A role for NEMO/IKKγ Ubiquitination in the activation of the IκB kinase complex by TNF-α

1Department of Biological Chemistry 2Program in Cellular and Molecular Biology 3Laboratory of Molecular Signaling and Apoptosis, Department of Biologic and Materials Sciences, School of Dentistry 4Institute of Gerontology University of Michigan Medical School Ann Arbor, MI 48109, USA 5Lineberger Comprehensive Cancer Center, Department of Biochemistry and Biophysics University of North Carolina at Chapel Hill Chapel Hill, NC 27599, USA


INTRODUCTION
The NF-κB/Rel family of transcription factors function in a wide range of biological activities including inflammation, immunity, and apoptosis and their activities are regulated by their interactions with IκB proteins (1)(2)(3)(4). In unstimulated cells, NF-κB is kept inactive in the cytoplasm by virtue of the masking of its nuclear localization sequence by bound IκB protein. Exposure of cells to proinflammatory stimuli triggers the activation of a multisubunit IκB kinase (IKK) complex that phosphorylates IκB proteins on two serine residues (5)(6)(7). Phosphorylation of IκB proteins triggers their polyubiquitination and their subsequent recognition and degradation by the proteasome. Destruction of IκB proteins liberates NF-κB to enter the nucleus and activate gene expression (2,3, 8).
The predominant IKK complex found in cell lines contains two catalytic subunits, IKKα (or IKK1) and IKKβ (or IKK2) (9)(10)(11)(12)(13), and a regulatory subunit, NEMO (or IKKγ) (14)(15)(16)(17). IKKα and IKKβ are serine/threonine protein kinases, whereas NEMO contains several protein interaction motifs but no apparent catalytic domains. Subunit reconstitution experiments in yeast and mammalian cells suggest that IKK is composed of a NEMO homodimer bound together with either an IKKα/IKKβ heterodimer or an IKKβ homodimer (18). Although structurally similar, IKKα and IKKβ have distinct cellular functions. IKKβ phosphorylates IκB and is critical for IKK and NF-κB activation in 6 7.5% polyacrylamide gel. Proteins were transferred to PVDF membrane and probed with Ub or NEMO antibodies. To minimize the detection of the IgG heavy chain, protein A conjugated to horseradish peroxidase (Zymed) was used as a secondary antibody for anti-NEMO immunoblots.
For coimmunoprecipitation assays, cells were lysed in EBC150 (50 mM Tris pH 8.0, 150 mM NaCl, 0.5% NP-40, 50 mM NaF, 0.1 mM Na 3 VO 4 , 1 mM DTT) supplemented with a cocktail of protease inhibitors (Roche) and lysates were incubated with the appropriate antibody for 2 hrs at 4 o C and with Protein G sepharose for 1 additional hr.
All immunoprecipitates were boiled in SDS-PAGE sample buffer and resolved on a 7.5% polyacrylamide gel.
RNAi knockdown experiments-For the knockdown of c-IAP1 expression, we employed a strategy similar to that described previously (33). An oligo corresponding to a hairpin siRNA targeting a unique 19 bp c-IAP1 mRNA sequence (GGAAAUGCUGCGGCCAACA) was subcloned into pTKO, a vector that directs the expression of siRNAs driven by the H1 RNA promoter and contains a puromycin selection marker (E.T. and K.G., manuscript in prep). One day following transfection of 293T cells with pTKO plasmids, transfected cells were selected for by incubation in media containing puromycin (3 ug/ml) (Sigma) for 72 hrs. Whole cell lysates were subsequently prepared for immunoblot or immunoprecipitation analysis.
Wild-type and mutant NEMO coding sequences were cloned into pMAL-c2 (NEB) for expression as MBP fusion proteins. c-IAP1 wild-type and mutant cDNAs were cloned into pGEX-KG for expression as GST fusion proteins. GST fusion, MBP fusion, and His 6tagged proteins were expressed in the E. Coli strain BL21(DE3)RIL (Stratagene) which harbors a helper plasmid which encodes the AGA-specific ArgU tRNA. His-tagged UbcH3, UbcH6, UbcH10, and UbcH13, and UbcH7 were purchased from Boston Biochemical.
Retroviral Infections -JM4.5.2 cells stably expressing NEMO proteins were created by retroviral infections. HA-tagged wild-type and mutant (C417R) NEMO cDNAs were PCR amplified and subcloned into the retroviral expression vector pPGS-CMV-CITEneo (gift of G. Nabel). Amphotropic Phoenix packaging cells (gift of G. Nolan) in 6-well plates were transfected with 2 µg of retroviral expression plasmid using Lipofectamine 2000 reagent (Gibco). 48 hrs following transfection, the viral supernatants were supplemented with polybrene (9 mg/ml) and filtered through a 0.45 µm filter. Viral supernatants were incubated with 1 x 10 6 JM4.5.2 cells and spun at 1800 rpm for 45 min at room temperature.
Infected cells were resuspended in regular growth media containing 1 µg/ml G418 (Gibco) and stably infected cells were selected for one week.

RESULTS AND DISCUSSION
Ubiquitination of NEMO in cells -An earlier report described the purification of a multiprotein IκBα kinase complex that could be activated in vitro by a ubiquitination event (34). One possibility is that ubiquitination of IKK may regulate its activity. In order to investigate whether IKK may be ubiquitinated during the course of its activation by TNF-α, we immunoprecipitated individual IKK subunits from Hela cells following stimulation for various time periods. Interestingly, TNF-α treatment induced the accumulation of a ladder of slower migrating forms of NEMO (Fig. 1A). These modifications on NEMO paralleled the phosphorylation and degradation of IκBα. On the other hand, we failed to detect similar modification of IKKα or IKKβ under the same conditions (data not shown).
Immunoblotting using an ubiquitin (Ub) antibody suggested that TNF-α stimulated the modification of NEMO by different numbers of Ub molecules (Fig. 1B). NEMO-Ub conjugates were also detected in transient transfection experiments using 293T cells in which both epitope-tagged Ub and NEMO were expressed (Fig. 1C). An "inactive" Ub mutant that contained a substitution of Ala for the terminal Gly residue (G76A) (35) was unable to support NEMO ubiquitination.
Because the experiments described above were performed in the absence of proteasome inhibitor, the NEMO-Ub conjugates seen in vivo were apparently not targets for immediate destruction by the proteasome. To examine this idea further, we compared NEMO-Ub conjugates with E2F-4-Ub conjugates, which are known to be targeted for degradation by the Ub-proteasome pathway (36). As expected, E2F-4-Ub conjugates were detectable only when cells were incubated with a proteosome inhibitor (Fig. 1D). In contrast, NEMO-Ub conjugates were readily visible in the absence of proteasome inhibitor and were only slightly altered in the presence of proteasome inhibitor. Thus, the ubiquitination of NEMO may have primarily a regulatory role, rather than a role in protein stability.
Ubiquitination of NEMO is dependent on its zinc finger -All NEMO orthologs identified to date contain predicted zinc finger motifs at their extreme C-termini (5-7) ( Fig.   2A). The zinc finger has been found to be the frequent target of mutations found in the NEMO gene loci of patients with the congenital disorder hypohidrotic ectodermal dysplasia with immunodeficiency (HED-ID) (27)(28)(29)(30)(31). Cells from these patients display defective NF-κB signaling (29,30). Recently, the zinc finger was reported to be partially required for IKK activation in response to TNF-α in mouse embryonic fibroblasts (32). We found, however, that deletion of the zinc finger in NEMO did not affect its binding to IKKα, IKKβ, RIP, or A20 in coimmunprecipitation assays (data not shown). To test whether the zinc finger might be important for NEMO ubiquitination, we examined two NEMO mutants. One mutant contained a deletion of the last 27 residues comprising the predicted zinc finger (∆ZF) and another had a substitution of Arg for Cys 417 (C417R) as seen in an HED-ID patient (29-31) ( Fig. 2A). Cys 417 is one of the predicted zinc coordinating residues of the zinc finger. When wild-type and mutant NEMO proteins were coexpressed with Ub in cells, the ubiquitination of wild-type NEMO was enhanced by TNF-α, as seen with endogenous NEMO (Fig. 2B). In contrast, the ubiquitination of neither of the two zinc finger mutants was enhanced by TNF-α treatment. An intact zinc finger in NEMO therefore appears to be necessary for the stimulation of NEMO ubiquitination by TNF-α.
Engagement of tumor necrosis factor receptor 1 (TNFR1) by TNF-α leads to the recruitment of several molecules to the receptor complex including TRAF2, c-IAP1 and the IKK complex (37)(38)(39). TRAF2 and c-IAP1 can both function as E3s (Ub ligases) via their RING fingers motifs (40-42). E3s function together with E2s (Ub-conjugating enzymes) to catalyze protein ubiquitination (43). We examined whether expression of either of these two Ub ligases could affect NEMO ubiquitination. Transfection of c-IAP1, but not TRAF2, markedly enhanced the levels of NEMO-Ub conjugates without decreasing the steady-state levels of NEMO (Fig. 3A). Another related inhibitor of apoptosis protein, XIAP, had a relatively smaller effect on NEMO ubiquitination. Since the E3 activity of c-IAP1 is dependent on its C-terminal RING finger domain (42), we examined whether the RING finger was required for c-IAP1 to enhance NEMO ubiquitination. Deletion of the RING finger prevented c-IAP1 from enhancing NEMO-Ub conjugates (Fig. 3B). A fragment of c-IAP1 containing the RING finger alone was significantly less active in elevating NEMO ubiquitination, suggesting that other sequences of c-IAP1 were important also. Thus, the stimulation of NEMO ubiquitination by c-IAP1 is RING finger-dependent and may depend on the E3 activity of c-IAP1.
Ubiquitination of NEMO in vitro by c-IAP1 and UbcH5C -Next, we investigated whether c-IAP1 might be able to function as an E3 to ubiquitinate NEMO using purified components in vitro. In our ubiquitination assay we used rabbit E1 (Ub activating enzyme), bovine Ub, and bacterially expressed UbcH5C, c-IAP1, and NEMO. A UbcH5 family member had been previously shown to both function as an E2 in conjunction with c-IAP1 to ubiquitinate proteins (42) and mediate IKK activation in vitro (34). Inclusion of E1, UbcH5C, and c-IAP1 in ubiquitination reactions was sufficient to catalyze the ubiquitination of NEMO (Fig. 3C). Ubiquitination was absolutely dependent on E1, UbcH5C, and c-IAP1 each because no ubiquitination was observed when any one of these components was omitted.
Given that an intact zinc finger in NEMO is required for its ubiquitination in response to TNF-α ( Fig. 2B), we examined whether the zinc finger might be important for ubiquitination in vitro by c-IAP1. We compared the ubiquitination of wild-type NEMO and zinc finger mutants by UbcH5C/c-IAP1. Like wild-type NEMO, both the ∆ZF and C417R mutants were also ubiquitinated by UbcH5C/c-IAP1, but less efficiently (Fig. 3D). Thus, an intact zinc finger appears to be important for NEMO to be efficiently recognized by attributable to the presence of inhibitors of ubiquitination in the bovine Ub purification (Fig.   3E). In order to determine whether NEMO was conjugated to polyUb chains in vitro by UbcH5C/c-IAP1, we tested a lysine-less Ub mutant (7R) in which all 7 Lys residues are replaced by Arg residues (Fig. 3F). In contrast to wild-type Ub, the 7R mutant cannot be used to extend a polyUb chain. Reactions using wild-type Ub contained higher molecular weight conjugates not seen when using the 7R mutant, suggesting that NEMO-polyUb conjugates were assembled (Fig. 3G). The ability of UbcH5C to catalyze polyubiquitination of NEMO was specific among several E2 enzymes that were tested (Fig.   3E). Also, unlike c-IAP1, the other RING finger E3s ROC1 and HDM2 failed to polyubiquitinate NEMO under similar conditions (data not shown).
As expected, the ability of c-IAP1 to function as an E3 was absolutely dependent on its RING finger because deletion of the RING finger (∆RING) or substitution of Cys 605 with Ala (C605A) rendered c-IAP1 inactive in ubiquitinating NEMO (Fig. 3H). The RING finger alone could also promote polyubiquitination, albeit less effectively than wild-type c-IAP1. We have found previously that the ROC1 RING finger alone is sufficient to activate UbcH5C in catalyzing polyubiquitin chains (44). Our results here suggest that the c-IAP1 RING finger may function in a similar manner. Taken together with our observation that the RING finger alone is also less effective than full length NEMO in stimulating ubiquitination in vivo (see Fig. 3B), it appears that sequences outside of the c-IAP1 RING finger are required for optimal substrate ubiquitination, perhaps by serving as a NEMO binding site (see below).

Ubiquitination of NEMO by non-Lys 48-linked polyUb chains -Most protein
ubiquitination studied to date involves the conjugation of Lys 48-linked polyUb chains to proteins, leading to their subsequent recognition and degradation by the proteasome. Given our findings that ubiquitinated NEMO is not targeted for degradation, we examined the polyUb chains assembled by c-IAP1 in more detail. ROC1 is a RING finger protein that is a component of the multisubunit E3 SCF complex that targets proteins such as IκB and β-catenin for degradation. We found previously that ROC1 alone is sufficient to function as an E3 with UbcH5C to catalyze polyUb chains (44). When c-IAP1 and ROC1 were tested for their abilities to function as E3s in in vitro assays, we found that Lys 48 in Ub was required for polyUb assembly by ROC1 but not by c-IAP1 (Fig. 3I). Thus, the identity of the particular RING finger E3 can dictate the structure of the polyUb chains that can be assembled in conjunction with UbcH5C.
Further experiments using single lysine Ub mutants (Fig. 3F) demonstrated that c-IAP1 could specifically assemble Lys 6-linked polyUb chains (Fig. 3J). When NEMO was included as a substrate, c-IAP1/UbcH5C was found to be able to catalyze the modification of substrate by Lys 6-linked polyUb chains (Fig. 3K). Of note, the assembly of Lys 6-linked chains was noticeably less efficient than the assembly of wild-type polyUb chains (Fig. 3J). This disparity can likely be attributed to the effect of the six lysine to arginine substitutions on protein structure. However, we cannot rule out the possibility from these experiments that other lysines in Ub other than Lys 6 can also be used in polyUb chains.
Role of c-IAP1 in IKK activation by TNF-α -Given the ability of c-IAP1 to stimulate NEMO ubiquitination, we wanted to examine if c-IAP1 could stimulate NF-κB activity as well. Overexpression of c-IAP1 in cells was able to activate NF-κB and this activity was dependent on the RING finger of c-IAP1 (Fig. 4A). Expression of the c-IAP1 RING finger mutants ∆RING and C605A had varying effects on NF-κB activation by TNF-α, with the latter mutant having a more potent dominant interfering effect (Fig. 4A).
Also, the ability of c-IAP1 to activate NF-κB was dependent on the presence of NEMO, as suggested in experiments with NEMO-deficient cells (Fig. 4B). When coexpressed in 293T cells, c-IAP1 and NEMO could coimmunoprecipitate together (Fig. 4C).
Endogenous c-IAP1 and NEMO were also found to associate together in Hela cells in a stimulus-independent manner (Fig. 4D). Past reports have suggested that c-IAP1 is recruited to the receptor complex by binding to TRAF2 (39, 45). We compared the abilities of TRAF2 and NEMO to bind to several deletion mutants of c-IAP1 in transient transfection assays. These results suggested that TRAF2 and NEMO bind to distinct domains outside of the RING finger of c-IAP1 (Fig. 4E). Experiments performed with these c-IAP1 deletion mutants suggested that both the RING finger and the ability to bind NEMO was required to activate NF-κB (Fig. 4A). Since TRAF2 directly binds to c-IAP1 and is required for recruitment of both c-IAP1 and NEMO to the signaling complex (38,45), we examined whether c-IAP1 could bridge an interaction between TRAF2 and NEMO. When NEMO and TRAF2 were coexpressed in the absence of transfected c-IAP1, a relatively small amount of NEMO was found bound to TRAF2 (Fig. 4F).
However, coexpression of c-IAP1 significantly enhanced the NEMO-TRAF2 interaction. A c-IAP1 mutant that was defective in binding to TRAF2 but not NEMO (∆121) (see Fig. 4E) failed to increase the interaction. Thus, c-IAP1 may mediate the recruitment of NEMO to TRAF2 and the TNFR1 signaling complex as well as regulate NEMO ubiquitination.
To see if endogenous c-IAP1 was required for IKK activation by TNF-α, we used RNA interference (RNAi) to deplete endogenous c-IAP1. We examined the TNF-α response of 293T cells that were transfected with a hairpin short interfering RNA (siRNA) expression plasmid that reduces the expression of c-IAP1 but did not effect the expression of the related proteins c-IAP2 or XIAP (Fig. 4G). Cells transfected with the c-IAP1 siRNA plasmid displayed impaired IKK activation and IκBα phosphorylation in response to TNFα (Fig. 4H, J). Expression of the RING finger C605A mutant of c-IAP1 was able to function as a dominant interfering mutant in suppressing IKK activation as well (Fig. 4I).
Furthermore, cells transfected with the c-IAP1 siRNA plasmid displayed impaired the ubiquitination of endogenous NEMO in response to TNF-α, suggesting that NEMO ubiquitination is mediated by c-IAP1 in vivo (Fig. 4J). Of note, we did not find any differences in rates of apoptosis between cells transfected with c-IAP1 siRNA plasmid and control cells (data not shown). These data together suggest an important role for c-IAP1 in mediating both IKK activation and NEMO ubiquitination in response to TNF-α.

Requirement of the NEMO zinc finger in TNF-α-induced IKK activation -Given
that cells from patients with HED-ID have been reported to display defective NF-κB signaling (29,30), we sought to determine if the NEMO zinc finger was important for TNFα-induced NF-κB activation. We compared the functional activities of wild-type NEMO and zinc finger mutants in JM4.5.2 cells, a T cell line that lacks expression of NEMO (46). These cells have been previously shown to be defective in activating NF-κB in response to PMA/ionomycin or the retroviral oncoprotein Tax (46,47). JM4.5.2 cells also failed to activate NF-κB in the response to TNF-α, and this defect was restored by the expression of wild-type NEMO (Fig. 5A). In contrast, both zinc finger NEMO mutants, ∆ZF and C417R, demonstrated impaired abilities to restore the activation of an NF-κB luciferase reporter by TNF-α, suggesting that an intact zinc finger is important for NF-κB activation (Fig. 5A).
To examine the importance of the NEMO zinc finger in IKK function, we stably expressed wild-type NEMO or the C417R mutant in JM4.5.2 cells by retroviral infection.
As expected, the C417R mutant displayed defective TNF-α-stimulated ubiquitination when compared to wild-type NEMO (Fig. 5B). Significantly, we found that in contrast to wildtype NEMO, the C417R mutant was unable to restore TNF-α-induced phosphorylation of IκBα ( Fig 5C). The activation of IKK was also not restored ( Fig. 5D) even though IKK complexes displayed similar levels of IKKα, IKKβ, and NEMO proteins (data not shown).
Thus, the zinc finger of NEMO is necessary for proper IKK activation, but not for its incorporation into the IKK complex. Furthermore, wild-type but not mutant NEMO protein could restore TNF-α-induced phosphorylation of IKKβ on Ser 181 in its activation loop (Fig. 5D). The defect in restoring NF-κB activation by the NEMO mutants can therefore be attributed to defective IKK phosphorylation and activation in these cells.
In summary, our results suggest that the conserved zinc finger of NEMO is required for both NEMO ubiquitination and IKK activation by TNF-α. The inhibitor of apoptosis protein c-IAP1, which is recruited to the TNF receptor complex, is an important mediator of by guest on March 24, 2020 http://www.jbc.org/ Downloaded from IKK activation and NEMO ubiquitination in response to TNF-α. Thus, NEMO ubiquitination by c-IAP1 appears to be a critical event in the activation of IKK in response to TNF-α. In addition, our data from in vitro Ub assays suggest that the identity of the RING finger E3 can dictate the lysine of Ub that is used in the polyUb chains that are assembled. A recent report found that mutation of two cysteines in the zinc finger of NEMO distinct from Cys 417 impaired TNF-α but not IL-1-induced IKK activation in MEFs (32). These results are in accordance with our findings and together with our data suggests that NEMO ubiquitination by c-IAP1 plays a specific role in TNF-α signaling.
Interestingly, previous work described the identification of a purified IKK complex that could be activated by a ubiquitination event involving a UbcH5 family member and an unidentified factor (34). Based on our findings, c-IAP1 may be this unidentified factor and NEMO may be the target for ubiquitination. TRAF2 is another E3 that plays an important role in IKK activation in response to TNF-α (38). Unlike, c-IAP1, TRAF2 does not appear to mediate NEMO ubiquitination. TRAF2 may instead play a role upstream of IKK activation in TNF-α signaling by ubiquitinating itself with Lys 63-linked polyUb chains, in a similar mechanism as has been postulated for TRAF6 in the IL-1 signaling pathway (48)(49)(50). Transfected cells were treated with TNF-α (10 ng/ml) prior to lysis. Immunoprecipitations were performed and analyzed as in Fig. 1D.