OTUB1 co-opts Lys48-linked ubiquitin recognition to suppress E2 enzyme function

SUMMARY

Ubiquitylation entails the concerted action of E1, E2 and E3 enzymes. We recently reported that OTUB1, a deubiquitylase, inhibits the DNA damage response independently of its isopeptidase activity. OTUB1 does so by blocking ubiquitin transfer by UBC13, the cognate E2 enzyme for RNF168. OTUB1 also inhibits E2s of the UBE2D and UBE2E families. Here we elucidate the structural mechanism by which OTUB1 binds E2s to inhibit ubiquitin transfer. OTUB1 recognizes ubiquitin-charged E2s through contacts with both donor ubiquitin and the E2 enzyme. Surprisingly, free ubiquitin associates with the canonical distal ubiquitin-binding site on OTUB1 to promote formation of the inhibited E2 complex. Lys48 of donor ubiquitin lies near the OTUB1 catalytic site and the C-terminus of free ubiquitin, a configuration that mimics the products of Lys48-linked ubiquitin chain cleavage. OTUB1 therefore co-opts Lys48-linked ubiquitin chain recognition to suppress ubiquitin conjugation and the DNA damage response.

INTRODUCTION

Conjugation of ubiquitin (Ub) onto substrates regulates the abundance, localization and activity of a large fraction of the proteome. Ubiquitylation is a multi-enzymatic process that first necessitates the activation of the terminal carboxyl group of Ub by an E1 enzyme (Pickart, 2001). Activated Ub is transferred to an E2 conjugating enzyme to form a high-energy thioester intermediate, denoted E2~Ub. E2~Ub is bound by an E3 Ub ligase to catalyze the formation of an isopeptide bond between an amino group, usually the ε-NH₂ of lysine, and the C-terminus of Ub. Ubiquitin possesses eight potential acceptor sites and therefore Ub conjugation can be repeated to form chains. Remarkably, the nature of the ubiquitylation products often dictates biological outcome. For example, mono-ubiquitylation of surface receptors is a signal for endosomal sorting (Raiborg and Stenmark, 2009), Lys48 (K48)-linked Ub chains target proteins for degradation by the 26S proteasome and Lys63 (K63)-linked Ub chains are non-proteolytic signals that nucleate protein-protein interactions (Behrends and Harper, 2011). Given that the complexity and importance of ubiquitylation rivals that of phosphorylation-based signaling networks, ubiquitylation must be highly regulated.

Ubiquitylation is a reversible post-translational modification. The Ub-substrate isopeptide bond can be cleaved by a group of peptidases called deubiquitylases (DUBs) (Komander et al., 2009). Sequence analysis and functional studies identified 95 potential DUBs encoded by the human genome grouped into five families: ubiquitin C-terminal hydrolases (UCHs), ubiquitin-specific proteases (USPs), ovarian tumor proteases (OTUs), the Josephins and Jab1/MPN/Mov34 metalloprotases (JAMM/MPN+). The first three groups are cysteine proteases whereas the latter two are zinc metalloproteases (Komander et al., 2009; Nijman et al., 2005). Of the 90 or so DUBs for which experimental evidence of expression exists, 79 are predicted to have deubiquitylase activity (Nijman et al., 2005). The relatively large number of putative DUBs without the required infrastructure to support proteolysis suggests that some DUBs have evolved protease-independent functions.

The response to DNA double-strand breaks (DSBs) is an example of a cellular pathway that relies extensively on protein ubiquitylation (Al-Hakim et al., 2010). Upon DSB detection, the ATM protein kinase initiates a cascade of protein recruitment on chromatin that can be visualized as discrete nuclear foci when examined by immunofluorescence microscopy (Lukas et al., 2011). The RING-type E3 RNF8 acts downstream of ATM to initiate a Ub-dependent cascade of protein recruitment (Al-Hakim et al., 2010; Huen et al., 2007; Kolas et al., 2007; Mailand et al., 2007). RNF8-dependent ubiquitylation of chromatin elicits the recruitment of a second RING-type E3, RNF168, which elaborates a K63-linked Ub chain on chromatin with its cognate E2, UBC13 (Doil et al., 2009; Stewart et al., 2009). This amplification of the ubiquitylation signal triggers the recruitment of additional DNA repair proteins such as RAD18, BRCA1 and 53BP1 to promote DSB repair (Bekker-Jensen et al., 2010; Huen et al., 2007; Kolas et al., 2007; Mailand et al., 2007; Stewart et al., 2009).

In a search for DUBs that oppose the action of RNF168, we identified OTUB1, an OTU family member, as a potent suppressor of ubiquitylation at DSB sites (Nakada et al., 2010). Depletion of OTUB1 resulted in a striking persistence of conjugated Ub and 53BP1 foci long after bulk DNA repair was completed (Nakada et al., 2010). This finding was initially surprising since OTUB1 is highly selective for K48-linked Ub chains, while RNF168 promotes the formation of K63-linked chains on chromatin. This paradox was solved when OTUB1 was found to act non-catalytically to inhibit the DNA damage response through the binding of UBC13. OTUB1 binds preferentially to the Ub-charged E2 in a manner that requires an N-terminal extension to its core OTU domain (Nakada et al., 2010; Wang et al., 2009). Furthermore, proteomic analyses indicated that OTUB1 binds to E2s of the UBE2D and UBE2E families including the E2 UbcH5 (Nakada et al., 2010; Sowa et al., 2009). Interestingly, OTUB1 was recently shown to modulate p53 stability through inhibition of UbcH5 (Sun et al., 2011).

In this study, we sought to elucidate the structural mechanism by which OTUB1 recognizes and inhibits UBC13, UbcH5 and related E2 enzymes. We present a genetically and biochemically validated X-ray crystallographic model that explains how OTUB1 recognizes Ub-charged E2s. Surprisingly, we found that in addition to binding the E2~Ub, OTUB1 also binds free Ub. Remarkably, the E2-linked and free Ub moieties mimic the configuration of cleaved K48-linked di-Ub chain, the canonical product of OTUB1 deubiquitylase activity. We conclude that OTUB1 co-opts a mechanism akin to product inhibition to inhibit another class of enzymes.

RESULTS

A genetic system to dissect E2 inhibition by OTUB1

To understand the mechanism of action of OTUB1 on UBC13, we introduced a plasmid expressing human OTUB1, tagged with the Flag epitope, in budding yeast under the control of the GAL1/10 galactose-inducible promoter. Our rationale was to test whether OTUB1 can inhibit yeast Ubc13. To our surprise, since Ubc13 is not essential in yeast, we observed that induction of OTUB1 expression severely curtailed yeast growth (Figure 1A). Growth inhibition was independent of the OTUB1 DUB activity, as the C91S mutation, which renders OTUB1 catalytically inert, also suppressed growth (Figure 1A). This growth inhibition phenotype was reminiscent of the isopeptidase-independent suppression of E2 enzyme action by OTUB1 (Nakada et al., 2010). In support of the possibility that the growth inhibition in yeast was due to E2 inhibition by OTUB1, we found that the OTUB1^ΔN mutant, which is unable to inhibit E2 enzymes, failed to block growth (Figure 1A). These results strongly suggested that OTUB1 expression in yeast shuts down the activity of one or more E2 enzymes required for growth.

Figure 1. A genetic system to dissect E2 inhibition by OTUB1 (see also Figure S1).

(A) Left panel: Five-fold serial dilutions of BY4741 strains transformed with the indicated pRS425 derivatives were spotted onto selective media containing either glucose (uninduced condition) or galactose (induced) as the carbon source. Right panel: Whole cell extracts (WCE) of the strains shown in the left panel were separated by SDS-PAGE and probed for Flag-OTUB1 expression. Pgk1 immunoblotting was used as a loading control.

(B) Left panel: Examination of yeast growth assay as in (A). Right panel: WCE of the strains shown in the left panel were separated by SDS-PAGE and probed for Flag-OTUB1 expression. Pgk1 immunoblotting was used as a loading control.

(C) Schematic of the genetic screen used to identify human OTUB1 mutants that fail to inhibit yeast growth.

(D) Examination of yeast growth as in (A).

(E) Whole cell extracts of the strains shown in (D) were separated by SDS-PAGE and probed for Flag-OTUB1 expression. Pgk1 immunoblotting was used as a loading control.

The growth suppression imparted by OTUB1 expression provided a convenient genetic assay for identifying mutations that disable the ability of OTUB1 to inhibit E2 enzymes. Indeed, we readily isolated two strains that were spontaneously resistant to OTUB1 expression. Sequence analysis of the recovered OTUB1 plasmids revealed that both strains contained the same mutation in the OTUB1 open reading frame, P263L (Figure 1B).

To uncover additional mutants, we carried out a random mutagenesis screen (Figure 1C) and screened over 3 × 10⁷ gap-repair transformants. We identified 104 colonies that were resistant to OTUB1 expression. We then verified Flag-OTUB1 expression to identify the clones most likely to contain missense mutations. After sequencing, individual mutations were generated anew by site-directed mutagenesis for confirmation. We ultimately identified 7 OTUB1 mutations that allow yeast to growth in spite of OTUB1 expression (Figure 1D). Mutations were scattered across the OTUB1 open reading frame, hinting that multiple surfaces on OTUB1 are required for its ability to inhibit E2 enzymes (Figure S1).

Characterization of OTUB1 mutants in human cells

We next sought to determine whether the OTUB1 mutations identified in yeast impacted the response to DSBs in human cells. We employed a gene-replacement strategy in which mutations were introduced in an OTUB1 expression vector that directs expression of an mRNA that is resistant to a potent OTUB1 siRNA (Nakada et al., 2010). These plasmids were transfected into U2OS cells in which endogenous OTUB1 expression was silenced by siRNA (Figure S2A). OTUB1 depletion resulted in hyper-activation of RNF168-dependent chromatin ubiquitylation, which can be detected by an increase in conjugated Ub foci at sites of DNA damage 24 h after 3 Gy of X-irradiation (Figure 2). Reintroduction of wild type OTUB1 restored normal levels of Ub foci kinetics indicating that the expression levels of transiently transfected OTUB1 recapitulated endogenous regulation of the DSB response by OTUB1 (Figure 2AB). Interestingly, reintroduction of OTUB1^C91S produced a stronger phenotype than wild type OTUB1, which was manifested by a reduced number of IR-induced conjugated Ub foci at the 4 h time point. OTUB1^C91S was expressed at the same level as OTUB1 wild type (Figure 2C) indicating that the C91S mutation results in a protein with enhanced potency in inhibiting the DSB response. Lastly, reintroduction of the OTUB1^ΔN mutant failed to rescue OTUB1 depletion, as high levels of IR-induced conjugated Ub foci were detected at 24 h post-irradiation. The OTUB1^ΔN mutant protein was expressed at the same level as wild type OTUB1 (Figure 2C) ruling out the possibility that the failure to rescue OTUB1 depletion was due to varying proteins levels. The activity profiles of transiently transfected OTUB1 wild type, C91S and ΔN expression vectors closely matched those of stably transfected cells that expressed the same proteins at levels that approached endogenous OTUB1 levels (Figure S2B). Together, these results indicated that the transient reintroduction of siRNA-resistant OTUB1 plasmids in OTUB1-depleted cells largely recapitulated endogenous regulation of the DSB response by OTUB1.

Figure 2. Regulation of RNF168-dependent ubiquitylation by OTUB1 (see also Figure S2).

(A–C) U2OS cells were first transfected with non-targeting (CTRL) or OTUB1 siRNAs and then transfected with the indicated pCDNA5-Flag derivatives. All OTUB1 plasmids were siRNA-resistant. Cells were then irradiated with X-rays (3 Gy) and processed for Flag or conjugated Ub (FK2) immunofluorescence 0, 4 and 24 h post-IR. DNA was counterstained with DAPI. Quantitation of the immunofluorescence data is shown in (A). Data is presented at the mean +/− S.D. n=3. The dashed line indicates the percentage of control-transfected cells with FK2 foci 24 h post-irradiation. Representative micrographs of immunofluorescence data 24 h post-irradiation are shown in (B). Scale bar=10 μm. Analysis of Flag-OTUB1 expression is shown in (C). Tubulin was used a loading control.

We next tested whether the Q33R, I37T, Q39L, A116T, T134R, D137G, F190S, Y261H and P263L OTUB1 mutants could rescue the OTUB1 depletion phenotype (Q39L was added to our list as a control mutation not predicted to inhibit the DSB response). We introduced these mutations in the C91S background to provide a greater dynamic range to our assay. We observed complete concordance between the results obtained in yeast and those observed in human cells: mutants that permitted yeast growth also lacked the ability to dampen RNF168-dependent chromatin ubiquitylation (for example, I37T and T134R; Figure 2). Conversely, mutants that ablated yeast growth also potently ablated the DNA damage response (for example, A116T; Figure 2). Each mutant was expressed at the same level as the wild type and C91S OTUB1 proteins (Figure 2C) ruling out the possibility that the activity profile was due to varying levels of protein expression. Together, these results identified the Q33, I37, T134, D137, F190, Y261 and P263 residues as being critical for the ability of OTUB1 to dampen RNF168-dependent Ub conjugation at DSB sites.

Crystallization of OTUB1 bound to a charged E2

To illuminate how the residues mapped above function in the inhibition of Ub conjugation, we initially carried out crystallization screens with a complex containing Ub-charged UbcH5b^C85S (Ub~UbcH5b) and free OTUB1. The C85S variant of UbcH5b was utilized because the oxyester linkage endowed by the serine residue stabilized the Ub-charged state. The trials yielded numerous hits corresponding overwhelmingly to Ub~UbcH5b alone. To assist identification of crystals containing the desired complex, we employed a UbcH5b-OTUB1 fusion construct (see Supplementary Methods for construct design and rationale). The fusion was subjected to Ub charging reactions with GST-Ub and E1 followed by purification using glutathione affinity chromatography. Following removal of the GST moiety by TEV cleavage, excess free Ub was separated from Ub~UbcH5b^C85S-OTUB1 by size-exclusion chromatography. Crystallization screens with this covalent complex yielded two related monoclinic crystal forms that allowed collection of 3.8 Å (C2) and 3.3 Å (P2₁) data sets (Table 1).

Table 1.

Data collection and refinement statistics

	Dataset 1	Dataset 2
Source	APS	CLS
Space Group	C 1 2 1	P 1 21 1
Unit Cell Parameters
a, b, c (Å)	173.74, 106.23, 134.74	134.57, 104.93, 148.48
α, β,γ (°)	90.00, 123.74, 90.00	90.00, 104.20, 90.00
Wavelength (Å)	0.97922	0.97949
Resolution (Å)	50.0 – 3.80 (4.02- 3.80)	49.29 – 3.30 (3.50 – 3.30)
Completeness (%)	96.7 (83.1)	99.7 (99.4)
Redundancy	3.56 (3.3)	4.25 (4.25)
I/σ(I)	9.44 (1.86)	9.62 (1.46)
Rmeas (%)	16.5 (86.6)	12.6 (121.6)
Refinement Statistics
Rfree/Rfactor	21.7/27.9	22.17/28.79
RMSD Bond Length (Å)	0.006	0.018
RMSD Bond Angle (°)	1.21	1.833
Average B-factor (Å2)	119.98	138.6
Number of protein atoms	9619	26574
Ramachandran
% preferred	92.5	89.3
% allowed	7.5	8.9
% outliers	0	1.8

Values in parentheses are for highest-resolution shell.

Structure determination and overview of binding topology

The C2 crystal contents consist of three biological complexes in the asymmetric unit. As shown in Figure 3A, each complex, denoted Ub~UbcH5b-OTUB1-Ub, contains one molecule of UbcH5b covalently linked to Ub, one molecule of OTUB1, and also one free Ub (coloured cyan, pink, green, and yellow, respectively).

Figure 3. Structure of the Ub~UbcH5b-OTUB1-Ub complex (see also Figures S1, S3, S4, S5).

(A) Schematic and ribbons representation of the Ub~UbcH5b-OTUB1-Ub complex with subunits coloured pink, cyan, green, and yellow, respectively. The side chains of the catalytic cysteine of OTUB1 and UbcH5 are shown as sticks. Helix αA of OTUB1 and the N- and C-termini of subunits are labeled where visible.

(B) Stereo view of the OTUB1-UbcH5b binding interface. Main chains are coloured as in (A). Contact residues are shown as sticks with carbon atoms coloured according to their respective main chain and oxygen, nitrogen, and sulfur atoms are coloured red, blue, and yellow, respectively. The positions of OTUB1 mutants tested in this study are labeled red.

(C) UbcH5c and UBC13 utilize similar surfaces to bind OTUB1. Significant perturbations (>1 S.D. from the mean) arising from OTUB1 binding to ¹⁵N-UbcH5c (top) or ¹⁵N-UBC13 (bottom) are mapped in red onto the respective E2 structures.

(D) NMR analysis shows binding of OTUB1 to UbcH5c in solution is consistent with X-ray data. Resonances in ¹⁵N-UbcH5c significantly affected by binding of OTUB1 are mapped (shown in red) onto the corresponding residues in the OTUB1:UbcH5b~Ub crystal structure.

The C2 crystal structure was solved by molecular replacement. OTUB1 (PDB 2ZFY) lacking 45 N-terminal residues (Edelmann et al., 2009) and UbcH5b (PDB 2ESK) (Ozkan et al., 2005) were used as search models and were unambiguously positioned in the unit cell in close contact using Phaser (McCoy et al., 2007). Following iterative rounds of restrained refinement, analysis of electron density maps revealed a 22-residue α-helix (denoted helix αA), corresponding to the missing N-terminus of OTUB1, in close contact with the E2-linked donor Ub. Inclusion of the donor Ub and the OTUB1 helix αA into model refinement revealed unexpected electron density for a free Ub molecule engaging the distal Ub binding site of OTUB1 (Messick et al., 2008). A near-complete model of the Ub~UbcH5b-OTUB1-Ub complex was refined at 3.8 Å resolution to an R_factor/R_free of 21.7/27.9 % with good geometry statistics (Table 1). Resolution was extended by refinement at 3.3 Å resolution against the P2₁ X-ray data set to an R_factor/R_free of 22.1/28.8%, with a doubling of the contents of the asymmetric unit

The fold architectures of UbcH5b, OTUB1 and Ub have been described previously (Edelmann et al., 2009; Ozkan et al., 2005; Vijay-Kumar et al., 1987) and do not deviate significantly in the Ub~UbcH5b-OTUB1-Ub complex (see Figure S1 for sequence alignments and secondary structure assignments). A detailed description of the binary interactions defining the overall topology of the Ub~UbcH5b-OTUB1-Ub complex follows.

Binding interface between OTUB1 and UbcH5b

UbcH5b engages two non-contiguous surfaces on OTUB1 (Figures 3AB and S3A) burying a total surface area of 2001 Ų. The larger surface on OTUB1 is centered on helices α5, α7, and the α9-α10 linker, which engages a surface composed of helix α1, the α2-α3 linker, and the β3-β4 linker on UbcH5b. Specific side-chains that mediate hydrophobic and charge-complementary interactions include Phe130, Thr131, Phe133, Thr134, Ile135, Asp137, Phe138, Asp169, Tyr170, Val173, Leu177, Gln206, Glu209, Pro210, Met211, and Lys213 on OTUB1 and Met 1, Arg5, Lys8, Asp12, Lys101, Pro61, Phe62, K63, Gln92, Ser94, Pro 95, Ala96, Thr98 and Lys101 on UbcH5b.

A smaller contact surface on OTUB1 is centered on the helix αA, which engages the catalytic cleft region of UbcH5b in a manner that partly shields the labile thioester (in this case oxyester) linkage to donor Ub (Figures 3AB and S3A). Specific residues that mediate this contact include Asp27, Glu28, Met31 and Asp35 on OTUB1 and K63, Lys66, Ser83, Arg90, and Ser91 on UbcH5b.

The surface engaged by OTUB1 overlaps with the E3-binding surface on UbcH5b (Figure S4AB). Occlusion of the E3-binding surface, along with the shielding of the Ub~UbcH5b ester linkage by OTUB1 likely accounts for the ability of OTUB1 to suppress E2 function. The surface of UbcH5b engaged by OTUB1 is relatively well conserved across the two E2 enzyme subfamilies to which OTUB1 was previously shown to bind (Figure S1). Thus we predict that the binding mode uncovered between OTUB1 and UbcH5b is of general relevance to all E2s targeted by OTUB1.

To verify that the interactions observed in the crystal structure are also observed in solution, the binding of the E2s UbcH5c (closely related to UbcH5b and also targeted by OTUB1) and UBC13 to full-length OTUB1 was examined by NMR spectroscopy. A series of (¹H, ¹⁵N)-HSQC spectra of UbcH5c~Ub in which the E2 subunit was ¹⁵N-labeled was mixed with increasing concentrations of OTUB1 (Figure S5). Chemical shift perturbations of backbone amide NMR resonances are sensitive indicators of their environment and can be used to map protein-interaction surfaces. Resonances corresponding to residues in α helix 1, in loops 4 and 7, and the UbcH5c active site (N77 and L86) are perturbed upon addition of OTUB1 (Figure 3C). The UbcH5 surface defined by the observed NMR resonance perturbations is the same as that observed in the crystal structure (Figure 3D). The presence of a conjugated Ub subunit does not appreciably alter E2 residues involved in binding OTUB1. NMR mapping experiments with both ¹⁵N-UbcH5c~Ub and free ¹⁵N-UbcH5c defined the same binding surface. Corresponding NMR experiments with¹⁵N-UBC13 showed that the same E2 structural elements are also used by UBC13 to bind OTUB1 (Figure 3C). These results strongly suggest that the closely related E2s that interact with OTUB1 do so in a similar manner. In contrast, UbcH7, which is not inhibited by OTUB1, does not interact with OTUB1 in our NMR experiments (Figure S5B).

When we model UBC13 binding to OTUB1, the surface of UBC13 engaged by its ancillary factor UEV2 (Mms2) (Eddins et al., 2006) is not occluded significantly by OTUB1 (Figure S4C). This is consistent with the observation that binding of UBC13 to UEV1A or OTUB1 is not mutually exclusive (Nakada et al., 2010). While the binding site on UbcH5b for E1 also overlaps with OTUB1 (Figure S4D), the strong binding preference of OTUB1 for the Ub-charged form of E2s could explain why OTUB1 does not inhibit charging of targeted E2s by E1. The binding surface of OTUB1 on UbcH5b does not overlap with either the previously characterized binding surface for non-covalently bound Ub (Brzovic et al., 2006) or with the predicted binding surface for donor Ub (Hamilton et al., 2001; Wickliffe et al., 2011) (Figure S4EF). Thus, direct occlusion of Ub-binding surfaces on the E2 is unlikely to contribute to OTUB1 E2 inhibitory function.

Binding interface between donor Ub and OTUB1

With the exception of the covalent tether between the C-terminus (Gly76) of the donor Ub and the catalytic cysteine (in this case Ser85) of UbcH5b, the donor Ub makes little contact with UbcH5b (total buried surface area of contact = 262Ų). Since the association of donor Ub with the donor Ub binding surface of E2s has been shown previously to be important for the Ub transfer function of some E2s (Saha et al., 2011; Wickliffe et al., 2011), this structural feature may also contribute to the E2 inhibitory function of OTUB1. Instead, the donor Ub engages two surfaces on OTUB1 (Figures 3A, 4A and S3B) burying a total surface area of 1700 Ų. The first contact surface involves loop regions (α-α2, β2-α3, β4-β5) on the core domain of OTUB1. Specific contact residues involved on OTUB1 include Tyr61, Ala62, Asp65, Tyr68, Pro87, Asp237, Pro263, and His265. The reciprocal contact surface on the Ub moiety is comprised by a segment spanning strands β3 to β5. Specific contact residues involved on Ub include Lys48, Arg54, Ser57, Asp58, Tyr59, Asn60 and Gln62. A second contact surface involves the N-terminal helix αA of OTUB1, which engages the β-sheet surface (β1, β3, β4 and intervening linkers) of Ub. Specific Ub-contacting residues on OTUB1 include Tyr26, Ile30, Gln33, Gln34, Arg36, Ile37, Gln38, and Ile41. Specific contact residues on the donor Ub include Lys6, Leu8, Arg42, Ile44, Gly47, His68, Val70, Leu71, and Leu73.

Figure 4. Binding mode of donor Ub to OTUB1 (see also Figure S1 and S3).

(A) Stereo view of the donor Ub-OTUB1 binding interface. Colouring as in Figure 3B. The positions of OTUB1 mutants tested in this study are labeled in red.

(B) Stereo comparison of Ub-UIM and donor Ub-OTUB1 helix αA binding modes. The UIM-Ub structures corresponding to VPS27 (PDBID 1Q0W) and Hrs (PDBID 2D3G) are coloured grey and orange respectively. The OTUB1-donor Ub structure is coloured as in Figure 3B. In all three complexes, Ub engages its target α-helices using a common surface centered on Ile44 (labeled). The N- and C-termini of the UIMs and OTUB1 α-helices are also labeled.

(C) A mixture of uncharged and Ub-charged UBC13^C87S that was either prepared with wild type Ub (UBC13~Ub) or UbI44A (UBC13~Ub^I44A) were incubated with either GST-OTUB1^C91S or, as control, GST-Pcc1. Proteins were separated by SDS-PAGE and stained with Coomassie Brilliant Blue.

The interaction of donor Ub with the N-terminal helix of OTUB1 is strikingly reminiscent of the interaction between Ub interaction motifs (UIMs) and Ub (Fisher et al., 2003). Both the OTUB1 αA- and UIM interactions involve the Ile44 hydrophobic patch of Ub engaging a single α-helix (Figure 4B). Small differences in binding mode include the interaction angle, which differs by ~23°, and the closest distance of approach, which is smaller in the Ub-UIM interaction by ~4 Å due to bulkier residues on the OTUB1 helix (for example Ile37) as compared to the corresponding residue (Ala) in canonical UIMs. In support of the Ile44 hydrophobic patch mediating important contact with OTUB1, UBC13 charged with an I44A Ub mutant (UBC13~Ub^I44A) was compromised for its ability to bind to OTUB1 (Figure 4C).

Binding interface between distal Ub and OTUB1

Perhaps the most surprising feature of the structure was the presence of a non-covalently linked Ub molecule bound to OTUB1 (Figures 3A, 5A, S3C). The free Ub moiety interacts with OTUB1 at its distal Ub-binding site in a manner nearly identical to yeast Otu1 covalently bound to Ub-Br3 (Messick et al., 2008) (Figure S6A) burying a total surface area of 2344 Ų. The interaction surface on the distally-bound Ub consists of the solvent exposed β-sheet (β1, β3, β4) and flanking loop regions which engage an extensive concave pocket on OTUB1 comprised of helix α10, strands β3, β3′, β4, and an irregular loop region connecting helices α7 and α9. Contact residues on OTUB1 include Phe189, His192, Phe193, Glu195, Glu214, Ser215, Asp216, His217, Ile218, Ile221, Gln225, Tyr235, Asp237, Asn245, His247, Phe249, Glu251, Tyr261, and Tyr266. Contact residues on Ub include Leu8, Thr9, Ile36, Pro37, Asp39, Gln40, Arg42, Ile44, Gly47, His68, Val70, Leu71, Leu73, Gly75 and Gly76.

Figure 5. Binding mode of free Ub to OTUB1 (see also Figure S1, S3 and S6).

(A) Stereo view of the free Ub-OTUB1 binding interface. Coloring is as in Figure 3B. The positions of OTUB1 mutants tested in this study are labeled in red. The side chain of Phe190 in OTUB1, immediately adjacent to Phe189 and Phe193, is not shown for sake of clarity. For the Phe190 position see Figure S5B.

(B) Stereo view of donor Ub and free Ub in the proximity of the catalytic cleft of OTUB1. Coloring is as in Figure 3B. Highlighted as sticks are the catalytic residues of OTUB1 (Cys91, His265 and Pro87), and residues in direct contact between donor and free Ub.

(C) Stereo view of donor Ub and free Ub in proximity of the catalytic cleft of OTUB1. Coloring is as in Figure 3B. Highlighted as sticks are the catalytic residues of OTUB1 (Cys91, His265 and Pro87), K48 of donor Ub and Gly76 of free Ub. Superimposed on the solved structure is a model of an isopeptide linkage between the K48 side chain of donor Ub and the Gly76 C-terminus of free Ub. A productive conformation of OTUB1 catalytic residues in the Ub~UbcH5b-OTUB1-Ub complex contrasts with a non-productive conformation observed in the apo-OTUB1 complex (PDB 2ZFY).

While the presence of free Ub in the OTUB1 complex was unexpected, reanalysis of our purification scheme revealed consistent levels of free Ub in our Ub~UbcH5b-OTUB1 preparations. Interestingly, the C-terminal tail of free Ub straddles the catalytic cleft of OTUB1 providing part of the platform engaged by the donor Ub (total buried surface area of contact = 378Ų). This affords van der Waals and salt bridge (between Glu51 in donor Ub and Arg72 and Arg74 in free Ub) interactions between Ub molecules (Figure 5B), which we suspected could confer interdependency of Ub~UbcH5b and Ub binding to OTUB1.

Molecular recognition of K48-linked di-Ub

The isopeptidase activity of OTUB1 is notable for its exquisite specificity for K48-linked Ub chains (Edelmann et al., 2009). In this context, the relative position of the two Ub molecules engaged by OTUB1 that positions K48 of the donor Ub in immediate proximity to the C-terminus of the distally bound Ub was remarkable (Figure 5BC). We reasoned that this binding configuration might reflect how OTUB1 recognizes its preferred substrate, K48-linked Ub chains. Indeed, when we model a K48 linkage between the Ub molecules, the structure is easily accommodated without changes to main chain conformation (Figure 5C). In addition, the active site cysteine is well positioned to attack the K48-linked isopeptide bond. Interestingly, the side chain of His265 in the OTUB1 active site hydrogen bonds to the side chain of Cys91, a characteristic feature of all papain-like protease active sites in their productive conformations. This is in contrast to the apo-OTUB1 structure, and suggests that the binding of the OTUB1 substrate may induce a productive conformation of the active site.

Mapping of OTUB1 mutations onto the Ub~E2-OTUB1-Ub structure

We next mapped the position of OTUB1 mutations identified in our genetic screen onto the structure of the Ub~E2-OTUB1-Ub complex. Each mutation identified was located on one of four distinct inter-subunit interaction surfaces, lending strong support to the functional relevance of our structural model (Figure 6A). Firstly, the I37T and Q33R mutations map to the OTUB1 helix αA-donor Ub interface (Figure 4A), indicating that at the genetic level, the OTUB1 UIM-like-donor Ub interaction is critical for the formation or function of the inhibitory complex. Consistent with this model, I37T and Q33R, but not Q39L (which does not inhibit the DNA damage response), inhibited di-Ub synthesis by UBC13 (Figure 6B). Secondly, the T134R and D137G mutations map to the E2-OTUB1 interface (Figure 3B), with the T134R mutation predicted to have a more drastic impact on the E2 interaction by virtue of insurmountable steric clashes. In support of this prediction, OTUB1^T134R is fully compromised in its ability to inhibit UBC13-dependent di-Ub synthesis (Figure 6B). OTUB1^D137G also failed to inhibit UBC13 (Figure 6B). Thirdly, the Y261H and P263L mutations map to the OTUB1 core domain-donor Ub interface (Figure 4A) further supporting the importance of OTUB1-donor Ub contacts for recognition and inhibition of the charged E2. Accordingly, P263L failed to inhibit Ub conjugation by UBC13 (Figure 6B). Fourthly, the F190S mutation maps to the OTUB1-free Ub interface where, in conjunction with Phe189, Phe193 and Ile221, it comprises an elaborate hydrophobic pocket that engages Leu8 of Ub (Figure S6B). Consistent with a functional importance of this pocket, the F190S mutation disrupted the ability of OTUB1 to suppress UBC13 activity in vitro (Figure 6B).

Figure 6. The OTUB1 mutants map to functionally important intermolecular interfaces.

(A) Position of the OTUB1 mutations identified in the yeast genetic screen on the structure-based model of the Ub~UBC13-OTUB1-Ub complex.

(B) Single-turnover ubiquitylation reactions were assembled by pre-incubating UBC13, E1, biotinylated Ub and ATP for 5 min with or without the indicated OTUB1 proteins. To the reaction mixtures were added excess unlabeled Ub, UEV1 and EDTA. The reactions were incubated for an additional 15 min, separated by SDS-PAGE, transferred onto nitrocellulose and stained with Ponceau S followed by streptavidin (SA)-HRP. Asterisk denotes a DTT-resistant ubiquitylated UBC13 species.

(C) Examination of yeast growth and protein expression as in Figure 1A.

(D–E) Graph of internally quenched K48-linked di-Ub cleavage reactions with the indicated OTUB1 proteins.

To further probe the importance of free Ub binding to OTUB1, we designed an E195R mutation, which we predicted to potently impair binding to free Ub. When introduced into yeast, OTUB1^E195R was unable to suppress yeast growth (Figure 6C) and when introduced in human cells, it failed to rescue conjugated Ub foci clearance caused by OTUB1 depletion (Figure S7A). These results further validate the functional importance of Ub engagement to the distal Ub-binding site on OTUB1. Finally, we carried out GST pull-down assays with the OTUB1 mutants that failed to inhibit UBC13 and found that all had reduced binding to Ub~UbcH5b and free Ub (Figure S7B).

Since the E2-linked donor Ub and distally bound free Ub occupy the inferred substrate recognition sites for K48 chain hydrolysis, we investigated the impact of OTUB1 mutations on its deubiquitylase activity. We employed a FRET-based assay with a substrate consisting of a K48-linked di-Ub with an internally quenched fluorophore. Cleavage results in a homogeneous fluorescence signal that can be followed in a time-resolved manner. In this assay, OTUB1 shows robust time-dependent activity whereas OTUB1^C91S is completely inactive (Figure 6DE). We first assayed OTUB1 mutations at the OTUB1-E2 interface (T134R and D137G). Both mutants displayed robust DUB activity (Figure 6D). Next, we examined the mutations that affect residues located at the various OTUB1-Ub interfaces (I37T, Q33R, P263L, F190S and E195R). In contrast to mutations at the OTUB1-E2 interface, mutations at each of the OTUB1-Ub interfaces greatly impaired OTUB1 deubiquitylase activity (Figure 6E). Together, the inhibitory profile of the OTUB1 mutants was fully consistent with the crystal structure of the Ub~E2-OTUB1-Ub complex and demonstrated (1) that the E2 interaction surface on OTUB1, remote from its catalytic site, was dispensable for its DUB function, and (2) that both Ub interaction surfaces on OTUB1 were required for its isopeptidase activity, as they likely engage K48-linked di-Ub substrates. Furthermore, the results identified the T134R mutation as a true separation-of-function mutation, as it impaired E2 inhibitory function without impairing the DUB activity of OTUB1.

The strong impact of the I37T and Q33R mutations on deubiquitylase activity was surprising since OTUB1^ΔN retained some level of catalytic activity against unlabeled di- or tetra-Ub substrates (Edelmann et al., 2009; Nakada et al., 2010). To probe this conundrum further, we assessed how OTUB1^ΔN behaved in our FRET-based DUB assay. As shown in Figure 6E, OTUB1^ΔN was largely inactive as a DUB, suggesting that the FRET-based assay was sensitized for perturbations in Ub-binding at the proximal site, possibly due to the presence of fluorescent and quencher probes on the substrate. While our use of a sensitized substrate to uncover perturbations in Ub-binding was serendipitous, these results are entirely consistent with the architecture of the Ub~E2-OTUB1-Ub model obtained by crystallography.

Insights into the regulation of the OTUB1-E2 interaction

How is the OTUB1-E2~Ub inhibitory complex regulated to release the E2 conjugating enzyme in its active form, for example during the DNA damage response, remains an open question. Our mutagenesis results and the structure of the Ub~E2-OTUB1-Ub complex however, offers two possible avenues for consideration. One avenue might entail disruption of the inhibitory complex by phosphorylation of OTUB1 on one or more of the nine conserved phosphorylatable residues located at essential interaction surfaces. Indeed, our genetic screen identified the T134R mutation, which targets one such residue in OTUB1 on its E2-binding surface. To extend this analysis, we mutated Thr134 to a phosphomimetic residue (T134E) or to a non-phosphorylatable alanine residue (T134A) and assayed the ability of OTUB1 to inhibit yeast growth and to rescue the OTUB1 depletion phenotype (in both cases, the mutations were introduced in the C91S background). Consistent with a possible phosphoregulatory role for Thr134, the non-phosphorylatable mutant OTUB1^T134A inhibited yeast growth (Figure 7A and Figure S7C) and was as potent as the C91S mutant in rescuing OTUB1 depletion in human cells (Figure 7B). In contrast, OTUB1^T134E was not functional in inhibiting either yeast growth or in restoring the normal kinetics of conjugated Ub foci in OTUB1-depleted cells (Figure 7AB). Furthermore, the OTUB1^T134E mutant failed to inhibit di-Ub synthesis by UBC13 in vitro (Figure 7C). These results demonstrate that phosphorylation could, in principle, regulate the E2 inhibitory function of OTUB1.

Figure 7. Potential modulation of the OTUB1-E2 interaction.

(A) Examination of yeast growth as in Figure 1A.

(B) Left panel: U2OS cells were first transfected with non-targeting (CTRL) or OTUB1 siRNAs and then transfected with the indicated pCDNA3-Flag derivatives and processed as in Figure 2. Quantitation of the immunofluorescence data is shown. Data is presented at the mean +/− S.D. n=3. Right panel: examination of Flag-OTUB1 expression. Tubulin was used as loading control

(C) Analysis of di-Ub synthesis by UBC13 in the absence (−) or presence of the indicated OTUB1 proteins exactly as Figure 6B.

(D) Titration analysis of OTUB1 binding to UbcH5b~Ub in the absence or presence of 1 or 10 μM free Ub using a time TR-FRET assay. The data presented as the mean +/− S.E.M. n=2.

(E) Pull down analysis of GST-OTUB1 binding to free Ub in the presence of 0, 5, 10 and 20 μM concentrations of UbcH5b~Ub. Proteins were separated by SDS-PAGE and stained with Coomassie Brilliant Blue.

A second possible avenue for regulation centres on the role of the distally bound free Ub in the formation of the OTUB1-E2~Ub complex. Based on the behaviour of the F190S and E195R mutants, which demonstrated an essential role for the distal Ub binding site in OTUB1 E2 inhibitory function, we reasoned that the cellular pool of free Ub could modulate inhibitory complex formation. To assess if free Ub could influence binding of OTUB1 to UbcH5b~Ub, we employed a time resolved (TR)-FRET-based assay where UbcH5b~Ub was labelled with Tb³⁺ cryptate and OTUB1 was labelled with Alexa 488. In the absence of free Ub, only weak non-saturating binding between OTUB1 and UbcH5b~Ub was observed. In contrast, complex formation was clearly stimulated in the presence of 1 or 10 μM free Ub (Figure 7D). Further demonstrating an interdependency of binding, we found that free Ub bound to OTUB1 only in the presence of UbcH5b~Ub (Figure 7E). In sum, these data support a potential role for free Ub or an alternate form of Ub, in the regulation of OTUB1’s E2 inhibitory function.

DISCUSSION

Despite the fundamental influence of ubiquitylation on most cellular processes, there is a paucity of mechanisms that have been reported to modulate the activity of E2 conjugating enzymes. One such mechanism is the regulation of a subgroup of E2s that includes UBC13 and UbcH5 by OTUB1, a DUB. We found that OTUB1 assembles a complex that places two Ub moieties in a configuration that is reminiscent of a cleaved K48 linked di-Ub chain. Our validation studies strongly support the hypothesis that this complex leads to suppression of E2 enzyme activity by blocking access to the E3, by hindering attack on the E2~Ub thioester bond, and possibly by hindering association of donor Ub with the donor Ub-binding surface of the E2.

Inhibition of an enzyme by its product, a phenomenon termed “product inhibition” in enzymology, has been recognized for over a century (Frieden and Walter, 1963). Product inhibition often provides negative feedback inhibition and has been manipulated by nature and chemists to produce selective enzyme inhibitors. In the case of the OTUB1-E2~Ub interaction, we have an unusual example of product inhibition mimicry; the assembly of the Ub~E2-OTUB1-Ub complex completes the formation of a pseudo DUB cleavage product that inhibits a second enzyme, in this case the E2.

Since the unique N-terminal extension of OTUB1 is a critical determinant of E2~Ub recognition, this explains why OTUB2, an OTUB1 paralog that lacks an analogous extension, cannot inhibit E2s. As the N-terminal region of OTUB1 also contributes to K48-linked Ub chain recognition, the lack of this extension in OTUB2 would also explain its weaker ability to discriminate between Ub chain linkages (Edelmann et al., 2009). Interestingly, TRABID, an OTU family member, appears to employ a similar strategy for K29 and K33 Ub-linkage recognition, whereby an ankyrin repeat domain serves an analogous function to helix αA in OTUB1 (Licchesi et al., 2011).

The mechanistic themes underlying K48-linked chain selection by OTUB1 are generally similar to those underlying K63-linked Ub chain selection by AMSH-LP (Sato et al., 2009) even though the underlying catalytic domain structures of AMSH-LP and OTUB1 are unrelated. Firstly, both DUBs recognize an extended conformation of a di-Ub unit characterized by few inter-Ub moiety contacts (Figure S6C). Secondly, both DUBs make extensive contacts to both proximal and distal Ub moieties (1,324 Ų and 472Ų, respectively in the case of AMSH-LP). The former point is somewhat surprising since K48-linked Ub chains, unlike K63-linked chains, adopt a more compact topology involving extensive inter Ub moiety contacts (Varadan et al., 2004). While K48-linked Ub chains may be more compact in solution, this feature might hinder accessibility to the scissile isopeptide bond. We posit that the need for accessibility to the scissile bond will impose a general requirement for an extended conformation of Ub chains irrespective of the preferred isolated chain conformation in solution.

One corollary of our model is that concentration of OTUB1 should be comparable to that of the Ub-charged E2s it inhibits. Indeed, we found that OTUB1 is more abundant than UBC13 in U2OS and 293T cells (Figure S7D). If one considers that OTUB1 binds primarily to the Ub-charged form of E2s, these OTUB1 levels are sufficient to regulate UBC13 action and the DSB response. Upon induction of the DSB response, the Ub~UBC13-OTUB1 interaction must be relieved to enable UBC13 to act with RNF168. Our mutagenesis and structural studies provide a few clues that will be the focus of future study. One avenue of regulation involves potential post-translational modification of OTUB1 at inter-subunit contact surfaces. While our data are consistent with a potential phosphoregulatory role at Thr134, phosphorylation at this site has yet to be detected experimentally. A second possible avenue of regulation involves engagement by OTUB1 of Ub at the distal Ub-binding site. It is conceivable that local or compartmentalized fluctuations of Ub levels might be achieved by bursts of DUB activity or, conversely, E3 ligase activity. In support of this possibility, FRAP experiments examining chromatin ubiquitylation at UV lesions indicated that Ub levels are highly dynamic at sites of DNA damage. (Marteijn et al., 2009).

EXPERIMENTAL PROCEDURES

Yeast strains and plasmids

Standard yeast methods were used for growth and genetic modification. All strains derive from BY4741. The OTUB1 plasmids used in this study were constructed by introducing a HindIII-XhoI fragment from pcDNA3-Flag-OTUB1 (Nakada et al., 2010) into pRS425 or pcDNA5/FRT/TO. Site-directed mutagenesis was carried using Quikchange (Agilent).

Protein expression and purification

UbcH5b^C85S, UbcH5b^C85S-OTUB^Δ1-24 and the OTUB1 proteins were expressed in Escherichia coli as TEV-cleavable GST fusions and purified by glutathione affinity, TEV cleavage, and size exclusion chromatography. The Ub-charged versions of UbcH5b^C85S and UbcH5b^C85S-OTUB^Δ1-24 were generated by incubating proteins with 0.5 μM E1 and a 3-fold molar excess of His₆-GST-TEV-Ub in 50 mM Tris-Cl (pH 8.5), 150 mM NaCl, 10 mM MgCl₂, 5 mM ATP and 1 mM DTT at 37°C overnight. The Ub-charged proteins were purified from uncharged proteins by glutathione affinity, TEV cleavage, size exclusion chromatography and concentrated to 5 mg/ml in buffer containing 20 mM HEPES (pH7.5), 150 mM NaCl.

Deubiquitylase assays

DUB assays were carried out with 100 nM of OTUB1 and 250 nM of the di-Ub (K48) IQF substrate #5 (Life Sensors) in a buffer containing 50 mM HEPES pH 7.5, 100 mM NaCl, 0.3 % Brij-35, 1 mM DTT, 0.1% BSA. TAMRA fluorescence was read on an Enspire 2300 plate reader (Perkin Elmer).