Comparison of native and non‐native ubiquitin oligomers reveals analogous structures and reactivities

Abstract

Ubiquitin (Ub) chains regulate a wide range of biological processes, and Ub chain connectivity is a critical determinant of the many regulatory roles that this post‐translational modification plays in cells. To understand how distinct Ub chains orchestrate different biochemical events, we and other investigators have developed enzymatic and non‐enzymatic methods to synthesize Ub chains of well‐defined length and connectivity. A number of chemical approaches have been used to generate Ub oligomers connected by non‐native linkages; however, few studies have examined the extent to which non‐native linkages recapitulate the structural and functional properties associated with native isopeptide bonds. Here, we compare the structure and function of Ub dimers bearing native and non‐native linkages. Using small‐angle X‐ray scattering (SAXS) analysis, we show that scattering profiles for the two types of dimers are similar. Moreover, using an experimental structural library and atomistic simulations to fit the experimental SAXS profiles, we find that the two types of Ub dimers can be matched to analogous structures. An important application of non‐native Ub oligomers is to probe the activity and selectivity of deubiquitinases. Through steady‐state kinetic analyses, we demonstrate that different families of deubiquitinases hydrolyze native and non‐native isopeptide linkages with comparable efficiency and selectivity. Considering the significant challenges associated with building topologically diverse native Ub chains, our results illustrate that chains harboring non‐native linkages can serve as surrogate substrates for explorations of Ub function.

Introduction

Post‐translational modifications (PTMs) dramatically expand the functions of proteins. Understanding the functional consequences of PTMs on a molecular level requires access to modified proteins for structural and biochemical studies—a challenge that is driving research in numerous labs. A PTM of particular interest, due to its involvement in virtually every biochemical pathway, is protein ubiquitination.1 Ubiquitin (Ub) is a small 76 amino acid protein that is covalently attached to substrates through an isopeptide bond between the C‐terminal glycine of Ub and the ε‐amino group of lysine residues. A set of dedicated enzymes (E1, E2, and E3) site‐selectively conjugates Ub to substrate proteins and builds Ub chains by connecting subunits via the seven internal lysine residues (K6, K11, K27, K29, K33, K48, K63) or the N‐terminus.2 Enzymatic synthesis can be applied in vitro to generate select Ub conjugates with varying topologies.3, 4, 5 However, the scope of enzymes identified for site‐selective ubiquitination currently does not approach the vast array of Ub conjugates suggested to regulate cellular pathways.6

To address this limitation, semi‐synthetic and total‐synthetic approaches have been developed to access native and non‐native Ub conjugates. Strategies to generate non‐native linkages benefit from their high‐yields and tunability for regiospecific modifications. There have been a number of success stories involving application of Ub conjugates with non‐native linkages to probe biological systems.7, 8, 9, 10 Notably, non‐native Ub chains were used to determine the minimal Ub signal required for efficient proteasomal targeting of substrate proteins.11 However, there is still some skepticism toward the use of non‐native mimics of isopeptide bonds for biochemical and biophysical studies to facilitate new discoveries. Such concerns are especially germane to the study of deubiquitinases (DUBs), a class of enzymes that directly target and cleave isopeptide bonds.12 To be recognized and processed by DUBs, Ub conjugates bearing non‐native linkages should recapitulate native structural features. The conformations and flexibility of select non‐native Ub linkages have previously been evaluated with computational methods.13 However, to effectively address whether non‐native linkages can adopt native chain conformations and functions, a solution‐based characterization is needed.

Our lab recently demonstrated that thiol‐ene coupling (TEC) could be used to rapidly build a diverse array of Ub chains with minimal synthetic manipulations.14, 15 TEC chemistry affords a thioether Gly‐Nɛ‐homothiaLys linkage that is 1.5‐Å longer than the native isopeptide bond (Fig. 1). We previously demonstrated that DUBs could still selectively cleave the artificial linkage in end‐point hydrolysis assays. To evaluate the extent to which the Gly‐Nε‐homothiaLys linkage can mimic the native isopeptide bond, we first sought to compare the Ub chain conformations generated by different linkage types. For this task, we selected small angle X‐ray scattering (SAXS), a popular technique for evaluating the conformations of flexible multi‐domain proteins in solution.16, 17 Among the different Ub chains, K48‐linked and K63‐linked Ub dimers (Ub2) are the best characterized in terms of structure, with more than 35 reported conformations for each Ub2 derived from X‐ray and NMR analyses.18, 19, 20, 21, 22, 23, 24, 25, 26, 27, 28, 29, 30, 31 We took advantage of this wealth of existing structural information to compare the SAXS‐derived structural properties of native and TEC‐derived Ub2. For both K48‐linked and K63‐linked Ub2, SAXS analysis of native and non‐native Ub2 reveals comparable scattering profiles that are matched to similar structures from a library of reported structures. Despite differences in the linkage itself, the conformations of native and non‐native Ub2 (K48 or K63‐linked) are markedly similar. As an alternative structural characterization protocol, multiple microsecond atomistic simulations were carried out for both K48‐linked and K63‐linked Ub2, and representative structures from the molecular dynamics trajectories were used to fit the experimental SAXS profiles. Because the low‐resolution nature of SAXS data prevents a precise determination of relative orientation of the two monomers, the MD derived ensemble of conformations differ somewhat from those fitted with the experimental structural library; however, the results again point to similar conformational ensembles for the native and TEC‐derived Ub2.

Figure 1.

Structures of native and non‐native isopeptide linkages. The distance from the Cα atom of lysine to the carbonyl carbon of Ub Gly76 is shown for each linkage. This is based on density functional theory geometry optimized structures.

In the second tack, we compared steady‐state kinetics of DUB cleavage reactions with TEC‐derived and native Ub dimers. There are five families of DUBs: the Ub C‐terminal hydrolases (UCHs), Ub specific proteases (USPs), ovarian tumor DUBs (OTUs), Machado‐Joseph disease proteases (MJDs), and the zinc‐dependent Jab1/Mpn/Mov34 proteases (JAMMs). The physiological substrates for many of these enzymes are unknown, but many DUBs exhibit selective hydrolysis activity toward Ub chains.32 For example, members of the JAMM family (e.g., AMSH and BRCC36) cleave K63‐linked chains exclusively,31, 33 and 10 out of the 16 human OTU DUBs preferentially to cleave only one or two linkage types.34 For TEC‐derived chains to be of use for systematic interrogation of DUB activity, hydrolytic cleavage of the nonnative linkage should occur with native‐like efficiency and selectivity. We demonstrate that these criteria are satisfied with several different DUBs and Ub dimers.

Results

SAXS analysis of native and TEC‐derived K48‐linked Ub2

Among the eight possible linkages between Ub subunits in a dimer, K48 has been extensively characterized using both X‐ray scattering and NMR analyses. The reported structures of K48‐linked Ub2 exhibit a range of conformations from compact (1AAR,21 2PE9,18 and 3AUL20) to extended (2KDF19). Consistent with this structural diversity, the conformational behavior of K48‐linked Ub2 is complex with multiple states coexisting in solution.35, 36, 37 To assess whether TEC‐derived K48‐linked Ub2 adopts a structure similar to that of the native dimers, we chose to use SAXS—an experimental technique for mapping the three‐dimensional conformations of macromolecules in solution.16, 17 The applicability of SAXS was recently demonstrated by work with native K63‐linked Ub dimers, which showed that the open and closed states of these oligomers could be distinguished.38 On the basis of these results, we surmised that SAXS would be sufficiently sensitive to detect any major structural differences between native and non‐native dimers.

The experimental SAXS profiles of native and TEC‐derived K48‐linked Ub2 reveal similar structural features for both dimers [Fig. 2(A)]. The radii of gyration, R _g, based on the Guinier approximation, are 19.35 ± 0.08 Å and 19.25 ± 0.12 Å for native and non‐native Ub2, respectively [Fig. 2(B)]. The distance distribution, P(r), curves for both samples show a major peak at 20 Å and a subtle shoulder after 40 Å, with maximum distance (D _max) of 70 Å for native Ub2 and 65 Å for non‐native Ub2 [Fig. 2(C)]. Lastly, Kratky curves for both native and nonnative dimer converge at high q, suggesting well‐folded structures [Fig. 2(D)].

Figure 2.

SAXS analysis of native and non‐native TEC‐derived K48‐linked (left) and K63‐linked (right) Ub2 reveals similar shapes. (A) SAXS intensity profiles of native (red) and TEC‐derived (blue) Ub dimers. The Kratky (B), Guinier (C), and distance distribution (D) plots suggest well‐folded structures with comparable dimensions. Guinier plots are staggered to clearly display both datasets.

The theoretical scattering profiles for the reported open and compact structures of K48‐linked Ub2 are distinct from the experimental data. With the open structure (2KDF19), the scattering profile curves downward in the low q region, whereas compact structures (1AAR,21 2PE9,18 3AUL20) give rise to positive curvature [Supporting Information Fig. S1(A)]. Within the same region, the scattering intensities for both the native and non‐native dimers occupy the space between the open and compact states, suggesting the average structure in solution is one that is a mixture of open and compact states. The differences between the experimental and theoretical scattering profiles are even more apparent when examining the P(r) curves [Supporting Information Fig. S1(B)]. In addition to the aforementioned compact and extended structures, there is a range of other intermediate K48‐linked Ub2 structures in the presence and absence of binding partners.22, 23, 37, 39, 40, 41, 42

To compare the experimental data to a library of extant structures, we compiled X‐ray and NMR structures of free Ub2 (PDBs: 1AAR,21 2PE9,18 and 3AUL,20 2BGF,39 3M3J,40 3NS841), and NMR structures of Ub2 bound to UBA2 from hHR23a (PDB: 1ZO622), UIM1 and UIM2 from S5a (PDBs: 2KDE and 2KDF19), and the CUE domain from the E3 ligase gp78 (PDBs: 2LVQ and 2LVP23). 2KDE and 2KDF each have seven different conformations for K48‐linked Ub2. We also extracted 20 different conformations from 2LVP, 24 from 2LVQ, and 10 from 1ZO6. The calculated SAXS profiles for the 85 structures exhibit a range of conformations from the most compact structure 2PE9 (R _g: 16 Å) to the most extended 2KDF‐6 (R _g: 23.9 Å). We organized the library of structures by R _g values and RMSD relative to 2PE9, 2KDF‐6, and the intermediate conformation 2LVP‐12 [Fig. 3(A)]. Differences in conformations can also be evaluated by comparing normalized P(r) plots [Fig. 3(B)]. Similar structures, such as 1AAR and 2PE9 (both compact) or 2LVP‐13 and 2KDF‐6 (both extended) have more than 90% overlap in their P(r) curves. A decrease in P(r) overlap corresponds to structural differences, and the two most distinct structures, 1AAR and 2KDF‐6, only exhibit 40% overlap (Supporting Information Fig. S2).

Figure 3.

A library of reported K48‐linked (top) and K63‐linked (bottom) Ub2 structures samples a range of conformations. (A) RMSD calculated for 85 K48‐linked Ub2 structures relative to 2PE9 (compact), 2LVP‐12 (intermediate) and 2KDF‐6 (extended); and 37 K63‐linked Ub2 structures relative to 3DVG (compact), 3JSV (intermediate) and 2ZNV‐1 (extended). (B) Heat map generated from absolute differences between P(r) curves to show structural variability among reported structures, arranged according to their R _g values. K48Ub2—orange/grey, K63Ub2—green/grey. (C) Evaluation of P(r) overlap between native and TEC‐derived Ub2 to reported structures presented in the heat map. The extent of overlap (from 50 to 100%) is illustrated with bar graphs.

Confident that the existing library of structures for K48‐linked Ub2 samples a broad conformational space, we compared the experimental SAXS patterns for native and TEC‐derived Ub2 to those of the 85 extant structures. Both experimental P(r) plots showed similar overlap with plots generated from the set of known structures [Fig. 3(C)]. Minimal ensemble search (MES) analysis identified 2LVQ‐24 and 2LVQ‐6 as the best single‐state representation for native and TEC‐derived Ub2, respectively. Both structures have intermediate R _g values, 19.3 Å for 2LVQ‐24 and 18.11 Å for 2LVQ‐6 [Fig. 4(A)].

Figure 4.

Minimal ensemble search for native (left) and TEC‐derived (right) K48‐linked Ub2. MES analysis identified (A) 2LVQ‐24 (for native) and 2LVQ‐6 (for TEC) as the single best‐fit structures; (B) 11% 3NS8, 89% 2LVQ‐24 (for native) and 39% 3AUL, 61% 2LVQ‐23 (for TEC) as the best two‐state ensemble; (C) 24% 2LVP‐12, 26% 3NS8, 5% 2LVQ‐21 (for native) and 33% 2LVQ‐24, 32% 2LVQ‐21, 36% 3NS8 (for TEC) as the best three‐state ensemble. (D) To show the range of fitting scores, χ values for the worst fit one‐state structure and the best MES fits are plotted. Hydrophobic patches centered on Ile44 and Ile36 are shown in blue and purple, respectively. Structures were visualized in Chimera.43

The best two‐state ensembles for native and TEC‐derived Ub2 both include a compact structure with a more extended structure [Fig. 4(B)]: 3NS8 with 2LVQ‐24 (native) or 3AUL with 2LVQ‐23 (TEC). The 2LVQ‐24 and 2LVQ‐23 have R _g values 19.3 Å and 19.4 Å, respectively. 3AUL and 3NS8 are nearly identical—they have the same R _g (16.6 Å), and RMSD difference between the two structures is 0.3 Å. For TEC‐derived Ub2, incorporating a third component to the ensemble did not enhance the MES fitting score [Fig. 4(C,D)], whereas a modest enhancement was calculated for native Ub2 [Fig. 4(C,D)]. All best‐fit ensemble results have >90% P(r) overlap with experimental curves for both native and TEC‐derived Ub2 (Supporting Information Fig. S3).

SAXS analysis of native and TEC‐derived K63‐linked Ub2

K63‐linked Ub2 has also been extensively characterized using X‐ray scattering and NMR analyses. Similar to K48‐linked Ub2, the conformational behavior of K63‐linked Ub2 is reported to be complex with multiple states coexisting in solution. The experimental SAXS profiles of native and TEC‐derived 63‐linked Ub2 reveal similar structural features for both dimers [Fig. 2(A)]. R _g values for native and non‐native Ub2 are 20.47 ± 0.07 Å and 21.58 ± 0.10 Å, respectively [Fig. 2(B)]. P(r) curves for both samples show a major peak at 20 Å and a relatively distinct shoulder at 40 Å [Fig. 2(C)]. Kratky curves for both native and non‐native dimer converge at high q, suggesting well‐folded structures [Fig. 2(D)].

To compare the experimental data to a library of extant structures, we compiled X‐ray structures of free K63‐linked Ub2 (PDBs: 2JF5,24 3H7P,25 and 3H7S25); X‐ray structures of Ub2 bound to the NZF domain of TAB2 (PDBs: 2WWZ,26 2WX0,26 2WX1,26 3A9J27), the NZF domain of TAB3 (PDB: 3A9K27), the CoZi domain of NEMO (PDB: 3JSV28), UIMs from RAP80 (PDB: 3A1Q29), a K63‐specific antibody (PDBs: 3DVG30 and 3DVN30), and the DUB AMSH (PDB: 2ZNV31); and NMR structures of Ub2 bound to UIMs (PDB: 2RR9). The 37 structures range from the most compact structure 3DVN‐2 (R _g: 17.3 Å) to the most extended 2ZNV‐1 (R _g: 25.1 Å). We organized the library of structures by R _g values and RMSD relative to 3DVG, 2ZNV, and the intermediate conformation 3JSV [Fig. 3(D)]. The diversity in Ub2 conformations were also demonstrated by comparing normalized P(r) plots [Fig. 3(E)].

With the library of structures, MES analysis identified 3A1Q‐1 and 3JSV as the best single‐state representation for native and TEC‐derived Ub2, respectively. Both structures have intermediate R _g values, 21.4 Å for 3A1Q‐1 and 21.3 Å for 3JSV [Fig. 5(A)]. The best two‐state ensembles for native and TEC‐derived Ub2 both include a relatively compact structure with a more extended structure [Fig. 5(B)]: 2WX1 with 2RR9‐17 (native) or 3A9J with 2RR9‐15 (TEC). The 2RR9‐17 and 2RR9‐15 have R _g values 23.0 Å and 23.3 Å, respectively, while 2WX1 and 3A9J have R _g values 17.7 and 18.3 Å. For both native and non‐native Ub2, incorporating a third component to the ensemble did not enhance the MES fitting score [Fig. 5(C,D)]. Overall, both native and TEC‐derived K63‐linked Ub2 adopt similar shapes, and the non‐native Ub2 is slightly more extended.

Figure 5.

Minimal ensemble search for native (left) and TEC‐derived (right) K63‐linked Ub2. MES analysis identified (A) 3A1Q (for native) and 3JSV (for TEC) as the single best‐fit structures; (B) 55% 2WX1, 45% 2RR9 (for native) and 51% 3A9J, 49% 2RR9‐15 (for TEC) as the best 2‐state ensemble; (C) 25% 2WXO‐1, 41% 2RR9‐17, 34% 2WX1 (for native) and 4% 2WXO‐2, 49% 2RR9‐15, 47% 3A9J as the best 3‐state ensemble. (D) To show the range of fitting scores, χ values for the worst fit 1‐state structure and the best MES fits are plotted. Hydrophobic patches centered on Ile44 and Ile36 are shown in blue and purple, respectively. Structures were visualized in Chimera.43

Atomistic simulations of conformational ensembles and comparison to experimental SAXS data

The use of an experimental structural library points to significant populations of open conformers for both native and TEC‐derived Ub2. The relative orientation of the two Ub monomers in the open conformer(s) cannot be assessed with SAXS data alone due to the low‐resolution nature of the analysis. Therefore, we have also carried out atomistic simulations for K48‐linked and K63‐linked Ub2 starting from many independent structures. By comparing calculated SAXS profiles for representative structures from the MD trajectories to measured scattering data, we characterized Ub2 structures and evaluated the robustness of findings from the previous sections.

For simulations with K48‐linked Ub2, all starting structures, even the most extended with R _g ∼25 Å, converged to modestly compact conformations R _g ∼18–19 Å. Compared to Ub2 simulations with a native linker, simulations with a TEC‐derived linker generated relatively more compact conformations with the average R _g slightly below 19 Å. Clustering analysis was done over all collected MD trajectories, regardless of the chemical nature of the linker. For the ten clusters obtained, SAXS profiles were computed for the representative members from each cluster (Supporting Information Fig. S3) and reweighted to best fit the experimentally measured SAXS profiles of the native [Fig. 6(A)] and TEC‐derived [Fig. 6(B)] dimers. This procedure led to clusters 8, 9, and 10 as representatives for the native Ub2 ensemble with a population of 25, 45, and 30%, respectively. For the TEC‐derived K48‐linked Ub2, clusters 8 and 9 were selected with a population of 28 and 72%, respectively. As shown in Table 1, clusters 8 and 9 have a similar level of compactness with R _g values of ∼18.5 Å, while cluster 10 is more open in nature and has an R _g of 21.1 Å. Accordingly, the ensemble‐averaged R _g values for the native and the TEC‐derived dimers are estimated to be 19.31 ± 1.31 Å and 18.51 ± 1.00 Å, respectively.

Figure 6.

Comparison of MD ensembles for native and TEC‐derived K48‐ and K63‐linked Ub2. SAXS profiles are calculated via reweighting the clusters. The experimental SAXS profiles (red, blue) and calculated fits (black) are shown for the native and TEC‐derived K48‐linked Ub2 (A) and native and TEC‐derived K63‐linked Ub2 (B). Each representative member is shown with its respective percentages in the population. Hydrophobic patches centered on Ile36 and Ile44 are shown in purple and blue, respectively.

Table 1.

Fitted Populations of Native and TEC‐derived Ub2 based on Conformations Obtained from Atomistic Simulations Matched to Experimental SAXS Profiles

R g	Cluster	Fitted	Score 1a	Score 2b	Score 3c
(Å)		Pop. (%)	(χ 1)	(χ 2 2<1E‐03)	(χ 3 2<20)
		Native K48 Ub2
18.45 ± 1.24	8d	25	1.3	0.8E‐04	1.2
18.53 ± 0.90	9d	45	2.0	0.5E‐04	2.8
21.17 ± 1.79	10d	30	1.8	2.2E‐04	6.8
		Non‐native K48C Ub2
18.45 ± 1.24	8d	28	1.0	3.6E‐04	1.3
18.53 ± 0.90	9d	72	0.9	2.2E‐04	0.7
21.17 ± 1.79	10d	–	–	–	–
		Native K63 Ub2
17.56 ± 0.78	3e	30	3.4	1.5E‐02	11.7
18.77 ± 0.98	7e	24	3.1	1.2E‐02	2.9
19.87 ± 2.03	9e	12	3.3	1.2E‐02	2.3
23.69 ± 1.65	10e	34	3.5	1.1E‐02	14.7
		Non‐native K63C Ub2
17.56 ± 0.78	3e	–	–	–	–
18.77 ± 0.98	7e	73	1.7	7.3E‐02	3.3
19.87 ± 2.03	9e	–	–	–	–
23.69 ± 1.65	10e	27	2.7	4.4E‐02	9.3

^aEquation (2).

^bEquation (3).

^cEquation (4).

^dMD clusters from K48‐Ub2 simulations.

^eMD clusters from K63‐Ub2 simulations.

MD simulations for K63‐linked Ub2 also show a wide range of conformational sampling. The simulations were initiated from extended structures 2JF524 and 3H7P25 with R _g values ∼23 Å, which converged to more compact structures. From the ten clusters for K63‐linked Ub2, clusters 3, 7, 9, and 10 comprise the best conformational ensemble for native K63‐Ub2 with populations of 30, 24, 12, and 34%, respectively. Clusters 3, 7, and 9 are relatively compact while cluster 10 is more extended (Table 1). R_g values for the selected clusters range from 17.6 to 23.7 Å. Clusters 7 and 10 are sufficient to represent the best fit conformational ensemble for TEC‐derived K63‐linked Ub2. The calculated ensemble‐averaged R _g values for native and non‐native Ub2 are 20.21 ± 0.70 Å and 20.10 ± 0.45 Å, respectively.

As shown in Table 1 (and SI Tables S8, S9) the general quality of fitting to the experimental SAXS profiles is similar regardless of the scoring function. Scoring functions 2 and 3 favor the MD ensembles, while scoring function 1 slightly favors the ensembles determined using the experimental structural library. For both K48‐ and K63‐linked Ub2 conformational ensemble analyses, the MD‐derived ensembles and ensembles obtained from extant structures vary in terms of the precise R _g values and relative orientations of Ub monomers. However, the overall shapes and composition of the ensembles are similar, reflective of similar SAXS scattering profiles for native and TEC‐derived Ub2.

Steady‐state kinetic analyses of Ub dimer hydrolysis by deubiquitinases

The SAXS patterns for native and non‐native Ub2 indicate that these dimers adopt similar conformations. However, it is possible that introduction of the unnatural linkage can affect recognition and hydrolysis by DUBs. To address this point, we compared hydrolysis of native and non‐native Ub2 by representative members from three different DUB families. Of the five DUB families, three are known to selectively dismantle Ub chains: USPs, JAMMs, and OTUs.12 We sought to investigate whether the selectivity and efficiency of these DUBs are preserved with non‐native substrates. We selected the DUBs USP15, AMSH, OTUB2, and OTUD7B as representative members of each family. To expand the scope of our analysis beyond K48‐ and K63‐linked Ub2, we also assessed the hydrolysis of K6‐linked and K11‐linked Ub2. Collectively, these Ub chain types are the most abundant Ub signals in asynchronous eukaryotic cells.44, 45, 46, 47

USPs are known to remove mono‐Ub from protein substrates, but a handful of DUBs from this family can also dismantle Ub chains in vitro with low selectivity.48 Using USP15, we assessed whether this type of DUB can differentiate between the native and TEC‐derived isopeptide linkage [Fig. 7(A)]. For hydrolysis of native and TEC‐derived K6‐linked Ub2, K _m and k _cat values were two‐fold higher for the native Ub2, translating to the same specificity constant (k _cat/K _m) on the order of 10³ M⁻¹ s⁻¹ for both native and TEC‐derived substrates [Fig. 7(C)]. With K48‐ and K63‐linked Ub2, a two‐fold higher K _m was observed with the native substrates while there was no significant difference in k _cat values. Overall, native and TEC‐derived K48‐ and K63‐linked Ub2 were similarly hydrolyzed by USP15, with specificity constants on the order of 10⁴ M⁻¹ s⁻¹ [Fig. 7(C)]. Hydrolysis of K6‐linked Ub2 is an order of magnitude lower compared to the other two linkages tested; both native and TEC‐derived substrates followed the same trend.

Figure 7.

Hydrolysis of native and TEC‐derived Ub2 by USP15 and AMSH. Michaelis–Menten analysis for hydrolysis of native (circles) and TEC‐derived (squares) Ub2 by (A) USP15, and (B) AMSH. (C) List of kinetic parameters measured for all experiments.

AMSH is a JAMM DUB known for specifically cleaving K63‐linked Ub chains. To direct linkage specificity, AMSH interacts with the surface surrounding K63 in the proximal Ub, which is linked to the C‐terminus of the distal Ub. Upon measuring the kinetics of AMSH‐catalyzed reactions, we find that the parameters are essentially the same for both dimer types, with k _cat/K _m on the order of 10² M⁻¹ s⁻¹ [Fig. 7(B)]. This relatively low catalytic efficiency is consistent with reported kinetic parameters for the AMSH catalytic domain.49 the kinetic parameters, K _m and k _cat, are similar for the hydrolysis of native and non‐native K63‐linked Ub2 suggests interactions important for both substrate binding and catalysis are preserved with non‐native Ub2.

OTUB1 is an OTU DUB known for specifically cleaving K48‐linked Ub chains.50 For this particular DUB, the rate of hydrolysis of TEC‐derived K48 linkage is slower compared to its native counterpart. To assess whether this was the case for all OTU DUBs, we also investigated the activity of OTUB2 and OTUD7B. OTUB2 is known to preferentially hydrolyze K63‐linked Ub chains,50 and our kinetic analysis also reports the same trend [Fig. 8(A)]. When OTUB2 was tested with native and TEC‐derived K48‐linked and K63‐linked Ub2, similar K _m values were observed for all four substrates and higher k _cat values were observed for K63‐linked over K48‐linked substrates [Fig. 8(C)]. The difference in k _cat values is four‐fold among native substrates and two‐fold with non‐native substrates. With both native and TEC‐derived substrates, the enzyme is able to preferentially hydrolyze K63‐linked Ub2 with higher efficiency than K48‐linked Ub2.

Figure 8.

Hydrolysis of native and TEC‐derived Ub2 by OTUB2 and OTUD7B. Michaelis–Menten analysis for hydrolysis of native (circles) and TEC‐derived (squares) Ub2 by (A) OTUB2, and (B) OTUD7B. (C) List of kinetic parameters measured for all experiments.

OTUD7B is an OTU DUB known to preferentially hydrolyze K11‐linked Ub chains.51 The k _cat and K _m values for native K11‐linked Ub2 were four to five fold higher compared to the TEC‐derived Ub2. However, the changes in both k _cat and K _m resulted in comparable specificity constants (k _cat/K _m) for hydrolysis of native and non‐native K11‐linked substrates [Fig. 8(B,C)]. To evaluate whether OTUD7B maintains this preference for the K11‐linkage with TEC‐derived substrates, we assessed hydrolysis of K48‐ and K63‐linked Ub2 [Fig. 8(B)]. A 10‐fold increase in enzyme concentration was necessary to induce appreciable hydrolysis of K48‐ and K63‐linked Ub2. For both native and TEC‐derived substrates, a nominal two‐fold higher k _cat/K _m is observed in hydrolyzing K63‐linked Ub2 over K48‐linked Ub2. More interestingly, for both native and TEC‐derived Ub2, k _cat/K _m dropped from 10⁴ M⁻¹ s⁻¹ with K11‐linked Ub2 to 10³ M⁻¹ s⁻¹ with K48‐ or K63‐linked Ub2. This preference for K11‐linked Ub2 can mainly be attributed to differences in k _cat values (40‐fold for native Ub2, 8‐fold for non‐native Ub2).

Discussion

Chemical strategies to generate Ub chains allow excellent control over chain length and regioselectivity, both of which are required for rigorous investigation of Ub chain structure and function. Whenever conjugation chemistry generates a non‐native linkage in place of the native isopeptide linkage, it is important to assess whether the chemical surrogate can direct Ub chain behavior in a native‐like manner. To evaluate the extent to which the TEC‐derived non‐native Gly‐Nε‐homothiaLys linkage mimics the native isopeptide bond, we compared both linkage types in the context of Ub dimers.

We used SAXS to assess the conformations of native and TEC‐derived K48‐linked Ub2 and K63‐linked Ub2 in solution. Because SAXS is a relatively low‐resolution technique, it is not possible to derive absolute structural information from SAXS scattering data alone. However, SAXS profiles can be compared to evaluate structural variability. Indeed, SAXS profiles calculated from libraries of extant Ub2 structures and atomistic molecular dynamics simulations show a range of overall shapes, reflecting the known structural diversity. Upon inspecting the experimental SAXS profiles for native and TEC‐derived K48‐linked Ub2, as well as for K63‐linked Ub2, it was evident that the average conformations for the dimers in solution do not match the most compact or extended possible structures. This finding is consistent with recent solution‐phase NMR and FRET measurements.35, 36 Computational studies with both K48‐ and K63‐linked Ub2 and a coarse‐grained potential function also reveal a multi‐state energy landscape with a number of intermediate conformers in addition to the most compact and extended structures.37

Integrating SAXS data with various sets of conformers enables further comparison of native and TEC‐derived Ub2. Using an existing structural library and MES, we find that not only are the average structures for native and TEC‐derived Ub2 quite similar, but there is also a strong resemblance in the two‐ and three‐state ensembles. Fitting representative structures from the MD trajectories to the experimental SAXS profile leads to slightly different structural ensembles compared to the MES/experimental structural library combination, especially in terms of the relative orientations of the monomers. This is not surprising due to the resolution constraints of SAXS measurements. Importantly, conformational ensembles generated from matching MD clusters to experimental scattering data are also similar for native and TEC‐derived Ub2.

Kinetic investigations of DUB activity have uncovered similar reactivity profiles between native and non‐native dimers. AMSH is a K63 linkage‐selective DUB that engages oligomers through multiple contacts within the proximal and distal units of Ub2. Upon comparing experimental SAXS profiles for native and TEC‐derived K63‐linked Ub2 with the calculated SAXS profile for K63‐linked Ub2 cocrystalized with AMSH (PDB: 2ZNV), both native and TEC‐derived Ub2 showed a moderate 70% P(r) overlap with the reported structure. 2ZNV captures an extended conformation of K63‐linked Ub2, with R _g values 24–25 Å, whereas the R _g values for free K63‐linked Ub2 in solution is smaller, 20.5 Å for native and 21.8 Å for TEC‐derived. As native and TEC‐derived K63‐linked Ub2 produced equivalent kinetic parameters, it suggests that they are both able to make the necessary contacts with AMSH and adopt the conformation needed for hydrolysis.

The OTU subfamily of DUBs is also known to be linkage selective. For example, OTUB1 prefers to hydrolyze K48‐linked Ub chains, OTUB2 prefers K63‐linked chains, and OTUD7B prefers K11‐linked chains. When OTUB1 was tested with native and TEC‐derived K48‐linked Ub2, the native Ub2 was hydrolyzed faster than the TEC‐derived substrate (data not shown). Because OTUB1 could not accommodate the non‐native linkage, either due to different recognition of the dimer structure or just of the linkage itself, we expanded our analysis to other OTU DUBs, including OTUB2 and OTUD7B. The catalytic domain of OTUB2 shares 58% sequence identity with that of OTUB1. OTUB2 is known to preferentially hydrolyze K63‐linked Ub chains, and our kinetic analysis confirms this trend for both native and TEC‐derived substrates. OTUB2 interacts with K48‐ and K63‐linked Ub2 similarly, and achieves selectivity for K63‐linked Ub2 in a k _cat dependent manner. Unlike OTUB1, OTUB2 processes native and non‐native Ub2 substrates with similar efficiency and selectivity.

We also evaluated cleavage activity of OTUD7B with native and TEC‐derived K11‐, K48‐ and K63‐linked Ub2. Similar to observations with OTUB2, k _cat is the major determinant for substrate selectivity. It is possible that selectivity arises from a necessary conformational change, in substrate and/or enzyme, which positions the isopeptide linkage into the proper orientation for catalysis. Because hydrolysis is best achieved with K11‐linked Ub2, both native and TEC‐derived, the positioning of this Ub2 is ideal for OTUD7B recognition and catalysis.

A similar k _cat effect was also observed with USP15, which interacts with K6‐, K48‐ and K63‐linked Ub2 with similar efficiency but shows relatively slower hydrolysis of K6‐linked Ub2. The recognition of K6‐linked Ub2 is unfavorable enough to lower k _cat/K _m, but does not significantly ablate hydrolysis. Consequently, unlike OTUB2 and OTUD7B, USP15 is promiscuous toward different Ub2 substrates and demonstrates a relatively subtle preference for K48‐ and K63‐linked Ub2. The same substrate preference was observed for both native and TEC‐derived substrates, suggesting that the non‐native substrates are efficacious mimics of their native counterparts.

Our comparative analyses of native and TEC‐derived Ub2 demonstrate that the non‐native Ub2 can be used to investigate DUB activity. As analogs of other post‐translational modifications, such as phosphorylation and glycosylation, have enabled leaps in mechanistic understanding in their respective fields, non‐native Ub chains have a promising role in Ub research. When possible, new discoveries acquired with non‐native Ub chains should be confirmed with the relevant native chains to better capture the molecular details influencing Ub functions.

Materials and Methods

Synthesis of non‐native and native Ub dimers

To generate non‐native Ub2 via thiolene coupling (TEC),15 Ub‐KxC‐D77 (1mM), Ub harboring a C‐terminal allylamine adduct (Ub‐AA, 1mM) and lithium acyl phosphinate (LAP) (0.5 mM) were combined in 250 mM NaOAc buffer pH 5. The reaction mixture (1.8 mL total) was cooled to 4°C and irradiated with 365 nM light for 30 min using an OmniCure series 1500 light source placed 15 cm above the sample. Native Ub2 derivatives were generated using human E1 Ub‐activating enzyme (0.1 μM) and linkage specific E2 conjugating enzymes as previously reported.3, 4, 5 For both the TEC and enzymatic reactions, the products were purified using a Superdex 75 HiLoad size exclusion column at a flow rate of 0.2 mL min⁻¹, collecting 2.5 mL fractions over 0.9 column volumes.

SEC‐SAXS measurements and data processing

SEC‐SAXS (size‐exclusion chromatography—small angle X‐ray scattering) experiments were performed at BioCAT (beamline 18‐ID, Advanced Photon Source at Argonne National Labs).52 The camera included a focused 12 KeV (1.03 Å) X‐ray beam, a 1.5‐mm quartz capillary sample cell, a sample to detector distance of ∼2.5 m, and a Mar165 CCD detector. The q‐range sampled was ∼0.0065–0.3 Å⁻¹. To ensure sample monodispersity, we used an in‐line SEC setup, which included an AKTA‐pure FPLC unit and a Superdex‐200 10/300 GL column (GE Healthcare Life Sciences). The column outlet was directly connected to the SAXS sample cell. Two second exposures were collected every 6 s during the gel‐filtration chromatography run. Samples were analyzed at room temperature in 10 mM Tris buffer pH 7.5, 50 mM NaCl, 1 mM DTT, and 1 mM EDTA. Exposures before and after the elution of the sample were averaged and used as the buffer curve, and the exposures during elution (coincident with the UV peak on the chromatogram) were treated as protein plus buffer curves. Data were corrected for background scattering by subtracting the buffer curve from protein plus buffer curves. P(r) functions and R _g values were determined from the scattering data using PRIMUS and GNOM from the ATSAS software package.53

Comparison of experimental SAXS profiles to calculated profiles generated from reported structures

PyMOL was used to determine root‐mean‐square‐deviation (RMSD) of all atomic positions between all reported K48‐ and K63‐linked Ub2 structures.54 The FoXS (Fast open‐source X‐ray Scattering) server was used to calculate and compare SAXS scattering profiles for the library of reported structures.55, 56 SAXS scattering profiles were computed based on the Debye Formula:

$$I\left(q\right)=\sum _{i=1}^N\sum _{j=1}^N{f}_i\left(q\right)f_j\left(q\right)\frac{\mathrm{s}\mathrm{i}\mathrm{n}\left(q{d}_{ij}\right)}{q{d}_{ij}}$$ (1)

The intensity, I(q), is a function of the momentum transfer, q; f _i (q) and f _j (q) represent the isotropic atomic form factors of the atoms i and j; d _ij is the distance between atoms i and j, and N is the total number of atoms in the molecule. SAXS profiles were computed with default settings assigned for the form factors—hydration layer, implicit hydrogens and excluded volume adjustments. Maximum q value was adjusted to match experimental data, q _max = 0.25 Å^−1, while profile size was left at a default value of 500. To assess the extent to which SAXS can be used to qualitatively distinguish different structures, we compared normalized P(r) curves generated from the library of K48‐linked Ub2 structures. One hundred percent overlap in the P(r) curves of two structures corresponded to identical shapes.

Experimental SAXS profiles were compared to FoXS computed profiles of reported structures. The quality of fit was evaluated by the χ ₁ function:

$${\chi }_1= \sqrt{\frac{1}{M}\sum _{i=1}^M{\left(\frac{I_{\mathrm{e}\mathrm{x}\mathrm{p}}{(q}_{\mathrm{i}})-cI(q_{\mathrm{i}})}{\sigma \left(q_{\mathrm{i}}\right)}\right)}^2}$$ (2)

M is the number of points in the profile, c is the scaling parameter and σ is the experimental error. The linear least squares minimization was performed to find the value of c that led to a low χ value, corresponding to a good fit of computed profile to experimental data. Minimal Ensemble Search (MES) was applied to determine two‐ and three‐states ensembles that best match experimental data.57 The genetic algorithm searched for ensembles from the library of reported structures, and results were ranked according to their χ ₁ values. The multi‐conformational scattering profile for each ensemble is a weighted average of the individual profiles.

Atomistic simulations and comparison of calculated and experimental SAXS profiles

As an independent structural characterization approach, atomistic simulations were carried out for Ub2. To facilitate sampling, independent MD simulations for Ub2 with a native K48‐linker started with four experimental structures that differ in the degree of compactness: 2O6V (R _g ∼ 16.4 Å),58 2PE9 (R _g ∼ 15.8Å),18 3AUL (R _g ∼ 16.7 Å),20 and 2KDF (R _g: 21.4–23.9 Å).19 Similarly, MD simulations for Ub2 with native K63‐linker started with 2JF5 (R_g ∼ 22.7Å),²⁴ and 3H7P (R_g ∼ 23.1Å).²⁵ Moreover, dimers bearing non‐native linkages were also simulated in which Gly‐Nε‐homothiaLys (thiol‐ene) was introduced at position‐48 or position‐63 of one Ub subunit to connect with the C‐terminus (Gly76) of another. With these different models, MD simulations were performed on the μs time‐scale to ensure sampling a diverse range of conformations. The Amber v12 Molecular Dynamics package59, 60, 61, 62, 63 was chosen for the simulations since the implicit solvent model implementation on a GPU affords a 100‐fold speed enhancement compared to a single CPU. Specifically, the ff14SBonlysc Amber force field was chosen64, 65, 66, 67 along with gb8 as the implicit solvent (Generalized Born) model.64 Production gb8 simulations were carried out for a minimum of 800 ns using a 1 fs time step (see Supporting Information Table SV for a summary). Langevin dynamics were used with a collision frequency of 20 ps⁻¹ at 300 K. The SHAKE algorithm was applied to bonds with hydrogen atoms with a tolerance of 10⁻⁵ Å.68 The non‐bonded cutoff was set at 9999 Å, and the maximum distance between atom pairs (rgbmax) for Born radii calculations was maintained at 12 Å. Debye–Hückel theory was applied to treat salt (implicitly) and the concentration was set to a physiologically relevant concentration (0.15 M).

The MD trajectories for different Ub2 systems were combined and then analyzed with a clustering protocol using the k‐clust command in MMTSB package69 with Cα‐RMSD as the similarity measure. SAXS profiles for each cluster centroid were calculated using Fast SAXS70, 71 and scored with respect to the experimental SAXS profile. SAXS fitting was carried out with two additional scoring functions that have been used in previous computational analyses72, 73, 74, 75; with both, a lower score means a better match. The second scoring function70 is given as,

$${\mathrm{\chi }}_2^2=\sum _{j=1}^K\frac{1}{\delta {\hspace{0.17em}\mathrm{l}\mathrm{o}\mathrm{g}\hspace{0.17em}I}^2{(q}_j)}{(\log I_{\mathrm{c}\mathrm{a}\mathrm{l}\mathrm{c}}\left(q\right)-\log I_{\mathrm{e}\mathrm{x}\mathrm{p}}\left(q\right)-\Delta )}^2$$ (3)

where q _min and q _max are the lower and upper limits of the q‐range from the experimental scattering profile I _exp(q) and δlog I(q) are the experimental uncertainties of log I _exp(q). The value of Δ corresponds to the offset between the theoretical and experimental SAXS profiles (log I _cal and log I _exp) at q = q _min. A good fit should have a score lower than 0.001.70, 72 The third scoring function is defined as the following,56

(4)

where K is the number of experimental points, σ(q) are the standard deviations, and μ is a scaling factor. A good fit should have a score lower than 2.000.56

With the computed χ ₃ ² values, optimal weights for different cluster centroids are computed based on the best fit of the ensemble‐averaged SAXS profile to the experimental data for q up to 0.20 Å⁻¹; the results are summarized in Table 1.

Steady‐state kinetic analyses of DUB‐catalyzed cleavage reactions

Stock solutions of DUBs and Ub dimers were prepared in a 50 mM Tris‐Cl buffer at pH 7.5 containing 150 mM NaCl and 5 mM DTT. Kinetic assays were performed by varying the concentration of Ub dimers while maintaining a constant concentration of the indicated DUB (0.2 µM of USP15, 0.2 µM and 2 µM of OTUD7B, 0.5 µM of AMSH and 0.74 µM of OTUB2) at 37°C. Reactions were quenched by addition of 6X Laemmli sample loading buffer. To visualize monoUb formation, the quenched reactions were run on a 15% denaturing SDS‐PAGE gel along with monoUb standard, followed by SYPRO® Ruby staining. ImageJ was used to quantify the appearance of monoUb over time. Experimental data were processed using Prism 5.02 (GraphPad Software).

Supporting information

Supporting Information