Heterologous SUMO-2/3-Ubiquitin Chains Optimize IκBα Degradation and NF-κB Activity

The NF-κB pathway is regulated by SUMOylation at least at three levels: the inhibitory molecule IκBα, the IKK subunit γ/NEMO and the p52 precursor p100. Here we investigate the role of SUMO-2/3 in the degradation of IκBα and activation of NF-κB mediated by TNFα. We found that under conditions of deficient SUMOylation, an important delay in both TNFα-mediated proteolysis of IκBα and NF-κB dependent transcription occurs. In vitro and ex vivo approaches, including the use of ubiquitin-traps (TUBEs), revealed the formation of chains on IκBα containing SUMO-2/3 and ubiquitin after TNFα stimulation. The integration of SUMO-2/3 appears to promote the formation of ubiquitin chains on IκBα after activation of the TNFα signalling pathway. Furthermore, heterologous chains of SUMO-2/3 and ubiquitin promote a more efficient degradation of IκBα by the 26S proteasome in vitro compared to chains of either SUMO-2/3 or ubiquitin alone. Consistently, Ubc9 silencing reduced the capture of IκBα modified with SUMO-ubiquitin hybrid chains that display a defective proteasome-mediated degradation. Thus, hybrid SUMO-2/3-ubiquitin chains increase the susceptibility of modified IκBα to the action of 26S proteasome, contributing to the optimal control of NF-κB activity after TNFα-stimulation.


Introduction
Protein modification with ubiquitin and the ubiquitin-like protein SUMO regulates a large diversity of cellular processes including cell cycle, apoptosis, DNA repair and signal transduction pathways [1]. The attachment of ubiquitin to a substrate, commonly known as ubiquitylation, involves the action of at least three enzymes, a ubiquitin activating enzyme or E1, a conjugating enzyme or E2 and a ubiquitin ligase or E3 [2]. The attachment of one of the three SUMO modifiers (SUMO-1, SUMO-2, SUMO-3) to a target protein (SUMOylation), is a biochemical process similar to ubiquitylation but involving SUMO specific E1, E2 and E3 enzymes [1]. Ubiquitin can be attached as a monomer in a single (monoubiquitylation) or multiple moieties (multiple monoubiquitylation). Ubiquitin can also form polymers of complex composition through the attachment of additional ubiquitin molecules on any of the seven lysine-residues present in each ubiquitin. Canonical functions have been attributed to some of these chain types. Chains linked through lysine 48 (K48) and 11 (K11) are mainly associated to protein degradation [2] meanwhile K63 and linear chains are associated to signal transduction [3,4,5,6,7,8]. However, chain composition appears to be more complex since mixed chains [9,10] as well as heterologous chains including other ubiquitin-like molecules such as SUMO-2/3 have been found [11,12]. Ubiquitylation and SUMOylation are highly dynamic reversible processes where deconjugation is mediated by a set of enzymes generically named deubiquitylating enzymes (DUBs) or SUMO-specific proteases (SUSPs) respectively [13,14,15].
The NF-kB pathway, one of the best-characterized signalling pathways regulated by ubiquitylation [16], leads to a variety of cellular responses, including the induction of pro-inflammatory and anti-apoptotic genes. One of the most abundant forms of NF-kB in mammals is a heterodimer composed of p65 and p50, whose activity is tightly controlled by a family of natural inhibitors named IkBs (a, b and e) [17]. In addition to ubiquitylation, this pathway is controlled by many other post-translational modifications including SUMOylation, NEDDylation, phosphorylation, and acetylation. These frequently have distinct, sometimes antagonistic, functional consequences [18,19,20,21]. Regulation by these posttranslational modifications can occur at different levels of the signalling cascade controlling NF-kB activation, including the activation of the essential IkB kinase IKK [22], maturation of the p50 precursor p105 [23], modification of NF-kB subunits and IkB molecules [16]. IkBa is modified with SUMO-1, which competes with ubiquitin for the same acceptor lysine (K21) during signalmediated stimulation [21]. While polyubiquitylation of IkBa depends on the IKK-mediated phosphorylation of serines 32 and 36 for its subsequent recognition by the ubiquitin-ligase (E3) SCF-bTrCP, IkBa SUMOylation with SUMO-1 does not depend on its phosphorylation [21]. A SUMO E3 ligase for IkBa has not been reported, but the unique E1 (SAE) and E2 (Ubc9) are sufficient for its SUMOylation in vitro [21]. As for other target proteins [13,15], IkBa ubiquitylation and SUMOylation are tightly controlled by the action of unidentified DUBs and SUSPs. In addition to IkBa, other proteins involved in the NF-kB signalling pathway are modified with SUMO (IKKc/NEMO, IKKe and p100) [22,24]. Recently, it has been demonstrated that SUMOylation with SUMO-2 can promote ubiquitylation by ubiquitin E3s and therefore simultaneous substrate modification with both ubiquitin and SUMO proteins is possible [11,12]. Using in vitro and ex vivo approaches, we investigate the role of SUMO-2 and SUMO-3 in the TNFa-induced IkBa degradation and the activation of the NF-kB transcription factor. We found that SUMO-2/3 forms heterologous chains with ubiquitin on IkBa, contributing to its optimal proteasomal degradation. This reveals an unsuspected importance of hybrid chains in TNFa mediated proteolysis of IkBa and subsequent activation of NF-kB promoted transcription.

SUMOylation Contributes to the Optimal TNFa-mediated NF-kB Activation and Degradation of IkBa
To evaluate the contribution of SUMOylation in TNFainduced activation of NF-kB, Ubc9 silencing experiments were performed in HeLa cells (Figure 1). A clear defect of NF-kB activation was observed using an NF-kB luciferase-reporter assay ( Figure 1A) or expression of NF-kB-dependent proteins such as A20 and IkBa ( Figure 1B). More importantly, we also observed a clear defect in TNFa-mediated phosphorylation and degradation of IkBa at the early stage of stimulation (around 5-15 minutes) just prior to the maximal degradation [25,26] ( Figure 1B). To investigate if other known SUMO substrates of the NF-kB were affected at the level of protein stability, western blots against p100 and NEMO/IKKc were performed ( Figure 1B). Our results indicate that IkBa is the only analysed SUMO target of the NF-kB pathway affected at the level of protein stability. Thus, attenuated IkBa SUMOylation correlates with a deficient TNFa-induced IkBa degradation and NF-kB activation.

IkBa is Modified by SUMO-2/3 in vitro and ex vivo
Since knockdown of Ubc9 does not provide information on which of the three SUMO molecules are involved in the regulation of IkBa, we investigated if in addition to SUMO-1 other SUMO molecules were able to modify IkBa. For this, we performed in vitro SUMOylation assays, which clearly indicate that IkBa WT, but not the IkBa mutant K21/22R, is modified by SUMO-1, SUMO-2 and SUMO-3 ( Figure 2A). Similar results were obtained ex vivo after transfection of HEK293 cells with HA-IkBa-SV5 WT and histidine-tagged ubiquitin, SUMO-1, SUMO-2 or SUMO-3 ( Figure 2B). In absence of stimulation, we observed a modification of IkBa by mono-modified forms of ubiquitin and all SUMO proteins. Akin to conjugation with SUMO-1 [21], these monomodified forms of IkBa are likely to be independent of phosphorylation of the serine residues 32 and 36 since these forms occur in absence of TNFa ( Figure 2B, left panel). After treatment with the proteasome inhibitor MG132 and stimulation with TNFa, we can observe an accumulation of high molecular weight bands, suggesting that polyubiquitylation and polySU-MOylation of IkBa (mainly with SUMO-2) are significantly enhanced in this condition ( Figure 2B right panel). Nevertheless, this approach does not exclude the possible integration of ubiquitin into SUMO chains and vice-versa, or simultaneous modification with both ubiquitin and SUMO on different lysines. Interestingly, the strongest band captured with His-ubiquitin, comigrating with SUMO bands under unstimulated conditions, is the only one that appears to promote chain extension after TNFa stimulation, as indicated by its disappearance (marked with an asterisk in Figure 2B). Altogether these results suggest that all three SUMO modifiers have the capacity to modify exogenous IkBa to different extents, with SUMO-2 being the most evident under these experimental conditions ( Figure 2B).

Endogenous IkBa is Modified by SUMO-2/3-Ubiquitin Hybrid Chains
A similar approach was designed to capture endogenous IkBa in HEK293 cells transiently expressing histidinylated versions of ubiquitin, SUMO-1, SUMO-2 or SUMO-3 ( Figures 3A, 3B). Under these conditions, we can observe, in the absence of TNFa stimulation, mono-modified forms of endogenous IkBa with SUMO-2 and SUMO-3 ( Figure 3A). Nevertheless, chain extension on IkBa was only evident in cells transiently expressing His 6ubiquitin after 15 min stimulation with TNFa and in the presence of MG132 ( Figure 3A). We interpreted these results as a handicap of His 6 -SUMO-2/3 to integrate into hybrid chains mainly composed of ubiquitin. We thus predicted that simultaneous coexpression of His 6 -tagged versions of both SUMO-2/3 and ubiquitin would lead to a cooperative effect during the purification, if hybrid chains exist on IkBa. Indeed, co-expression of His 6ubiquitin together with a mix of His 6 -SUMO-2/3 led to a clear increase of modified IkBa compared to expression of His 6ubiquitin or His6-SUMO-2/3 alone, indicating that both ubiquitin and SUMO-2/3 were simultaneously modifying this NF-kB inhibitor. The same results were obtained using two plasmid concentrations ( Figure 3B). SUMOylated forms of IkBa captured with His 6 -SUMO-2/3 were only observed when the film was overexposed (data not shown). In agreement with our in vitro observations (Figure 2A), we found that K21 and K22 are the main residues involved in these modifications, since we observed a significant reduction of the capacity of the mutant IkBa K21/22R to capture SUMO-2/3-Ubiquitin heterologous chains ( Figure 3C). The use of denaturing conditions in this protocol indicates that both modifiers are covalently linked to IkBa. These results were also confirmed by immunoprecipitations using anti-ubiquitin, anti-IkBa and anti-SUMO-2/3 antibodies followed by western-blot detection of IkBa ( Figure 3D). Using this approach it can be observed that in the presence of the proteasome inhibitor MG132 and after 20 minutes of TNFa stimulation, IkBa accumulated as both ubiquitylated and SUMOylated forms. However, after 60 minutes of TNFa stimulation the levels of ubiquitylated IkBa dramatically decreased, while SUMOylated IkBa maintained a modest but consistent increase as compared to the unstimulated condition ( Figure 3D).
To further explore the role of heterologous SUMO2/3ubiquitin chains in the regulation of IkBa stability after cell activation with TNFa, we used a tool recently developed by our group to capture endogenous ubiquitylated proteins. These ubiquitin-traps named TUBEs (Tandem Ubiquitin Binding Entities), specifically capture ubiquitin chains and do not directly bind to SUMO-1, SUMO-2, SUMO-3 or NEDD8 [27]. This technique respects the endogenous level of ubiquitin and ubiquitin-like molecules and therefore do not generate disequilibrium in other cellular functions when over-expression of these protein modifiers is used. Furthermore, this approach allows recovery of samples for further analysis when using a protocol coupled to an IkBa immunoprecipitation [28]. Our results show that the TUBEs-IP procedure efficiently purifies endogenous polyubiquitylated IkBa in a TNFa-mediated time-course response in HEK293 cells ( Figure 4). Consistent with previously published observations [26], a peak of polyubiquitylated IkBa is detected after 20 minutes of TNFa stimulation ( Figure 4C). However, polyubiquitylated IkBa is importantly reduced after 60 minutes of TNFa stimulation, even in the presence of proteasome inhibitors ( Figure 4C). The observed signal is specific as GST or control antibody cannot capture polyubiquitylated IkBa ( Figures 4A and  4B). To our surprise, under unstimulated conditions a basal level of polymodified IkBa was detected with the antibodies recognizing . Cells were co-transfected with a NF-kB-luciferase reporter plasmid (3EnhancerConA) and bgalactosidase reporter. Twenty-four hours later cells were stimulated with TNFa (15 ng/ml) as indicated and luciferase and b-galactosidase activities measured as previously described [50]. The graph corresponds to the mean of three independent experiments. (B) HeLa cells were transfected during 72 h with control or Ubc9 siRNA (100 nM) and stimulated with TNFa (15 ng/ml) as indicated. Western-blot analyses were performed with the indicated antibodies. doi:10.1371/journal.pone.0051672.g001 IkBa, ubiquitin and SUMO-2/3 antibodies (Figures 4C). Under these conditions, no significant signal was detected with anti SUMO-1 antibody (data not shown). The basal levels of IkBa ubiquitylation or SUMOylation are not artefacts of the TUBEsmediated capture, as we do not observe this in non-activated rat tissues [29] (data not shown). Interestingly, antibodies recognizing SUMO-2/3 are able to specifically detect high molecular weight molecules progressively integrated within the chain architecture of polyubiquitylated IkBa ( Figures 4C and 4D). It is important to note that while the increase of SUMOylation, as detected with anti-SUMO2/3 antibody appears to be modest, it does not decrease after 60 minutes of TNFa-stimulation, as is the case for the signal detected with the anti-ubiquitin antibody (Figures 4C). These findings also confirm the earlier results obtained by immunopreciptiation ( Figure 3D) and suggest that IkBa captured after 60 minutes of TNFa-stimulation could correspond to ubiquitin chains enriched in SUMO-2/3. Thus taken together our data indicate that IkBa is modified by hybrid chains composed of SUMO-2/3 and ubiquitin.

SUMO Molecules are Integrated within Ubiquitin Chains
In order to evaluate the contribution of all SUMO molecules in the architecture of ubiquitin chains modifying IkBa, we silenced Ubc9 and performed a TUBE-IkBa immunoprecipitation. This was done after 20 minutes of TNFa-stimulation thus coinciding with the peak of ubiquitylated IkBa, allowing us to analyse the contribution of SUMO-2/3 in the formation of hybrid chains. This timing will also favour the TUBEs-mediated capture of ubiquitylated species of IkBa. It was first confirmed that, while Ubc9 silencing reduces the input of SUMOylated molecules, it does not significantly affect the input of ubiquitylated substrates ( Figures 5B and 5C). However, knockdown of Ubc9 led to decreased amounts of modified IkBa recovered with the TUBE-IP method, as detected with anti-IkBa and anti-ubiquitin antibodies ( Figures 5A and B). Under the same experimental conditions western blot detections with anti SUMO-2/3 antibody show reduced levels of hybrid SUMO-2-ubiquitin chains ( Figure 5C) despite the fact that the peak of IkBa SUMOylation is at 60 minutes of TNFa-stimulation ( Figure 3D and Figure 4). Propor- tionally, the silencing of bTrCP has a higher impact on the capture of both ubiquitylated and SUMO-2-ubiquitin modified IkBa (data not shown), indicating that ubiquitin is a major component of the hybrid chains and confirming the heterologous nature of these chains.

Ubiquitin and SUMO-2 Promote Efficient Chain Extension on IkBa
To investigate if SUMO or ubiquitin molecules could promote chain extension on IkBa, chimeric proteins were obtained by fusing these modifiers to the N-terminus of IkBa. Since the conjugation site is very close to the IkBa N-terminus, we expect that the fusion protein will behave similarly to the endogenous SUMOylated protein. This strategy has been successfully used by several groups exploring the stability, localization and activity of diverse cellular factors [22,30,31]. Due to the high homology between SUMO-2 and SUMO-3 (86% identity) as well as the capacity of SUMO-2 to favour chain extension (Figure 2), we decided to develop a chimeric protein where IkBa was fused to SUMO-2 ( Figure 6A). To make fusion proteins, the double Cterminal glycines (GG) were replaced by double alanines (AA) to avoid cleavage by DUBs or SUSPs. HA-N-terminal and SV5-Cterminal tags were included to monitor the integrity of the fusion proteins when expressed ex-vivo. To analyse the contribution of the fused moiety in the chain composition, chain extension and halflife of IkBa, fusions were made without the two lysines 21 and 22 necessary for IkBa ubiquitylation and SUMOylation. However, the presence of these lysines does not significantly alter the results obtained (data not shown). To determine the capacity of each fusion protein to be further modified by ubiquitin or SUMO, the different IkBa fusions were co-expressed with a histidinylated version of ubiquitin ( Figure 6B) and SUMO-2 ( Figure 6C) in HEK293 cells. As shown in Figure 6B, TNFa drives an efficient ubiquitin chain extension of both ubiquitin-IkBa and SUMO-2- Empty vector (EV) was also used to compensate plasmid DNA to final concentration of 2 or 4 mg respectively. Cells were pre-treated with MG132 and stimulated with TNFa during the indicated times. His 6 -ubiquitylated or SUMOylated proteins were purified using denaturing conditions and Ni 2+ chromatography. Captured material was analysed by western-blot with the indicated antibodies. (C) HEK293 cells were transfected with IkBa-SV5 WT or mutated on K21 and K22 in the presence of His 6 Ubiquitin, His 6 -SUMO2 and His 6 -SUMO3, pre-treated with MG132 and stimulated with TNFa. His 6 -modified proteins were purified using denaturing conditions and Ni 2+ chromatography procedure (D) Time-course modification of IkBa after TNFa-stimulation analysed by immunoprecipitations with anti-IgG control, anti-ubiquitin, anti-IkBa and anti-SUMO-2/3 antibodies. Cells were treated with MG132, stimulated with TNFa and lysates were submitted to immunoprecipitation experiments as indicated. Precipitated material was analysed by western-blot with anti-IkBa antibody. doi:10.1371/journal.pone.0051672.g003 IkBa fusions, far superior to the one observed with IkBaWT or SUMO-1-IkBa fusion. Similar observations were obtained when capturing SUMO-2 modified IkBa-fusions ( Figure 6C). Whereas IkBaWT and SUMO-1-IkBa appear to be mainly mono-modified with SUMO-2, ubiquitin-IkBa and SUMO-2-IkBa are also polySUMOylated after 1hr pre-treatment with MG132 and TNFa stimulation ( Figure 6C). However, a significant difference in the levels of ubiquitin-IkBa and SUMO-2-IkBa captured by His6-ubiquitin and His6-SUMO-2 indicate again that ubiquitin is a major component of the ubiquitin-SUMO-2 hybrid chains (compare Figures 6B and 6C).
The capacity of the different IkBa-fusion proteins to be modified with SUMO-1 and SUMO-2 was also analysed in vitro using SUMOylation assays. Although these assays lack any potential SUMO E3 for IkBa, resulting in a less efficient modification than ex vivo approaches, some differences can be observed. Only SUMO-2-IkBa was efficiently modified by SUMO-1 and SUMO-2 ( Figure 6D). While a modest modification was observed for IkBaWT and SUMO-1-IkBa, no modification was detected on the ubiquitin-IkBa fusion. As IkBa lysines 21 and 22 are absent from all fusion proteins, the observed polySUMOylation of SUMO-2-IkBa might occur on the SUMO consensus site of SUMO-2 moiety [12]. The fact that the SUMO consensus is absent from SUMO-1 [32], suggests that when using overloaded in vitro conditions, additional lysine residues on SUMO-1 could be involved in this SUMOylation process [33]. To be able to capture ubiquitin-SUMO-2 hybrid chains from the distinct IkBa-fusions, we performed a TUBEs capture experiment. Results were similar to those obtained with the nickel beads approach using His6ubiquitin or His6-SUMO-2, excepting that in the presence of TUBEs, very high molecular weight forms of SUMO-1-IkBa and SUMO-2-IkBa fusions were captured ( Figure 6E). Under these  experimental conditions, the bulk of ubiquitylated forms of IkBa appear to be shorter with ubiquitin-IkBa than the one obtained with the SUMO-IkBa fusions ( Figure 6E). Thus, single ubiquitin and SUMO-2 moieties consistently promote further modification of IkBa with both ubiquitin and SUMO-2.

SUMO2/3-Ubiquitin Heterologous Chains Drive an Efficient 26S Proteasomal Degradation of IkBa
As SUMO-2/3 appears to play a role in the formation of ubiquitin chains on IkBa, we decided to set up in vitro conjugation assays using all recombinant components and in vitro translated IkBa. In these assays, cell extracts from TNFa-stimulated HEK293 cells were used as source of E3s (see materials and methods). Using suboptimal conditions of conjugation, we observed by western blot (Figure 7A) or S 35 labelled IkBa ( Figure 7B) that the simultaneous conjugation with ubiquitin and SUMO-2/3 allowed a more efficient hybrid chains formation on IkBa. However, the use of saturating conditions of conjugation does not always allow evaluation of cooperative effects of SUMO-2/3 on ubiquitin chain extension ( Figure 7C and 7D). To investigate the role of ubiquitin-SUMO hybrid chains in proteasomal degradation of IkBa, in vitro modified material was submitted to degradation by the 26S proteasome. Different molar ratios of Ubiquitin: SUMO were tested to identify the optimal condition for IkBa modification and proteasomal degradation in vitro. We found that the ratios 2:1/1 or 1:1.5/1.5 of Ubiquitin: SUMO-2/SUMO-3 were the most efficiently modified ( Figure 7C  lanes 3 and 4). However, the ratio 2:1/1 (lane 3) showed the best 26S-mediated proteasomal degradation of the modified IkBa ( Figure 7C, bottom panel). Using these settings, we performed similar reactions with the same molar concentrations of Ubiquitin, SUMO-2/SUMO-3 or the combination of both. In this assay the abundant recombinant material (E1, E2 and modifiers) aims to out-compete but does not exclude the integration of SUMO or ubiquitin present in the transcription/translation reaction. While degradation of ubiquitin or SUMO-2/3 modified IkBa appears to be modest under these conditions (with only around 15% of the modified material been degraded), degradation of ubiquitin-SUMO hybrid chains is significantly more efficient and goes up to 46% ( Figure 7D upper and bottom panels). These observations are also reflected in the amount of unmodified IkBa in a lower exposed film ( Figure 7D middle panel). To confirm the contribution of SUMO-2/3 in the formation of ubiquitin chains that are driven to proteasomal degradation, SUMO-2/3-ubiquitin chains were purified using a TUBEs-IP IkBa protocol from control and siRNA Ubc9 cells. Immunopurified material was exposed to an in vitro degradation assay in the presence of purified 26S proteasome. Our results indicate that hybrid chains on IkBa promote a more efficient degradation of this protein after incubation with purified 26S proteasome as compared to the siRNA Ubc9 conditions ( Figure 7E). These results also indicate that direct effects of SUMO-2/3 deficient ubiquitin chains on IkBa degradation can be evaluated using this method. Furthermore, under siRNA Ubc9 condition, we do not observe an in vitro deconjugation of IkBa, suggesting that de-modifying enzymes associated with the proteasome, are likely more efficient when SUMO molecules are integrated within the ubiquitin chains ( Figure 7E). Cumulatively, these results are compatible with the proposed role of SUMO-2/3 molecules in the formation of ubiquitin chains and suggest that IkBa modification with SUMO-2/3 contribute to its optimal ubiquitin-dependent degradation by the proteasome in a similar manner as reported for PML [11,12].

Discussion
Considerable evidence underscores the significance of SUMOylation in the regulation of the transcription factor NF-kB [21,22,24]. Here, we demonstrate the importance of SUMOubiquitin hybrid chains in the TNFa-induced degradation of endogenous IkBa and activation of the NF-kB transcription factor ( Figure 8). Multiple in vitro and ex vivo approaches such as the use of TUBEs and TUBEs/IP, Ubc9 silencing, Ni 2+ : NTA chromatography and reconstituted in vitro systems support these conclusions. We have shown that silencing of Ubc9 leads to loss of phosphorylation of IkBa, attenuation of SUMO-2-Ubiquitin heterologous chains on IkBa, decreased proteasomal degradation of IkBa and a delay in NF-kB activation. Nevertheless, the role of SUMO in NF-kB signalling is difficult to integrate into a simple model, mainly due to its capacity to act at different levels of this pathway [21,22,24]. Although the stability of other known SUMO targets such as p100 or IKKc/NEMO were not affected by the Ubc9 silencing, we cannot exclude that the function of these SUMO targets is altered or that other unidentified SUMO substrates regulate the stability of IkBa and NF-kB-dependent transcription. For instance, upon genotoxic stress the SUMO ligase PIASy induces the modification of IKK with SUMO-1 but not SUMO-2/3, resulting in an increased NF-kB activity [34]. One could speculate that the inhibition of Ubc9 could have affected IKK activity by reducing its SUMOylation, repressing IkBa degradation and consequently acting on NF-kB activity. However, SUMOylation has also been associated to repression of RelA/p65 nuclear translocation [35], making difficult to conclude that Ubc9 silencing will always promote repression of NF-kB activity. An important piece of evidence is that TUBE captures endogenously modified IkBa, mainly composed by ubiquitylated and SUMO-2/3 conjugated forms of IkBa. TUBEs-captured IkBa behave as previously reported with a peak of ubiquitylated IkBa after 20 minutes of TNFa stimulation followed by a reduction at 60 minutes, even in the presence of the proteasome inhibitor MG132 [26]. Remarkably after 60 minutes of TNFa stimulation and proteasome inhibitor treatment, the proportion of IkBa modified by SUMO-2/3 is not significantly reduced. These results suggest that the fraction of IkBa modified by hybrid SUMO-Ubiquitin chains is resistant to deubiquitylating enzymes (DUBs). Consistent with these results IkBa mainly modified with SUMO-2/3 chains is a poor 26S proteasome substrate.
Although our results clearly demonstrate that SUMO-2/3 favours the covalent modification of IkBa with ubiquitin, the order of chain formation needs to be further investigated. Due to the inefficiency of ubiquitin to prime SUMO-2 chain extension in vitro, one can speculate that SUMO-2 could be attached first to somehow predispose the modified IkBa to ubiquitylation, playing a potential role of ubiquitin-chain extender. The SUMO-2priming ubiquitylation model is also consistent with the fact that a monoubiquitylation of IkBa induces time-dependent resistance to proteolysis [29]. However, the fact that SUMO-2/3 better integrates ubiquitin chains than ubiquitin integrates SUMO-2/3 chains, suggest that both protein modifiers have the potential to prime and extend chains. Conclusions on the order of integration into the ubiquitin chains have to be taken with caution as information, recently published, indicates that over-expression of ubiquitin-like proteins or stresses such as inhibition of the proteasome activity can force the ubiquitin conjugating enzymes to incorporate ubiquitin like proteins into ubiquitin chains [36,37,38]. The existence of one main lysine receptors for SUMOylation (K21) indicates that if SUMO-2/3 is attached first to IkBa, ubiquitin should be attached on it to generate chain extension. If ubiquitin is attached first to IkBa, two ubiquitin-SUMO mixed chains are most likely to be formed on K21/K22. However, one other hypothetical possibility is that there is a mixed population of independently SUMOylated (on K21) and Ubiquitylated (on K22) IkBa as it is illustrated in Figure 8. Mass spectrometry data should put some light on the architecture of Ubiquitin-SUMO-2/3 chains modifying IkBa. A major technical bottleneck is the identification of SUMO peptide signatures attached to endogenous proteins. This is reflected by the fact that only a bit more than a hundred SUMOylation sites have been identified [39,40].
All together our data suggest that SUMO2/3 can be incorporated into ubiquitin-chains to regulate IkBa proteasomal degradation. Beyond the observation that the optimal proteasome degradation of IkBa is mediated by hybrid SUMO-ubiquitin chains conditioning the activity of NF-kB, the integration of SUMO-2/3 into the hybrid SUMO-ubiquitin chains could simultaneously favours ubiquitin chain extension and recycling of IkBa (Figure 7). In apparent contradiction, chains enriched with SUMO-2/3 also appear to promote proofreading/recycling of IkBa. Therefore our data propose the existence of a mechanism regulating the balance between SUMO-2/3 and ubiquitin into the same chain, offering the possibility to regulate proteasomemediated proteolysis and resistance to the action of DUBs. Although some aspects of this speculative working model (Figure 8) require further investigation, it suggests the presence of a dynamic way to regulate protein degradation, proofreading, recycling of molecules and recovery of the cytoplasmic reservoir of NF-kB/ IkBa, typical of this highly dynamic system.
Under hypoxic conditions SUMOylation of IkBa is also regulated [41,42] indicating that different types of stimulation might influence the level of IkBa SUMOylation. Other SUMO substrates are regulated under different stress conditions such as heat shock, proteasome inhibition or chemotherapeutic drugs [39,43,44]. The best-known protein regulated by SUMO-2/3 is the promyelocytic leukemia protein PML whose degradation is induced by Arsenic Trioxide (ATO) [11,12]. As occurs for IkBa, the formation of chains containing Ubiquitin and SUMO-2/3 drive PML to proteasomal degradation [11,12]. The mechanism driving the formation of these hybrid chains on PML is well characterized and involves the action of the SUMO-dependent ubiquitin ligase RNF4 [11,12]. However, in the case of IkBa it is not clear if the mechanism is similar since SUMO-2/3 does not favour the integration of ubiquitin moieties after TNFa treatment as ATO does it for PML (Data not shown). Furthermore, RNF4 SIM domains do not capture SUMOylated IkBa after proteasome inhibition with MG132 and TNFa-stimulation (data not shown). Whether the mechanism of IkBa degradation involves the participation of a different SUMO-dependent ubiquitin-protein ligase or ubiquitin-dependent SUMO-ligase will have to be investigated.
Thus, the proteolytic mechanisms involving hybrid SUMO-Ubiquitin chains is not restricted to PML. The evidence presented here underlines the contribution of SUMO-2/3 in the control of IkBa degradation. This time-dependent formation of high molecular weight ubiquitin-SUMO-2/3 chains likely optimises IkBa proteasomal degradation and controls NF-kB activity (Figure 8). The mechanism of SUMO-ubiquitin hybrid chains could be more general and might affect more protein targets than initially suspected [45].

Cloning
Ubiquitin, SUMO-1, SUMO-2 (accession numbers CAA44911, NM-003352 and NM-006937 respectively) were used to generate IkBa fusions and cloned into BamHI/Not1 restriction sites of a modified pcDNA3 vector containing a N-terminal HA tag and Cterminal SV5 tag. The C-terminal glycine residues (GG) of SUMO-1, SUMO-2 or ubiquitin were changed to alanine (AA), and lysine 21 and 22 of IkBa were mutated to alanine to avoid respectively the action of DUBs and additional attachment of moieties at the N-terminus of IkBa using the following oligonucleotides: for ubiquitin fusion proteins 59-ctc cgt ctt aga gct gcg gag cgg cta ctg gac gac-39 and 59-gtc gtc cag tag ccg ctc cgc agc tct aag acg gag-39, for SUMO1 59-cag gaa caa acg gcg gct gag cgg cta ctg gac gac-39, and 59-gtc gtc cag tag ccg ctc agc cgc cgt ttg ttc ctg-39 and for SUMO-2 59-caa cag gag acg gca gct gag cgg cta ctg gac gac-39and 59-gtc gtc cag tag ccg ctc agc tgc cgt ctg ctg ttg-39. As a control, a construct of IkBa WT, containing an N-terminal HA tag and C-terminal SV5 tag, was used ( Figure 6A). All constructs have been verified by DNA sequencing. His-6-Ubiquitin, His-6-SUMO-1, His-6-SUMO-2 and His-6-SUMO-3 have been previously reported [47].

Purification of SUMO and Ubiquitin Chains
His 6 -ubiquitylated or SUMOylated proteins were purified using denaturing conditions and Ni 2+ chromatography as previously described [48]. The use of low-density nickel beads (QLNI-100, ABT) reduced the capture of sticky unmodified IkBa ( Figure 3B) compared to high-density beads (QLNI-25, ABT) ( Figures 2B and  3A). However, it also reduces the purification of poorly expressed IkBa monoSUMOylated forms ( Figure 3A). To capture ubiquitin chains using TUBEs, the lysis buffer was supplemented either with 3.5 mM of TUBEs hHR23A or GST as previously described [27,28]. Lysates were clarified by cold centrifugation, and added to glutathione agarose beads (Sigma). Glutathione beads were eluted and bound material was submitted to western blot analysis or to IkBa, ubiquitin or SUMO2/3 immunoprecipitations.
In vitro 26S-mediated Degradation Assay of TUBEcaptured IkBa Polyubiquitylated proteins were captured using TUBEs as described [28]. Samples were briefly eluted from TUBEs using a glycine buffer pH 4 and equilibrated with Tris pH 7.5 to a final concentration of 100 mM. Eluted material was immunoprecipitated using a specific IkBa antibody. Ubiquitylated IkBa was eluted from specific antibodies with glycine pH 2.5, neutralized as indicated previously, before being submitted to an in vitro degradation assay as reported [27] using 2 mg of 26S proteasome (ENZO) for 30 and 120 minutes at 37uC.
In vitro SUMOylation, Ubiquitylation and Hybrid Ubiquitylation-SUMOylation Assays incubated at 30uC for 2 hours and stopped by addition of SDS sample buffer. Reaction products were resolved by SDS-PAGE and dried gels analysed by phosphorimaging.