Ubiquitination is reversed by the activity of deubiquitinase enzymes (DUBs) (Figure 1). About 100 human DUBs are known, and they are divided into seven major families: the cysteine proteases of the USP (ubiquitin-specific proteases), UCH (ubiquitin C-terminal hydrolases), OTU (ovarian tumor), MJD (Machado-Joseph domain-containing proteases), MINDY (motif interacting with the Ub-containing novel DUB family), and ZUFSP (zinc finger with the UFM1-specific peptidase domain protein) families and the Zn-dependent metalloproteases of the JAMM (JAB1/MPN/MOV34 domain-associated) family [4,16,17,18,19]. The USP family is the largest, with 58 constituents. The members of each family are related through structurally homologous catalytic domains but can vary greatly by the presence of insertions, deletions, and additional domains serving diverse functions. Cysteine protease DUBs utilize a catalytic triad or dyad to facilitate a nucleophilic attack and hydrolysis of the scissile bond connecting the C-terminal Gly-76 of Ub to a substrate or other Ub of a poly-Ub chain. Contrarily, zinc-dependent metalloprotease DUBs (metalloDUBs) utilize a coordinated zinc ion and an activated, nucleophilic water molecule for catalysis [4].

DUBs are essential for maintaining Ub homeostasis by processing Ub precursors (UBC, UBB, UBA52, and UBA80) [20] and unconjugated chains [17]; recycling Ub at the proteasome [21]; and deubiquitinating various targets. Additionally, they function as interpreters and editors of a highly complex code. E3-DUB complexes are common and couple conjugation and deconjugation to facilitate chain editing and fine-tuning of the molecular output [22]. In the rare case of A20 (an OTU family member), both E3 and DUB activities are encoded into a single multifunctional enzyme [23].

The factors dictating DUB activity and substrate specificity include the structure of the catalytic domain, the presence of additional domains for substrate recruitment or Ub binding [24,25,26], alternative splicing [27], allostery [28], oligomerization [29], PTMs [30], and subcellular localization [4]. Importantly, not all DUBs are catalytically active. Approximately 10% are pseudo-enzymes currently found in the USP, OTU, JAMM, and MINDY families [31]. Pseudo-DUBs are critical components of large macromolecular complexes or as allosteric modulators of active DUBs. Most famously, PSMD7, a pseudo-DUB, dimerizes with the active DUB PSMD14. This complex is a critical component of the 19S proteasome regulatory particle [32,33,34]. Without detectable activity, pseudo-DUBs are most readily identified by sequence and structural similarity to catalytically competent homologs.

DUBs can discriminate the linkage type, chain length, modifier (Ub or Ubl), and/or the substrate to which Ub is conjugated [35]. Furthermore, they can remove poly-Ub chains in different modes: iteratively at the distal end (exo-cleavage), from within a chain (endo-cleavage), or all at once (en bloc-cleavage) [36]. Members of the USP family are generally linkage-nonspecific and often referred to as promiscuous [24,37,38]; however, using tool reagents, including small molecule compounds and activity-based probes, our group and others have revealed surprising linkage-dependent processing activities for USP9X [39] and USP7 [40], two USPs previously believed to be nonspecific. Whether or not other members of the USP family have a similar specificity remains to be explored. Contrarily, members of the OTU, MINDY, ZUFSP, and JAMM families tend to be linkage-specific [4,17,25,38,41,42,43,44,45]. For OTUD1, a C-terminal Ub-interacting motif (UIM) endows specificity for K63 linkages by forming a proximal Ub-binding site (S1’) to orient the scissile bond across the catalytic site [25]. The structural determinants of the linkage specificity can also be leveraged for chain length specificity. The five Ub-binding sites of the catalytic domain of MINDY1/2 render strict K48 linkage specificity and the ability to sense the poly-Ub chain length to modulate the exo- (<6 Ub) and endo-cleavage (>5 Ub) modes [46]. UCH DUBs preferentially cleave Ub from small or unstructured C-terminal leaving groups. This represents a sort-of substrate specificity that results from an active site crossover loop limiting the size of ubiquitinated substrates able to occupy the catalytic cleft [47].

Since ubiquitination is involved in almost all biological processes, including protein homeostasis, transcription regulation, DNA repair, endocytosis and endolysosomal sorting, autophagy, immune response, and stem cell renewal, unsurprisingly, DUBs are also major players in these diverse cellular functions. Therefore, their dysregulation is associated with many diseases. This has been the subject of multiple recent reviews, either about their general functions [17,18,53] or on specific disease areas such as neuronal [54], cardiovascular [55], developmental [56], autoimmune [57] diseases, and cancers [58,59] and is beyond the scope of this review. Instead, we highlight the importance of using quality tool reagents when exploring the biology of DUBs at the cellular level.

Due to the scarcity of quality DUB inhibitors, studies often use unvalidated, weak, or semi-selective inhibitors when probing for cellular functions. These practices can lead to spurious conclusions and add to the reproducibility crisis [60,61]. For example, compounds WP1130 and G9 have been used to examine the biological roles of USP9X, a large, 250-kDa USP family DUB. Studies using these compounds have claimed to elucidate numerous interactors and functions [62]; however, many of these interactors (for example, E3 ligases ITCH [63] and FBXW7 [64]; Halabelian and Tong unpublished data) and functions (Michael Clague personal communication) cannot be validated. One inhibitor, G9, which cross-reacts with USP5 and USP24 [65], has been used by many separate studies to assess the functions of USP9X [66,67,68], USP5 [69,70], or USP24 [71]. It is difficult to exclude the possibility of off-target effects, crosstalk, or functional compensation between these DUBs. On the other hand, FT709, a probe-quality inhibitor with a nanomolar affinity for USP9X, has provided much clearer evidence of its involvement in ribosomal stalling and confirmed its interaction with two other E3 ligases, ZNF598 and MKRN2 [72]. Another example is spautin-1 [73], which is widely used as a specific autophagy inhibitor for its inhibition of both USP10 and USP13. These two DUBs have very different structures. Whereas USP10 has a typical USP fold without insertions, USP13 has two ubiquitin-associated (UBA) domains inserted in the catalytic domain. USP13 is closer in sequence and domain architecture to USP5 than USP10. No orthogonal biophysical methods, such as the surface plasmon response (SPR) or isothermal titration calorimetry (ITC), were used to validate the binding of spautin-1 with USP10 or USP13.

2. Small Molecule DUB Inhibitors