Covalent modification of proteins with ubiquitin (Ub) is widely implicated in the control of protein function and fate. Over 100 deubiquitylating enzymes rapidly reverse this modification, posing challenges to the biochemical and biophysical characterization of ubiquitylated proteins. We circumvented this limitation with a synthetic biology approach of reconstructing the entire eukaryotic Ub cascade in bacteria. Co‐expression of affinity‐tagged substrates and Ub with E1, E2 and E3 enzymes allows efficient purification of ubiquitylated proteins in milligram quantity. Contrary to in‐vitro assays that lead to spurious modification of several lysine residues of Rpn10 (regulatory proteasomal non‐ATPase subunit), the reconstituted system faithfully recapitulates its monoubiquitylation on lysine 84 that is observed in vivo. Mass spectrometry revealed the ubiquitylation sites on the Mind bomb E3 ligase and the Ub receptors Rpn10 and Vps9. Förster resonance energy transfer (FRET) analyses of ubiquitylated Vps9 purified from bacteria revealed that although ubiquitylation occurs on the Vps9‐GEF domain, it does not affect the guanine nucleotide exchanging factor (GEF) activity in vitro. Finally, we demonstrated that ubiquitylated Vps9 assumes a closed structure, which blocks additional Ub binding. Characterization of several ubiquitylated proteins demonstrated the integrity, specificity and fidelity of the system, and revealed new biological findings.
Ubiquitin (Ub) is a key post‐translational modifier in all eukaryotes. The process of ubiquitylation, in which Ub attaches to a lysine residue of a substrate protein, regulates the function of thousands of proteins and is involved in numerous cellular processes such as protein degradation, DNA remodelling and repair, and protein trafficking. Malfunctions in the Ub system underlie various human disorders, including cancer and neurodegenerative, metabolic and infectious diseases (Ciechanover and Brundin, 2003; Nalepa et al, 2006; Mizushima et al, 2008). Ubiquitylation involves the concerted action of Ub‐activating, Ub‐conjugating, and Ub‐ligase enzymes (E1, E2 and E3, respectively); the latter conclude the process by promoting the covalent attachment of Ub to a lysine residue of a specific target protein (Hershko et al, 1980; Wilkinson et al, 1980; Bachmair et al, 1986; Li et al, 2008; Deshaies and Joazeiro, 2009). Some forms of ubiquitylation serve as canonical signals for rapid degradation by proteasomes and lysosomes, while others perform various signalling functions.
Ubiquitylated proteins are recognized by a large set of cognate Ub receptors (Hicke et al, 2005; Hurley et al, 2006). These Ub receptors decrypt the Ub signal by tethering a Ub‐binding domain (UBD) to a functional domain, thus linking the ubiquitylated target to a specific function in trans. Many Ub receptors undergo coupled monoubiquitylation themselves, presumably in order to impose a closed (cis) conformation that prevents binding of ubiquitylated targets (Di Fiore et al, 2003; Prag et al, 2003; Shih et al, 2003; Hoeller et al, 2006). Deubiquitylating enzymes (DUBs) are able to rapidly reverse this auto‐regulatory signal by opening the cis conformation. Hence, the current models suggest that Ub receptors may exist in three states: (i) in the free, unmodified apo form, (ii) in the ubiquitylated and closed cis form, which is inactive due to intramolecular Ub–UBD interactions and (iii) in the trans form, which binds ubiquitylated target proteins through the UBD. The rapid dynamics of ubiquitylation/deubiquitylation has impeded the structure–function analysis of Ub receptors. Indeed, of the >100 UBD structures in the Protein Data Bank (PDB; January 2011), none is of a ubiquitylated cis form.
To bypass this obstacle, we reconstructed the entire eukaryotic ubiquitylation system in Escherichia coli. The system is built modularly, facilitating the straightforward production of ubiquitylated proteins. Two affinity tags, one on Ub and the other on the substrate, enable the ubiquitylated proteins to be purified in quantities and homogeneity suitable for biochemical, biophysical and crystallographic analyses. This system will facilitate the determination of the structures of ubiquitylated proteins, and will enable high‐resolution insight into cellular mechanisms for interpreting Ub signals. Moreover, permutations of the system can be devised to screen for targets of E3 enzymes and UBDs or to study in detail the ubiquitylation of human disease proteins, including the facile identification of their modified residues.
In this work, we demonstrate the utility of this expression system by reporting several new biological findings. First, we demonstrate E2‐specific auto‐ubiquitylation of the E3 enzyme, Mind bomb. Next, we show that Epsin proteins undergo E3‐independent ubiquitylation. Finally, we focus on the ubiquitylation of the yeast Ub receptor Vps9, in which a UBD is tethered to a guanine nucleotide exchanging factor (GEF) domain. We characterize in vitro the GEF activity and Ub‐binding functions of this Ub receptor in its apo, trans and cis forms. A cycle of Ub binding and monoubiquitylation was postulated to regulates the association of the human orthologue of Vps9, Rabex‐5, with endosomes (Mattera and Bonifacino, 2008). Our study provides a direct explanation for the regulatory function of Ub in controlling the cellular localization of Vps9.
Construction of an integrated recombinant ubiquitylation system in bacteria
We developed a ubiquitylation system in bacteria that permits the purification of stably modified ubiquitylated proteins. The system consists of two compatible expression vectors. One is a generic plasmid (pGEN) that harbours His6–Ub, E1‐activating enzyme and E2‐conjugating enzyme (Figure 1). We cloned these genes into a modified pHis6‐parallel2 vector in which the β‐lactamase gene was replaced with a kanamycin resistance cassette. A second plasmid (pCOG) encodes a selected substrate for ubiquitylation and its cognate E3 ligase. In this vector, the substrate is typically fused to glutathione S‐transferase (GST) or to maltose‐binding protein (MBP) affinity tags. We constructed each type of vector such that it expressed its corresponding genes from a single promoter (pT7 or pTac), giving rise to a polycistronic mRNA. We produced a large collection of pGEN and pCOG vectors expressing different E2s, E3s, substrates, and their mutants (Supplementary Table SIV). Co‐expression of the proteins encoded by these vectors yields large quantities of ubiquitylated substrates (0.5–1.0 mg of purified ubiquitylated Vps9 or Rpn10 per 1 l culture). Affinity tags such as His6 fused to Ub, and MBP or GST fused to the substrate facilitate separation of the ubiquitylated from the unmodified substrate (as demonstrated for Rpn10 and Vps9, respectively). Moreover, tobacco etch virus (TEV) protease or rhinovirus protease recognition sites located between the tags and the ubiquitylated substrate enable tag removal, which, upon further affinity chromatography, facilitates the separation of the ubiquitylated product from the non‐modified form. Using this method, the ubiquitylated protein is highly purified and ready for biochemical and biophysical studies.
To assess the functionality of the system, we first produced pGEN vectors co‐expressing Ub and wheat Uba1 with either human UbcH5b or yeast Ubc5, the most promiscuous E2s with regards to association with E3s. Since most E3 ligases undergo auto‐ubiquitylation in the absence of a specific substrate (Lorick et al, 1999), we simplified the system by first testing whether E3 ligases would undergo auto‐ubiquitylation when expressed in bacteria from pCOG. E3 ligases fall into two large groups, U‐Box/RING (really interesting new gene) and HECT (homologous to E6‐AP carboxyl terminus) domain enzymes (Lorick et al, 1999; Hatakeyama et al, 2001; Hagglund and Roizman, 2002). To assess the general applicability of the reconstituted system, we first asked whether RING‐ and HECT‐domain E3 ligases undergo auto‐ubiquitylation in bacteria (Figure 2).
The RING‐containing E3 Mind bomb (Mib) of Danio rerio (zebrafish), which is known to promote endocytosis in the Delta/Notch pathway (Itoh et al, 2003), was chosen as a representative protein. A GST–Mib construct containing its three RING domains (a.a. 738–1030) was co‐expressed with His6–Ub, E1 and UbcH5b, sequentially purified on reduced glutathione‐ and Ni+2‐affinity matrices and resolved by sodium dodecyl sulphate–polyacrylamide gel electrophoresis (SDS–PAGE). The purified protein displayed a laddering pattern that entirely depended on the co‐expression of Ub, indicating that Mib underwent multi‐monoubiquitylation or polyubiquitylation (Figure 2A). A similar pattern was observed by western blot analysis with an anti‐His antibody recognizing the His6–Ub fusion (Figure 2A). The ubiquitylation pattern of Mib observed here was consistent with the pattern previously found to occur in vitro (Itoh et al, 2003), confirming the reliability of the system.
We next tested whether the ubiquitylation signal on Mib is K48‐ or K63‐polyUb, the most abundant polyUb modifications (Peng et al, 2003). We found that replacing the wild‐type Ub with a point mutant in which lysines 48 and 63 were replaced by arginine residues (K48R, K63R double mutant) did not affect the ubiquitylation pattern, suggesting that Mib undergoes either multi‐monoubiquitylation or polyubiquitylation through lysines other than the K48 or K63. We then used Ub K0 (in which all Lys residues were mutated to Arg). The ladder pattern was retained, demonstrating that Mib is multi‐monoubiquitylated in the bacterial system. Using mass spectrometry analysis, we identified five distinct lysine residues that underwent ubiquitylation (Figure 2D; Supplementary Table SI; Supplementary Figure S1).
Since Mib is expressed as GST fusion, we tested whether ubiquitylation takes place on Mib or on the GST moiety. The linker between the fusion partners was cleaved and the proteins were separated by chromatography on a GSH‐affinity matrix. As shown in Figure 2B and C, the flow through (FT) contained heavily ubiquitylated Mib. Using anti‐GST antibody, we found no evidence for ubiquitylation of GST (Figure 2B, Elution). However, blotting with anti‐His antibody detected a faint band at ∼39 kDa which may represent a minor fraction of monoubiquitylated GST (see Elution in Figure 2C). Thus, the majority of the Ub modification was attached to Mib not to GST. To determine whether HECT‐containing E3 ligases are also active in the reconstituted ubiquitylation system, we cloned a representative HECT protein, yeast Rsp5, as MBP fusion, into pCOG. We found that, like Mib, Rsp5 underwent auto‐ubiquitylation when co‐expressed with its cognate E2s yeast Ubc4 or Ubc5 (Figure 2E–G). HECT proteins contain a catalytic cysteine that transfers the Ub to the substrate. When we substituted the predicted catalytic cysteine by alanine (C777A), the ligase failed to charge with Ub (Figure 2G), although the expression levels of ubiquitylation components were equal in both bacterial lysates (Figure 2E; Supplementary Figure S4). Interestingly, the level of Ub‐charged E2 was very low apparently because Ub is rapidly transferred to the E3 ligase (Figure 2G). However, in the presence of catalytically inactive Rsp5, accumulation of Ub‐charged E2 was apparent (compare WT and C777A lanes in Figure 2F). Nevertheless, Ub‐charged E3 did not accumulate when a substrate such as Rpn10 was added to the reconstituted system, ostensibly because of rapid Ub transfer to the substrate (Supplementary Figure S4), since Ub was transferred to it. Taken together, this series of experiments demonstrates our successful reconstitution of the entire ubiquitylation cascade for RING and HECT E3 ligases in E. coli.
The reconstituted system faithfully recapitulates Rpn10 ubiquitylation
It has recently been shown that Rpn10 (regulatory proteasomal non‐ATPase subunit) is ubiquitylated by Rsp5 in vivo at a single lysine residue (K84) (Isasa et al, 2010). However, in‐vitro assays performed by the same team and others showed that Rpn10 and its human orthologue S5a undergo multi‐monoubiquitylation or polyubiquitylation (Lu et al, 2008; Uchiki et al, 2009; Isasa et al, 2010). To test whether, unlike an in‐vitro ubiquitylation assay, the reconstituted bacterial system would faithfully recapitulate Rpn10 ubiquitylation observed in S. cerevisiae in vivo, we studied the dependence of Rpn10 ubiquitylation on the concerted action of E1, E2 and E3 enzymes. As shown in Figure 3A, co‐expression of His6–Ub, Uba1, Ubc5 and Rsp5 was necessary and sufficient for Rpn10 monoubiquitylation. To assess the requirement of individual components, we systematically omitted each of the enzymes and the substrate (Figure 3A and B; for typical expression levels of E1, E2 and E3, see Supplementary Figure S4). As expected, Rpn10 monoubiquitylation was only observed when all components were present. Applying the scheme depicted in Figure 1, we purified ubiquitylated Rpn10. Sequential purification via the two affinity tags, MBP on Rpn10 and His6 on Ub, efficiently isolated an apparently homogenous species corresponding to monoubiquitylated Rpn10 (Figure 3C). Size‐exclusion chromatography upon enzymatic removal of the MBP and the His6 tags resulted in a highly symmetric peak at the elution volume expected for correctly folded Ub‐Rpn10 (Figure 3C). The yield of pure Ub‐Rpn10 obtained from 1 L of bacterial culture was ∼1.0 mg.
Using an in‐vitro ubiquitylation assay followed by mass spectroscopy analysis, Crosas and co‐workers identified four lysine residues in Rpn10 that were modified by Ub (K71, K99, K84 and K268), although K84 was found as the major site (52%). Using tandem mass spectrometry, we only identified K84 as ubiquitylated in Ub‐Rpn10 produced in the bacterial system (Figure 3D; Supplementary Figure S2; Supplementary Table SII). Although the mass spectrometry on an LTQ‐Orbitrap instrument employed here is exquisitely sensitive, we cannot entirely exclude the possibility of minor modifications of other lysine residues in Rpn10.
Rpn10 contains a Ub interaction motif (UIM), a domain that in some cases can mediate the ubiquitylation of a fusion partner such as GST (Oldham et al, 2002). Our mass spectrometry data revealed a similar phenomenon as we identified ubiquitylation of MBP within the context of the MBP–Rpn10 fusion protein (Supplementary Table SII). This finding corroborates the suggestion that UIMs can function to direct protein ubiquitylation.
Specificity and fidelity of the bacterial reconstituted ubiquitylation system
In the ubiquitylation process, particular protein–protein interactions govern several levels of specificity, including the interactions of E1:Ub‐like protein (UbL), E1:E2, E2:E3 and E3:substrate. For example, the E1 Ub‐activating enzyme (Uba1) associates with Ub but not with the UbL SUMO; whereas the E1‐like enzyme of the autophagy system, ATG7 cooperates with two different UbLs, ATG12 and LC3 (Schulman and Harper, 2009). E. coli possesses at least two UbLs that serve as sulphur donors in the thiamine (vitamin B1) and the molybdenum cofactor biosynthesis pathways (Gutzke et al, 2001; Lake et al, 2001; Wang et al, 2001). We, therefore, asked whether bacterial E1‐like enzymes interact with the eukaryotic factors in the context of the reconstituted system. The lack of Rpn10 ubiquitylation in the absence of Uba1 (Figure 3A) suggests that, assuming their constitutive expression under our culture conditions, E. coli E1‐like proteins do not participate in reactions observed with the reconstituted ubiquitylation system.
We demonstrated the E2:E3 specificity of the system by showing that Mib auto‐ubiquitylation activity occurred in conjunction with UbcH5b, a cognate E2 (Itoh et al, 2003), but not in conjunction with Cdc34, a non‐cognate E2, known to be involved in cell‐cycle control (Goebl et al, 1988; Figure 4A and B). The functionality of Cdc34 is supported by the formation of free polyUb chains (Gazdoiu et al, 2005). Similar E2 specificity was found for the Rsp5‐dependent ubiquitylation of Rpn10, which was clearly observed with the cognate E2 Ubc4 (Huibregtse et al, 1995), but barely with Cdc34 (Figure 4C and D). This indicates that Rps5 can associate with Cdc34, but with a significant lower efficiency compared with its cognate E2, Ubc4.
To examine E3:substrate specificity, we used the yeast biosynthetic multi‐vesicular body (MVB) cargo, Carboxypeptidase‐S (Cps1p), as the substrate. Cps1p is a type II transmembrane protein in which a short N‐terminal tail is facing the cytosol. Monoubiquitylation of the cytosolic tail was demonstrated to serve as crucial signal for its MVB sorting (Katzmann et al, 2001). Emr and co‐workers identified Rsp5 as the cognate E3 ligase for Cps1p (Katzmann et al, 2004). Recently, Katzmann and co‐workers established that a C‐terminal GST fusion of the Cps1p tail mimics its membrane localization and identified residues that contribute to the specific interaction with Rsp5 (Oestreich et al, 2007; Lee et al, 2009). We found that co‐expressing Ub, Uba1, Ubc4, Rsp5 and the Cps1pTail–GST fusion in the bacterial system was sufficient for the ubiquitylation of Cps1p (Figure 4E). In contrast, GST alone was not ubiquitylated (Figure 4E, lane 2).
PolyUb chains linked through different Ub lysine residues encode distinct cellular signals (Peng et al, 2003; Fushman and Walker, 2010). It seems that E2s determine the lysine specificity of these polyUb chains. For example, Cdc34 or E2‐25K form free polyUb chains conjugated via lysine 48 (K48 polyUb; Gazdoiu et al, 2005), whereas UBC13/UEV complex forms K63 polyUb (Hofmann and Pickart, 1999). Co‐expression of His6–Ub, E1 and Cdc34 specifically yielded K48 polyUb (Figure 4F). We purified the resultant His6‐tagged proteins on Ni2+ beads and subjected them to western blot analysis using anti‐His antibody. Bands corresponding to di‐, tri‐, tetra‐ and penta‐Ub were evident (Figure 4F). Replacing wild‐type Ub with the K48R mutant abolished this ubiquitylation. Similar results were obtained by replacing Cdc34 with E2‐25K (Figure 4G) Moreover, mass spectrometry analysis of trypsin‐digested peptides confirmed that these bands contained Ub. Similarly, co‐expression of UBC13 together with its partner UEV1a, along with Ub and E1, resulted in free chains of K63 polyUb (Figure 4H). The reconstituted system allowed the purification 15–22 mg of K48 or K63 diUb chains from 1 L of E. coli culture. Additional steps of purification (Pickart and Raasi, 2005) resulted in the isolation of apparently homogenous K48‐ and K63‐linked diUb from the polyUb chains as shown by Coomassie blue staining (Figure 4I and J). In conclusion, the reconstituted bacterial ubiquitylation system fully maintains the lysine linkage specificity of all E2 enzymes tested here.
E3‐independent auto‐ubiquitylation of Ub receptors
Many Ub receptors are regulated by auto‐monoubiquitylation, presumably to promote intramolecular binding of the UBD to the attached Ub moiety thus leading to a closed, inactive conformation of the receptor (Shih et al, 2000; Prag et al, 2003; Hoeller et al, 2006; Mosesson et al, 2009). Dikic and co‐workers showed that several Ub receptors, including signal‐transducing adaptor molecule (STAM), undergo auto‐ubiquitylation in an E3‐independent manner (Hoeller et al, 2007). Mouse STAM2 was chosen to test whether the reconstituted ubiquitylation system maintains this E3‐independent auto‐ubiquitylation. GST‐STAM was co‐expressed in E. coli with its cognate E2s (UbcH5b or C). Upon purification on GSH‐affinity resin, the protein was subjected to western blot analysis using anti‐His or anti‐GST antibodies. In agreement with a previous report (Hoeller et al, 2007), STAM underwent auto‐monoubiquitylation in the presence of either UbcH5b or C (Figure 5A).
We next examined auto‐ubiquitylation of an additional Ub receptor. Some Ub receptors, including Sts1/2, Hrs and STAM, undergo E3‐independent auto‐ubiquitylation, whereas others such as Vps9, Rabex5, GGA3, Rpn10 and human Epsin were shown to require E3 ligases (Shih et al, 2003; Timsit et al, 2005; Mattera et al, 2006; Yogosawa et al, 2006; Isasa et al, 2010). The reconstituted system provides a straightforward assay to test the E3 dependency of Ub receptor auto‐ubiquitylation. We addressed this for yeast and D. rerio (zebrafish) Epsin proteins (Eps15 interacting protein). Epsin tethers UBDs, a membrane‐binding domain and a clathrin‐binding motif through flexible hinges. By binding to ubiquitylated transmembrane proteins such as receptor tyrosine kinases, Epsin acts as sensor that binds Ub thereby mediating endocytosis. A His6–MBP fusion of the yeast Epsin, Ent1p was co‐expressed along with pGEN vectors harbouring either yeast Ubc5 or Cdc34 (Supplementary Table SIV). The proteins were enriched on Ni+2‐affinity matrix, resolved on SDS–PAGE and subjected to western blot analysis using anti‐His tag antibody, which recognized both His6–MBP–Ent1 and His6–Ub. We found that yeast Ubc5 was sufficient to promote Ent1p auto‐ubiquitylation (Figure 5B), whereas Cdc34 did not (Figure 5B). Thus, unlike human Epsin, Ent1 underwent auto‐ubiquitylation in an E3‐independent manner. We also found that a yet uncharacterized D. rerio Epsin undergoes auto‐ubiquitylation in an E3‐independent manner (manuscript in preparation). As yeast and zebrafish are evolutionarily distant, E3‐independent ubiquitylation may be a general mechanism common to the Epsin protein family.
Mapping the ubiquitylation sites on Vps9
Vps9 (vacuolar protein sorting) is a Ub receptor that promotes the trafficking of ubiquitylated transmembrane cargo on early endosomes. Vps9 contains a GEF domain and a UBD named CUE (conjugation of Ub to endoplasmic‐reticulum degradation; Figure 6). In Vps9, Ub binding and intramolecular monoubiquitylation are coupled (Shih et al, 2003). To test if Vps9 is ubiquitylated in the bacterial reconstituted ubiquitylation system, we co‐expressed GST–Vps9 with its cognate ubiquitylation proteins (Ub, Uba1, Ubc4/5 and Rsp5). The system yielded large quantities of monoubiquitylated Vps9 (Figure 6A and B). Comprehensive mass spectrometry analysis of purified ubiquitylated Vps9 identified at least five modified lysine residues (K83, K109, K144, K204 and K328), all but K328, within the GEF domain (Figure 6C and D; Supplementary Figure S3 and Supplementary Table SIII in Supplementary data show the fragmentation spectra). The ubiquitylated peptides of Vps9 carried the GG or LRGG tags that characteristically result from trypsin digestion (Peng et al, 2003). We mutated these lysine residues to arginine residues, either individually or all five residues together. Neither of the single mutants showed a significant decrease in the amount of ubiquitylation (Figure 6E). In fact, the quintuple mutant still underwent ubiquitylation, although to a lesser degree, suggesting that other lysine residues become modified in the absence of the major sites we identified by mass spectrometry. A similar observation has been reported for the ubiquitylation of β‐Arrestin2 where only after substitution of all 31 lysine residues ubiquitylation was abolished (Shenoy et al, 2007).
Does ubiquitylation regulate the GEF activity of Vps9?
How auto‐ubiquitylation regulates the activity of Ub receptors remains to be explored. In an attempt to shed light onto this question, we chose Vps9 as a model Ub receptor. As shown above, most ubiquitylation takes place in the Vps9‐GEF domain. We found that a point mutation (M419D) in the CUE domain, which is known to abolish Ub binding and Vps9 ubiquitylation (Prag et al, 2003), elicits higher GEF activity (Figure 7A and B). Interestingly, the human orthologue of Vps9, Rabex5, harbours a coiled‐coil element that confers auto‐inhibitory activity on its GEF domain (Mattera et al, 2006; Delprato and Lambright, 2007). Moreover, interaction of this element with Rabaptin5 reverses the inhibition. To date, no Rabaptin5 orthologue has been identified in yeast. Taking together the higher GEF activity of the M419D mutant and the apparent lack of a Rabaptin5 orthologue in yeast, we investigated the possibility that Vps9–GEF activity is downregulated by auto‐ubiquitylation. Purified ubiquitylated Vps9 enabled us to test this question directly. Using Förster resonance energy transfer (FRET) between the yeast Rab5 GTPase Vps21 and mantGDP as the nucleotide probe (Davies et al, 2005), we monitored the nucleotide exchange rate (GEF activity) of the unmodified apo and the ubiquitylated cis forms of Vps9 (Figure 7A and B and see also model in Figure 8). To our surprise, apo‐Vps9 and cis‐Ub‐Vps9 presented similar levels of GEF activity that were significantly lower than the maximal activity measured with the M419D mutant. These results indicate that ubiquitylation of Vps9 does neither impair nor enhance its GEF activity in vitro, suggesting that another factor, such as a yet undefined Rabaptin5 orthologue, regulates the Vps9‐GEF.
Direct evidence for cis and trans Ub binding by Vps9
Previous work has postulated that the activity of a ubiquitylated Ub receptor to bind ubiquitylated target proteins is inhibited owing to cis interaction of the UBD with the conjugated Ub moiety (Prag et al, 2003; Hurley et al, 2006). The bacterial purification system for ubiquitylated proteins established here allowed us to directly address this proposition for Vps9.
Vps9 is a homodimer that undergoes coupled monoubiquitylation (Prag et al, 2003). In this process, Ub binding to the CUE domains is a prerequisite of ubiquitylation (Shih et al, 2003). In the ubiquitylated Vps9 dimer, only one Vps9 protomer is ubiquitylated, as is evident by the two bands in Figure 6A. Nevertheless, according to the model proposed here, both CUE domains are occupied with the same Ub moiety. To assess this hypothesis, we performed pull‐down and crosslinking assays with Ub and Vps9 (both the apo form and the ubiquitylated Vps9 form). This was made feasible using the bacterial reconstituted ubiquitylation system that produced large amounts of purified ubiquitylated Vps9. Purified apo and ubiquitylated MBP–Vps9 proteins were incubated with Ub‐agarose beads and washed extensively before loading onto SDS–PAGE. The Coomassie blue‐stained gel clearly showed that the unmodified Vps9 (apo) bound to the Ub agarose, whereas the ubiquitylated MBP–Vps9 (cis) did not (Figure 7C). To verify if the cellular environment affects the binding, we performed the same assay but in bacterial cell extracts rather than with purified proteins (Figure 7D). Furthermore, to confirm that the fusion partner does not affect the interactions, we replaced MBP–Vps9 by GST–Vps9. E. coli lysates containing GST–Vps9, ubiquitylated GST–Vps9 or GST were incubated with Ub‐agarose beads and washed extensively followed by separation on SDS–PAGE. We found that the apo‐GST–Vps9 bound to the Ub agarose, but that the ubiquitylated GST–Vps9 did not (Figure 7D). Only a faint band of Vps9 was visible, a finding that is consistent with an excess amount of apo‐Vps9 that is present in the bacteria. To verify that ubiquitylated Vps9 was indeed present in the lysate, we further purified a fraction of the same lysates and showed that similar amounts of proteins were expressed and modified in the tested cultures (see input in Figure 7D).
While Ub agarose may mimic ubiquitylated transmembrane conjugates, we also devised a crosslinking assay with free, soluble Ub (that mimics cytosolic Ub). Moreover, here we used purified apo and ubiquitylated Vps9 proteins whose tags were removed (as shown in the purification scheme of Figure 1). Equal amounts of apo and ubiquitylated cis‐Vps9 proteins were incubated with excess Ub, followed by crosslinking with disuccinimidyl suberate (DSS). Figure 7E shows that apo‐Vps9 bound Ub, whereas the ubiquitylated cis‐Vps9 did not. Interestingly, in the ubiquitylated sample we observed a substantial shift in the stoichiometric apo:cis ratio after crosslinking, reflecting the dynamic nature of the complex ubiquitylated Vps9 dimer. The latter observation corroborates our earlier findings that the CUE domain can bind Ub as a monomer with low affinity and as a dimer with high affinity (Prag et al, 2003). While it appears that some of the ubiquitylated Vps9 undergoes limited degradation in the crosslinking buffer, thus complicating the experiment, the data clearly show that monoubiquitylation of Vps9 inhibits its Ub‐binding function.
E. coli is the established work‐horse for producing large quantities of recombinant proteins. Despite the availability of advanced cloning and purification techniques, relatively few eukaryotic post‐translational modification systems have been implemented in E. coli (Cronan, 1990; Mencia and de Lorenzo, 2004; Baba et al, 2005; Su et al, 2006); and as a result, the utility of this expression system has been limited. Two studies used E. coli expression of HA‐tagged Ub, Uba1, E2 and E3 to detect a specific substrate for a given E3 ligase (Su et al, 2006; Rosenbaum et al, 2011). Here, we present a modular bacterial system not just for identification, but also for the purification and characterization of ubiquitylated proteins. A main advantage of E. coli is the absence of DUBs, which renders the modified product fully stable. As a result, the system yields large amounts of ubiquitylated proteins that can be easily purified. It provides tools for identifying new substrates and for swiftly deciphering the enzymatic requirements and the specific amino‐acid residues subjected to modification.
Due to the close physical coupling of mRNA transcription and translation, bacterial polycistronic systems may facilitate protein–protein interactions of the nascent products. Therefore, the reconstituted ubiquitylation system described here used polycistronic vectors. To allow for different combinations of its various components, we designed the system as sets of two modular plasmids, pGEN and pCOG. Currently, the system includes over 20 plasmids harbouring different E2s and Ub mutants, and >30 vectors harbouring cognate substrates and E3s. The advantage of this system lies in the ease and swiftness of combining and transforming the experimental vectors into bacteria.
Two studies reported a chemical biology approach for Ub modification of H2B and PCNA by mutating target lysine into cysteine residues that enabled disulphide exchange and intein chemistry (Chatterjee et al, 2010; Chen et al, 2010). However, the limitation of this chemical approach is that the modified protein is not in its natural form, since a short disulphide bond is formed rather than an isopeptide bond. Indeed, in the case of H2B, the authors showed that shortening the Ub∼H2B linker reduces the signal decoding by 40% (in terms of the resulting methyltransferase activity) (Chatterjee et al, 2010). Several other chemical biology approaches to generate Ub chains were recently reported (El Oualid et al, 2010; Kumar et al, 2010; Virdee et al, 2010; Fushman et al, 2011). These applications are currently feasible for small well‐folded proteins (such as Ub chains) that could be synthesized in vitro. A genetic incorporation of δ‐thio‐lysine combined with native chemical ligation and followed by desulphurization was demonstrated to yield diUb and Ub‐SUMO conjugates (Virdee et al, 2011). While these chemical approaches produced only free Ub or UbL chains, using the reconstituted system we also obtained conjugated E1, E2, E3 and ubiquitylated substrates. Our system ensures production of the native form of ubiquitylated proteins, as it relies exclusively on an enzymatic cascade.
The reconstitution of Rpn10 monoubiquitylation is one example that showcases this authenticity. Isasa et al (2010) recently showed that Rpn10 is monoubiquitylated in vivo in yeast. In‐vitro ubiquitylation of Rpn10, however, results in multi‐monoubiquitylated or polyubiquitylated substrate, as reported by several groups, including Isasa et al (Lu et al, 2008; Uchiki et al, 2009; Isasa et al, 2010). Why is the bacterial system more successful in recapitulating the eukaryotic process than an in‐vitro assay? We suggest that the high protein concentration in the cytosol of E. coli (∼500 mg/ml; Luria, 1960) better mimics the crowded eukaryotic cell environment than the dilute conditions commonly used for in‐vitro ubiquitylation assays.
While purifying ubiquitylated proteins using the reconstituted system, we could not detect background of endogenous ubiquitylated proteins from E. coli (Supplementary Figure S4). The latter finding demonstrates that the high specificity of the eukaryotic ubiquitylation system is fully maintained upon reconstitution in E. coli.
Many Ub receptors undergo auto‐monoubiquitylation. A parsimony model suggests that this mechanism serves to block the UBD in cis (Figure 8; Hurley et al, 2006). However, many Ub receptors, including the human orthologue of Vps9, Rabex5, harbour multiple UBDs that bind the same patch on the Ub surface. Although only one of the UBDs will be blocked by auto‐monoubiquitylation of Rabex5, the modification is sufficient to promote the dissociation of Rabex5 from early endosome membranes (Figure 8). Vps9 differs from its mammalian orthologue in several important aspects. A conserved GEF domain present in both Rabex5 and Vps9 acts on the Rab5 protein family. However, in Vps9 a C‐terminal CUE domain functions as a single UBD, while in Rabex5 two different UBDs (A20‐like ZnF and IUIM) are located at the N‐terminus (Figure 8; Prag et al, 2003; Lee et al, 2006; Penengo et al, 2006). Whereas in yeast the CUE domain is responsible both for binding ubiquitylated cargo and for auto‐ubiquitylation, in humans these functions are divided between the ZnF and IUIM domains (Lee et al, 2006). The combined results for both the human and the yeast GEF‐Ub receptors, Rabex5 and Vps9 suggest a mechanism of convergent evolution. The principle that ubiquitylation acts to sequester membrane‐associated proteins from their targets appears to be conserved in evolutionarily unrelated Ub receptors (Mattera and Bonifacino, 2008; Mosesson et al, 2009).
Previous studies have provided indirect evidence that ubiquitylated Ub receptors form a closed structure (cis) (Hoeller et al, 2006; Mattera and Bonifacino, 2008). The reconstituted system has enabled us to demonstrate that auto‐monoubiquitylation of Vps9 reduces its ability to bind Ub in trans. The fact that human Rabex5 and yeast Vsp9 developed virtually the same mechanism to regulate GEF localization, notwithstanding their unrelated UBDs, suggests a general paradigm by which the cell interprets the Ub signal. Given the huge spectrum of ubiquitylation signalling, it will be interesting to explore the possibility that different mechanisms decoded by Ub receptors (other than shown here for Vps9) govern the association of proteins with membranes.
In conclusion, we have reconstituted in bacteria a pathway that specifically and accurately generates ubiquitylated eukaryotic proteins. The availability of the bacterial system, its ease of use, and its large yields should greatly facilitate future crystallographic, biophysical and biochemical analyses of ubiquitylated proteins.
Materials and methods
cDNAs were PCR amplified, digested and subcloned into the parallel system vectors (Sheffield et al, 1999). To create the pGEN plasmid, the β‐lactamase resistance gene in the pHIS‐parallel2 vector was replaced with a kanamycin resistance cassette. Then, the cDNA of human Ub was subcloned into the BamHI and EcoRI sites to create His6–Ub. The cDNA of wheat Uba1 was subcloned between EcoRI and BstBI sites. Finally, the cDNAs of the different E2s were subcloned into the EcoRI and AscI endonuclease recognition sites (the AscI site was inserted by PCR). Shine‐Dalgarno sequences derived from the Lac operon of the E. coli strain K12/W3110 (accession #AP009048) were for the E2 5′‐AGGAAATCCATTATG‐3′ and for E1 5′‐AGGAAACAGCTATG‐3. Ub was cloned downstream to the Shine‐Dalgarno sequence that already existed in the pHis‐parallel vector. These sequences were added to the forward primer of E2 and E1 to form a polycistronic expression vector.
The His–Ub‐E2/Ubc5‐E1 segment was PCR amplified, digested and subcloned into the pCDF‐duet vector (Novagen) between AvrII sites.
To create the pCOG plasmid, the ORF of yeast Vps9 encoding amino acids 55–451 was PCR amplified, digested and subcloned into a pMBP‐Rsp5 vector (Davies et al, 2003) between SpeI and SphI sites. The same was done for yeast Rpn10. Mind bomb and mSTAM were subcloned into the pGST‐parallel2 vector between BamHI and EcoRI endonuclease recognition sites. The ORF of MBP was inserted into the pCDF‐duet vector between EcoRI and AscI to create pCDF‐His6–MBP. Ent1p and Vps9 were subcloned into the pCDF‐duet vector as His6–MBP fusions. UEV1A was subcloned into the pHis‐parallel2 vector, UBC13 was subcloned downstream of the UEV1A ORF with a preceding Shine‐Dalgarno sequence.
Point mutations were introduced using the ExSite or QuickChange (Stratagene) protocols. All constructs were verified by DNA sequencing to ensure introduction of the desired mutations and that no other mutations were introduced.
Protein expression and purification
Rosetta2(λDE3)BL21 E. coli cells were co‐transformed with pGEN and pCOG vectors. Cultures were grown in Terrific Broth medium at 37°C. Each culture was induced by addition of 0.5 mM isopropyl β‐d‐1‐thiogalactopyranoside (IPTG) and was further grown at 16°C for 16–20 h. Cells were harvested by centrifugation and were resuspended in buffer with lysozyme and AEBSF (4‐(2‐Aminoethyl) benzenesulphonyl fluoride hydrochloride). Complete lysis and DNA shearing were achieved by sonication following centrifugation to isolate the soluble fraction. Proteins were affinity purified using reduced glutathione, nickel, (GE‐Amersham Biosciences) or amylose‐affinity matrixes (New England Biolabs) according to the manufacturer's instructions. Affinity tags were removed with His6‐tagged TEV proteases and with GST‐tagged rhinovirus proteases. Size‐exclusion chromatography was used for further purification of the affinity‐purified proteins.
Samples were transferred onto a nitrocellulose membrane and incubated with rabbit anti‐His6 epitope tag antibody (1:20 000 dilution, Rockland), mouse anti‐Ub antibody, or mouse anti‐GST antibody (1:200, Santa Cruz), and infrared dye coupled goat anti‐mouse secondary antibody (1:12 000, LI‐COR). Scanning was performed with the Odyssey infrared imaging system (LI‐COR Biosciences) in accordance with the manufacturer's instructions at 700 and 800 nm.
In all, 100 ml of E. coli BL21(lDE3) cultures expressing GST, GST–Vps9(55−451) and GST–Vps9(55−451) with E1, Ubc4 Rps5 and His6–Ub was harvested and prepared described. In all, 10% of the lysates were used for binding assays and the rest were further purified. For binding assays, equal protein concentrations were used diluted with binding buffer containing 25 mM Hepes‐Na pH 7.4, 75 mM NaCl and 0.1% Triton X‐100 to a final volume of 0.5 ml. In all, 20 μl of activated Ub‐agarose beads (Sigma) was incubated with the lysates for 2 h at 4°C. Unbound fraction was removed by centrifugation and beads were washed three times with 1 ml of binding buffer. The bound and unbound fractions were analysed by SDS–PAGE. The remaining lysate fractions were batch purified using nickel beads followed by GSH beads (for Ub‐Vps9) or GSH beads only (for apo‐Vps9 or GST tag). Equal amounts were analysed by SDS–PAGE. For the pull down with purified proteins, purified MBP, MBP–Vps9 and ubiquitylated MBP–Vps9 were used. Buffer was exchanged with AmiconUltra Centricon (Millipore) to binding buffer (25 mM Hepes‐Na pH 7.4, 75 mM NaCl, 1 mM DTT). Samples were concentrated to ∼15 mg/ml and 20 μl was used with 80 μl of binding buffer supplemented with 0.1% BSA and 0.25% Triton X‐100. The mixture was incubated with 30 μl Ub agarose, washed and analysed by SDS–PAGE.
Crosslinking assays with purified proteins were carried out according to the method described by Azem et al (1998). Reactions contained 0.5 mM DSS in 25 mM Hepes buffer pH 7.4, 150 mM NaCl and 10 μM of the indicated proteins. The Vps9 to Ub or ubiquitylated Vps9 to Ub ratios were 1:1. The reaction was stopped after 10 min by boiling the solution in SDS‐loading buffer and samples were analysed by SDS–PAGE.
Mapping ubiquitylation sites by RP/LTQ‐OrbiTrap mass spectrometry
Mass spectroscopy protocol is described in detail in Supplementary data. Briefly, gel bands containing ubiquitylated GST–Vps9, GST–Mib or MBP–Rpn10 were excised and subjected to reduction and alkylation. Samples were digested with either trypsin or Asp‐N (and in the case of Mib also with GluC). Protease‐digested samples were extracted, concentrated and desalted. Eluates were vacuum dried and redissolved in LC/MS loading buffer (2% acetonitrile in 0.1% formic acid in water). ADVANCE low‐flow electrospray ionization source was used to load the digested samples onto a Magic C18 AQ reversed‐phase column (Michrom Bioresources Inc.) with a 2‐h LC gradient coupled to the OrbiTrap mass spectrometer. To avoid the iodoacetamide artifact that may mimic ubiquitylation, cysteine residues were alkylated at low temperature and concentration (40°C, 15 mM iodoacetamide) for 40 min (Nielsen et al, 2008). Search results were viewed, sorted, filtered and analysed using comprehensive proteomics data analysis softwares.
GEF activity analysis
GEF activity analysis was carried out according to a published protocol (Davies et al, 2005). Briefly, wild‐type proteins and the D25A1/E288A and M419D mutants were expressed as His6–MBP fusions. Proteins were purified by Ni+2 chromatography, cleaved with TEV protease and resubjected to Ni+2 chromatography, and then fractionated by size‐exclusion chromatography on a Superdex 75 column. Vps21 was purified by a similar procedure. Vps21 was then loaded with mantGDP, and excess nucleotide was removed via size‐exclusion chromatography. Vps21 release assays containing 2.5 μM Vps21•mantGDP and 100 nM Vps9 (wild‐type, mutant or ubiquitylated Vps9) in release assay buffer (20 mM Hepes pH 7.5, 5 mM MgSO4) were initiated by addition of GTP to 1 mM. FRET (ex. 290 nm, em. 440 nm) was measured continuously in a Photon Technology International Quanta Master 2001 fluorometer with 1 per second time points for 15 min. Data were graphed in Prism (GraphPad Software, Inc.) and rates of release determined by linear regression.
Supplementary data are available at The EMBO Journal Online (http://www.embojournal.org).
Conflict of Interest
The authors declare that they have no conflict of interest.
We are very thankful to A Weissman and A Chitnis for the fruitful discussions and providing us with cDNAs for UbcH5b, c and zebrafish Mib and Epsin. We thank R Vierstra, I Dikic, D Katzmann and S Misra for providing us plasmid cDNA of Uba1, STAM, Cps1, UBC13 and UEV1a, respectively. We gratefully acknowledge B Zion for technical assistance. We are thankful to A Azem for the kind help with the crosslinking assay. This research was supported by grants from the ISRAELI SCIENCE FOUNDATION (grant no. 1695/08); from the EC FP7 Marie Curie International Reintegration Grant (PIRG03‐GA‐2008‐231079), from the Israeli Ministry of Absorption and from the Israeli Ministry of Health (5108) to GP and by NIH grant GM59780 to DAW.
Author contributions: TKK designed, performed, analysed and implemented the ubiquitylation system, including the Rpn10, Rsp5 and Vps9 functionality experiments and participated in writing the manuscript. IA performed and analysed the Mib and Cps1 experiments. BD and BH designed, performed and carried out data analysis for the Vps9‐GEF activity. NT performed the studies on yeast Epsin. YR performed the studies with Cdc34 and zebrafish Epsin. OLK and EL carried out the studies on STAM. OK and MG performed the mass spectrometry analysis of Rpn10. MK subcloned the Ubc13/UEV. KM and DAW designed, performed and analysed the mass spectrometry of Mib and Vps9; GP conceived the idea of the reconstituted ubiquitylation system, designed and planned experiments, analysed data and wrote the manuscript.
- Copyright © 2012 European Molecular Biology Organization