium inhibited the accumulation of nuclear IRF3, infection with ⌬escN and ⌬nleB did not (Fig. 2C). We reached similar conclusions after using immunofluorescence microscopy and scoring the number of cells with predominantly nuclear IRF3 (Fig. 2, D and E).
NleB Reduces the Nuclear Abundance of STAT1-IFN-␤ stimulation results in the activation of the JAK-STAT pathway, which has a critical role in regulating the immune response to bacterial infection. After cytokine stimulation and activation of JAK, STATs, including STAT1, become phosphorylated and translocate to the nucleus, where they can activate the expression of immune response genes (22). To determine whether NleB inhibits p-STAT1 nuclear translocation, we evaluated the relative abundance p-STAT1 in HeLa cells infected with C. rodentium strains with or without IFN-␤ stimulation. We fractionated the cells to separate nuclear from cytoplasmic components. As expected, p-STAT1 nuclear abundance significantly increased after stimulation with IFN-␤ (Fig. 2, F and G). p-STAT1 nuclear abundance was reduced in samples infected with WT C. rodentium even after stimulation with IFN-␤ (Fig. 2, F and G). By contrast, ⌬nleB C. rodentium failed to inhibit p-STAT1 nuclear translocation, indicating that NleB inhibits IFN-␤-induced p-STAT1 nuclear translocation (Fig. 2, F and G). NleB Inhibits Lys 63 Ubiquitination of TRAF3-TLRs are pattern recognition receptors that sense microorganisms and are critical for the host innate immune system. A subset of liganded TLRs activates intrinsic signaling pathways to induce type I IFN production. We found that NleB also inhibited the activation of an IFN-␤-dependent luciferase reporter in the presence of poly(I:C), a synthetic double-stranded RNA used for TLR3-dependent stimulation of type I IFN, in an NleB dose-dependent manner (Fig. 3A).
Differential ubiquitination of TNF receptor (TNFR)-associated factors can selectively activate or suppress the expression of different groups of cytokines (23). For example, TRAF3 Lys 63 ubiquitination typically activates IFN responses, whereas TRAF3 Lys 48 ubiquitination mediated by TRAF6 typically induces pro-inflammatory cytokine expression (24). Lys 48ubiquitin linkages typically result in proteasome-mediated protein degradation, whereas Lys 63 -ubiquitin linkages are generally associated with regulating protein subcellular localization and activation (24).
We then hypothesized that NleB might inhibit TRAF3 ubiquitination. We transfected RAW264.7 cells with plasmids expressing TRAF3-FLAG and with ubiquitin-HA mutants that contain lysine only at a single position (e.g. Lys 48 or Lys 63 ). After 36 h, cells were treated with LPS (100 ng/ml, 30 min) to induce TRAF signaling. TRAF3-FLAG was immunoprecipitated, and these samples were immunoblotted for HA to detect ubiquitin conjugates. As expected, LPS treatment induced significant Lys 48 -and Lys 63 -linked TRAF3 ubiquitination (Fig. 3B). Quantification of these data is shown in Fig. 3, C and D.
Performing these studies with cells that were subsequently infected with C. rodentium strains possessing or lacking NleB revealed the selective impact of NleB on Lys 63 -linked TRAF3 ubiquitination. Although Lys 48 -linked TRAF3 ubiquitination did not differ among infection conditions, Lys 63 -linked TRAF3 ubiquitination was significantly inhibited in cells infected with either WT or ⌬nleB/pnleB C. rodentium in contrast to its ubiquitination after infection with ⌬nleB C. rodentium (Fig. 3,  B-D). We corroborated these data by transfecting NleB in the absence of bacterial infection. Transfecting NleB into RAW264.7 cells inhibited Lys 63 -linked but not Lys 48 -linked TRAF3 ubiquitination (Fig. 3E). Overall, these data suggest that NleB selectively inhibits Lys 63 -linked TRAF3 ubiquitination. NleB Targets GAPDH to Regulate Lys 63 Ubiquitination of TRAF3-We showed previously that NleB functions as a translocated glycosyltransferase enzyme that covalently modifies host proteins with GlcNAc to subvert their normal functions (25). Specifically, we found that NleB disrupts TRAF2 signaling, leading to inhibition of the pro-inflammatory NF-B pathway (25). We also revealed that the glycolysis enzyme GAPDH functions as a co-activator of TRAF2 activity. The modification of GAPDH with GlcNAc by NleB prevents GAPDH from binding to and activating TRAF2, leading to a reduced NF-B response to infection (25).
To determine the extent to which the phenotypes we observed with TRAF3 were also dependent upon GAPDH, we first determined whether GAPDH interacts with TRAF3. By transfecting TRAF3-FLAG, we observed that GAPDH immunoprecipitated with TRAF3, similar to its ability to immunoprecipitate with TRAF2 (Fig. 4A). Co-incubation of GAPDH with NleB almost abolished the binding of GAPDH to TRAF3, at low concentrations of TRAF3, as shown in pulldown assays and in ELISAs (Fig. 4, B and C), similar to our previous observations with GAPDH binding to TRAF2 (25). Although a ϳ20% complex formation in the presence of NleB was achieved, this is contrast to ϳ100% complex formation in the absence of NleB (Fig. 4C).
Then we investigated whether the GAPDH-TRAF3 interaction is important for TRAF3 activity in response to LPS. We transfected GAPDH siRNA targeting the GAPDH 3Ј UTR and complemented these knockdown cells with two GAPDH-Myc plasmids, either WT or a C150S GAPDH mutant (which does not interact with TRAFs (25)). After LPS treatment, Lys 63 -Ub levels on TRAF3 were significantly elevated, as expected (Fig.  4D). Complementing GAPDH knockdown cells with WT, but not with C150S GAPDH, supported TRAF3 ubiquitination. Co-transfecting WT NleB, but not the glycosyltransferase-deficient mutant NleB(AAA) (25), significantly reduced Lys 63linked TRAF3 ubiquitination and inhibited the association of GAPDH with TRAF3 (Fig. 4D).

Discussion
NleB is a translocated glycosyltransferase enzyme that was originally studied for its ability to suppress NF-B activation (26,27). We found previously that NleB disrupts TRAF2 signaling by modifying GAPDH with GlcNAc and preventing GAPDH from binding to and activating TRAF2 (25). The EPEC NleB1 effector modifies arginine residues of host proteins (N-GlcNAcylation) rather than the more typical targeting of serine/threonine residues (O-GlcNAcylation (28,29)). In particular, it was found that EPEC NleB1 glycosylates the death domains of TRADD, FADD, RIPK1, and TNFR1, blocking the assembly of the TNFR1 complex. This N-GlcNAcylation disrupts TNF signaling in EPEC-infected cells, thus impacting NF-B activation, apoptosis, and necroptosis (28,29). A recent paper defined numerous amino acids that are important to the N-GlcNAcylation of FADD by EPEC NleB1 and also used the C. rodentium infection model to show the role of these amino acids in virulence (30). To date, the studies that have evaluated the arginine-GlcNAcylation of substrates by NleB orthologs have only described a single addition site for each protein. This modification occurs on Arg 117 in FADD (29) and on Arg 235 in TRADD (28). Labeling of an arginine residue by NleB1 was also detected for TNFR1 and RIPK1 (28). It is conceivable that some NleB orthologs could label other substrates, including GAPDH, on single or multiple arginine residues, a topic we are investigating in ongoing studies.
Here we observed an additional role of C. rodentium NleB affecting type I IFN signaling by inhibiting Lys 63 -linked TRAF3 ubiquitination. EPEC suppresses type I IFN production using the T3SS effector NleD (19), which we did not observe in our screening of C. rodentium mutants. We used the VSV-G/GFP system as the readout for type I IFN production, in which VSV replication is very sensitive to type I IFN secreted from bacterially infected cells. By contrast, Long et al. (19) measured IFN-␤ mRNA levels in Caco2 cells 3 h after EPEC infection. Because the expression of innate immunity genes is dynamic, mRNA levels assayed at a single time point might not fully reflect type I IFN production during bacterial infection. Additionally, the activities of the EHEC, EPEC, and C. rodentium NleB proteins may be functionally distinct (25,28,29).
TRAF3, a ubiquitin ligase, together with TRAF6, are recruited by the mitochondrion adaptor mitochondrial antiviral-signaling protein to activate the non-canonical IKK-related kinase TBK1 with various adaptor proteins, including TRAF family member-associated NF-B activator, NAP1, and NF-B essential modulator. The TBK1 complex induces the phosphorylation and dimerization of IRF3, which is required for IRF3 nuclear translocation. IRF3 is a master transcriptional regulator of type I IFN gene expression. TLR4 activation induces Lys 63linked TRAF3 ubiquitination, which is required for IRF3 activation.
GAPDH is a stress sensor involved in many biological processes, including immune signaling and apoptosis (31). We showed that the interaction between GAPDH and TRAF2 is enhanced by TNF and is required for maximal NF-B activation after TNF stimulation (25). Here we found that GAPDH also interacts with TRAF3 and enhances TRAF3 Lys 63 -linked ubiquitination. Given that TRAF family members are widely involved in mammalian immune signaling pathways, our data suggest that GAPDH may have an unappreciated role in regulating the immune system. There are six distinct TRAF proteins in mammals, each containing a ring finger domain (except TRAF1), a zinc finger motif, an amino-terminal TRAF domain, and a carboxy-terminal TRAF domain (TRAF-C). TRAF-C is highly conserved among all TRAFs and is responsible for the association of TRAFs with other proteins. It will be interesting to investigate whether GAPDH interacts with other TRAFs under different conditions. NleB inhibited the TRAF3 and GAPDH interaction by modifying GAPDH with GlcNAc. Other studies have shown that EPEC NleB1 glycosylates the death domains of TRADD, FADD, and other death domain-containing proteins (28,29). Although these results are able to explain NleB function in the TNF/ NF-B signaling pathway, they fail to explain the function of NleB in the TLR4/type I IFN pathway we report here. The role of GAPDH in T cell activation has been described recently (32). GAPDH functions as a "switcher" of T cell activation by binding to AU-rich elements within the 3Ј UTR of IFN-␥ mRNA to suppress the translation of IFN-␥ (32). Therefore, selective targeting of GAPDH might provide an interesting framework with which to understand the multiple effects of NleB on immune signaling pathways. Overall, our data suggest that C. rodentium NleB affects both NF-B and interferon signaling.