Site-Specific Protein Ubiquitylation Using an Engineered, Chimeric E1 Activating Enzyme and E2 SUMO Conjugating Enzyme Ubc9.

Ubiquitylation-the attachment of ubiquitin (Ub) to proteins in eukaryotic cells-involves a vast number of enzymes from three different classes, resulting in heterogeneous attachment sites and ubiquitin chains. Recently, we introduced lysine acylation using conjugating enzymes (LACE) in which ubiquitin or peptide thioester is site-specifically transferred to a short peptide tag by the SUMO E2 conjugating enzyme Ubc9. This process, however, suffers from slow kinetics-due to a rate-limiting thioester loading step-and the requirement for thioesters restricts its use to in vitro reactions. To overcome these challenges, we devised a chimeric E1 containing the Ub fold domain of the SUMO E1 and the remaining domains of the Ub E1, which activates and loads native Ub onto Ubc9 and obviates the need for Ub thioester in LACE. The chimeric E1 was subjected to directed evolution to improve its apparent second-order rate constant (k cat/K M) 400-fold. We demonstrate the utility of the chimeric E1 by site-specific transfer of mono- and oligo-Ub to various target proteins in vitro. Additionally, the chimeric E1, Ubc9, Ub, and the target protein can be coexpressed in Escherichia coli for the facile preparation of monoubiquitylated proteins.

Introduction

Ubiquitylation—the attachment of ubiquitin (Ub) to proteins—regulates a vast number of functions in eukaryotic cells, such as localization, protein degradation, and DNA repair.[1−3] Ubiquitin is a small, stably folded protein with a molecular weight of 8.5 kDa. Typically, the C-terminal carboxylic acid of Ub is activated and attached to lysine residues of substrate proteins, or to one of the seven lysine residues or the N terminus of Ub itself to form polyUb chains. Ubiquitylated proteins are recognized by downstream pathways where various chain isoforms are associated with different functions including proteosomal degradation and recognition by ubiquitin-binding proteins to form larger protein assemblies. The study of these biochemical processes requires access to homogeneously ubiquitylated proteins, but the complex topologies and enzymes involved in ubiquitylation make their preparation a formidable challenge. To date, site-selective ubiquitylation of proteins has been achieved by total synthesis or semisynthesis,[4−7] sortylation,[8,9] or semisynthetic methods generating lysine surrogates.[10−17] However, the reliance of these methods on the incorporation of unnatural amino acids or unnatural Ub linkages renders them a specialized technique not readily accessible to a broad community of researchers. On the other hand, employing natural enzymes of the Ub cascade for the site-selective installation of Ub is rather challenging. A recently reported prudent strategy utilizes canonical E1, E2, and E3 enzymes and a deubiquitylating enzyme (DUB) with an E3-binding peptide tag to monoubiquitylate single-lysine variants of a protein.[18]

The canonical Ub cascades rely on a relay of different classes of enzymes: 2 homologues of E1 activating enzymes, ∼40 homologues of E2 conjugating enzymes, and 600–1000 different E3 ligases. Other ubiquitin-like proteins (Ubls), such as SUMO and NEDD8, utilize homologues of the same class of enzymes.[19] Ub and Ubl pathways are mostly orthogonal, in that each E1 and E2 enzyme only serves to conjugate its cognate Ubl. For example, an E1 activating enzyme adenylates the C-terminal carboxylic acid of the Ubl and captures it with an active-site cysteine residue, giving rise to a Ubl–E1 thioester. A Ubl-specific E2 attacks this complex to afford an E2–Ubl thioester. Some E2s are capable of transferring their Ubl cargo to protein substrates directly, but most require E3 ligases that either provide a scaffold to the E2–Ubl thioester or themselves form an E3–Ubl thioester. In this manner, substrate/sequence specificity is determined by the E2 in some cases and the E3 in most cases. Orthogonality between different Ubl cascades is governed largely by the specific activation of their cognate Ubl by the E1 and subsequent formation of the correct E1–E2 as well as E2–E3 interactions.

Ubc9 is the only known E2 of the small ubiquitin-like modifier (SUMO) pathway. This E2 is notable among Ubl E2s for having a consensus sequence for the recognition of protein substrates, namely, the SUMOylation motif ΨKxD/E, where Ψ is a hydrophobic residue, and x is any residue. In addition, Ubc9 is active in the absence of E3s, which makes it an appealing enzyme to exploit for a general protein modification method. Recently, our group developed lysine acylation using conjugating enzymes (LACE, Figure a)[20] using Ubc9 and Ub-derived thioesters. LACE can transfer Ub–Mes thioester onto the lysine residue in a LACE tag—itself derived from the SUMOylation motif—enabling direct, site-specific Ub transfer onto both natural and non-natural SUMOylation substrates. Although quite selective, the reaction is somewhat sluggish due to the rate-limiting Ub loading step. Furthermore, the Ub–Mes and Ub–Ubc9 thioesters are prone to hydrolysis or nonspecific reactions, rendering the process inefficient.

Figure 1

Strategies for site-selective ubiquitylation using SUMO conjugating enzyme Ubc9. Ubc9–Ub thioester transfers Ub to the lysine residue in the LACE tag sequence, IKxE (x is any amino acid). (a) Lysine acylation using conjugating enzymes (LACE) using Ub–Mes thioester as an acyl donor. (b) Native Ub monomers or oligomers activated by an engineered E1 as an acyl donor.

To overcome these challenges, we sought to identify an E1 activating enzyme capable of loading Ub onto Ubc9, thereby circumventing the need for the prior formation of Ub thioesters and enabling in cellulo LACE reactions. In this report, we describe the development of an engineered and evolved chimeric E1 activating enzyme that activates the Ub C-terminal carboxylic acid and loads it onto Ubc9 (Figure b). This allows for site-specific ubiquitylation using readily available Ub monomers and Ub oligomers. All of the reaction components can be easily produced in Escherichia coli (E. coli), which also allows a facile recombinant expression of monoubiquitylated proteins.

Results

Design and Directed Evolution of Chimeric E1

E1s have two interaction surfaces for E2s: the second catalytic cysteine half (SCCH) domain and the ubiquitin fold domain (UFD) (Figure a). Of the two, UFDs are known to play an important role in discriminating the E1s’ cognate and noncognate E2s. We reasoned that a chimeric E1 possessing the UFD of SUMO activating enzyme Sae1/Sae2 and the remaining domains from Ub activating enzyme Uba1 could activate the Ub C-terminal carboxylic acid with its adenylation and SCCH domains. It could then load the Ub onto Ubc9 by a favorable UFD–Ubc9 interaction, in a way analogous to that previously demonstrated for the Nedd8 E1 (Figure a,b).[21] The UFD is connected to the rest of the domain by a long linker in most Ubl E1s, and the linker in Sae2 has a zinc-binding motif that locks the UFD in the vicinity of the adenylation domain. On the other hand, Uba1 is devoid of this zinc-binding motif, and the linker is predicted to be rather flexible.[22] Since a favorable linker architecture for the chimeric E1 was unknown, we prepared chimeric E1 variants by modulating the linkage position on the linker and tested whether they activated and loaded Ub onto Ubc9 (Figure c). In an E2 thioester formation assay, Ub was activated by all E1 variants but was loaded onto Ubc9 only by the variants having Sae2 UFD, suggesting that Ubc9 needs to be positioned for the Ub loading with the aid of the Sae2 UFD (Figure d). Chimeric E1 (ChE1) v0.1, which has a part of the linker region that contributes to the zinc-binding motif in Sae2, was less reactive than ChE1 v0.2, which had a linker entirely from Uba1.

Figure 2

Design and directed evolution of chimeric E1. (a) Structure of the spUba1–spUbc4 complex (PDB ID 4ii2)[23] and Sae1/Sae2 heterodimer (PDB ID 1y8r).[24] The E1s are shown in a surface model, and Ubc4 is shown in a ribbon model. Catalytic cysteine residues are highlighted in yellow. (b) Design of the chimeric E1, in which the UFD of Uba1 is replaced by that of Sae2. The interaction between SCCH and Ubc9 was further enhanced by directed evolution. The protein representation is based on Uba1 (PDB ID 6dc6).[25] (c) Domain architecture of Uba1 ΔUFD, chimeric E1 (ChE) v0.1 and v0.2, in comparison to Sae1/Sae2 and Uba1. For detailed linking positions, see the Supporting Information. (d) E2 thioester formation assay using the E1 variants in part c. The reactions contained 0.2 μM E1, 5 μM Ubc9 Y134A,[23] 5 μM biotin–Ub (Ubbt), 1 mM DTT, 5 mM ATP-Mg, 50 mM HEPES pH 7.5, 50 mM KCl and were incubated at 37 °C for 1 h. The reactions were resolved by SDS-PAGE and visualized by Western blot using streptavidin-HRP. (e) β-Lactamase exporting assay of representative chimeric E1 variants from each round of the directed evolution. (f) Locations of the 7 mutations in ChE1 v4.5. The protein representation is based on Uba1 (PDB ID 6dc6). (g) E2 thioester formation assay using Uba1, ChE1 v0.2 and v4.5, and reversion mutants of ChE1 v4.5. Conversion to Ubc9–Ub thioester was quantified by gel densitometry using Stain-Free imaging (BioRad). The reactions contained 0.5 μM E1, 3 μM Ubc9 K14R, 30 μM Ub, 0.5 U/mL inorganic pyrophosphatase, 0.5 mM DTT, 5 mM ATP-Mg, 50 mM HEPES pH 7.5, and 50 mM KCl and were incubated at 37 °C for 4 min. We could not test S748A since the mutant could not be expressed in E. coli. Data are presented as average values ± s.d. n = 3 independent experiments; individual data points are shown.

With the improved interaction between the chimeric E1 UFD and Ubc9 in hand, we wished to enhance the other important interaction, that is, the chimeric E1 SCCH–Ubc9 interaction. To this end, we sought to employ directed evolution of ChE1 v0.2 using a β-lactamase exporting assay inspired by a system reported by Guntas and Kuhlman.[26] In this assay, a plasmid coding for the components of the ubiquitylation reaction is constitutively expressed in E. coli (Figure S1). A periplasm exporting signal peptide is conjugated to the N terminus of Ub, and Ubc9 is fused to the N terminus of β-lactamase. An expression of these two proteins, as well as a chimeric E1 in E. coli, results in Ub loading onto Ubc9, whose complex is then transported to the periplasm, giving the cell ampicillin/carbenicillin resistance.

In a preliminary assay, one of the clones coding for ChE1 v0.2 showed a weak carbenicillin resistance in a manner dependent on the Ubc9 active-site cysteine (Figure e). Next, we randomly mutated the SCCH domain of the chimeric E1 by error-prone PCR and selected for clones showing higher ampicillin/carbenicillin resistance. Three rounds of directed evolution enriched mutations on the E1–E2 interaction surface (Table S1). Of the mutations observed, we were particularly interested in S579N, N688Y, and F689I because of their early prevalence. We therefore created a library for the fourth round of evolution by a combination of saturation mutagenesis at these three sites and gene shuffling of the third-round variants. Mutants from this round revealed that S579 tolerates diverse residues, whereas N688 and F689 are preferably mutated to tyrosine and isoleucine, respectively. From the round 4 variants, we chose an isoform 4.5 containing 7 mutations as the final variant (Figure f). This enzyme showed a significant improvement in carbenicillin resistance in the β-lactamase exporting assay compared to the original chimeric E1, ChE1 v0.2 (Figure e).

Steady-state kinetic parameters of ChE1 v4.5 and v0.2 and Uba1 for E2–Ub thioester formation were determined at pH 7.5 and 37 °C (Table , Figure S3). For these studies, we employed Ubc9 K14R mutants, which suppressed the automodification of Ubc9.[27,28] The optimized E1 loads Ub to Ubc9 K14R with a turnover number (kcat) of 0.15 ± 0.02 s–1 and a Michaelis constant (KM) of 0.42 ± 0.12 μM. A comparison between ChE1 v4.5 and v0.2 shows that the directed evolution increased kcat 75-fold and decreased KM 6-fold—thereby increasing the apparent second-order rate constant (kcat/KM) 400-fold. Although kcat of ChE1 v4.5 is 10–30 times lower than that of Uba1 in conjunction with its canonical E2s Ube2K and Ube2G1, its apparent second-order rate constant of kcat/KM, 3.6 ± 0.8 × 105 M–1 s–1, at low E2 concentration is comparable to those of canonical Ub E1–E2 pairs.

Table 1

Kinetic Constants for E2–Ub Thioester Formationa

E1	E2	kcat (s–1)	KM (μM)	kcat/KM (M–1 s–1)
ChE1 v4.5	Ubc9 K14R	0.15 ± 0.02	0.42 ± 0.12	3.6 ± 0.8 × 105
ChE1 v0.2	Ubc9 K14R	2.0 ± 0.1 × 10–3	2.4 ± 0.5	8.8 ± 2.2 × 102
Uba1	Ube2K	1.4 ± 0.2	9.6 ± 4.9	1.6 ± 0.5 × 105
Uba1	Ube2G1	4.7 ± 1.2	3.9 ± 0.6	1.2 ± 0.2 × 106

See Figure S3 for Michaelis–Menten plots. Each value represents mean ± s.d. of 3 independent experiments.

To examine the contribution of the individual mutations to the activity of ChE1 v4.5, we prepared reversion mutants in which one of the mutations is reverted to the original residue and then assayed by E2 thioester formation (Figure g). The mutations at the putative Ubc9-interacting surface, A640T, N688Y, and F689I were found to be indispensable for the activity of ChE1 v4.5, while the other mutations distant from the interaction surface had no contribution or a negative contribution to the activity. Particularly, the reversion of either N688Y or F689I drastically decreased the activity, supporting the results of saturation mutagenesis. Interestingly, it has been suggested by a computational model of Schizosaccharomyces pombe Uba1–Ubc9 complex that spUba1 F689, which corresponds to hUba1 F689, might have a steric clash with Ubc9, contributing to the orthogonality between the Ub pathway and the SUMO pathway.[23] This prediction has never been empirically supported, as such a complex would not be possible without a favorable UFD–Ubc9 interaction realized in our chimeric E1 variants.

In Vitro Site-Selective Ubiquitylation Mediated by Chimeric E1 and Ubc9

The selected chimeric E1 v4.5 was subjected to an in vitro monoubiquitylation assay using a model substrate GFP-I, a superfolder green fluorescent protein (GFP) variant having an internal LACE tag (IKQE) (Figure a). We were pleased to find that 15 μM GFP-I was fully converted (>95% by gel densitometry) when treated with 0.5 μM ChE1 v4.5, 30 μM Ubc9 K14R, and 75 μM Ub at pH 7.5 and 37 °C for 2 h (Figure b). The reaction was dependent on the lysine residue in the LACE tag, and both the chimeric E1 and Ubc9 were required. An apparent first-order reaction half-life was 15 min, which was significantly shorter than reactions with chimeric E1 v0.2 under the same condition (Figure c).

Figure 3

In vitro ubiquitylation. (a) Monoubiquitylation of GFP-I mediated by chimeric E1 and Ubc9. The LACE tag sequence is shown, in which the hydrophobic residue Ψ (bold), acceptor lysine K (blue), and acidic residue E (red) are highlighted. (b) Coomassie-stained SDS-PAGE of the reaction time-course and control reactions under the standard conditions using the acceptor lysine mutant K6R and omitting ChE1 v4.5 or Ubc9. (c) Quantification of the reaction conversion over time by gel densitometry using ChE1 v4.5 and v0.2 and Uba1. The apparent reaction half-time (t1/2) for ChE1 v4.5 is given. n = 2 independent experiments; individual data points are shown. (d) PolyUb transfer mediated by chimeric E1 and Ubc9. Linkage-defined Ub oligomers are prepared by using canonical E1, E2s, and DUBs (K11, K48, and K63) or by bacterial expression (M1) as previously reported. Transfer of Ub dimer (e) or trimer (f) to GFP-I under the standard conditions. (g) In-gel GFP fluorescence of transfer of K63 Ub hexamer to GFP-I under modified conditions (7.5 μM GFP-I, 0.5 μM ChE1 v4.5, 15 μM His6-Ubc9 K14R, 10 μM K63 Ub hexamer). Samples were not boiled to prevent smeared bands of polyUb, resulting in multiple bands of GFP and the different mobility of GFP variants. For a corresponding Coomassie-stained gel, see Figure S4. (h) Transfer of K63 Ub dimer to SUMO2 dimer.

We next investigated the suitability of these conditions for ubiquitin chain transfer. In Nature, polyUb chains are built by oligomerization onto a single Ub that is already attached to a substrate. Chain formation in vivo is inherently heterogeneous, controlled by a variety of E2, E3, and DUBs. However, specific chain lengths are expected to have different affinities to downstream proteins,[13] and it is unclear if there are unique physiological roles for respective Ub chains. Such a question has been partly addressed by making Ub chains of defined length by enzymatic stepwise chain elongation using proximally and distally blocked Ub monomers and deblocking thereof or chemical bioconjugation using synthetic Ub monomers.[10] A recent report employs orthogonal sortase variants and C-terminally mutated Ub monomers.[9] We hypothesized that the chimeric E1 could activate and load Ub oligomers to Ubc9, which would subsequently transfer the defined Ub chains onto a LACE-tag on a substrate protein (Figure d). This approach would make use of natural linkage-defined Ub chains, which are commercially available or can be readily prepared and purified to homogeneity using canonical linkage-specific E2s, E3s, and DUBs.[29] To our delight, all K11, K48, and K63 dimers and trimers—which are most abundant in Nature—as well as linear M1 chains prepared by reported methods[30,31] were efficiently transferred onto the LACE tag in GFP-I under the standard conditions (Figure e,f). Commercially available K63 hexamer was also transferred to GFP-I under a modified condition (Figure g and Figure S4), although the reaction was slower due to the lower concentration of the proteins.

We further tested the transfer of Ub chains to a SUMO dimer. It is known that SUMO2 forms polymers at K11 and that Ub chains can be attached to the end of this SUMO2 chain to give Ub–SUMO hybrid chains.[32] SUMO2 monomer, having a mutation of V10I introduced to generate an optimal LACE tag,[20] was allowed to react in the presence of Sae1/Sae2 and Ubc9 to afford a SUMO2 dimer (diSUMO2) conjugated at K11.[33] K63 Ub dimer was transferred to diSUMO2 under the standard conditions (Figure h). The SUMO dimer remained unreacted in the absence of Ub dimer, indicating that the chimeric E1 activates only the Ub dimer and not the SUMO dimer.

Monoubiquitylation in E. coli

The development of a chimeric E1 makes it possible to produce all of the components and perform site-specific ubiquitin transfer in cells. Inspired by reports on ubiquitylation and SUMOylation systems in E. coli,[35−40] we incorporated His6-tagged Ub, Ubc9, ChE1 v4.5, and an MBP-tagged substrate into two plasmids (Figure a). The proteins are produced in E. coli, and ubiquitylation reactions take place in the cells (Figure S5a). After His6- and MBP-tag purification, followed by tag cleavage with TEV protease, the monoubiquitylated protein is obtained. This strategy obviates the need for purifying each protein recombinantly and enables straightforward purification of monoubiquitylated protein. In the case of a model substrate GFP-I, we obtained 0.5 mg/L culture of monoubiquitylated GFP-I after purification (Figure b and Figure S5b). The K6R variant of GFP-I was not monoubiquitylated in an appreciable level, and instead, the ubiquitylation of MBP was predominantly observed (Figure S5b). As MBP is known to contain SUMOylation sites, it can be ubiquitylated by Ubc9 in our system.[37]

Figure 4

Monoubiquitylation in E. coli. (a) Constructs coding for MBP-substrate (pSub), His6-Ub, Ubc9, and chimeric E1 v4.5 (pUb). Monoubiquitylation and purification of GFP-I (b) and of HRas (c). Deconvoluted ESI-MS of the products (calc., calculated; obs., observed) and Coomassie-stained SDS-PAGE are shown. HRas–Ub is partly glycosylated (+162 Da). Protein bands marked with an asterisk are TEV protease. The predicted structure of HRas was obtained from the AlphaFold Protein Structure Database.[34] A detailed purification scheme and the complete SDS-PAGE and anti-Ub Western blot images, including GFP-I tag K6R and HRas K170R mutants, are available in Figure S5.

As a second example, we selected human HRas as a naturally monoubiquitylated substrate. HRas is monoubiquitylated at K117, K147, and K170 in cells to influence its subcellular localization and activity.[41−44] We chose K170, whose ubiquitylation is regulated by leucine zipper-like transcriptional regulator 1 (LZTR1). Dysregulation of the LZTR1-mediated ubiquitylation leads to human disease or drug resistance.[43,44] To the substrate plasmid pSub, we subcloned the HRas gene with the mutations of R169I and N172E to create an optimal LACE site at K170 and introduced an Asp8 tag to the N terminus of MBP to facilitate the final purification by ion-exchange chromatography. Monoubiquitylated HRas (2 mg/L culture) was obtained after purification with a small amount of doubly ubiquitylated product (Figure c and Figure S5c). The K170R variant was not a competent substrate, which is consistent with the conjecture that monoubiquitylation by Ubc9 occurs at K170 (Figure S5c).

Discussion

Developing a general approach to anchor ubiquitin to substrate proteins at a specific site with a defined chain length contributes to the interrogation of the complex ubiquitin code. Here, we overcome the requirement of LACE for Ub–thioester and the limitations attributable to it in our previous report by developing a chimeric E1 that allows Ub with a free C terminus to serve as the acyl donor. This was made possible by swapping the E2-discriminating UFDs of the Ub and SUMO E1s, retaining Uba1’s ability to activate free Ub. The chimeric E1 was further evolved to give better catalytic efficiency using a β-lactamase exporting assay. This strategy of employing Ubc9 for site-selective Ub transfer and chimeric E1 for activation and loading of Ub onto Ubc9 should be applicable to other Ubls, as long as Ubc9 can be loaded with the Ubl.[20]

Obviating the need to isolate the semisynthetic Ub thioester allows readily available Ub oligomers with a free C terminus to be conveniently employed for the Ub transfer mediated by the chimeric E1 and Ubc9. We showed that the reactions with K11, K48, and K63 oligomers—the most abundant Ub chains in cells—and M1 all worked well, but we expect that other minor chains or heterotypic unanchored Ub-chains would also be suitable for this method, although some loss of reactivity due to steric clash may be anticipated. K11 and K48 chains were transferred efficiently in this system with lower reaction rates, although both K11 and K48 are located at the E1–Ub interface.[45]

With all of the components available in cells by protein expression, monoubiquitylated proteins could be obtained from a reconstituted cascade in E. coli. The model substrate GFP-I as well as HRas were site-selectively monoubiquitylated and purified to homogeneity.

The present method requires a minimal (up to 3 residues) alteration of sequence in a flexible region on a target protein to dictate the specific lysine to be ubiquitylated. In this sense, it complements the method of ubiquitylation of proteins bearing a Lys(ε-Gly–Gly) residue recently reported by Lang and co-workers, which introduces two mutations at the C terminus of Ub to convert it to a sortase-compatible sequence.[8,9]

In conclusion, we have engineered and evolved a chimeric E1 that activates and loads Ub onto Ubc9 at a rate approaching that of canonical Ub E1–E2 pairs. We demonstrated that this E1-mediated LACE approach realized the efficient transfer of not only Ub monomers but also Ub oligomers to substrates with a LACE tag in a site-selective manner. The Ub transfer can be easily performed in vitro or in E. coli, enabling the convenient preparation of homogeneously ubiquitylated proteins.