OTU deubiquitinases reveal mechanisms of linkage specificity and enable ubiquitin chain restriction analysis

Abstract

Sixteen ovarian tumor (OTU) family deubiquitinases (DUBs) exist in humans, and most members regulate cell-signaling cascades. Several OTU DUBs were reported to be ubiquitin (Ub) chain linkage specific, but comprehensive analyses are missing, and the underlying mechanisms of linkage specificity are unclear. Using Ub chains of all eight linkage types, we reveal that most human OTU enzymes are linkage specific, preferring one, two, or a defined subset of linkage types, including unstudied atypical Ub chains. Biochemical analysis and five crystal structures of OTU DUBs with or without Ub substrates reveal four mechanisms of linkage specificity. Additional Ub-binding domains, the ubiquitinated sequence in the substrate, and defined S1' and S2 Ub-binding sites on the OTU domain enable OTU DUBs to distinguish linkage types. We introduce Ub chain restriction analysis, in which OTU DUBs are used as restriction enzymes to reveal linkage type and the relative abundance of Ub chains on substrates.

Graphical abstract

Figure 1

Human OTU DUBs and Reactivity of Analyzed Constructs (A) Phylogenetic tree of human OTU domain DUBs. HIN1L (^∗) is a pseudogene, and FAM105A (^∗∗) is lacking active site residues. (B) Domain composition in human OTU DUBs (updated from Komander et al. [2009]). (C) Constructs analyzed in this study. Full-length proteins not used in this study are shown in gray. (D) Purified OTU proteins according to (C) resolved on a Coomassie-stained 4%–12% SDS-PAGE gradient gel. M, marker. Asterisks (^∗) indicate purified constructs. (E) Reactivity of analyzed constructs against the suicide probe Ub propargylamide (Ub-PA, upper panel) and Cy5-labeled Ub-PA (lower panel). Asterisks (^∗) indicate the modified form of the OTU DUB. See also Figure S1.

Figure 2

Linkage Specificity of Human OTU DUBs (A) Purified OTU DUBs (constructs according to Figure 1C) were incubated with diUb of all linkage types for the indicated times and resolved on silver-stained SDS-PAGE gradient gels. Enzyme concentration is as indicated and differs for each DUB. See Figure S2 for additional experiments. (B) OTU DUB linkage specificity against diUb substrates can be grouped to enzymes cleaving one linkage type (group I), two linkage types (group II), three or more linkage types (group III), or inactive enzymes (group IV).

Figure 3

Roles for UBDs in OTU Specificity (A) Surface representation of an OTU domain (blue) bound to a distal Ub molecule (yellow) with its C terminus reaching to the active site. The proximal Ub in the dimer needs to bind such that only the preferred linkage point(s) (indicated in red on Ub surface) are presented to the active site. (B) DUB assays performed as in Figure 2A with OTUD1 aa 287–481 (OTU+UIM, top) and 287–437 (OTU, bottom). The construct lacking the UIM domain is nonspecific and less active (14.5× higher enzyme concentration used in gel below). (C) Specificity analysis of different OTUD2 constructs. Top, OTUD2 lacking the UBX-like domain. Second from top, OTUD2 lacking the ZnF domain. Third from top, OTUD2 isolated OTU domain. Bottom, OTUD2 with a mutation in the ZnF domain. The ZnF affects the ability of OTUD2 to cleave Lys27-, Lys29-, and Lys33-linked diUb. See Figure S3 for additional experiments. (D) Specificity assays of OTUD3 for constructs including the OTU and UBA domains (top) and the catalytic domain alone (bottom). The UBA domain has no influence on diUb hydrolysis. (E) Mechanism 1, positioning and orientation of the proximal Ub is achieved by its binding to a UBD present in the OTU enzyme.

Figure 4

Linkage Specificity Determinants in the Proximal Ub (A) Schematic representation (left) and sequence of generated fluorescent ubiquitinated Ub peptides. The red K indicates the ubiquitination site in the peptide. TAMRA refers to the fluorescent group appended to the N terminus of the peptide. (B–E) OTUD1 (B), OTUD3 (C), and OTUD2 (D and E) used at the different concentrations (indicated to the right) cleaved the indicated peptides over time. OTUD1 (B) and OTUD3 (C) (as well as OTUB1 and Cezanne2, see Figure S4) hydrolyzed most of or all the peptides similarly, indicating a lack of sequence preference and a requirement for other regions in the proximal Ub to recover specificity. OTUD2 hydrolyzed all peptides if used at high enzyme concentrations (D) yet showed the highest activity against the K6 and K11 peptide that were already hydrolyzed at the start of the measurement. Dilution of OTUD2 to picomolar concentrations (E) revealed that the enzyme was sequence specific for a ubiquitinated peptide based on the Ub Lys11 context. (F) Alanine scanning mutagenesis of the K11 peptide and assay with OTUD2 at 1 nM concentration as performed in (E). The y axis scale is the same in all graphs except F4A. Residues affecting OTUD2-mediated hydrolysis are indicated in red in the sequence alignment (right). Leu15, not present in the K6 peptide, explains the difference in sequence specificity between these similar peptides. Mutation of Lys6 to Ala resulted in an insoluble peptide, and Gly10 was not mutated. See also Figure S4. (G) Mechanism 2, OTUD2 is able to read the sequence context of the ubiquitination site, bind, and cleave in a sequence-specific fashion.

Figure 5

Structural Studies on OTUD1, OTUD2, and OTUD3 Reveal a Conserved S1’ Site (A–C) Crystal structures of the OTU domains of OTUD1 (A), OTUD2 (B), and OTUD3 (C). A cartoon representation in identical orientation is shown. The S1 Ub-binding site, N and C termini, and N- or C-terminal α helix are labeled. (D) Structure of inactive OTUD2 catalytic domain (C160A) in complex with the ubiquitinated K11 peptide (orange, see Figure 4) bound across the active site of the enzyme (boxed) shown as in (B). The inset shows a stick model of the ubiquitinated peptide. (E–G) Surface residues of OTUD1 (E), OTUD2 (F), and OTUD3 (G) are colored according to conservation of the protein throughout evolution (on the basis of the alignments in Data S1). (H) Top view of the putative S1’ site in OTUD2. The peptide structure in (D) reveals how the isopeptide is bound across the active site of an OTU DUB. Cys, His, and V loops as well as the C-terminal helix are indicated. (I) Putative S1’ site in the structure of OTUD2 bound to the ubiquitinated K11 peptide. An arrow indicates the scissile bond. (J) The same view as in (I) for the OTUB1 structure with two Ub moieties bound in S1 and S1’ sites (Wiener et al., 2012). The proximal Ub contacts the Cys and His loops and also a dedicated S1’ binding site in a protruding N-terminal helix unique to OTUB1. (K) Superposition of (I) and (J) showing the compatibility of S1’ binding sites. (L) Sequence of Cys, His, and V loops in the human OTUD enzymes, yOtu1, and dmOtu1. Residues in red are “anchor” points of conserved structural residues. An asterisk (^∗) indicates catalytic Cys or His. (M) A His loop mutation in OTUD3, R¹⁷⁸YGE to LSNG, creates a less active OTUD3 variant in which Lys11-diUb activity is more strongly affected than Lys6-diUb activity (see also Figure S5K). Note the differences in enzyme concentration used in the assays. (N) Mechanism 3, a conserved S1’ Ub-binding site on OTU DUBs positions the proximal Ub toward the catalytic center. See also Figure S5.

Figure 6

Complex Structures of OTUD2 Reveal an S2 Ub-Binding Site (A) Structure of the inactive OTUD2 OTU domain (C160A) bound to Lys11-linked diUb in the S1 and S2 site of the enzyme. The orientation of the Ub molecules is compatible with Lys11 linkage, although the linker sequence is not resolved in the electron density maps (indicated by arrows). (B) Close-up image of the hydrophobic S2 site on the α6 helix formed by Ile292 and Val295, which interact with the Ile44 patch of Ub. An alignment shows conservation of this sequence in different species (see also Data S1). (C) Structure of inactive OTUD2 in complex with the ubiquitinated K11 peptide as in Figure 5D. A second Ub for which the peptide is disordered is bound in the S2 site (see Figure S5G). (D) DUB assays with Lys11-linked Ub chains. Assays comparing isolated catalytic domains of WT OTUD2 (aa 147–314) and S2 site mutant (MutS2, aa 147–314, I292Q, V295Q) as well as dmOtu1 (aa 143–313) toward Lys11-diUb (top), Lys11-triUb (middle), and Lys11-tetraUb (bottom). Human OTUD2 hydrolyzed tri- and tetra-Ub immediately, and this depended on the S2 site of the enzyme. (E) Cleavage of differently linked triUb chains. In comparison to (D) and Figure 2A, a 4-fold lower enzyme concentration was used. (F) Mechanism 4, an S2 Ub-binding site on OTU DUBs allows the DUB to target and specifically hydrolyze longer Ub chains. See also Figures S6A–S6E.

Figure 7

Exploiting OTU DUBs in Ub Chain Restriction Analysis (A) Schematic of the principle of Ub chain restriction analysis. (B–I) Ub chain restriction analysis against the indicated substrates. SDS-PAGE gradient gels were silver-stained (B–E, G, and H) or western blotted with anti-Ub (F) or anti-RIP1 (I). M, marker; Control, ubiquitinated protein without DUB treatment. Enzyme bands are highlighted in silver-stained gels (green boxes). (B) Enzyme input reference gel. (C) GST-TRAF6, UBE2N, and UEV1A generated free and attached Lys63-linked polyUb. See Figure S6F for an anti-Ub western blot of this gel. (D) Lys63-autoubiquitinated GST-tagged NEDD4 HECT domain with UBE2L3. See also Figure S6G. (E) Lys48-autoubiquitinated GST-E6AP with UBE2L3. See also Figure S6H. (F) Lys11-autoubiquitinated UBE2S containing contaminating Lys63 linkages (Bremm et al., 2010). (G) Met1-linked polyUb generated by a minimal HOIP construct with UBE2L3. (H) OTUD2 released polyUb chains from GST-E6AP and GST-NEDD4 compared to free Lys48- and Lys63-polyUb. (I) Ub chain restriction analysis of polyubiquitinated RIP1 generated by FLAG-TNFα mediated purification of TNF-RSC from human embryonic kidney 293T cells.

Figure S1

Reactivity of Human OTU DUBs with Probes Derived from Ub and Ub-like Modifiers, Related to Figure 1 (A) OTU DUBs shown alone and after incubation with the suicide probe Ub-propargylamide (Ub-PA, 1 hr reactions). Asterisks (^∗) indicate modified forms of DUB. (B) Reactivity of ALG13 and A20, which did not react with Ub-PA (Figures 1E and S1A 1 hr reactions) against Ub-PA and Ub bromoethylamine (C2Br). (C) Reactivity of OTU DUBs against ISG15-C2Br (3 hr reactions). Only vOTU is modified, indicated by an asterisk (^∗). (D) Reactivity of OTU DUBs against NEDD8-C2Br (3 hr reactions). Asterisks (^∗) indicate NEDD8 modified enzymes. (E) Fluorescence anisotropy assays of selected OTU domains (boxed in D) against fluorescent ubiquitinated or neddylated KG peptides. The raw data are shown at different enzyme concentrations.

Figure S2

Additional DUB Assays for Selected OTU DUBs, Related to Figure 2 (A) OTUD1 constructs tested at 22× concentration (top) or at 10× concentration and longer time course (24 hr, bottom) compared to Figure 2A showing that under these conditions, other linkages are hydrolyzed. (B) OTUD2 DUB assay at long time points (24 hr). (C) DUB assay of Cezanne at 10× higher concentration reveals that all isopeptide-linked diUb are hydrolyzed. (D) Linkage specificity of S. cerevisiae Otu1 (yOtu1), aa 91–301 (Messick et al., 2008). yOTU1 has a similar linkage specificity profile as its human counterpart OTUD2. (E) DUB assay of D. melanogaster Otu1 (dmOtu1), aa 143–347, shows a similar cleavage pattern compared to human and yeast orthologs but additionally cleaves K6-linked diUb.

Figure S3

Impact of the ZnF Domain on dmOtu1 Specificity and the Ability of the Human OTUD2 ZnF Domain to Bind monoUb, Related to Figure 3 (A) Specificity analysis of different dmOtu1 constructs. Top, dmOtu1 lacking the UBX-like domain. Bottom, isolated catalytic OTU domain. As in the human ortholog OTUD2, the ZnF domain affects the enzyme’s ability to cleave Lys27-, Lys29- and Lys33-linked diUb. (B) Sequence of the analyzed human OTUD2 ZnF construct (aa 314–348). (C) Fully assigned ZnF spectrum derived from 3D NMR experiments. (D) Ub binding experiment with OTUD2 ZnF construct. Shown are overlaid spectra of 50 μM OTUD2 ZnF alone (blue) and with addition of 1 mM unlabeled Ub (red). The boxed section is shown in close-up view below. The lack of chemical shift perturbations indicate no binding of the ZnF to monoUb. (E) The reverse experiment from (C), using 80 μM labeled Ub and 400 μM unlabeled OTUD2 ZnF. Again, no Ub residue is perturbed, hence there is no binding under these conditions.

Figure S4

Additional Fluorescence Polarization Assays and Ala Scan of the K11 Peptide, Related to Figure 4 (A and B) Fluorescence polarization assays as in Figure 4B–4D for OTUB1 (A) and Cezanne2 (B). (C and D) Alanine scanning mutagenesis of the K11 peptide. (C) Sequences of mutated fluorescent DUB substrates. (D) Bar graph representation of graphs shown in Figure 4F. Error bars represent SD from the mean.

Figure S5

Structural Studies on OTUD1, OTUD2, and OTUD3 Reveals a Conserved S1’ Site, Related to Figure 5 (A–C) Close-up view of OTUD1 (A), OTUD2 (B) and OTUD3 (C) with 2|Fo|-|Fc| electron density contoured at 1σ covering catalytic triad residues. (D) Superposition of OTUD1, OTUD2, OTUD3, OTUD5/DUBA (Huang et al., 2012) and yOtu1 (Messick et al., 2008). Structures are highly similar with low rmsd values (∼0.8 Å). (E) Structure of phosphorylated human OTUD5 bound to a Ub suicide probe (pdb-id 3tmp) (Huang et al., 2012). (F) 2|Fo|-|Fc| electron density contoured at 1σ covers the isopeptide and peptide bound across the active site of OTUD2. (G) Arrangement of molecules in the OTUD2 C160A complex with ubiquitinated K11 peptide. Two OTUD2 molecules are present, one of which interacts with two Ub moieties in S1 and S2 sites. 2|Fo|-|Fc| electron density contoured at 1σ covers the Ub molecules in the complex, indicating that the molecule in the S2 site is less well ordered. A beige OTUD2 from a neighboring asymmetric unit forms a crystal contact with the complexed OTUD2 molecule, affecting the peptide binding site of OTUD2. (H–J) OTUD2 C160A complex with ubiquitinated K11 peptide (H), the OTUB1 structure with two Ub moieties bound in S1 and S1’ sites (Wiener et al., 2012) (I) and OTULIN in complex with Met1-linked diUb (J) in identical orientations. Putative S1’ site elements (Cys, His, and V loop C-terminal helix in OTUD2) are indicated. (K) A His-lop mutation in OTUD3, R¹⁷⁸YGE to LSNG creates a OTU DUB with reduced activity against Lys11-linked diUb. Comparison of a fine time course of Lys6- and Lys11-linked diUb. Note differences in enzyme concentration used in assays.

Figure S6

Additional Assays for S2 Site Characterization and Ub Chain Restriction Analysis, Related to Figures 6 and 7 (A) Reactivity of wild-type OTUD2 catalytic domain (aa 147–314), S2 site mutant (MutS2, aa 147–314, I292Q, V295Q), S1 site mutant (MutS1, aa 147–314, AI200-201DD) and corresponding inactive variants (C160A) against Ub-PA. (B) Specificity analysis of OTUD2 MutS2. (C) DUB assay with constructs in (A) against K11-linked diUb. (D) Cleavage of triUb substrates as in (B). (E) Specificity analysis of OTUD1 using triUb substrates. (F–H) Western blotted restriction analysis gels (anti-Ub) of GST-TRAF6 assembled free and attached Lys63-linked polyUb (F), autoubiquitinated GST-NEDD4 (G) and GST-E6AP (H). Compare silver-stained counterparts in Figures 7C–7E.