YopJ targets TRAF proteins to inhibit TLR-mediated NF-κB, MAPK and IRF3 signal transduction

Authors

  • Charles R. Sweet,

    1. Division of Infectious Diseases and Immunology, Departments of Medicine, University of Massachusetts Medical School, Worcester, MA 01605, USA.
    Search for more papers by this author
  • Joseph Conlon,

    1. Division of Infectious Diseases and Immunology, Departments of Medicine, University of Massachusetts Medical School, Worcester, MA 01605, USA.
    Search for more papers by this author
  • Douglas T. Golenbock,

    1. Division of Infectious Diseases and Immunology, Departments of Medicine, University of Massachusetts Medical School, Worcester, MA 01605, USA.
    Search for more papers by this author
  • Jon Goguen,

    1. Department of Molecular Genetics and Microbiology, University of Massachusetts Medical School, Worcester, MA 01605, USA.
    Search for more papers by this author
  • Neal Silverman

    Corresponding author
    1. Division of Infectious Diseases and Immunology, Departments of Medicine, University of Massachusetts Medical School, Worcester, MA 01605, USA.
    Search for more papers by this author

*E-mail neal.silverman@umassmed.edu; Tel. (+1) 508 856 5826; Fax (+1) 508 856 5463.

Summary

The Yersinia pestis virulence factor YopJ is a potent inhibitor of the NF-κB and MAPK signalling pathways, however, its molecular mechanism and relevance to pathogenesis are the subject of much debate. In this report, we characterize the effects of this type III effector protein on bone fide signalling events downstream of Toll-like receptors (TLRs), critical sensors in innate immunity. YopJ inhibited TLR-mediated NF-κB and MAP kinase activation, as suggested by previous studies. In addition, induction of the TLR-mediated interferon response was blocked by YopJ, indicating that YopJ also inhibits IRF3 signalling. Examination of the NF-κB signalling pathway in detail suggested that YopJ acts at the level of TAK1 (MAP3K7) activation. Further studies revealed a YopJ-dependent decrease in the ubiquitination of TRAF3 and TRAF6. These data support the hypothesis that YopJ is a deubiquitinating protease that acts on TRAF proteins to prevent or remove the K63-polymerized ubiquitin conjugates required for signal transduction. Our data do not directly address the alternative hypothesis that YopJ is an acetyltransferase that acts on the activation loop of IKK and MKK proteins, but support the conclusion that the critical function of YopJ is to deubiquinate TRAF proteins.

Introduction

The Gram-negative bacterium Yersinia pestis, the etiologic agent of plague (Perry and Fetherston, 1997), requires a type III secretion system for virulence. This molecular syringe is responsible for the delivery of six Yersinia Outer Proteins (Yops) directly into the cytoplasm of host cells, in particular innate immune cells such as macrophages (Marketon et al., 2005; Cornelis, 2006). The Yops collaborate to potently inhibit many aspects of the innate immune response, including phagocytosis and cytokine production, allowing the bacteria to thrive in extracellular tissue (Kerschen et al., 2004; Viboud and Bliska, 2005). YopJ (YopP in Yersinia enterocolitica) is thought to interfere with immune signal transduction and thereby inhibit the production of pro-inflammatory cytokines (Boland and Cornelis, 1998; Palmer et al., 1998; 1999; Schesser et al., 1998), induce macrophage apoptosis (Mills et al., 1997; Denecker et al., 2001; Ruckdeschel et al., 2001; Zhang and Bliska, 2003), and prevent CD8 T-cell responses (Trulzsch et al., 2005), thus contributing to successful infection (Monack et al., 1997; 1998; Trulzsch et al., 2004).

YopJ is homologous to Clan CE cysteine proteases, a family of proteases composed of de-ubiquitinating and de-SUMOylating enzymes (Orth, 2002; Rawlings et al., 2004). An alanine substitution mutant of the catalytic cysteine (YopJC172A) is unable to inhibit signalling (Orth et al., 2000). Previously, YopJ was proposed to function as a de-SUMOylating protease, although SUMO is not known to be required in the signal transduction pathways targeted by YopJ (Orth et al., 2000; Orth, 2002; Yoon et al., 2003). More recently, it was demonstrated that recombinant YopJ has de-ubiquitinating (DUB) activity, not de-SUMOylating activity, and two reports concluded that YopJ functions by de-ubiquitinating a variety of signal transduction proteins (Haase et al., 2005; Zhou et al., 2005). On the other hand, two recent reports concluded that YopJ is an acetyltransferase that acetylates host mitogen-activated protein kinase (MAPK) kinases (MKKs), thereby inactivating innate immune pathways (Mittal et al., 2006; Mukherjee et al., 2006). These apparently contradictory findings have not yet been resolved, and it remains possible that YopJ has both protease and acetyltransferase activities.

Innate immune signalling is often initiated by Toll-like receptors (TLRs). (Janeway and Medzhitov, 2002). Humans express 10 TLR proteins that have evolved to recognize conserved microbial components derived from bacteria, viruses, fungi or parasites (Akira, 2006). For example, TLR4 is required for the response to lipopolysaccharide (LPS) from the outer membrane of Gram-negative bacteria, TLR2 is required for the response to various bacterial lipopeptides, and TLR3 responds to dsRNA, a signature of viral infection. Stimulation of TLRs results in the activation of nuclear factor kappa B (NF-κB) (Silverman and Maniatis, 2001) and some MAPK pathways (Dong et al., 2002). In addition, some TLRs (TLR3, 4, 7, 8 and 9) also activate type I interferon responses (Kawai and Akira, 2006; Seth et al., 2006) through activation of the interferon response factor (IRF) transcription factors (Taniguchi et al., 2001).

It has been assumed and inferred from earlier experiments that YopJ inhibits TLR-mediated NF-κB and MAPK activation, but this issue has not been examined directly. Moreover, the effect of YopJ on TLR-induced IRF activation has not been examined. In this report, we present a detailed analysis of the effect of YopJ on TLR-mediated signalling, showing that YopJ interferes with ubiquitination of the TNF receptor-associated factor (TRAF) proteins and effectively inhibits NF-κB, MAPK and IRF activation. Based on these data and the known protein homology of YopJ to deubiquitinating enzymes, we conclude that YopJ inhibits innate immune signalling pathways by removing ubiquitin from TRAF6 (in NF-κB and MAPK signalling) and TRAF3 (in interferon signalling).

Results

YopJ inhibits cytokine production upon activation of TLR signalling

In order to determine if YopJ inhibits TLR-induced cytokine production, flow cytometry was used to assay levels of the cytokine IL-8 in HEK293 cells that express TLR2 or TLR4. These cells were transiently transfected with a Y. pestis YopJ-YFP expression plasmid, then stimulated with either Pam2CysK4 or LPS, as appropriate, and assayed for intracellular IL-8 production. As shown in Fig. 1A and B, YopJWT prevented IL-8 production in response to stimulation of either TLR2 or TLR4; significant TLR-stimulated IL-8 production was observed in the cells transfected with the catalytically inactive YopJC172A but not in the cells transfected with YopJWT. Note, these data examine only the YopJ-expressing cells, selected for analysis by flow cytometric detection of the YFP-tag. Control samples assessing cell viability via the early apoptotic marker Annexin V or the cell death marker To-Pro-3 show that transient expression of YopJ does not induce cell death during the time-course of these experiments (data not shown). Also, TLR expression levels were not decreased by the presence of YopJ, as measured by flow cytometry of enhanced cyan fluorescent protein (CFP)-tagged TLRs (data not shown). Furthermore, extent of stimulation of IL-8 or phospho-p38 was the same in empty-vector transfection as in C172A transfection (data not shown). These results directly demonstrate that YopJ strongly inhibits TLR-mediated cytokine production, although the exact signalling pathways affected are not clear from these data, because IL-8 expression is dependent on both NF-κB and MAPK signalling cascades (Hoffmann et al., 2002).

Figure 1.

YopJ inhibits TLR-mediated hIL-8 production and phosphorylation of pp38 – HEK293 cells stably expressing (A) TLR2 or (B) TLR4 were transfected with expression plasmids encoding YopJWT or YopJC172A covalently tagged with YFP. After treatment with Pam2CysK4 or LPS and Brefeldin A for 5 h, fixed and permeabilized cells were stained with anti-hIL-8 antibody. All samples shown have been gated for YFP-positive cells to examine only those cells expressing YopJ. The stimulated cells expressing YopJC172A (blue), demonstrate hIL-8 production whereas the cells expressing YopJWT (red) do not. Grey denotes unstimulated cells. In C and D, HEK293 cells stably expressing TLR4 were transfected with YopJWT or YopJC172A covalently tagged with YFP. After treatment with LPS for 2 h (green) or anisomycin for 30 min (blue), fixed and permeabilized cells were stained with antiphospho-p38 antibody. The LPS and anisomycin stimulated cells expressing YopJC172A (C), demonstrate phosphorylation of p38, whereas the cells expressing YopJWT (D) do not. Red denotes unstimulated cells. The small activated populations in D represent residual YopJ-YFP negative cells that were not excluded from the gate.

YopJ inhibits TLR4-induced activation of p38 MAPK

In order to directly examine the effect of YopJ on TLR-induced MAPK signalling, a similar flow cytometric approach was used. For these experiments, HEK293 cells expressing TLR4 were again transiently transfected with YopJ conjugated with enhanced yellow fluorescent protein (YFP) and activation of the MAPK p38 following LPS stimulation was monitored with a phospho-specific antibody. As a positive control, cells were stimulated with anisomycin, a cell stressor and established TAK1-dependent activator of p38 (Ninomiya-Tsuji et al., 2003). Both LPS- and anisomycin-induced p38 phosphorylation were inhibited by YopJWT (Fig. 1D), while p38 phosphorylation was unaffected in cells expressing YopJC172A (Fig. 1C). Note that some residual activity (the small positive peak) is still present in 1D, presumably due to the inability to completely exclude YFP-negative cells from the YFP-positive gate.

YopJ inhibits TLR-mediated NF-κB activation

In addition to MAPK signalling, TLR stimulation activates NF-κB. To examine the effect of YopJ on TLR-mediated NF-κB activation, we employed a dual luciferase reporter approach (Radons et al., 2002). In this assay, expression plasmids for YopJWT or YopJC172A were transiently transfected along with κB and control reporters. To establish this system, the inhibitory effect of YopJ on TNF-induced NF-κB signalling was first examined. As suggested in previous studies (Orth et al., 1999; Palmer et al., 1999; Spiik et al., 1999), YopJWT dose-dependently inhibited TNF-mediated NF-κB activation (Fig. 2A–C, middle panels). A similar effect was observed on NF-κB stimulation by IL-1β (data not shown).

Figure 2.

YopJ inhibits TLR-mediated NF-κB activation – HEK293 cells stably transfected with (A) TLR2, (B) TLR3 or (C) TLR4 were transfected with expression plasmids encoding YopJWT (black) or YopJC172A (grey) and assayed for NF-κB induction using a dual luciferase assay system, as described in Experimental procedures. All data points were gathered in triplicate and averaged, then converted to fold-activation. Error is reported as standard deviation. YopJ dose-dependently inhibits NF-κB activation induced by a TLR receptor ligand or TNFα stimulation, but not stimulation by transfection of 50 ng of p65 (RelA) expression plasmid.

Having established the dual luciferase reporter system as an effective assay to monitor YopJ-mediated inhibition of NF-κB, we then examined TLR signalling. YopJ significantly and dose-dependently inhibited NF-κB activation following stimulation of TLR2 (Fig. 2A), TLR3 (Fig. 2B) or TLR4 (Fig. 2C) with Pam2CysK4, poly[I:C] or LPS respectively. As reported previously (Collier-Hyams et al., 2002), YopJ did not prevent the activation of the κB-reporter mediated by overexpression of the NF-κB subunit p65/RelA. Together with the IL-8 and phospho-p38 data, these results demonstrate that YopJ potently inhibits signalling events driven by TLR stimulation, and does so prior to activation/translocation of the transcription factor itself.

YopJ blocks NF-κB prior to TAK1 (MAP3K7)

In order to further characterize the NF-κB inhibitory activity of YopJ, its ability to inhibit NF-κB activation induced by expression of TLR signalling pathway components was similarly assayed. YopJ blocked the activation of NF-κB induced by expression of MyD88, MAL, TRAF6 or TAB2 (Fig. 3A–D), however, it did not block activation caused by expression of IκB kinase (IKK) β or the p65/RelA subunit of NF-κB itself (Fig. 3F and G). These results are consistent with previously published reports (Orth et al., 1999; 2000) and expand upon them using bona fide components of the TLR-NF-κB pathway.

Figure 3.

YopJ acts to prevent the activation of TAK1 – HEK293 cells were transfected with expression plasmids encoding YopJWT (black) or YopJC172A (grey), and assayed for NF-κB induction by dual luciferase system, as described in Experimental procedures. All data points were gathered in triplicate and averaged, then converted to fold-activation. Error is reported as standard deviation. YopJ potently inhibits NF-κB stimulation induced by transfection of 20 000 cells with (A) 50 ng MyD88, (B) 50 ng MAL, (C) 50 ng TRAF6, (D) 50 ng TAB2, but not by (E) 50 ng TAK1/50 ng TAB2, (F) 50 ng of IKKβ or (G) 50 ng of p65 (RelA).

While expression of the MAPK kinase kinase (MAP3K) TAK1 did not activate NF-κB signalling (data not shown), induction of this pathway by cotransfection of TAB2 and TAK1 was insensitive to YopJ (Fig. 3E). This result suggests that overexpression of both TAK1 and TAB2 together, but not TAB2 alone, circumvents the inhibitory activity of YopJ. Another group has recently reported that TAB2-induced NF-κB activation is dependent on TRAF6 ubiquitination and the subsequent assembly of TRAF6 with the TAK1 and IKK complexes (Kishida et al., 2005). This observation is consistent with our data that TAB2-induced activity is blocked by the proposed deubiquitination of TRAF6 by YopJ. In addition, our data suggest that coexpression of TAK1 and TAB2 alleviates the dependence on TRAF ubiquitination, perhaps by increasing the levels of both proteins such that assembly of TRAF6-ubiquitin complex is no longer required.

The literature contains contradictory reports regarding the ability of YopJ to inhibit NF-κB induced by the expression of IKKβ. Some studies have concluded that YopJ does not block IKKβ-induced signalling (Collier-Hyams et al., 2002; Haase et al., 2005) an observation supported by our results (Fig. 3F). On the other hand, other recent reports found that YopJ could inhibit IKKβ-induced signalling (Zhou et al., 2005; Mittal et al., 2006; Mukherjee et al., 2006). We speculated that these conflicting results could be due to different relative expression levels of IKKβ or YopJ used in these publications. To address this, we transfected various doses of the IKKβ and YopJ expression vectors, and found that YopJ expressed from pSFFV inhibited IKKβ signalling when the relative dose of IKKβ is substantially lower than the dose used in Fig. 3 (Fig. 4A–D). However, immunoblot analysis of protein expression suggests that the inhibition observed at low doses of IKKβ is artefactual, as coexpression with YopJ caused a significant drop in the levels of IKKβ protein at these dosages (when expressed behind the standard CMV promoter on pCDNA3). Similar observations were also made using CMV-driven p65/RelA (data not shown). In fact, we have consistently observed that YopJ interferes with expression from CMV-based expression plasmids. However, this problem is only significant at plasmid doses well below those typically used in our experiments. To ensure this confounding factor did not affect our studies, most experiments use the pEF-Bos vector based on the YopJ-insensitive elongation factor 1α (EF1α) promoter (Mizushima and Nagata, 1990). In addition protein expression levels were rigorously controlled in all experiments, particularly in the few cases where a CMV-based vector was necessary. To wit, TAK1, IKKβ and p65 shown in Figs 3 and 4 and TRAF6 shown in Fig. 6A and B were expressed from CMV promoter plasmids (TRAF6 in Figs 3 and 6C, however, was expressed from pEF-Bos). Control immunoblots showing expression level are shown in Figs 4 (IKKβ) and 6 (TRAF6) and in supplemental figure S1 (TAK1, IKKβ and p65).

Figure 4.

Low dose CMV-based expression of IKKβ is suppressed by YopJ – HEK293 cells were transfected with expression plasmids encoding YopJWT (black) or YopJC172A (grey), and assayed for NF-κB induction by dual luciferase system, as described in Experimental procedures. All data points were gathered in triplicate and averaged, then converted to fold-activation. Error is reported as standard deviation. YopJ potently inhibits NF-κB stimulation induced by transfection of (A) 2.5 ng of IKKβ or (B) 10 ng of IKKβ but not (C) 25 ng of IKKβ or (D) 50 ng of IKKβ (per 20 000 cells). This inhibition is caused not by an effect on signalling pathways but by a striking YopJWT-dependent loss of IKKβ protein expression at low doses of IKKβ plasmid transfection, as shown by immunoblotting lysates for FLAG-tagged IKKβ protein (E) (*ng plasmid DNA per 20 000 cells, as in A–D).

Figure 6.

TRAF6 lacks ubiquitination in the presence of YopJ – HEK293 cells were transfected with expression plasmids encoding TRAF6, YopJWT or YopJC172A Western blots of immunoprecipitated TRAF6 were probed with anti-ubiquitin and anti-FLAG antibodies as shown in the figure. TRAF6 ubiquitination is demonstrated both by anti-ubiquitin (A) and anti-FLAG (B), and is prevented by YopJ cotransfection. (C) HEK293 cells were cotransfected with TRAF6, YopJWT or YopJC172A, and HA-tagged ubiquitin K63R or K48R mutants. Western blots of immunoprecipitated TRAF6 were probed with anti-HA antibody, revealing that YopJ targets only K63-polymerized ubiquitin for removal from TRAF6.

YopJ inhibits TLR3-mediated IRF3 activation

In addition to NF-κB and MAPK signalling, TLR3 also induces interferon-mediated responses through activation of the transcription factor IRF3. To examine the effect of YopJ on this signalling pathway, dual–luciferase reporter assays were performed using a 5X-ISRE (interferon sensitive response element) reporter. (Note, this reporter responds to both IRF3, stimulated by TLR3, and to ISGF3, a STAT1/STAT2/IRF9 complex stimulated by the type I interferon receptor). Transient coexpression of YopJWT, but not YopJC172A, resulted in inhibition of TLR3-mediated IRF activation (stimulated by extracellular poly[I:C]), but not ISGF3 activation (stimulated by IFNβ ( Fig. 5A and B). This inhibition of TLR-interferon signalling was unexpected, as the TLR3-induced interferon signalling pathway has significant differences from the NF-κB and MAPK pathways, diverging from these pathways at the most receptor-proximal step.

Figure 5.

YopJ inhibits TLR-mediated IRF activation – HEK293 cells or HEK293 cells stably transfected with TLR3 were transfected with expression plasmids encoding YopJWT (black) or YopJC172A (grey) and assayed for ISRE induction by dual luciferase system, as described in Experimental procedures. All data points were gathered in triplicate and averaged, then converted to fold-activation. Error is reported as standard deviation. YopJ potently inhibits ISRE activation by (A) poly[I: C], but not by (B) IFNβ or (C) Sendai virus.

Sendai virus stimulation of the interferon response is not mediated by TLR3, but instead relies on the intracellular receptor RIG-I (Yoneyama et al., 2004; Kato et al., 2005; Li et al., 2005), a helicase protein that recognizes 5′-triphosphorylated RNA (Hornung et al., 2006; Pichlmair et al., 2006). RIG-I also activates IRF3-dependent ISRE activity, however, YopJ is unable to inhibit this response (Fig. 5C). The RIG-I competitor Lgp2 was used as a control inhibitor in this assay (Rothenfusser et al., 2005). This observation is a critical finding, demonstrating a fundamental difference between TLR3-mediated and RIG-I-mediated IRF activation.

YopJ de-ubiquitinates TRAF3 and TRAF6

The results presented above are consistent with the hypothesis that YopJ functions by de-ubiquitinating the TRAFs. TRAF6 autoubiquitination (with activating polyubiqutin chains polymerized through lysine 63) is critical for NF-κB signalling (Deng et al., 2000; Wang et al., 2001; Sun et al., 2004), and TRAF3 is likely to function similarly in the IRF pathway. To address the effect of YopJ on TRAF3 and TRAF6 directly, we expressed FLAG-tagged TRAFs in HEK293 cells. Ubiquitination of these proteins was analysed by immunoprecipitation followed by immunoblotting for endogenous ubiquitin or for the TRAF protein itself. Ubiquitinated forms of TRAF6 were readily observed in the absence of YopJ and were strongly diminished in the presence of YopJWT but not YopJC172A (Fig. 6A and B). Likewise, ubiquitination of TRAF3 was also diminished in the presence of wild-type YopJ (Fig. 7A and B). Next, we examined the ubiquitination of TRAF6 and TRAF3 with HA-tagged mutant versions of ubiquitin; either R48 ubiquitin (incapable of forming K48-polymerized ubiquitin) or R63 ubiquitin (incapable of forming K63-polymerized ubiquitin) was used. This experiment revealed that the presence of YopJ inhibits R48 mutant polyubiquitination, but not R63 mutant polyubiquitination (Figs 6C and 7C), consistent with the hypothesis that YopJ prevents signal transduction by interfering with K63 ubiquitination.

Figure 7.

TRAF3 lacks ubiquitination in the presence of YopJ – HEK293 cells were transfected with expression plasmids encoding TRAF3, YopJWT or YopJC172A Western blots of immunoprecipitated TRAF3 were probed with anti-ubiquitin, anti-HA or anti-FLAG antibodies, as shown. TRAF3 ubiquitination is demonstrated both by anti-ubiquitin (A) and anti-FLAG (B), and is prevented by YopJ cotransfection. The bottom panel of (B) shows the identical immunoblot as presented in the top panel, with the contrast adjusted digitally to accentuate TRAF3 polyubiquitination. (C) HEK293 cells were cotransfected with TRAF3, YopJWT or YopJC172A, and HA-tagged ubiquitin K63R or K48R mutants. Western blots of immunoprecipitated TRAF3 were probed with anti-HA antibody, revealing that YopJ targets only K63-polymerized ubiquitin for removal from TRAF3.

Discussion

YopJ in NF-κB and MAPK signalling

YopJ has been the focus of much research because of its ability to inhibit numerous signal transduction pathways. Recent biochemical analysis has demonstrated that YopJ is a de-ubiquitinating protease (Zhou et al., 2005) capable of targeting multiple distinct ubiquitination-dependent steps in NF-κB activation, such as the TRAF proteins, IKKγ/NEMO and IκBα. In this work, we directly examined the effect of Y. pestis YopJ specifically on discrete TLR-mediated signalling events. The use of HEK293 cell culture allows us to examine the activation of individual TLRs in isolation. This effort revealed that YopJ prevents the production of IL-8 in response to TLR activation, demonstrating that YopJ does indeed effectively target the signalling pathways downstream of these critical innate immune receptors.

To further explore this inhibition, we examined the effect of YopJ on TLR-induced MAPK activation and NF-κB activation, two pathways necessary for the production of IL-8 and other inflammatory cytokines. It is unclear which TLR receptor is most important during Yersinia infection. YopJ-induced apoptosis is mediated by TLR4 signalling but not TLR2 signalling in Yersinia pseudotuberculosis infection. This suggests TLR4 is the critical immune receptor in infection with this bacterium (Zhang and Bliska, 2003), although this situation may not translate directly to Y. pestis as Y. pestis has a modified lipid A structure at 37°C that prevents activation of TLR4 (Montminy et al., 2006). It is possible, however, that inhibition of TLR4 signalling by YopJ is still crucial in very early stages of bubonic infection, where bacteria grown at flea temperature would possess immunostimulatory LPS. Another study reported that LPS stimulation of splenocyte (presumably B-cell) proliferation is inhibited by YopJ (Meijer et al., 2000), but proinflammatory cytokines and NF-κB signalling were not examined in their work. Additionally, TLR2 may be an important target for reducing inflammation, regardless of its role in apoptosis.

We found that YopJ effectively inhibited both p38 MAPK activation and NF-κB activation in response to TLR stimulation, suggesting that the target of YopJ is prior to the branch point (at TAK1) that leads to activation of these distinct responses. In addition, the inhibition of TLR2, TLR3 and TLR4 signalling suggests that the target of YopJ is not a receptor-proximal component of the pathway, as these receptors differ in their requirements for TIR adapters and IRAK kinases during signal transduction. Thus, it is likely that YopJ inhibits the pathway between the receptor complex and the activation of TAK1. YopJ also inhibits NF-κB activation induced by TNFα or IL-1α stimulation. The IL-1 receptor shares its signal transduction machinery with the TLRs. TNFR, on the other hand, recruits a different receptor-proximal complex involving TRADD, FADD, RIP1 and TRAF2 (Silverman and Maniatis, 2001). The first common component of the TNFR and TLR pathways is TAK1, the MAP3K implicated in the activation of both MAPK and NF-κB responses to TLR stimulation. The ability of YopJ to inhibit pathways that activate NF-κB and MAPKs in response to a diversity of stimuli further supports the hypothesis that the earliest signalling activity YopJ could target is the activation of TAK1 by ubiquitination of the various TRAF proteins.

The data examining the ability of YopJ to inhibit NF-κB activation mediated by overexpression of various components of the TLR signalling pathways (Fig. 4) are also consistent with the hypothesis that YopJ targets the activation of TAK1. YopJ inhibited NF-κB activation induced by various TIR-containing adapters and TRAF6 (Fig. 3), suggesting that YopJ does not act upstream of TRAF6. Conversely, NF-κB activation by expression of IKKβ and p65/RelA was insensitive to YopJ (when the protein expression levels of IKKβ and p65/RelA are appropriately controlled, as shown in Fig. 4), suggesting that YopJ functions upstream of these proteins.

Toll-like receptor signalling is thought to induce TRAF6 aggregation resulting in TRAF6 auto-ubiquitination, a required event for activation of TAK1 and downstream signal transduction (Lamothe et al., 2006). It is likely that the role of TRAF6 ubiquitination is to provide an assembly scaffold for downstream components known to bind K63 ubiquitin, such as IKKγ/NEMO and the TAB2/3 proteins (Kanayama et al., 2004; Wu et al., 2006). Expression of TAK1 itself results in no measurable stimulation of NF-κB, but overexpression of the TAK1 accessory protein TAB2, which has been shown to induce TRAF6 ubiquitination and formation of the TRAF6/TAK1/IKK signalling complex (Kanayama et al., 2004; Kishida et al., 2005), leads to NF-κB activation that is strongly inhibited by YopJ (Fig. 3D). Curiously, coexpressing TAK1 with TAB2 activates NF-κB in a YopJ-insensitive manner. TAB2 and TAK1 co-overexpression appears to create conditions that overcome the necessity for TRAF6-regulated assembly of a signalling scaffold, perhaps through the presence of high concentrations of both TAB2 and TAK1. Furthermore, the insensitivity of the TAB2/TAK1 complex when co-overexpressed suggests that the relevant target of YopJ in signal transduction is an upstream component directly involved in the assembly of this activated complex, and not a modification of TAK1 itself. YopP has also been shown to block the activation of TAK1 during IL-1 signalling (without disrupting the interaction of TAK1 with TAB2) (Thiefes et al., 2006), an observation that is entirely consistent with our data and the hypothesis that TRAFs are the target molecules of YopJ and inconsistent with the hypothesis that YopJ is capable of independently inhibiting IKK activity and IκB degradation, as some have proposed (Zhou et al., 2005; Mittal et al., 2006; Mukherjee et al., 2006).

In addition, IKKγ/NEMO has also been shown to be a target of TRAF6-mediated ubiquitination (Lamothe et al., 2006; Sun et al., 2004). Targeting of IKKγ by YopJ, as recently suggested (Haase et al., 2005), provides an additional mechanistic explanation for the inhibition of NF-κB activation, however, it is not possible to determine if the effect of YopJ on IKKγ ubiquitination is direct or a consequence of TRAF6 inactivation. Another article has shown that IKKβ is mono-ubiquitinated after stimulation by strong overexpression of the T-cell oncogene Tax, and that this ubiquitination is blocked by YopJ (Carter et al., 2003). However, this study did not address the target of YopJ in this atypical constitutive Tax-induced NF-κB activation. Tax itself is ubiquitinated (but not degraded) and this conjugation is associated with its activation of NF-κB (Harhaj et al., 2007); it is possible that Tax itself or some other upstream component is the target of YopJ in this system.

To validate TRAF6 as a target of YopJ, we examined the ubiquitination of FLAG-tagged TRAF6 when YopJ was coexpressed. We observed a reduction in TRAF6 ubiquitination in the presence of YopJ (Fig. 6A and B), an observation that has recently been independently confirmed by two other investigators (Haase et al., 2005; Zhou et al., 2005). Because K63 ubiquitination of TRAF6 (and presumably TRAF3) is essential for signalling (Wang et al., 2001; Lamothe et al., 2006), removal of ubiquitin from these molecules by YopJ is likely to make an important contribution to the interference of signal transduction by YopJ. Indeed, when TRAF6 is conjugated with a mutant of ubiquitin that does not contain K63, the resulting ubiquitination is not sensitive to YopJ (Fig. 6C), but conjugation with a mutant of ubiquitin that does not contain K48 (but has K63) is strongly inhibited by the presence of YopJ (Fig. 6C). This observation is in contrast to that of Zhou et al. (2005), who suggested that YopJ inhibits both K63 and K48 ubiquitination. The reason for this difference is unclear; however, their work lacked two controls that are important for full interpretation of this experiment. Specifically, they did not present data with the mutant YopJ in their experiments with ubiquitin variants, nor did they demonstrate that the level of HA-ubiquitin expression was unaffected during coexpression of YopJ. We have found that at least one of these controls is critical to rule out expression artefacts, as coexpression of other proteins (regardless of activity) can often alter the amount of HA-ubiquitin that is made, making the empty vector control insufficient.

YopJ in IRF3-mediated signalling

The interferon response is a primary mechanism by which mammalian cells respond to infection by viral pathogens. TLRs 3, 4, 7/8 and 9, along with intracellular receptors such as RIG-I and Mda5, activate type I interferons through the transcription factors IRF1, 3, 5 and/or 7 (Schoenemeyer et al., 2005; Schroder and Bowie, 2005; Negishi et al., 2006). Surprisingly, we found that YopJ is capable of blocking the activation of an interferon-sensitive reporter element (ISRE) reporter by the TLR3 ligand poly [I:C]. The YopJ-mediated inhibition of TLR3–induced interferon signalling has important implications for this signalling cascade. YopJ has de-ubiquitinating activity, implicating a ubiquitin-dependent signalling event in the activation of IRF3 by TLR3. The suspected targets of YopJ-mediated de-ubiquitination in TLR- and TNFR-induced NF-κB signalling (TRAF2/6 activation of TAK1) do not play a role in TLR3-mediated IRF activation, however, two recent reports demonstrate that the putative RING-finger containing E3 ubiquitin ligase TRAF3 is critical for this pathway (Hacker et al., 2005; Oganesyan et al., 2005). Current understanding of the TLR3-induced interferon pathway posits that the signal is transduced via TRAF3 from the TIR-adapter TRIF to the MKK/IKK superfamily member TANK-binding kinase (TBK) 1 (Fitzgerald et al., 2003; Sato et al., 2003; Hemmi et al., 2004; Hacker et al., 2005; Oganesyan et al., 2005), which then activates IRF3 by phosphorylation. Consistent with the lack of a MAP3K in this pathway, the activation loop of TBK1 contains a phosphomimetic glutamic acid at residue 168, in place of serine 177 in the IKKβ‘activation loop’ which is phosphorylated by TAK1 during NF-κB signalling (Wang et al., 2001; Kishore et al., 2002). YopJ potently inhibits this pathway, as shown, even though it may not require a MAP3K for activation.

Based on the data presented here, we propose that once activated, TRAF3 functions as an E3-ligase required for activation of the IKK-like kinase TBK1. Based on analogy to TRAF2/6, it is likely that TRAF3 ubiquitinates itself and/or other proteins [perhaps the NEMO analogue NAP1 (Sasai et al., 2006)] to serve as a scaffold for further signalling complex assembly. In Fig. 7, we show that TRAF3 is ubiquitinated when it is expressed in HEK293 cells, and that YopJ removes this modification (Fig. 7A and B). As with TRAF6, YopJ only affects modification of TRAF3 by the activating K63 polymer of ubiquitin, not through lysine 48 (Fig. 7C). This further supports the hypothesis that the inhibitory mechanism of YopJ is the proteolytic cleavage of K63-polymerized ubiquitin from TRAF proteins. Based on these results, we would also predict that TLR4-mediated interferon induction, which also proceeds through the TIR adapter TRIF, will be blocked by YopJ. However, TLR4-mediated IRF3 activation is difficult to observe in the HEK293 model system used in this study and was therefore not addressed.

Detection of intracellular viral RNA by RIG-I or Mda5 induces IRF3 through a pathway distinct from the TLR3 signalling pathway. These receptors activate IRF3 through the adapter MAVS (also known as IPS-1, VISA or Cardif) (Kawai et al., 2005; Meylan et al., 2005; Seth et al., 2005; Xu et al., 2005). Through a mechanism that is still poorly understood and may involve RIP-1 and FADD (Balachandran et al., 2004), MAVS activates TBK1 which can phosphorylate and activate IRF3 (Fitzgerald et al., 2003). This pathway may also use another IKK-related kinase, IKKε, for the activation of IRF3. In Fig. 5 we show that YopJ can inhibit TLR-mediated, but not RIG-I-mediated, ISRE activation. It remains to be determined how MAVS stimulates TBK1 or IKKε and subsequent IRF3 activation. Although TRAF3 is believed to function in this pathway (Saha et al., 2006), the lack of YopJ inhibition suggests that K63 ubiquitination is not involved in this signal transduction. Alternatively, it is possible that the involvement of IKKε, which is not important for TLR3 responses but serves as an alternative mechanism of IRF activation in the RIG-I pathway (Hemmi et al., 2004; McWhirter et al., 2004; Perry et al., 2004), somehow renders the RIG-I pathway refractory to YopJ.

Although the implications of IFN inhibition in Yersinia pathogenesis are unclear, recent evidence suggests that type I interferons are involved in the response to bacterial as well as viral infections (Fehr et al., 1997; O'Connell et al., 2005; Opitz et al., 2006; Stetson and Medzhitov, 2006; Henry et al., 2007). It is possible that Y. pestis gains an advantage from the inhibition of IRF3 activation. In contrast, cytosolic intracellular bacteria such as Francisella and Listeria stimulate an IFN response that potentiates inflammasome-mediated apoptosis, an effect which at least in Listeria contributes to microbial pathogenesis (O'Connell et al., 2004; Henry et al., 2007). It is also possible that IRF inhibition is simply an ‘off-target’ effect caused by the similarity of TLR-mediated IRF signalling to other TLR-induced pathways, in particular the recently discovered requirement for a TRAF family member, TRAF3, in TLR3-induced IRF activation.

Two recent articles make the surprising claim that YopJ is a bacterial acetyltransferase that modifies and inactivates MKK6 and IKKα/β by blocking the sites of activating phosphorylation with acetyl group attachment (Mittal et al., 2006; Mukherjee et al., 2006). We do not see any evidence in support of this hypothesis when YopJ is expressed in mammalian cells; activation by IKKβ expression is not inhibited by YopJ unless there is an artefactual decrease in protein expression level (Figs 3 and 4). Furthermore, this hypothesis does not account for our observation that YopJ inhibition is circumvented by coexpression of TAB2 and TAK1. In addition, this proposed mechanism is unlikely to explain how YopJ inhibits TLR-mediated IRF3 activation, yet has no effect on RIG-I induced activation. The data presented here are most consistent with the proposed de-ubiquitinating activity of YopJ and of TRAFs as the critical targets, however, it remains possible that YopJ has both protease and acetyltransferase activities.

Experimental procedures

Cell lines and constructs

Human embryonic kidney 293 (HEK293) cell lines and HEK293 expressing hTLR2-CFP, hTLR3-YFP or hTLR4-CFP (with or without retrovirally transduced hMD-2) were described previously (Latz et al., 2002; Fitzgerald et al., 2003). Expression constructs pRK5/TRAF6 (Cao et al., 1996), pFLAG/TAK1 (Sakurai et al., 1999), pRK7/IKKβ (Woronicz et al., 1997) and the 5X-κB-luc reporter plasmid (Radons et al., 2002) have been reported previously; interferon-sensitive response element (ISRE)-luc is from Stratagene; phRL-TK is from Promega. The plasmids pEF-Bos/MyD88, pEF-Bos/MAL and pEF-Bos/TRIF were gifts from Kate Fitzgerald; pEF-Flag/TAB2, pEF/ubiquitin, pEF/Ub-K48R and pEF/Ub-K63R were given by James Chen; and pCDNA3/p65(RelA) was provided by Tom Maniatis.

Cloning

Yersinia pestis KIM5 YopJ was cloned by polymerase chain reaction amplification from pCD1 isolated from Y. pestis using pfx platinum DNA polymerase (Invitrogen) and inserted into pBluescript, then reamplified and ligated into pCDNA3 with an N-terminal FLAG tag or a C-terminal enhanced yellow fluorescent protein fusion tag. Stratagene ‘Quikchange’ site directed mutagenesis kit was used to create the C172A mutation in this construct, incorporating a novel Eco57I site for ease of screening. Both YFP-tagged YpYopJ and YpYopJ C172A constructs were transferred into pSFFV (a gift from Kim Orth) with standard techniques. Correct constructs of pSFFV/YpYopJ-YFP, and pSFFV/YpYopJ-YFP C172A were designated pSCS96 and pSCS151 respectively. The FLAG-tagged YopJ wild-type and mutant constructs were also moved to pEF-Bos (pSCS283 and pSCS293 respectively). TRAF6 and TRAF3 were subcloned into pEF-Bos (Mizushima and Nagata, 1990) from pRK5/TRAF6 or pSG5/TRAF3 (Miller et al., 1997); positive clones were designated pSCS313 and pSCS321 respectively. All constructs were verified by DNA sequencing.

Cell culture and transfection

HEK293 cells were maintained in Dulbecco's modified Eagle's medium (DMEM) mammalian media from Biowhittaker, with 10% fetal bovine serum. 500 μg ml−1 geneticin from Invitrogen was used for stable cell selection. The cells were grown at 37°C in 5% CO2. HEK293 cells were transfected using Genejuice (Novagen) as indicated by the manufacturer. Escherichia coli O111:B4 lipopolysaccharide was from Sigma. Human TNFα was obtained from PreproTech, and human IFNβ and IL-1β from R&D systems.

Flow cytometry

These experiments were performed using a Becton Dickenson LSR or LSR II flow cytometer to detect CFP, YFP, Alexa647, allophycocyanin (APC), cyanin 5 (Cy5) or ToPro-3.

Intracellular hIL-8 assays

HEK293 cells were transiently transfected with pSFFV, pSCS96 or pSCS151. After 40 h of transfection, the tranfection media was replaced with fresh media (in the case of TLR4 experiments, it was instead replaced with conditioned media from a 60–80% confluent flask of HEK293 cells expressing retrovirally transduced hMD-2, an accessory protein for TLR4 receptor recognition). Cytokine export was blocked with 0.5 μg ml−1 Brefeldin A and the cells were simultaneously stimulated with 5 nM synthetic lipopeptide Pam2CysK4, 10 μg ml−1 synthetic double-stranded RNA poly[I:C], 10 ng ml−1E. coli lipopolysaccharide (LPS) or 50 ng ml−1 tumour necrosis factor (TNF) α for 5 h. Cells were fixed in phosphate buffered saline (PBS)/4% paraformaldehyde at room temperature and permeabilized in Hank's Buffered Salt Solution (HBSS) with 0.1% saponin and 0.05% sodium azide (SAP buffer) at 4°C, stained in SAP buffer containing 2 μg of anti-hIL-8 (BD Bioscience #554717) and/or the secondary antibody anti-mouse IgG-APC (Caltag M30005) on ice and assayed by flow cytometry. Results are typical of multiple experiments. Cell viability assays were performed similarly, except the antibody staining was done with the directly conjugated anti-annexin V-Cy5 (MBL international) to detect exposed phosphoserine or with the To-Pro-3 stain (Molecular Probes) to detect access to DNA. Both were used according to the manufacturer's protocols. Dead cell controls were generated by washing the cells in 20% EtOH.

Intracellular pp38 assays

HEK293 cells were transiently transfected with pSFFV, pSCS96 or pSCS151. After 20 h of transfection, the tranfection media was replaced with fresh media (or MD2 media as above). The cells were then treated with 5 nM Pam2CysK4, 10 μg ml−1 poly[I:C], 100 ng ml−1E. coli LPS for 2 h or with 2 μg ml−1 anisomycin for 10 min. Cells were harvested immediately after stimulation by trypsinization for 1 min at 37°C, fixed by PBS/1.5% paraformaldehyde at room temperature, and permeabilized in ice-cold 90% MeOH. Cells were stained in ice-cold PBS with 1% bovine serum albumin containing 10 μl of anti-pp38-Alexa647 (Cell Signaling Techology #4552) and analysed by flow cytometry. Results are typical of multiple experiments.

Luciferase reporter gene assays

To analyse NF-κB or ISRE promoter activity by luciferase reporter, HEK293 cells (20 000 cells per sample) were transfected for 16–24 h with various expression plasmids, firefly luciferase under the control of the 5X-κB or 5X-ISRE promoter, and Renilla luciferase under the control of the housekeeping TK promoter and treated with TLR ligands and cytokines. After all transfections and treatments, cells were assayed using the Promega Dual Luciferase Assay System. To optimize protein expression levels, YopJ was expressed from pSFFV in the experiments shown in Fig. 2 and from pEF-Bos in the transfection-induced experiments shown in Fig. 3. Expression control immunoblots for the stimulating plasmids in Fig. 3 are shown in Fig. S1. All samples were run at least in triplicate. Error is reported as standard deviation.

Immunoblotting

HEK293 cells were transfected with various expression constructs. After 24 h, cells were washed with PBS and then harvested by direct lysis on ice in 10% glycerol, 1% Triton X-100, 20 mM Tris pH 7.6, 150 mM NaCl, 25 mM β-glycerolphosphate, 2 mM EDTA, 1 mM dithiothreitol, 1 mM orthovanadate, 2 mM iodoacetic acid, with Sigma mammalian protease inhibitor cocktail. Western blots and were performed according to standard laboratory procedures. Anti-HA (HA-7) and anti-FLAG (M2) are from Sigma, anti-ubiquitin (P4D1), anti-TRAF3 (G-6) and anti-TRAF6 (D-10) are from Santa Cruz. Anti-GFP (JL-8) is from BD Bioscience. Secondary antibodies were either anti-rabbit IgG-HRP (Bio-Rad) or anti-mouse IgG-HRP (Amersham). The blots were developed with Pierce West Pico luminol reagent. Immunoprecipitation experiments used anti-FLAG M2-agarose (Sigma) or Protein A Sepharose CL-4B (Amersham Biosciences) with anti-TRAF3 (Santa Cruz M-20) or anti-TRAF6 (Santa Cruz H-274) to precipitate the target proteins. The membranes were stripped, if necessary, with Chemicon Re-Blot Plus Strong.

Acknowledgements

The authors would like to thank Kate Fitzgerald for critical reading of this manuscript, and Nicholas Paquette, Sara Montminy and Egil Lien for insightful discussions. This work was supported in part by the following NIH Grants: AI053809 to N.S., GM54060 and AI52455 to D.T.G. and AI060301 to C.R.S. This work was also supported by the Ellison Medical Foundation 2004 New Scholar Award in Global Infectious Disease to N.S.

Ancillary