Activation and selectivity of OTUB-1 and OTUB-2 deubiquitinylases

The ovarian tumor domain (OTU) deubiquitinylating cysteine proteases OTUB1 and OTUB2 (OTU ubiquitin aldehyde binding 1 and 2) are representative members of the OTU subfamily of deubiquitinylases. Deubiquitinylation critically regulates a multitude of important cellular processes, such as apoptosis, cell signaling, and growth. Moreover, elevated OTUB expression has been observed in various cancers, including glioma, endometrial cancer, ovarian cancer, and breast cancer. Here, using molecular dynamics simulation approaches, we found that both OTUB1 and OTUB2 display a catalytic triad characteristic of proteases but differ in their configuration and protonation states. The OTUB1 protein had a prearranged catalytic site, with strong electrostatic interactions between the active-site residues His265 and Asp267. In OTUB2, however, the arrangement of the catalytic triad was different. In the absence of ubiquitin, the neutral states of the catalytic-site residues in OTUB2 were more stable, resulting in larger distances between these residues. Only upon ubiquitin binding did the catalytic triad in OTUB2 rearrange and bring the active site into a catalytically feasible state. An analysis of water access channels revealed only a few diffusion trajectories for the catalytically active form of OTUB1, whereas in OTUB2 the catalytic site was solvent-accessible, and a larger number of water molecules reached and left the binding pocket. Interestingly, in OTUB2, the catalytic residues His224 and Asn226 formed a stable hydrogen bond. We propose that the observed differences in activation kinetics, protonation states, water channels, and active-site accessibility between OTUB1 and OTUB2 may be relevant for the selective design of OTU inhibitors.

Ubiquitinylation is a post-translational modification and involves one or more covalent additions to the lysine residues of the target protein that is mediated by sequential action of ubiquitin ligases E1 (ubiquitin-activating enzyme), E2 (ubiquitinconjugating enzyme), and E3 (ubiquitin ligase). It can be reversed by removal of ubiquitin (deubiquitinylation) catalyzed by ubiquitin hydrolases or deubiquitinylating enzymes (DUBs) 2 (1). DUBs play an essential role in the regulation of cellular processes like apoptosis, putative cell growth, cell cycle control, and DNA repair (2,3). 102 DUBs were identified and can be grouped into five subfamilies (4,5). The 18 OTUs identified in human can further be grouped into four categories: OTUBs, OTUDs, A20s, and OTULINs (6,7). Cysteine proteases, such as ovarian tumor deubitquitinylases Otubain-1 (OTUB1) and Otubain-2 (OTUB2) belong to the ovarian tumor protease (OTU) subfamily of DUB proteins (8) and were the first two OTU proteins to show DUB activity (8,9). Elevated OTUB1 expression is associated with glioma (10), esophageal squamous cell carcinoma metastasis (11), endometrial cancer (12), hepatocellular carcinoma (HCC) (13), ovarian cancer (14), colorectal cancer (15,16), breast cancer (17), liver cancer (13), and gastric cancer (14). A recent study by Li et al. (18) have identified the role of OTUB2 and its overexpression in non-small cell lung cancer tissues. Several studies have described the therapeutic potential of DUBs for cancer treatment.
OTUB1 and OTUB2 belong to the cysteine protease class with a sequence identity and similarity of 48 and 70%, respectively. A 37-residue-long N-terminal region in OTUB1 is not present in OTUB2. Despite the considerable sequence conservation and structural overlap, there is perceivable substrate specificity. OTUB1 has a slower cleavage kinetics and favor Lys 48 -linked polyubiquitin (19), whereas OTUB2 cleaves differently linked polyubiquitin chains.
Cysteine proteases have a Cys-His-Asn/Asp catalytic triad as the active site. When the cysteine sulfur atom is deprotonated, the histidine residue acts a base and becomes positively charged. The thiolate's high nucleophilicity leads to an attack of the carbon of the substrate peptide bond to produce a tetrahedral intermediate; then the active-site histidine acts as a proton donor to release the N-terminal fragment of the substrate (20). The intermediate is further stabilized by hydrogen bonding between the substrate oxyanion and a conserved glutamine residue (21). However, this catalytic triad is performing the proteolytic activity, and the protonation states (neutral or zwitte-rionic form) of cysteine and histidine are highly debated (22)(23)(24)(25). The third residue (Asp/Asn) of the catalytic triad is required for a correct alignment and polarization of catalytic histidine (5). The Asp in the catalytic triad plays an role in the charge relay systems and is involved in the stabilization of the imidazolinium intermediate and the positioning of the catalytic site histidine (26). The role of Asn in the catalytic triad is to orient the side chain of the catalytic histidine to its optimum position for various steps of the catalytic mechanism (27).
OTUB1 in the absence of ubiquitin (crystal structure PDB code 2ZFY) is catalytically incompetent because of long distances between the catalytic triad residues histidine (His 265 ) and cysteine (Cys 91 ) and aspartic acid (Asp 267 ) of 5.5 and 4.6 Å, respectively (19). Access to the active site of OTUB1 is blocked by Glu 214 of the ␣9-␣10 loop and forms a backbone hydrogen bond with the catalytic Cys 91 . The surface charge property of the ubiquitin-binding site (PЈ side) of OTUB1 is different from OTUB2 located within the close proximity of Cys 91 , which sterically restricts active-site access. Conformational changes and specificity binding studies (19) show that OTUB1 and OTUB2 may adopt two conformational states and display different lysine linkage specificities, probably because of structural differences in their catalytic centers. Although OTUB1 and OTUB2 are close homologs (8), OTUB1 is highly specific for Lys 48 cleavage, whereas OTUB2 has a broad cleavage profile (28). OTUB1 is autoinhibited in the ubiquitin-free (apo) state and is activated upon ubiquitin binding, whereas the OTUB2 is catalytically active even in the absence of ubiquitin (29). Structural analysis of OTUB1 shows the preference of ubiquitin (Ub)-charged E2 acting as an N-terminal extension of OTU (30,31) and also binding to free ubiquitin, both mimicking the configuration of the canonical product of its DUB activity, the Lys 48 -linked di-Ub chain (32).
A long N-terminal helix of OTUB1 is critical for E2-ubiquitin recognition and inhibition; OTUB2 thus fails to recognize and inhibit E2 because it lacks N-terminal helix (32). The proximal ubiquitin binds with OTUB2 in a different manner because of the striking structural differences in the N-terminal region (absence of N-terminal helix in OTUB2), and several reports (7,32,33) show that the absence of N-terminal helix may be the reason for the nonselective deubitquitinylase activity of OTUB2. The short loop (residues 44 -48) preceding the activesite helix different from all other cysteine proteases and in the crystal structure is shifted toward the catalytic residues to generate a spatially restricted active site and stabilize the oxyanion intermediate (29). Interestingly, in OTUB2 only subtle changes were observed in the catalytic residues, but larger conformational changes can be seen in the disordered region 197-204 (turned into ordered region upon ubiquitin binding), helices 1 and 2, and ␤-sheets 3 and 4 (28).
Deubiquitinylating enzymes have recently been identified as relevant cancer drug targets (34 -36). Molecular simulations have become essential tools in identifying protein conformational dynamics and rationalizing drug selectivity. For USP7 (37) two druggable binding sites were identified of which one was essential for the inactive to active state conformational transition. MD studies elucidated the open to close conformation transition in viral DUB (turnip yellow mosaic virus) (38) in WT and mutants. The drug selectivity of zinc-dependent human DUBs of the JAMM family Rpn11 and CSN5 toward capzimin and CSN5i-3 could be explained because of formation of heterodimeric protein-protein complexes (39).
In this work, we focus on OTUB1 and OTUB2 DUBs to understand their differences in structural stability in the apo and holo forms and the role of water molecules in stabilizing the catalytic triad conformations using molecular dynamics simulation. Analyzing the conformational stability of the catalytic triad residues and the access and passage and pathways of solvent molecules helps us to understand the role of water molecules in stabilizing the active site. Our findings suggest that OTUB1 achieves its productive conformation in its charged state only upon ubiquitin binding, whereas in OTUB2 the productive conformation already prevails in its holo form conformation in the neutral protonation state. A very low number of water trajectories toward the catalytic state from the surface were observed in the productive conformation of OTUB1 in contrast to the very large number observed in the productive conformation of OTUB2. A number of conserved water molecules are located close to His 265 and Asp 267 in OTUB1 and in vicinity of Cys 51 in OTUB2.

Results and discussion
The comparison of OTUB1 and OTUB2 sequences shows 48% identity and 70% similarity (Fig. S1). Larger structural differences are apparent in the catalytic site and in the ubiquitinbinding pockets (19). A long N-terminal ␣1-helix is present in OTUB1 but missing in OTUB2. In our nomenclature, we refer to the apo OTUB1 structure in charged and neutral states as O1U 0 C and O1U 0 N, respectively; likewise the holo OTUB1 structures in charged and neutral states are O1U 1 C and O1U 1 N. Similarly, apo and holo OTUB2 charged and neutral states will be referenced as O2U 0 C/O2U 0 N and O2U 1 C/ O2U 1 N. For representational purpose, we show a zwitterionic catalytic triad (in O1U 1 C) and a neutral charge catalytic site (in O2U 0 N) in Fig. 1.

Absence of Ub: apo OTUB1 (O1U 0 C, O1U 0 N) and OTUB2 (O2U 0 C, O2U 0 N)
Molecular dynamics simulations of OTUB1 (charged and neutral) and OTUB2 (charged and neutral) were replicated three times. The RMSD (C␣) plots of all apo OTUB1 and OTUB2 simulations were analyzed and show that the structures are stable throughout the simulations (Figs. S2 and S3). In addition, the global protein RMSFs reveal no large fluctuations. Only the loop region connecting sheets 2 and 3 in OTUB2 (O2U 0 C and O2U 0 N) shows a high degree of flexibility ( Fig. S4 and S5).

Structural characteristics of the catalytic triad in ubiquitin-free OTUB1 and OTUB2
Catalytic protease activity requires an accurate positioning of the Cys/His residues to perform their catalytic function. The interatomic distances of the catalytic residues of both OTUB1 and OTUB2 were monitored. The average distances and their standard deviations are given in  (Table 1).
Apparently, there is no conformational change of the catalytic residues associated upon activation of OTUB2. The differences in structural parameters such as interatomic distances and dihedral angle are only minor and show that this is in agreement with experimental results that demonstrate the prevalence of the neutral state of apo OTUB2 to be catalytically active.

The Ub-bound protein conformation of OTUB1 and OTUB2
Simulations of the charged and neutral states of the catalytic triad residues of OTUB1 and OTUB2 in the ubiquitin-bound conformation were performed in triplicate. The RMSD (C␣) plots of all holo OTUB1 and OTUB2 show the stability of the charged and neutral states of OTUB2 but not for OTUB1.
Whereas the OTU of OTUB1 is rather rigid with a RMSD of 2 Å, it is the flexible N terminus that is stabilized by interactions with the E2-linked ubiquitin binding and responsible for the very large RMSD of 6 -8 Å in OTUB2 (Fig. S6). The RMSF analysis further confirms that the large fluctuations originate from the N-terminal helical region (first 21 amino acids of the protein), and the rest of the protein is less flexible (Fig. S7). The analysis of the RMSF further also shows local residue fluctuations in the loop connecting the helices 2 and 3 in both charge states. Less pronounced flexible regions are helices 8 and 9 in the charged state (O1U 1 C) and the loop connecting sheets 3 and 4 in the neutral state (O1U 1 N) (Fig. S7).
OTUB2, which does not possess the N-terminal helix, reveals a smaller RMSD in both charge states (O2U 1 C and O2U 1 N) during the entire simulations (Fig. S8). The analysis of the RMSFs of holo OTUB2 shows that the neutral state O2U 1 N is structurally more stable than the O2U 1 C except that the latter one shows sharp fluctuations in the short N-terminal helix and the small loop connecting sheet 5 and 6 ( Fig. S9).

Analysis of catalytic triad conformation in ubiquitin-bound OTUB1 versus OTUB2
In O1U 1 C, the average contact distance in the MD trajectories between His 265 ND1 . . Cys 91 SG is 3.3 Ϯ 0.2 Å, and the His 265 NE2 maintains a close contact distance of 2.7 Ϯ 0.1 Å with Asp 267 OD1 ( Figs    In OTUB1, throughout the total simulation time of the O1U 1 C state, the strong salt-bridge interaction between the catalytic site His 265ϩ and Asp 267Ϫ is persistent. Conformational restrictions of the active-site residues in OTUB1 are also due to Pro 87 that is located in close proximity of Cys 91 , and also the small loop before the His 265 has an effect on the Cys 91 orientation. The role of ␣9-␣10 loop to influence the conformational change on Cys 91 and His 265 upon the ubiquitin binding was only inferred from a static picture in the X-ray structure (19). A hydrogen-bonding interaction between the C-terminal end of ␣3 and the backbone of the catalytic Cys 91 was speculated to inhibit substrate access to the active site and be the reason for an OTUB1 ligand-induced activation step.

Water access of active site in Ub-free in apo OTUB1 and OTUB2
Water molecules in the active site play an important role in regulating the activity, stability and selectivity of enzymes (40 -42). The buried active sites in OTUB1 and OTUB2 are connected with the bulk solvents by water tunnels. The molecular properties of the amino acids of the water tunnels control the access of the water molecules. Studying the water tunnels is important, because it is the major route of the substrate delivery to the active-site center. The analysis of water molecules diffusing from the surface of the protein to the active site is a versatile tool to discriminate structurally similar proteins. Despite our

OTUB-1 and OTUB-2 activation and selectivity
special focus on dissecting the active state conformations O1U 1 C and O2U 0 N, we have also analyzed the complementary conformations of the OTUB1 and OTUB2 proteins in the different charge states.
In O1U 0 C, the number of water inlets is 1125, of which 471 are incoming paths and 654 are outgoing paths. The localization of water molecules in the vicinity of the active-site residues His 265 and Cys 91 is shown in Fig. 4A. A high density of water molecules is also observed close to these residues with longer times of residence caused by strong hydrogen bonding and electrostatic interactions (Fig. 4C). In O1U 0 N, the number of inlets is smaller with 776, of which 387 are incoming paths and 389 are outgoing paths. The water inlets are mostly located close to the region of His 265 and Asp 267 (Fig. 4B). The local distribution of water molecules is high near the side chain oxygen of Asp 267 , low close to His 265 , and not present in the vicinity of Cys 91 (Fig. 4D). This shows that the catalytic aspartate residue plays a vital role in trapping water molecules close to the active-site region of OTUB1.
In O2U 0 C, 260 water inlets can be identified, and the numbers of incoming and outgoing channels are almost evenly distributed with 135 and 123, respectively. In this case, the number of water molecules that enter or reside close to the active-site His 224 is high compared with two other active-site residues, cysteine and aspartate (Fig. 5A). The local distribution of water molecules is high near the His 224 side chain compared with the other active-site residue, Asn 226 (Fig. 5C). In O2U 0 N, on the other hand, 1499 inlets are observed, of which 833 are incoming paths and 666 are outgoing paths (Fig. 5B). O2U 0 N has numerous water trajectory paths with more incoming and outgoing paths compared with a small number of water molecules residing trapped in the active-site region (Fig. 5D). In O2U 0 N, the absence of the long N-terminal helix allows the opening of two large cavities located near the entrance of the catalytic triad, which act as the main route of access for the water molecules in this truncated structure. The water molecules are not densely located close to one catalytic residue but sparsely distributed around the whole catalytic triad. More specifically, analysis of the local water distribution density shows that the water molecules reside longer in the vicinity of the catalytic cysteine residue and can less frequently be found surrounding the catalytic histidine and asparagine residues.

Water access in the Ub-bound conformation of OTUB1 and OTUB2
In the catalytically active state of OTUB1 (O1U 1 C), the water access to the active site is limited with a total of just 19 inlets comprising of 10 outgoing and 9 incoming paths (Fig. 6A). In O1U 1 C, the highly fluctuating extended N-terminal helix may acts as a swinging gate and not enable many water molecules to diffuse from the surface to the active-site region and vice versa. In O1U 1 C, very few molecules enter and leave the active site; compared with the number of incoming and outgoing water molecules' trajectories, more water molecules seem to be trapped inside the catalytic triad. This can be attributed to the strong salt bridge and stable interactions observed in the charged catalytic site state and also the presence of the long N-terminal helix. The local distribution density of water molecules enables the identification of essential residues capable of trapping water molecules, and it can also be used for conserved water identification or regulation of ligand access (43). In O1U 1 C, a very low water distribution density is found. In O1U 1 C, the water molecules are localized around the catalytic residues His 265 and Asp 267 , especially close to His 265 (Fig. 6C). In contrast to the charge state of OTUB1, in the neutral state (O1U 1 N), the total number of inlets is larger with 361, of which almost equal halves of them are incoming and outgoing paths (Fig.  6B). Analysis of the water residence time distribution shows that the water molecules are trapped far from the catalytic triad and remain in that pocket during the entire simulation. Although O1U 1 N has a larger number of water sites compared with O1U 1 C, they are not located in the vicinity of the catalytic sites, and hence the access to the catalytic site is restricted. The water distribution density also shows that during the simulations the water molecules are found distant from the catalytic sites, and only very few approach the catalytic cysteine residue (Fig. 6D).

OTUB-1 and OTUB-2 activation and selectivity
In the charged state of the Ub-bound conformation of OTUB2 (O2U 1 C), 586 water inlets were identified, of which 307 are incoming and 279 are outgoing (Fig. 7A). The local distribution density of water molecules is higher for the catalytic residue histidine (His 224 ) than for the other two residues (Fig. 7C). In the neutral state of OTUB2 (O2U 1 N), of 896 water inlets identified, the numbers of incoming and outgoing paths are equally distributed with 450 and 446, respectively (Fig. 7B). The local water distribution density is not high close to the catalytic residues but at a certain distance from the active-site histidine and asparagine residues and remote from the active-site cysteine (see Fig. 7D).

Comparative analysis of water trajectories of OTUB1 in Ub-free versus Ub-bound protein conformations
The comparison of the OTUB1 zwitterionic state in the ubiquitin-free (apo) and ubiquitin-bound (holo) protein conformations shows that there are thousands of water inlets, and the outgoing paths are larger in number in the free protein. The holo state shows fewer water inlets with just ϳ20, which are equally distributed among the incoming and outgoing. If you compare the local water distribution density, both apo and holo conformations show a preferred water distribution density close to the catalytic histidine and cysteine 91 in the apo form and aspartate 265 in the holo form.
As an aside, the speculative neutral protonation states of OTUB1 in the apo and holo protein conformations show very different results. In the apo state, ϳ780 inlets are equally distributed between incoming and outgoing paths. In the holo state, there is a larger number of water inlets compared with the charged state; there are ϳ360, which are equally distributed as in the apo state. The major site of localization is remote from the catalytic region and not significant for the stabilization of the catalytic triad residues.

Comparative analysis of water trajectories of OTUB2 in Ub-free versus Ub-bound protein conformations
The comparison of charged states of apo and holo structures of OTUB2 shows a significant difference between Ub-free and Figure 5. Analysis of water access to the active site in ubiquitin-free OTUB2 structures. Top panels, water inlets to the catalytic site for incoming and outgoing water molecules. Bottom panels, local distribution density of water molecules to identify stable water molecules. A and C, charged states of catalytic triad residues O2U 0 C. B and D, neutral charge state of catalytic triad residues O2U 0 N.

Figure 6. Water trajectory analysis of ubiquitin-bound OTUB1 structures.
Top panels, water inlets to the catalytic site for incoming and outgoing water molecules. Bottom panels, local distribution density of water molecules to identify stable water molecules. A and C, charged state of catalytic residues O1U 1 C. B and D, neutral state of active-site O1U 1 N.

OTUB-1 and OTUB-2 activation and selectivity
Ub-bound protein conformations. In the apo conformation, only half of the water inlets are present compared with the holo conformation. In the apo state (O2U 0 C), 260 water inlets are almost equally distributed among incoming and outgoing channels (only slightly less outgoing paths), and the large number of 586 water inlets in the holo form (O2U 1 C) are also almost equally distributed among the incoming and outgoing (slightly less in outgoing). In both cases, the local water distribution density is high in the catalytic site.
The same observation can also be made for the neutral states. In the OTUB2 apo conformation (O2U 0 N), of the total 1500 water inlets, there are more (833 inlets) incoming and outgoing (666 inlets) identified. However, in the holo conformation (O2U 1 N), ϳ900 inlets are almost equally distributed between incoming and outgoing. Interestingly in the Ub-free conformation, the local water distribution density is high in the vicinity of the side-chain atoms of all three catalytic residues, whereas as in the holo conformation state, it is remote from the active site.

Catalytically active forms OTUB1 (O1U 1 C) versus OTUB2 (O2U 0 N)
OTUB1 is catalytically active only when ubiquitin is bound, which points to a substrate-induced activation process. The active site in OTUB1 is not competent for catalysis in its apo state because the short loop connecting ␤4-␤5 and the loop between helices 9 and 10 have an influence on the catalytic cysteine orientation and mediate the conformational change upon ubiquitin binding. Based on the careful analysis of catalytic site geometry from our molecular dynamics simulations, we infer that the zwitterionic state of OTUB1 in the Ub-bound conformation (O1U 1 C) represents the catalytically active state.
In contrast, OTUB2 is already active in the Ub-free apo state, and the active-site residues are in a prearranged, catalytically active conformation. However, according to our results, we stress that the neutral charge state of OTUB2 (Cys-SH, His 0 , Asn 0 , Asp Ϫ ) is in a catalytically active conformation (O2U 0 N), and only minor proton transfer events are required to initiate the protease activity.
The comparison of the water trajectories between the catalytically active OTUB states, however, shows that there are significant differences between the OTUs. OTUB2 (O2U 0 N) possesses a huge number of water inlets compared with OTUB1 (O1U 1 C), and the clustering of water molecules is close to the catalytic site region in OTUB2 but remote from the catalytic site in OTUB1. The preferred sites of localization are all three catalytic residues in OTUB2 but only histidine and aspartate in OTUB1. The inner pocket volumes calculated based on the water molecule distribution densities reveal a smaller pocket of just 62 Å 3 in OTUB1, whereas in OTUB2 it is nearly 6-fold larger with a volume of 362 Å 3 (Fig. 8).

Water-mediated interactions in the catalytic triad residues
After the completion of pathway analysis of water molecules and its distribution in the active-site region, we also analyzed the hydrogen-bonding interactions of the conserved water molecules with the key catalytic residues. We only mention the water-mediated hydrogen bond interactions between the catalytic residues for the productive conformations of OTUB1 and OTUB2.
In O1U 1 C, in only 1.4% of the simulation time, direct watermediated interactions between the ND1 of His 265 and the sulfur atom of cysteine residue (Cys 91 ) can be observed (Fig. 9). No extended water-mediated stabilizations of two or more water molecules are found. The NE2 of His 265 and OD1 of Asp 267 show water-mediated interactions for ϳ11% of the simulation time. For only 1% of the simulation time, extended water-mediated contacts are observed. In O1U 1 C, the salt-bridge interaction between the positively charged His 265 and the negatively charged Asp 267 is stable and likewise a hydrogen bond interaction between the proton at the NE2 atom of His 265 and OD1 atom of Asp 267 . The water molecule also makes a hydrogen bond interaction with OD2 atom of Asp 267 and backbone oxygen atom of the His 265 (Fig. 10A).
In the case of O2U 0 N, ND1 of His 224 and the sulfur atom of Cys 51 show direct water-mediated interactions for 11.4% of the simulation time and furthermore extended water-mediated

OTUB-1 and OTUB-2 activation and selectivity
contacts for nearly 2% of the simulation time. Similarly, the NE2 of His 224 has direct and extended water-mediated interaction for nearly 9 and 4.2%, respectively, of the simulation time with OD1 of Asn 226 (Fig. 9).
In case of O2U 0 N, the water molecule mediates the interaction between histidine and asparagine only when the distance between these residues increases sufficiently during the simulation. Then the water molecule bridges the OD1 atom of Asn 226 and NE2 atom of His 224 (Fig. 10B). If the distance further increases, extended water-mediated interactions can be observed in which more than one water molecule form an interaction bridging the OD1 atom of Asn 226 and the NE2 atom of His 224 (Fig. 10C). The results show that the persistent and strong electrostatic interactions between the zwitterionic cata-lytic residues in O1U 1 C and the presence of a long flexible N-terminal helix restrict and control the entry and egress of water molecules to the catalytic triad.

Conclusions
The detailed structural comparison of the zwitterionic and neutral states of Ub-free and Ub-bound protein conformations of OTUB1 and OTUB2 shows that the apo OTUB1 structure with a charged catalytic site (O1U 1 C) is in a stable but catalytically incompetent conformation. OTUB1 requires substrateinduced protease activation. A possible explanation is that OTUB1 is in inactive conformation in the absence of ubiquitin and a conformational change may induce a movement of His 265 in closer proximity of Cys 91 upon ubiquitin substrate binding.

OTUB-1 and OTUB-2 activation and selectivity
Also the DUBs USP7 (44) and USP12 crystallize in an autoinhibited state; USP12 requires an extra allosteric activation upon UAF1 and WDR20 binding (45). The allosteric binding to remote binding sites increases the catalytic rate, k cat , by a factor of 20, whereas the substrate affinity (K m ) remains almost unchanged. OTUB1 is highly selective for Lys 48 -linked polyubiquitin chains, but not cleaving Lys 63 -, Lys 29 -, Lys 6 -, or Lys 11linked polyubiquitins (31). The selectivity could be rationalized by a bidentate ubiquitin-binding model with distinct sites for catalytic activity (Cys 95 ) and a second distal binding site. Also, the role of the N-terminal extension of its core OTU domain is not fully resolved yet (32). According to our work, the long extended N-terminal helix resolved in the O1U 1 C is flexible throughout the simulation and acts as a gate of specificity regulation to control substrate and solvent access to the catalytic site.
In OTUB2, the catalytic triad is stabilized by unique hydrogen-bonding network configuration (29). The similarity in the side-chain orientations of Cys 51 and His 224 in OTUB2 to those of other cysteine protease structures implies that they are in a catalytically productive geometry even in the absence of ubiquitin binding. On the basis of sequence analysis of OTUs, the active-site catalytic triad is commonly annotated as Cys 51 / His 224 and Asp 48 , with the latter stabilizing the orientation of the protonated histidine. In the apo and holo crystal structures and in our simulations, we can show that this aspartate is too far distant to directly interact with His 224 , but Asn 226 forms hydrogen bonds with His 224 and nearby other residues. This configuration is stable in both Ub-free and Ub-bound states and requires only a minor proton transfer event from Cys 51 to His 224 without any configurational reorientation to initiate the protease activity. The role of Asn 226 in the isopeptidase reactions is corroborated by the absence of catalytic activity upon an N226A mutation. Based on the water trajectory analysis, OTUB2 is more prone to enable water diffusion and passage to and from the active site compared with OTUB1.
DUBs and OTUs have recently emerged and promising new drug targets for treating various types of cancer. The role of an asparagine residue in a stable configuration to replace an aspar-tate in cysteine proteases confirms that there are clearly distinct functional subclasses within the OTU family. The difference in protonation states of the residues of the catalytic triad (diad) plus selective ubiquitin-binding sites proximal and distal to the active site will allow the design of selective inhibitors.

Structure preparation
Crystal structures of OTUB1 and OTUB2 in the apo and co-crystallized ubiquitin-bound states are available from the PDB for OTUB1 structures in the absence of ubiquitin (code 2ZFY) and in the ubiquitin-bound state (code 4DDG); and for OTUB2 (ubiquitin-free structure, code 1TFF) and co-crystallized with the ubiquitin (code 4FJV) ( Table 2). Downloaded structures were preprocessed to add missing loops, add amino acid side chains, and revert introduced mutations (in the crystal structure 4DDG-S1091C) using Prime module (Schrodinger) and then further processed to assign bond orders, add hydrogen atoms, and identify hydrogen bond interactions using the protein preparation wizard. In the resulting structures, only hydrogen atoms were minimized using the Epik module (Schrodinger). To carry out an unbiased study, the ubiquitin molecule, the E2 ligase and other structural units and solvent molecules were removed from both OTUB1 and OTUB2 structures. The cysteine and histidine residues in the catalytic triad were prepared in neutral (protonated Cys-SH/His 0 N⑀ and charged (deprotonated Cys-S ؊ /His ϩ imidazolium cation) using the 3D builder (Schrodinger).

Molecular dynamics simulation
Every simulation was performed in triplicate of all-atom 100-ns simulations using Gromacs 2018.3 (46) and the CHARMM36 force field (47). The TIP3P water model was used, and 0.15 M of salt (Na ϩ Cl Ϫ ) was added to neutralize the system. Energy minimization was carried out using the steepest descent algorithm for 5000 steps. Equilibrations were carried out in NVT and NPT ensembles for 1 and 2 ns, respectively. Temperature (310 K) and pressure (1 atm) were controlled by the velocity rescaling thermostat and a Parrinello-Rahman barostat  (48), respectively. For the treatment of long-range electrostatic interactions, the particle-mesh Ewald summation was used. The three independent production runs for 100 ns each were carried out without any position restraints in an NPT ensemble. A time step of 2 fs was used, and the coordinates were saved every 2 ps.

Water pathway and analysis
The paths of the solvent diffusion to and from the active-site area (within a radius of 4 Å from the catalytic triad) to the surface of the protein were analyzed using AQUA-DUCT software (49). AQUA-DUCT is a tool to perform a detailed analysis of solvent access and channels in proteins. Each water molecule, which diffuses between a defined active-site sphere to the surface of the protein, is individually tracked. The raw paths connecting the surface of the protein and the active site are determined by looping over all frames of the simulation trajectories. A separate path for each water molecule that enters and leaves the surface of the protein and active-site sphere during the simulations is calculated. A clustering of the inlets (points where the traceable water molecules enters or leave the surface of the protein) is performed using different clustering algorithms. Finally, the results can be visualized using PyMOL (50). It also allows classification of pockets as "inner" or "outer" when analyzing the distribution of local water molecules on a grid spanning all paths with a default size of 1 Å. The maximal space explored by traced water molecules represents the outer pocket, whereas the inner pocket is the area, which is easily accessible by water molecules. Along the entire trajectory, the number of paths crossing the grid cells divided by number of frames gives the average density of traced residues. It can also detect high-density water points (also called "water hot spots") from the distribution of local densities on the grid, traced water molecules trapped in hydrophobic cages or attracted by favorable interactions with nearby amino acids and long residence times. The results were visualized using PyMOL software (50).

Computation of direct and water-mediated interactions in the active site
The HBonds plugin in VMD was used to calculate the hydrogen bonds along the molecular dynamics simulations (51), and the geometric criteria of donor-acceptor distance should be Ͻ3.5 Å with a donor-hydrogen-acceptor angle of Ͼ100°. A water-mediated interaction between two residues is calculated when one water molecule is bridging a hydrogen-bond donor and acceptor. An extended water-mediated interaction is defined as a water-mediated hydrogen bond formed between two residues by two bridging water molecules. Although we have analyzed the entire structure, the focus in this work is on the water-mediated interaction between the key atoms of the catalytic triad residues.

Data availability
All prepared protein structural files in PDB format and topology files to reproduce the molecular dynamics simulations with GROMACS are freely available at Zednodo (3741678).