|Home | About | Journals | Submit | Contact Us | Français|
Lysine-63 (K63)-linked polyubiquitination has emerged as a mechanism regulating diverse cellular functions, including activation of the protein kinase IKK in the NF-κB pathways. However, genetic evidence for a key role of K63 polyubiquitination in IKK activation is lacking. Here we devise a tetracycline-inducible RNAi strategy to replace endogenous ubiquitin with a K63R mutant in a human cell line. We demonstrate that K63 of ubiquitin and the catalytic activity of Ubc13, an E2 that catalyzes K63 polyubiquitination, are required for IKK activation by IL-1β, but surprisingly, not by TNFα. We further show that IKK activation by TNFα requires Ubc5, which functions with the E3 cIAP1 to catalyze polyubiquitination of RIP1 not restricted to K63 of ubiquitin. These results indicate that distinct ubiquitin-dependent mechanisms are employed for IKK activation by different pathways. The ubiquitin replacement methodology described here provides a means to investigate the function of polyubiquitin topology in various cellular processes.
Ubiquitin is a central regulator of cellular functions (Hershko, 1983; Pickart, 2004). The best known function of ubiquitin is to target protein degradation by the proteasome through covalent attachment of polyubiquitin chains that are usually linked through Lys-48 of ubiquitin (Chau et al., 1989). Ubiquitination also has numerous regulatory functions independent of proteasomal degradation (Chen and Sun, 2009). In particular, monoubiquitination and K63-linked polyubiquitination have been shown to regulate a variety of cellular functions, including chromatin dynamics, membrane trafficking, DNA repair and protein kinase activation.
Both the proteolytic and non-proteolytic functions of ubiquitin are critically involved in the signaling pathways leading to activation of NF-κB, a dimeric transcription factor consisting of members of the Rel family of proteins (Krappmann and Scheidereit, 2005). NF-κB is normally sequestered in the cytosol through association with a member of the IκB family proteins. Stimulation of cells with many different agents, including microbial pathogens and inflammatory cytokines, leads to the rapid phosphorylation of IκB by the IκB kinase (IKK) complex, which contains the catalytic subunits IKKα and IKKβ, and an essential regulatory subunit known as NEMO or IKKγ. Phosphorylated IκB is polyubiquitinated by a ubiquitin ligase complex consisting of Skp1, Cul1, Roc1 and βTrCP and subsequently degraded by the proteasome, allowing NF-κB to enter the nucleus to regulate a plethora of target genes.
The non-proteolytic function of ubiquitin in NF-κB pathways was discovered in the course of isolating the IκB kinase complex and studying its regulation by TRAF6, a signaling protein essential for NF-κB activation by several pathways, including those emanating from interleukin-1 (IL-1) and Toll-like receptors (TLRs) (Chen et al., 1996; Deng et al., 2000). Initial biochemical studies showed that IKK could be activated in vitro by polyubiquitination independent of proteasomal degradation. Subsequent studies showed that TRAF6 is a RING domain ubiquitin ligase that functions together with a ubiquitin E2 complex consisting of Ubc13 and Uev1A to catalyze the synthesis of K63-linked polyubiquitin chains. Recent studies have shown that unanchored K63 polyubiquitin chains can directly activate a protein kinase complex composed of TAK1, TAB1 and TAB2 (or TAB3) (Wang et al., 2001; Xia et al., 2009). TAB2 and TAB3 are homologous proteins containing an evolutionarily conserved zinc-finger-type ubiquitin-binding domain known as NZF (novel zinc finger), which is important for TAK1 activation (Kanayama et al., 2004). After TAK1 is activated, it phosphorylates IKKβ at two serine residues within the activation loop, resulting in IKK activation. The activation of IKK requires NEMO, which contains a coiled-coil-type ubiquitin-binding domain (UBD) known as NUB (Nemo ubiquitin binding, also called UBAN, CoZi or NOA) (Ea et al., 2006; Israel, 2006; Wu et al., 2006). In addition, NEMO contains a C-terminal zinc finger-type UBD that confers specificity for K63 polyubiquitin chain binding and is required for NF-κB activation (Laplantine et al., 2009). The binding of NEMO to polyubiquitin chains recruits IKK to the TAK1 complex, facilitating IKK phosphorylation by TAK1. TAK1 can also phosphorylate members of the MKK family, such as MKK6, leading to activation of JNK and p38 kinase cascades (Ninomiya-Tsuji et al., 1999; Wang et al., 2001).
Recent studies have suggested an expanding role of K63 polyubiquitination in diverse pathways leading to NF-κB activation (Chen, 2005; Chen and Sun, 2009; Krappmann and Scheidereit, 2005). One of these pathways that have been extensively studied is the tumor necrosis factor (TNF) pathway. The binding of TNFα to its receptor leads to membrane recruitment of several signaling proteins, including the RING domain proteins TRAF2, TRAF5, cIAP1 and cIAP2 and the protein kinase RIP1. Both TRAF2 and RIP1 are polyubiquitinated in response to TNFα stimulation. In particular, ubiquitination of RIP1 at K377 is important for the activation of IKK by TNFα (Ea et al., 2006; Li et al., 2006). Polyubiquitinated RIP1 binds to TAB2 and NEMO, thereby recruiting the TAK1 and IKK complexes, respectively, to the membrane receptor complex to facilitate the activation of these kinases. The polyubiquitin chains on RIP1 are thought to be linked through K63 of ubiquitin based on transfection of ubiquitin mutants and an antibody specific for the K63 linkage of ubiquitin (Ea et al., 2006; Li et al., 2006; Newton et al., 2008; Wertz et al., 2004).
While accumulating evidence supports an important role of K63 polyubiquitination in IKK activation in several signaling pathways, conflicting data regarding the role of Ubc13 in IKK activation in vivo have been reported (Fukushima et al., 2007; Yamamoto et al., 2006a; Yamamoto et al., 2006b). In addition, genetic evidence that K63 of ubiquitin is important for IKK activation is still lacking, owing to the technical difficulty of mutating endogenous ubiquitin genes in vivo. From yeast to human, eukaryotic cells contain four highly conserved ubiquitin genes; two encode linear polyubiquitin in which ubiquitin is linked to each other from “head-to-tail”, and the other two encode ubiquitin fused to ribosomal subunits (Finley et al., 1987). Here we report a new strategy to replace endogenous ubiquitin with a K63R mutant using a tetracycline-inducible system. Using this system, we demonstrate for the first time that K63 polyubiquitination and the catalytic activity of Ubc13 are essential for IKK activation by IL-1β. Surprisingly, the mutation of K63 of ubiquitin does not impair IKK activation by TNFα. Furthermore, we show that Ubc5, but not Ubc13, is required for IKK activation by TNFα. Biochemical experiments show that Ubc5 functions with cIAP1, but not TRAF2, to catalyze the conjugation of RIP1 by polyubiquitin chains that are not restricted to the K63 linkage. These results suggest that distinct signaling pathways can employ unique sets of ubiquitination enzymes to catalyze different forms of polyubiquitination capable of activating IKK.
To investigate the function of K63 polyubiquitination, we set up a tetracycline-inducible RNAi system to knock down the expression of all four ubiquitin genes in a human cell line, and simultaneously replace them with RNAi-resistant K63R ubiquitin genes (Ubr) whose expression is also induced by tetracycline (Fig. 1A). To knock down all four ubiquitin genes, we identified a sequence (designated as sh-Ub1; see Experimental Procedures) complementary to those of human UBC and UBA52, and another sequence (designated as sh-Ub2) targeting both human UBB and RPS27A. Four copies of sh-Ub1 and six copies of sh-Ub2 were subcloned into a shRNA vector in such a way that each shRNA sequence is controlled by a tetracycline-inducible promoter (see Supplementary Figure S1 for the shRNA vector construction strategy) (Zhong et al., 2005). The shRNA vector, which harbors a puromycin-resistance gene, was stably integrated into U2OS, a human osteosarcoma cell line engineered to express a tetracycline repressor. Stable cell clones in which endogenous ubiquitin could be efficiently depleted (by ~85% based on immunoblotting and densitometry) following three days of tetracycline induction were selected for further transfection with ubiquitin rescue constructs (Fig. 1A). These constructs contained two expression cassettes in which RNAi-resistant wide type or K63R ubiquitin was expressed under the control of the tetracycline-inducible promoter. Since RNAi also knocks down the expression of two ribosomal subunits (L40 and S27A) that are fused to ubiquitin, the “rescue” expression construct allows for expression of both ribosomal subunits fused to ubiquitin. The second ubiquitin fusion gene is driven by the IRES sequence and contains an N-terminal HA epitope to facilitate detection. The rescue constructs also contained a neomycin-resistance gene, allowing for selection of stable cell clones in which endogenous ubiquitin can be knocked down and replaced with the wide type or mutant ubiquitin upon tetracycline induction. These stable clones are designated as U2OS-shUb-Ub(WT) and U2OS-shUb-Ub(K63R), respectively, whereas the clones containing Tet-inducible shUb alone are designated as U2OS-shUb.
To characterize the efficiency of ubiquitin knockdown and replacement in our engineered cell lines, we carried out RT-PCR analyses using primers that distinguish endogenous and exogenous ubiquitin genes. When the cells were growing in the presence of tetracycline for 4 days, the expression of all four endogenous ubiquitin genes was reduced by 80–95% (Fig. 1B). Concomitantly, the expression of the ubiquitin transgenes was dramatically induced. To further evaluate the efficiency of ubiquitin replacement at the protein level, we used semi-quantitative mass spectrometry to determine the relative abundance of wild type and K63R ubiquitin during the time course of tetracycline induction. As shown in Figure 1C, ,55 days after tetracycline induction, the ratio of a tryptic peptide from T55 to K63 of endogenous ubiquitin (m/z=541.3) to a reference peptide (E64-R72; m/z=534.3), which is present in both endogenous and mutant ubiquitin, decreased by about 90%. In contrast, the ratio of a peptide from T55 to R63 of the mutant ubiquitin (m/z=555.3) to the reference peptide increased dramatically. As a control, the ratio of a peptide from L43 to K48 of ubiquitin (m/z=324.7), which is present in both the endogenous and mutant protein, to the reference peptide remains relatively unchanged after tetracycline induction. As expected, no peptide containing R63 of ubiquitin was detected in U2OS-shUb-Ub(WT) cells before or after tetracycline induction, whereas the peptide containing K63 (T55-K63) remained largely constant (Supplementary Figure S2A). These results demonstrate that endogenous ubiquitin was largely replaced with K63R ubiquitin after tetracycline induction in U2OS-shUb-Ub(K63R) cells. To determine if K63 polyubiquitination was blocked in cells expressing K63R ubiquitin, we immunoprecipitated HA-Ub or HA-Ub (K63R) from cells treated with tetracycline, then examined polyubiquitinated proteins by immunoblotting with an antibody specific for the K63 linkage of ubiquitin chains (Fig. 1D). While K63 polyubiquitination was readily detectable in cells expressing the wild type ubiquitin, it was markedly reduced in cells expressing K63R ubiquitin, indicating that K63 polyubiquitination was defective when most of the endogenous ubiquitin was replaced with the K63R mutant in these cells. The total ubiquitin conjugate levels detected with the HA antibody were similar in the cells expressing the wild type and K63R ubiquitin (Fig. 1D, lower panel), suggesting that the K63R mutant could form polyubiquitin chains through other lysines. After tetracycline treatment, U2OS-shUb cells were viable for three days, and the U2OS-shUb-Ub(WT) and U2OS-shUb-Ub(K63R) cells were viable for at least 5 days, allowing experiments to be performed within these time frames (see Supplementary results and Figure S2).
To determine if knockdown of endogenous ubiquitin is sufficient to inhibit IKK activation, we stimulated U2OS-shUb cells with IL-1β or TNFα, then measured IκBα phosphorylation and degradation by immunoblotting (Fig. 2A). In control cells, IL-1β and TNFα triggered rapid phosphorylation and degradation of IκBα. In cells treated with tetracycline for three days, which effectively reduced both monomeric and conjugated forms of ubiquitin, the phosphorylation and degradation of IκBα was significantly impaired. Direct kinase assays confirmed that IKK activation by both IL-1β and TNFα were strongly inhibited in the ubiquitin knockdown cells (Fig. 2A), emphasizing the importance of ubiquitination in IKK activation. On the other hand, the ubiquitin RNAi did not inhibit the phosphorylation of STAT1 and ERK by IFN-γ and EGF, respectively, suggesting that the silencing of ubiquitin expression under these experimental conditions did not have a general adverse effect on cell signaling (Supplementary Figure S3A & S3B). Taken together, these results suggest that ubiquitin is limiting in the IKK activation pathway and that reduction of ubiquitin expression is sufficient to selectively inhibit IKK activation without affecting other signaling pathways. Consistent with an important role of ubiquitination in IKK activation, tetracycline-induced knockdown of ubiquitin E1 blocked IKK activation by both IL-1β and TNFα (Figure S3C & S3D).
To determine the role of K63 polyubiquitination in IKK activation, U2OS-shUb-Ub(WT) and U2OS-shUb-Ub(K63R) cells growing in the presence or absence of tetracycline were stimulated with IL-1β or TNFα, then IKK activation was measured by immunocomplex kinase assay or direct immunoblotting of IκBα (Fig. 2B & 2C). In the cells in which endogenous ubiquitin was replaced with wide type ubiquitin [U2OS-shUb-Ub(WT)], IKK activation by IL-1β and TNFα was restored (compare Fig. 2B and 2C to Fig. 2A), confirming that the blockade of IKK activation observed in Figure 2A resulted from ubiquitin depletion. In contrast, in U2OS-shUb-Ub(K63R) cells, IL-1β-induced IKK activation was blocked, demonstrating that K63 polyubiquitination is critical for IKK activation in the IL-1β pathway. Control experiments showed that the activation of ERK and STAT1 by EGF and IFN-γ, respectively, was not compromised in U2OS-shUb-Ub(K63R) cells (Supplementary Figure S3E & S3F). Surprisingly, expression of K63R ubiquitin in the ubiquitin-knockdown cells restored TNFα-induced IKK activation (compare Fig. 2C to Fig. 2A; right panels), suggesting that K63 of ubiquitin is dispensable for TNFα-induced IKK activation.
The Ubc13/Uev1A E2 complex is known to be highly specific in catalyzing K63 polyubiquitination (Deng et al., 2000; Hofmann and Pickart, 1999; VanDemark et al., 2001). To investigate the function of Ubc13 in IKK activation, we established a U2OS stable cell line in which the expression of endogenous Ubc13 was silenced in the presence of tetracycline. The silencing of Ubc13 completely blocked IKK activation by IL-1β, but not TNFα, again supporting the requirement of K63 polyubiquitination in the IL-1β, but not TNFα, pathway (Fig. 3A). To determine whether the catalytic activity of Ubc13 is required, plasmids for tetracycline-inducible expression of wide type Ubc13 or its active site mutant (C87A) were stably integrated into U2OS-shUbc13 cells. Treatment of these cells with tetracycline allows for replacement of endogenous Ubc13 with Ubc13(WT) or Ubc13(C87A) (Fig. 3B & 3C). Wide type Ubc13, but not the C87A mutant, was able to rescue IKK activation by IL-1β, demonstrating that the catalytic activity of Ubc13 is required for IKK activation in this pathway (Fig. 3B). In contrast, the cells expressing Ubc13(C87A) were still capable of activating IKK in response to TNFα stimulation (Fig. 3C), suggesting that another E2 may be required in this pathway.
Biochemical studies have shown that an E2 of the Ubc5 family can support IKK activation in crude cell extracts (Chen et al., 1996). There are three Ubc5 genes encoding different isoforms, Ubc5a, b and c, which are highly homologous to each other. To determine if Ubc5 is involved in IKK activation in the TNFα pathway, we established a U2OS stable cell line in which Ubc5b and Ubc5c were knocked down using tetracycline-inducible shRNA. The knockdown of these Ubc5 isoforms significantly inhibited IKK activation by TNFα, whereas IL-1β-induced IKK activation was only modestly affected (Fig. 4A). The activation of IKK by TNFα was restored by replacing endogenous Ubc5 with wide type Ubc5c, but not the catalytically inactive mutant C85A (Fig. 4B). The residual IKK activation, which occurred with a delayed kinetics, may be due to incomplete knockdown of Ubc5b and/or Ubc5c, or the presence of Ubc5a, which could also support IKK activation and RIP1 ubiquitination in vitro (Chen et al., 1996)(data not shown).
TNFα-induced polyubiquitination of RIP1 recruits the TAK1 and IKK complexes to the receptor complex, resulting in the activation of these kinases (Ea et al., 2006; Kanayama et al., 2004; Li et al., 2006). To determine if Ubc5 is required for RIP1 polyubiquitination, we stimulated U2OS-shUbc5 cells with TNFα, then examined ubiquitination of RIP1 in the cell lysates by immunoblotting. A small fraction of RIP1 was polyubiquitinated in response to TNFα stimulation, and this ubiquitination was abrogated in cells depleted of Ubc5 with tetracycline (Supplementary Fig. S4A). To enrich for ubiquitinated RIP1, we immunoprecipitated NEMO and analyzed the presence of ubiquitinated RIP1 in the NEMO complex by immunoblotting with a RIP1 antibody. Tetracycline-induced silencing of Ubc5, but not Ubc13, markedly inhibited TNFα-induced polyubiquitination of RIP1 (Fig. 5A & 5D). RNAi of Ubc5 also prevented the recruitment of polyubiquitinated RIP1 to the TNF receptor in U2OS cells (Fig. S4B) and the association of polyubiquitinated RIP1 with NEMO in HEK293T cells (Fig. S4C). RIP1 ubiquitination was restored when endogenous Ubc5 was replaced with the wide type Ubc5c but not its C85A mutant (Fig. 5C). In sharp contrast, IL-1β-induced polyubiquitination of IRAK1 depended on Ubc13 but not Ubc5 (Fig. 5B & 5E; Supplementary Figure S4D).
While Ubc13 specifically synthesizes K63 polyubiquitin, Ubc5 is known to make ubiquitin chains linked through various lysines of ubiquitin. Therefore, we examined the ubiquitination of RIP1 or IRAK1 in cells where endogenous ubiquitin was depleted or replaced with K63R ubiquitin. In ubiquitin knockdown cells, both IRAK1 and RIP1 ubiquitination were blocked (Figure 6A &6B). When K63R ubiquitin was expressed in cells depleted of endogenous ubiquitin, polyubiquitination of RIP1, but not IRAK1, was restored (Figure 6C &6D), indicating that ubiquitination of IRAK1, but not RIP1, requires K63 of ubiquitin. Similarly, TNFα stimulation led to association of TAB2 with polyubiquitinated RIP1 in U2OS-shUb-Ub(K63R) cells (Supplementary Fig. S4E). Titration of TNFα concentrations showed that cells expressing endogenous ubiquitin (-Tet) and those expressing K63R ubiquitin (+Tet) were equally sensitive to TNFα stimulation, as judged by IκBα phosphorylation and RIP1 polyubiquitination (Fig. S5A). This K63-independent activation of IKK was likely mediated through TNF receptor-1 (TNFR1) because U2OS cells express TNFR1, but no detectable level of TNFR2 (Fig. S5B). To further confirm that ubiquitinated RIP1 observed in U2OS-shUb-Ub(K63R) cells contains K63R ubiquitin instead of residual endogenous ubiquitin, we took two approaches. In the fist approach, ubiquitinated RIP1 associated with NEMO was first eluted with 1%SDS, then immunoprecipitated with a HA antibody that recognizes HA-Ub(K63R). Immunoblotting of the HA immunoprecipitates with a RIP1 antibody clearly showed that RIP1 was conjugated by polyubiquitin chains containing HA-Ub(K63R) in U2OS-shUb-Ub(K63R) cells (Fig. 6E). In the second approach, we immunoprecipitated K63 polyubiquitinated proteins from U2OS-shUb-Ub(K63R) cells stimulated with TNFα, and probed for the presence of RIP1 (Fig. 6F). Consistent with a previous report (Newton et al., 2008), TNFα stimulated K63 polyubiquitination of RIP1 in cells containing endogenous ubiquitin (−Tet). In contrast, K63 polyubiquitinated RIP1, but not total RIP1 ubiquitination, was markedly reduced in cells expressing K63R (+Tet). Collectively, these results strongly suggest that TNFα induces conjugation of RIP1 by polyubiquitin chains that are not linked exclusively through K63 of ubiquitin.
The RING domain ubiquitin ligases TRAF2, cIAP1 and cIAP2 have been implicated in NF-κB activation in the TNF pathway, but whether loss of cIAPs enhances or reduces NF-κB activation has been a subject of debate (Bertrand et al., 2008; Mahoney et al., 2008; Varfolomeev et al., 2008; Vince et al., 2007). To determine if cIAPs are the E3 for RIP1, we incubated U2OS cells with a SMAC mimetic, a compound known to trigger autoubiquitination and degradation of both cIAP1 and cIAP2 by the proteasome (Wang et al., 2008). The cells were then stimulated with TNFα, followed by immunoprecipitation of ubiquitinated RIP1 through its association with NEMO. As shown in Figure 7A, TNFα-induced polyubiquitination of RIP1 and activation of IKK were significantly reduced in cells depleted of cIAPs as compared to control cells.
The kinase activity of RIP1 is dispensable for NF-κB activation but required for cell death induced by TNFα (Holler et al., 2000; Ting et al., 1996). As overexpression of RIP1 causes cell death, we expressed and purified a catalytically inactive mutant of RIP1 (D138N) from Sf9 cells using the baculovirus expression system. We also expressed and purified cIAP1 from E. coli to determine if cIAP1 could directly catalyze ubiquitination of RIP1 in vitro. Recombinant cIAP1 protein was incubated with E1, Ubc5c, and ubiquitin together with RIP1 in the presence of ATP. Wild type Ubc5c, but not Ubc5(C85A), could support RIP1 polyubiquitination together with cIAP1 (Fig. 7B). Neither Ubc13/Uev1A nor TRAF2 was able to support RIP1 ubiquitination under the same condition, whereas TRAF6 was only weakly active (Fig. 7C & 7D). The polyubiquitin chains on RIP1 synthesized by cIAP1 and Ubc5c were not restricted to K48 or K63 of ubiquitin, as both K48R and K63R mutants of ubiquitin were capable of supporting RIP1 polyubiquitination (Fig. 7E). Taken together, these results suggest that cIAP1 (and likely cIAP2) functions together with Ubc5 to catalyze polyubiquitination of RIP1 that is not restricted to K63 of ubiquitin.
In summary, as shown in Supplementary Table 1, we have provided multiple lines of evidence demonstrating that IKK activation by IL-1β requires Ubc13 and K63 polyubiquitination. In contrast, IKK activation by TNFα requires Ubc5 and cIAPs but not Ubc13 or K63 polyubiquitination.
Biochemical studies have led to the discovery of K63 polyubiquitination as an important mechanism of IKK activation by TRAF6. The evidence in support of this model includes: 1) in vitro reconstitution of TAK1 and IKK activation by TRAF6-catalyzed K63 polyubiquitination (Deng et al., 2000; Sun et al., 2004; Wang et al., 2001); 2) discovery of ubiquitin-binding domains in TAB2 and NEMO, which are important for the activation TAK1 and IKK, respectively (Ea et al., 2006; Kanayama et al., 2004; Wu et al., 2006); 3) discovery of deubiquitination enzymes such as CYLD and A20 that remove K63 polyubiquitin chains to inhibit IKK (Boone et al., 2004; Brummelkamp et al., 2003; Kovalenko et al., 2003; Trompouki et al., 2003; Wertz et al., 2004). However, conflicting results regarding the role of Ubc13 in IKK activation have recently been reported. One study showed that conditional deletion of Ubc13 in several cell types such as B cells and fibroblasts did not completely block IKK activation (Yamamoto et al., 2006a), although a follow-up study by the same group showed that Ubc13 deficiency in T cells severely impaired the activation of TAK1, IKK and MAP kinases by T cell receptor stimulation (Yamamoto et al., 2006b). Another study found that cells lacking one copy of Ubc13 were strongly compromised in IKK activation in response to inflammatory stimuli (Fukushima et al., 2007). As Ubc13 is the only E2 known to specifically synthesize K63 polyubiquitin chains, these results raise the question of whether K63 polyubiquitination is important for IKK activation in vivo.
In this study, we demonstrate that the catalytic activity of Ubc13 is essential for IKK activation by IL-1β. Moreover, we show that the replacement of endogenous ubiquitin with the K63R mutant blocks IKK activation by IL-1β, formally demonstrating the key role of K63 polyubiquitination in this pathway. It is not clear why conditional deletion of Ubc13 in some mouse cells did not completely block IKK activation in one of the previous studies (Yamamoto et al., 2006a). One possibility is that the Cre-mediated deletion of Ubc13 was not complete in some cells, and the residual Ubc13 was sufficient to support IKK activation. Another possibility is that another E2 may substitute for the loss of Ubc13 in certain pathways, as is discussed below for the TNFα pathway.
Through overexpression of ubiquitin mutants in mammalian cells, we and others have obtained results suggesting that K63 polyubiquitination of RIP1 is important for IKK activation by TNFα (Ea et al., 2006; Wertz et al., 2004). Indeed, TNFα-induced K63 polyubiquitination of RIP1 has been confirmed with an antibody specific for K63 polyubiquitin chains (Newton et al., 2008). However, it remains possible that other ubiquitin linkages may be present on RIP1, and an obligatory role of K63 polyubiquitination in the TNFα pathway has not been proven, largely owing to the difficulty of mutating complex ubiquitin genes in cells. Our study provides several lines of evidence suggesting that K63 polyubiquitination is in fact dispensable for IKK activation by TNFα: 1) The K63R substitution of ubiquitin does not impair IKK activation by TNFα (Fig. 2); 2) TNFα-induced IKK activation is normal in cells lacking Ubc13 or expressing a catalytically inactive mutant of Ubc13 (Fig. 3); 3) TNFα-induced polyubiquitination of RIP1 is normal in cells lacking Ubc13 or expressing a K63R mutant of ubiquitin, but not in those depleted of Ubc5 (Fig. 5 & 6); 4) Ubc5 and cIAP1 catalyze polyubiquitination of RIP1 in vitro even when K63 of ubiquitin is mutated (Fig. 7). Although K63 polyubiquitination does not appear to be required for IKK activation in the TNFα pathway, ubiquitination is required, as RNAi of E1, ubiquitin or Ubc5 prevents TNFα-induced activation of IKK (Figure 2, ,44–6; Supplementary Fig. S3).
Why do the IL-1β and TNFα pathways employ two distinct ubiquitin-dependent mechanisms to regulate IKK? The answer may lie in the choice of different ubiquitin E3s in these pathways. In the IL-1β pathway, TRAF6 is clearly essential for the activation of NF-κB and MAP kinases. In the TNFα pathway, although TRAF2 and TRAF5 are required for signaling, the ubiquitin ligases directly involved in RIP1 polyubiquitination are cIAP1 and cIAP2 [Fig. 7; see also (Bertrand et al., 2008; Mahoney et al., 2008; Varfolomeev et al., 2008)]. Varfolomeev et al also found that Ubc5 and cIAP1 could catalyze polyubiquitination of RIP1 in vitro. However, they concluded that RIP1 was conjugated specifically by K63 polyubiquitin chains although the data showed that a ubiquitin mutant containing K48-only was still capable of supporting RIP1 ubiquitination. Our results not only showed that Ubc5 and cIAP1 could support conjugation of RIP1 by non-K63 polyubiquitin chains, but also provided the first functional evidence that Ubc5 and non-K63 polyubiquitination of RIP1 mediate IKK activation in the TNFα pathway.
The role of ubiquitination in IKK activation is conserved in the Drosophila IMD pathway, a TNF-like pathway that utilizes dIAP2 instead of dTRAF2 as the ubiquitin ligase to activate the Drosophila homologue of IKK (Gesellchen et al., 2005; Kleino et al., 2005). Interestingly, RNAi of Drosophila Ubc13 only partially inhibited IKK activation by peptidoglycans, suggesting that other E2s may also be involved in the IMD pathway (Zhou et al., 2005). Indeed, knockdown of both Ubc13 and Ubc5 led to a much stronger inhibition of the IMD pathway than RNAi of either E2 alone (N. Silverman, personal communication). Although mammalian TRAF2 and TRAF6 have similar domain organizations, only TRAF6 has been shown to activate IKK in vitro, suggesting that the functions of these two RING domain proteins may be quite different (Deng et al., 2000). Similar functional distinction between TRAF2 and TRAF3 has recently been reported in the non-canonical pathway of NF-κB activation (Vallabhapurapu et al., 2008; Zarnegar et al., 2008). Thus, the TNFα pathway appears to use cIAPs instead of TRAF proteins as the E3 for RIP1 ubiquitination. cIAPs then function together with Ubc5 instead of Ubc13/Uev1A to catalyze polyubiquitination in which the polyubiquitin chains can be linked through multiple lysines of ubiquitin.
To determine if a lysine other than K63 of ubiquitin is important for IKK activation by TNFα, we have generated a panel of U2OS cell lines in which endogenous ubiquitin is replaced with a ubiquitin mutant containing a single lysine substitution (K6R, K11R, K27R, K29R and K33R). TNFα-induced polyubiquitination of RIP1 was reduced but not eliminated in these cells (Supplementary Figure S6), suggesting that ubiquitination of RIP1 may occur on more than one lysine (K48R cells were not tested because these cells were not viable after tetracycline induction). To determine if simultaneous mutations of two lysines of ubiquitin impair IKK activation, we generated U2OS cell lines in which endogenous ubiquitin was replaced with Ub(K6R/K63R) or Ub(K11R/K63R) in the presence of tetracycline. As shown in Supplementary Figure S7, ubiquitin containing double mutations on the lysines were still capable of supporting RIP1 ubiquitination and IKK activation induced by TNFα. In contrast, the same mutations completely abolished IKK activation by IL-1β.
Recent studies suggest that linear polyubiquitination of NEMO may play a role in IKK activation (Tokunaga et al., 2009). In addition, the NUB/UBAN domain of NEMO binds to linear di-Ub with higher affinity than it does to K63 di-Ub (Lo et al., 2009; Rahighi et al., 2009). However, full-length NEMO binds to K63 polyubiquitin chains with high affinity due to the presence of the C-terminal zinc finger-type ubiquitin-binding domain, which confers K63 specificity and is important for NF-κB activation (Laplantine et al., 2009). Nevertheless, to test if RIP1 is modified by linear polyubiquitin chains, we used siRNA to knock down the expression of HOIP, a component of the E3 complex (HOIP-HOIL-1L) responsible for linear polyubiquitin chain synthesis. Supplementary Figure S8 showed that RNAi of HOIP did not impair TNFα-induced polyubiquitination of RIP1, consistent with the previous finding that the HOIP-HOIL-1L E3 complex does not ubiquitinate RIP1 in vitro (Tokunaga et al., 2009). We also found that linear polyubiquitin chains synthesized by HOIP-HOIL-1L do not activate IKK, whereas the polyubiquitin chains synthesized by TRAF6 and Ubc13 or Ubc5 directly activate TAK1 and IKK (Xia et al., 2009). Taken together, these results do not support a role of linear polyubiquitination of RIP1 in IKK activation by TNFα. Rather, our results suggest that modification of RIP1 with heterogenous polyubiquitin chains by Ubc5 and cIAPs is sufficient to recruit and activate the TAK1 and IKK complexes in the TNFα pathway.
Previous studies on the role of linkage-specific polyubiquitination were limited to overexpression of ubiquitin mutants in cell lines containing abundant endogenous ubiquitin. Recent advances in mass spectrometry and the development of ubiquitin linkage-specific antibodies allow for the detection of polyubiquitin linkages in endogenous proteins. However, the complexity of ubiquitin gene organization makes it difficult to interrogate the ubiquitin system to study its function by traditional genetic methods. In this study, we have demonstrated the utility of inducible RNAi and transgene expression to study the function of ubiquitin linkages in a mammalian cell culture system. The advantages of this system include the ability to rapidly exchange endogenous ubiquitin with a specific ubiquitin mutant in a controllable manner, analogous to an inducible “knock-in” system. Using this system, we not only demonstrate the essential role of K63 polyubiquitination in IKK activation by IL-1β, but also make the surprising discovery that TNFα activates IKK through a ubiquitin and Ubc5-dependent, but K63 polyubiquitination-independent, mechanism. Just as ubiquitin-linkage specific antibodies significantly facilitate the detection of polyubiquitinated proteins, we believe the ability to replace endogenous ubiquitin with a specific ubiquitin mutant in a controllable manner will facilitate investigating the function of polyubiquitin topology in diverse cellular processes.
Specific primers for shRNA (the sequences are shown in Supplementary Experimental Procedures) were annealed and subcloned into pSUPERIOR.puro vector according to the manufacturer’s instruction (OligoEngine). The tetracycline-inducible shRNA transcription unit was isolated by digestion with HindIII and EcoRI and subcloned into a modified pBluescript vector where BglII and BamHI sites are flanking the shRNA transcription unit (Supplementary Figure S1). The shRNA transcription unit was then isolated by digestion with BamHI and BglII and inserted into the same plasmid that was linearized with BamHI so that two tandem shRNA transcription units could be obtained. This process was repeated to get the transcription unit that contains multiple tetracycline-inducible shRNAs with the same or different sequences. Then the whole transcription unit was isolated by BamHI and BglII digestion and subcloned into a mammalian expression vector containing a puromycin-resistance gene. For the tetracycline-inducible expression of exogenous genes, silent mutations in the cDNA were generated to render them RNAi-resistant. The cDNAs were then subcloned into a modified pcDNA3 vector where two TetO2 sites were inserted between the CMV promoter and the start codon.
U2OS stable cells were treated with tetracycline (1 μg/ml) for 3 to 7 days, then stimulated with IL-1β (10 ng/ml), GST-TNFα (1 μg/ml), or IFNγ (1000 U/ml) for different lengths of time as indicated. The cells were lysed in Buffer A (20 mM Tris-HCl, pH7.5, 150 mM NaCl, 25 mM β-glycerolphosphate, 1 mM Na3VO4, 10% glycerol, 0.5 mM DTT, 1% Triton-X100, 1 mM PMSF and 10 μg/ml leupeptin). In some experiments, N-ethylmaleimide (NEM; 1 mM) was added to Buffer A to inhibit deubiquitination during the preparation of cell lysates. Crude cell lysates were used directly for immunoblotting or subjected to immunoprecipitation according to standard protocol. For immunoprecipitation with the K63-Ub antibody, U2OS cells were lysed in a buffer containing 6M urea, 20 mM HEPES pH 7.5, 150 mM NaCl, 1.5 mM MgCl2, 2 mM EDTA, 2 mM DTT, 0.5% Triton X-100, 1mM PMSF, 10 μg/ml leupeptin and 2mM NEM, at room temperature for 30 minutes. The cell lysate was diluted to reduce urea concentration to 0.3M, and pre-cleared with protein A/G agarose for 1 hour before the K63-Ub antibody (0.5 μg/ml; Millipore) was added to enrich K63 polyubiquitinated proteins.
For IKK activity assay, NEMO antibody (200 ng/ml) and 10 μl of Protein A/G-Sepharose were incubated with cell lysates for 4 hours before the beads were washed with Buffer B (20 mM Hepes-OH, pH7.5, 20 mM NaCl, 20 mM β-glycerolphosphate, 1mM Na3VO4 and 10mM MgCl2). The washed beads were incubated with the N-terminus of IκBα (GST-IκBα-NT, 0.1 mg/ml), 0.1mM ATP and γ-32P-ATP at 30°C for 30 minutes. For EGF stimulation, U2OS cells were serum-starved overnight before treatment with EGF (25 ng/ml).
We thank Dr. Xiaodong Wang (UT Southwestern) for providing tetracycline-inducible shRNA and cDNA expression vectors as well as the SMAC mimetic; Dr. Hongtao Yu (UT Southwestern) for the Cyclin B1 antibody; Dr. Alan Weissman (NIH) for the UbcH5 antibody and Dr. Allen Taylor (Tufts University) for the E1 antibody. This work was supported by grants from National Institute of Health (RO1-AI09919 and RO1-GM63692) and the Robert Welch Foundation (I-1389). M.X is a postdoctoral fellow of Leukemia and Lymphoma Society (5337-08). B. S is supported by an NIH pre-doctoral training grant (GM007062). Z.J.C is an Investigator of Howard Hughes Medical Institute.
Publisher's Disclaimer: This is a PDF file of an unedited manuscript that has been accepted for publication. As a service to our customers we are providing this early version of the manuscript. The manuscript will undergo copyediting, typesetting, and review of the resulting proof before it is published in its final citable form. Please note that during the production process errors may be discovered which could affect the content, and all legal disclaimers that apply to the journal pertain.