Main

The sequential enzymatic processes that covalently attach ubiquitin, a 76-residue polypeptide, to target proteins — a process known as ubiquitylation — are now well understood1 (Fig. 1a). In some cases, a single ubiquitin is attached to the target protein, whereas in others, multiple monoubiquitin adducts are conjugated to different residues of the target. In many instances, various types of ubiquitin chains are produced, wherein one ubiquitin moiety is attached to a free amino group of another. This leads to linear ubiquitin chains and chains involving internal ubiquitin lysine residues K6, K11, K27, K29, K33, K48, K63, as well as mixed ubiquitin chains containing different linkages, or linkages between ubiquitin and ubiquitin-like proteins (UBLs) that include small ubiquitin-like modifier (SUMO) and neuronal precursor cell-expressed developmentally downregulated protein 8 (NEDD8).

Figure 1: The ubiquitylation cascade and the deubiquitylase family of proteins.
figure 1

a | Schematic of key events in ubiquitylation and deubiquitylation. The E1 enzyme activates ubiquitin in an ATP-dependent manner, resulting in a covalent thioester linkage between ubiquitin and the E1 cysteine residue. Ubiquitin is then transferred to an E2 conjugating enzyme, forming a thioester linkage with the catalytic cysteine. Finally, an E3 ligase assists or directly catalyses the transfer of ubiquitin from the E2 to a substrate, usually via a lysine side chain. An example of a HECT (homologous to the E6AP carboxyl terminus) or RBR (RING-between-RING) E3 ligase is shown. In subsequent rounds, ubiquitin molecules can be conjugated to the N-terminal amino group or lysines on ubiquitin itself to form chains. Deubiquitylating enzymes (DUBs) remove ubiquitin molecules from substrates or process ubiquitin precursors to generate free ubiquitin pools. b | DUB phylogenetic tree. Sequences for full-length DUB and SENP (sentrin/SUMO-specific protease) proteins were aligned with COBALT (constraint-based multiple alignment tool), a computational tool for multiple protein sequences, and subsequently visualized with FigTree v1.4.3. In regard to USP17-like, note that various related human USP17-like DUBs exist. AMSH, associated molecule with the SH3 domain of STAM; AMSHLP, AMSH-like protease; ATXN3, ataxin 3; BAP1, BRCA1-associated protein 1; CEZANNE, cellular zinc finger anti-NF-κB protein; CSN, COP9 signalosome complex subunit; CYLD, cylindromatosis; DESI, desumoylating isopeptidase; EIF3, eukaryotic translation initiation factor 3; JAMM, JAB1, MPN, MOV34 family; JOSD, josephin domain; MINDY, motif-interacting with ubiquitin-containing novel DUB family; MJD, Machado–Josephin domain-containing protease; OTUD, OTU domain-containing protein; PRPF8, pre-mRNA-processing splicing factor 8; UCHL, ubiquitin carboxy-terminal hydrolase-like; USP, ubiquitin-specific protease; VCPIP1, valosin-containing protein p97/p47 complex-interacting protein 1.

PowerPoint slide

Full size image

These different types of ubiquitin and UBL modifications, sometimes referred to as 'the ubiquitin code', have specific and diverse effects on protein and cell physiology. For example, such modifications can target proteins that are damaged or improperly folded, or that have intrinsically short half-lives for degradation via the ubiquitin–proteasome system (UPS)2. Appropriately polyubiquitylated proteins are recognized and degraded by the 26S macromolecular proteasome complex3 via mechanisms that have been extensively reviewed elsewhere4,5. In other instances, ubiquitylation regulates protein interactions, localization and enzymatic activities, thereby affecting cellular processes, including transcription, DNA damage signalling and DNA repair, cell cycle progression, endocytosis, apoptosis and various other processes6,7,8,9. Such control mechanisms often involve ubiquitin-binding proteins, many of which exist in eukaryotic cells10. The recent demonstration of post-translational modification of ubiquitin itself provides an additional layer of regulation that affects various cellular processes11.

Like other post-translational modifications, ubiquitylation is reversible: peptidases termed deubiquitylating enzymes (DUBs) can cleave ubiquitin from substrate proteins, edit ubiquitin chains and process ubiquitin precursors12. Some DUBs and related enzymes are involved in editing or processing UBLs and their conjugates13; prime examples of these being the SENP (sentrin/SUMO- specific protease) proteins that process SUMO precursors and SUMO conjugates14. DUBs are classified into six families based on sequence and domain conservation (Fig. 1b): USPs (ubiquitin-specific proteases), UCHs (ubiquitin carboxy-terminal hydrolases), MJDs (Machado–Josephin domain-containing proteases), OTUs (ovarian tumour proteases), MINDYs (motif-interacting with ubiquitin-containing novel DUB family) and JAMMs (JAB1, MPN, MOV34 family). SENPs and the first five DUB families are cysteine peptidases, whereas JAMMs are zinc metallopeptidases.

Ubiquitylation and related processes control myriad aspects of human cell biology and physiology, and defects in such processes contribute to many diseases. Accordingly, DUB deregulation contributes to various sporadic and genetic disorders. Notable examples include: the UCH family member BRCA1-associated protein 1 (BAP1), mutated in melanoma, mesothelioma and renal cell carcinoma15; USP6, translocated in aneurysmal bone cysts16; USP7, mutated in neurological disorders17; USP8, whose mutations cause Cushing disease18,19; USP9X, whose mutations cause developmental disorders20 and whose expression is dysregulated in cancer21; USP15, amplified in certain glioblastoma, breast and ovarian cancers22; and CYLD, commonly mutated in cylindromatosis23. Deregulation of MJD-family DUBs has also been linked to diseases associated with polyglutamine amplification. For example, expansion of DNA 'CAG' trinucleotide repeats in ataxin 3 (ATXN3) causes Machado–Joseph disease (also known as spinocerebellar ataxia 3)24. Furthermore, mutations in the JAMM family member associated molecule with the SH3 domain of STAM (AMSH; also known as STAMBP) cause microcephaly–capillary malformation syndrome25.

There has been growing interest in exploiting components of the ubiquitylation machinery as therapeutic targets26. Although there has been strong progress in developing small-molecule inhibitors of ubiquitin and UBL E1 enzymes27, the highly pleiotropic nature of E1s means that such drugs will likely be confined to acute settings, such as in the treatment of aggressive cancers. Given their greater numbers and diversity, E2s, E3s and DUBs offer the potential for developing drugs with more specific effects. In particular, being a group of diverse enzymes with well-defined catalytic clefts, DUBs are intrinsically attractive as potential drug targets26. However — as we discuss further below — until recently, the development of selective DUB inhibitors has been limited by insufficient understanding of DUB biology, difficulties in establishing robust biochemical assays suitable for compound screening, limitations in cellular and in vivo models to assess DUB activity or inhibition, and the pleiotropic nature of various small-molecule DUB inhibitors. With many of these issues now being largely overcome, the rate of progress in DUB drug discovery has accelerated over the past few years, with various selective compounds being described and characterized by both academic groups and companies.

In this Review, we discuss how DUBs and their deregulation affect human diseases, particularly cancer, neurodegeneration and inflammation (Table 1), and highlight the therapeutic potential for pharmacological modulation of DUB activities. Recent advances in assay development and screening technologies, which are enabling researchers and drug developers to overcome recurrent challenges in the clinical translation of DUB inhibitors, are also discussed.

Table 1 DUBs connected with human diseases
Full size table

DUBs in oncology

Accumulating evidence implicates DUBs in tumorigenesis at multiple levels (Fig. 2). First, DUBs such as BAP1, UCHL1 and CYLD have been described as displaying intrinsic oncogenic or tumour suppressor activities28. Second, some DUBs, such as USP22, are connected to controlling key epigenetic changes that promote tumour development29. Third, through their deubiquitylating activities, various DUBs, such as USP7 and USP28, have been reported to regulate the levels and/or activities of various oncogene or tumour suppressor proteins30,31. Fourth, DUBs modulate other therapeutically relevant cellular components and processes, such as the UPS (for example, USP14 and UCHL5 (also known as UCH37))32, stem cell renewal (for example, USP16 or USP22)29,33, DNA damage response (DDR) and DNA repair (for example, USP1 or USP11)9, immuno-oncology (for example, USP7)34 or receptor tyrosine kinases (for example, USP8 or USP9X)35,36. Consequently, and as described in more detail below, various DUBs are emerging as attractive targets for the development of novel cancer therapies.

Figure 2: Various roles of DUBs in oncology.
figure 2

Selected, representative examples of deubiquitylating enzymes (DUBs; light blue ovals) involved in distinct cellular pathways and regulation of various ubiquitylated substrates (dark blue ovals) related to oncology. The proteasome and associated DUBs facilitate protein turnover and recycle ubiquitin. Ubiquitin-specific protease 28 (USP28) regulates turnover of the oncogene product MYC, ataxin 3 (ATXN3) controls the stability of the tumour suppressor p53, and USP7 regulates p53 and its E3 ubiquitin ligase human double minute 2 (HDM2). USP1, USP4 and USP11 have important roles in DNA damage repair, whereas USP9X regulates claspin and is linked to replication stress and checkpoint signalling. BRCA1-associated protein 1 (BAP1) and USP22 participate in chromatin remodelling by deubiquitylating histones, and ubiquitin carboxy-terminal hydrolase-like 1 (UCHL1) plays a part in AKT signalling. These are representative examples only and not meant to be exhaustive. Examples of small-molecule compounds targeting these DUBs are shown. BRCA1, breast cancer type 1 susceptibility protein; CTIP, C-terminal-binding protein-interacting protein; FANCD2, Fanconi anaemia group D2 protein; GFR, growth factor receptor; MRE11, meiotic recombination 11 homologue 1; NBS1, Nijmegen breakage syndrome protein 1; PALB2, partner and localizer of BRCA2; PCNA, proliferating cell nuclear antigen; PI3K, phosphoinositide 3-kinase.

PowerPoint slide

Full size image

Proteasomal DUBs

The successful targeting of the proteasome for cancer therapy is underlined by the clinical success of bortezomib, a broadly acting proteasome inhibitor, in refractory multiple myeloma37 or mantle cell myeloma38. However, three DUBs associated with proteasome functions — PSMD14 (also known as POH1), USP14 and UCHL5 — may represent more specific anticancer targets. To facilitate the degradation of proteasome-targeted substrates, these specialized DUBs remove ubiquitin moieties that would otherwise impede entry into the 20S proteasome catalytic core39.

The JAMM metalloprotease PSMD14 has been highlighted as a potential therapeutic target through studies showing that its levels inversely correlate with survival of patients with multiple myeloma and that its depletion impairs proliferation of multiple myeloma cells40. In addition, nuclear PSMD14 is elevated in hepatocellular carcinomas and correlates with E2F1 overexpression and tumour growth41. PSMD14 has also been reported to regulate the ubiquitylation and stability of the oncogene receptor tyrosine kinase ERBB2 (Ref. 42). Furthermore, as PSMD14 has been connected to promoting cellular responses to DNA double-strand breaks, particularly by the process of homologous recombination, PSMD14 inhibition could potentially sensitize cancer cells to DNA-damaging agents and/or preferentially kill cancer cells that rely strongly on homologous recombination43.

Another potential anticancer therapeutic target is USP14, which is primarily associated with the proteasome 19S regulatory subunit, where it potentiates ubiquitin recycling44. USP14 is not constitutively active but reversibly associates with the 19S RPN1 subunit (also known as PSMD2), which enhances its activity45. USP14 inhibits proteasomal degradation of ubiquitin–protein conjugates by trimming ubiquitin chains on protein substrates before their degradation46. USP14 expression is upregulated in non-small-cell lung cancer, especially in adenocarcinoma47, and its levels are reportedly elevated in ovarian cancer samples48. In line with this, USP14 is connected to several important signalling pathways, for example, as a substrate of AKT mediating intracellular signalling for growth factors49 and a modulator of Dishevelled proteins, key positive regulators of WNT signalling50.

Like USP14, the DUB UCHL5 reversibly interacts with the proteasome51, binding to the RPN13 receptor (also known as ADRM1)52 in a manner that enhances UCHL5 isopeptidase activity51,53. A key function of UCHL5 is to remove distal ubiquitin moieties from polyubiquitylated proteins, thereby liberating proteins from destruction54, or facilitating destruction of certain substrates, as described for inducible nitric oxide synthase and nuclear factor-κB inhibitor-α (IκBα)55. It therefore seems that, like USP14, UCHL5 suppresses the destruction of certain proteins, while promoting the degradation of others. Notably, RNA interference studies showed that depletion of either USP14 or UCHL5 alone had no detectable effect on cell growth, proteasome structure or proteolytic capacity but did accelerate cellular protein degradation53. By contrast, depletion of both DUBs decreased protein degradation, suggesting that they have overlapping functions. UCHL5 is overexpressed in epithelial ovarian cancer, which is associated with advanced tumour progression and poor clinical outcome56. UCHL5 is also overexpressed in hepatocellular carcinoma, and was shown to promote cell migration and invasion57.

These proteasome-associated DUBs represent attractive drug targets, as their inhibition might have substantial effects on cancer cell physiology but with fewer toxicities than are seen with drugs targeting core proteasome catalytic function58. Indeed, VLX1570 (Table 2), the most advanced reported DUB inhibitor, which was recently in phase I trials (now suspended) for treatment of multiple myeloma and solid tumours59, has been described to target USP14 and UCHL5 (Ref. 60). VLX1570 is a ring-expanded version of the compound b-AP15 (also known as VLX1500) that was identified from cell-based screens looking for compounds inducing p53-independent apoptosis. Cells treated with b-AP15 accumulate polyubiquitin chains61, and it has been claimed that b-AP15 targets USP14 and possibly also UCHL5 (Ref. 60). This compound was reported to be reversible and reasonably selective against other DUBs60 in a cell-based activity probe assay, with a median inhibitory concentration (IC50) of ~2 μM against purified 19S proteasome DUB activities. b-AP15 displayed strong activity when tested in various in vivo solid tumour models59, including multiple myeloma62, but it remains to be seen whether VLX1570 selectivity will be sufficient to deliver on its promise as a next-generation proteasome inhibitor. Cleave Biosciences has also published a series of patent applications describing compounds that inhibit JAMM proteases, providing potential angles for developing selective PSMD14 inhibitors63,64,65 (Table 2).

Table 2 DUB inhibitors in development
Full size table

DUBs linked to DNA repair

One hallmark of cancer is the downregulation, loss or deregulation of certain DNA repair and DDR pathways and/or strong reliance on such pathways66,67. DNA repair and DDR mechanisms are regulated by post-translational modifications, such as ubiquitylation, and many DUBs are strongly linked to such processes9,68.

One example of this is USP1, a DUB identified as a regulator of Fanconi anaemia group D2 protein (FANCD2) ubiquitylation, a key protein involved in the Fanconi anaemia pathway of DNA crosslink repair69,70. USP1 influences accumulation of the Fanconi anaemia core complex at DNA damage sites and deubiquitylates FANCD2–FANCI in a cell cycle-dependent manner69. USP1 also removes monoubiquitin from proliferating cell nuclear antigen (PCNA), a DNA replication component that also functions in DNA repair by translation synthesis71. Other USP1 activities include involvement in a feedback loop to limit DDR CHK1 protein kinase activity72, and the regulation of cellular differentiation in osteosarcoma cells by deubiquitylating and hence affecting the stability of inhibitors of DNA-binding proteins73. In vitro, USP1 activity is greatly stimulated by USP1-associated factor 1 (UAF1; also known as WDR48), enhancing USP1 catalytic turnover (kcat) but not affinity (Km) for monoubiquitylated substrates74. Selective USP1 inhibitors with submicromolar potency have been identified75, and one of these, pimozide, was shown to re-sensitize platinum-resistant non-small-cell lung cancer cells and promote FANCD2 and PCNA monoubiquitylation75. However, although these studies indicated on-target effects, DUB selectivity profiling suggested that pimozide might be less selective than initially described76. Optimization of certain USP1 screening hits has generated additional molecules77, most notably a selective pyrimidine core compound, ML323 (Table 2). This molecule allosterically blocks complex formation between UAF1 and USP1 (Ref. 78), potentiates cisplatin cytotoxicity, and increases PCNA and FANCD2 monoubiquitylation in cells77. So far, however, little progress has been made in advancing selective USP1 inhibitors into clinical development.

Another DUB linked to DNA repair is USP11, which was initially described to form a complex with the DDR tumour suppressor breast cancer type 2 susceptibility protein (BRCA2) to promote the DNA double-strand break repair pathway of homologous recombination79. Depletion of USP11 has been shown to sensitize cells to olaparib (also known as AZD2281), which inhibits the DDR enzyme poly(ADP-ribose) polymerase 1 (PARP1)80. Recently, an interaction between BRCA1 and partner and localizer of BRCA2 (PALB2) — which functionally cooperates with BRCA2 in DNA repair — was shown to be under ubiquitin control, with PALB2 ubiquitylation suppressing its interaction with BRCA1 in a manner counteracted by USP11 (Ref. 81).

The only currently reported USP11 inhibitor is the topoisomerase inhibitor mitoxantrone82 (Table 2). Although the authors reported low nanomolar potency in a pancreatic ductal adenocarcinoma cell survival model, no further development of this compound has been reported. Given the apparent amenability of USP11 to small-molecule inhibition, it is notable that USP4, a DUB closely related to USP11, was recently shown to be involved in the DDR through promoting early stages of homologous recombination83.

USP9X21, which maintains DNA replication fork stability and DNA damage checkpoint responses by regulating the protein claspin during S phase84, may represent another potential therapeutic target. USP9X has been shown to affect radiosensitivity in glioblastoma cells by myeloid cell leukaemia 1 (MCL1)-dependent and -independent mechanisms85. The best-described USP9X inhibitor is WP1130 (Table 2), identified in a screen for Janus kinase 2 (JAK2) inhibitors, which was shown to inhibit USP9X as well as other DUBs (USP5, USP14 and UCHL5)86,87. It was shown by mass spectrometry that the covalent mechanism of action of this compound is reversible76.

Regulation of oncogenes and tumour suppressors

Various DUBs have been reported to have connections to tumour-suppressing or oncogenic functions, and may therefore represent potential therapeutic targets88.

p53 regulation. Several DUBs have been linked to regulation of the tumour suppressor protein p53, which has pivotal roles in cellular stress responses and is lost or mutated in many cancers89. Human double minute 2 (HDM2; also known as MDM2) is a RING-type ubiquitin E3 ligase and key negative regulator of p53 through its ability to ubiquitylate p53 and target it for degradation90. By cleaving ubiquitin chains on HDM2, USP7 counteracts HDM2 proteasomal degradation, leading to p53 suppression through increased ubiquitylation and degradation91,92. In theory, USP7 inhibition should therefore trigger HDM2 degradation, p53 stabilization and ultimately activation of apoptotic pathways in tumour cells93. Additional USP7 targets have also been described, such as phosphatase and tensin homologue (PTEN), forkhead box protein O4 (FOXO4) and FOXP3 (Refs 34,94,95), suggesting alternative therapeutic mechanisms for USP7 inhibitors. USP7 has also recently been shown to promote DNA replication via acting as a DUB for the UBL SUMO96.

The first published submicromolar USP7 inhibitor HBX41108 (Ref. 97) was shown to be a rather nonspecific inhibitor of DUBs76. Recently, more-selective amidotetrahydroacridine derivatives such as HBX19818 and HBX28258 were identified, although these exhibited fairly low potency98. Despite this, HBX19818 was shown to covalently bind to the catalytic Cys residue of USP7 in preference to other cysteinyl groups, and to stabilize p53 and promote G1 arrest and apoptosis in cells98. Progenra's thiophene chemical series also provided relatively nonspecific USP7 inhibitors, including the compounds P5091 and P22077 (Ref. 99). In multiple myeloma cells, P5091 stabilized p53 and inhibited tumour growth, whereas in animal models, P5091 was well tolerated, inhibited tumour growth and prolonged survival99. More recent in vivo studies using P22077 within an orthotopic neuroblastoma mouse model showed statistically significant inhibition of xenograft growth100. While these findings are encouraging, little is known about the binding modes of these compounds and whether they can be further optimized into more 'drug-like' entities. Recently, Almac Discovery and Genentech reported that fragment-based screens provided hits as starting points for USP7 discovery programmes101. Optimization of one hit, ADC-01, assisted by X-ray crystallography, produced the non-covalent, highly selective USP7 inhibitor ADC-03 (Table 2).

The stability of p53 has also recently been reported to be regulated by the DUB ATXN3 (Ref. 102). ATXN3 was shown to bind to and deubiquitylate p53, resulting in p53 stabilization. Deletion of ATXN3 resulted in destabilization of p53, whereas ectopic expression of ATXN3 induced expression of p53 target genes and promoted p53-dependent apoptosis. How and whether ATXN3 inhibitors could be exploited to treat cancer or other diseases remains to be established.

USP28 is another DUB that has recently been connected to p53, which functions together with TP53-binding protein 1 (TP53BP1) to promote p53-mediated transcriptional responses103. Furthermore, USP28 is mutated in human cancer cells and is reported to antagonize the tumour suppressor F-box and WD40 domain-containing protein 7 (FBW7)31, highlighting the potential for USP28 inhibitors in various tumours, especially colorectal cancer104. USP28 has also been reported to antagonize ubiquitin-dependent degradation of the oncogene product MYC as well as JUN and Notch105. Although no USP28 inhibitors have yet been reported, it seems likely that drug discovery activities are underway.

HIF1α and USP20. Another tumour suppressor that has been linked to DUB activity is the von Hippel–Lindau (VHL) protein, which ubiquitylates hypoxia-inducible factor 1α (HIF1α) when cellular oxygen levels are normal, leading to the degradation of HIF1α. USP20 (also known as VDU2), is reported to deubiquitylate a number of proteins, including HIF1α. USP20-mediated deubiquitylation of HIF1α prevents proteasomal degradation, allowing for transcription of hypoxic response genes. Thus, inhibition of USP20 has potential for suppressing proliferation of hypoxic tumour cells. GlaxoSmithKline (GSK) presented brief details of its search for USP20 inhibitors at a conference in 2012 (Ref. 106) (Table 2).

EGFR and USP8. Ubiquitylation serves as a signal that delivers membrane receptors from the cell surface to lysosomes, and in mammalian cells, this mechanism has been most intensively studied for epidermal growth factor receptor (EGFR). Upon EGF binding, activated EGFR is rapidly internalized and transported, via early and late endosomes to lysosomes, where EGFR is degraded. USP8 (also known as UBPY) deubiquitylates EGFR on early endosomes, rescuing EGFR from degradation107,108. In several tumours — including glioblastoma, lung and breast cancer — EGFR is amplified or mutated in the tyrosine kinase domain, resulting in deregulation of receptor signalling that drives uncontrolled proliferation of tumour cells109. USP8 inhibitors (for example, HBX90659) of a similar structural class to those identified for USP7 (Ref. 110) have been reported (Table 2). Moreover, a derivative of these compounds was shown to be efficacious in mouse models of lung cancer111.

TGFβ and USP15. USP15 regulates the transforming growth factor-β (TGFβ) pathway and is thought to be important for the proliferation of glioblastoma cells22. USP15 binds to the SMAD7–SMAD-specific E3 ubiquitin protein ligase 2 (SMURF2) complex, and deubiquitylates and stabilizes the type I TGFβ receptor, leading to enhanced TGFβ signalling. The USP15 gene is amplified in glioblastoma, breast and ovarian cancers, and high expression of USP15 correlates with high TGFβ activity22. Depletion of USP15 reduces the oncogenic capacity of patient-derived glioma-initiating cells owing to diminished TGFβ signalling, suggesting therapeutic potential for development of USP15 inhibitors. In addition, USP15 has been shown to deubiquitylate receptor-regulated SMADs (R-SMADs)112, another set of TGFβ signalling pathway components.

UCHL1. The DUB UCHL1, normally expressed predominantly in neurons and neuroendocrine tissues113,114, is highly expressed in many cancers, and its expression correlates with poor prognosis115. Although there are reports that UCHL1 has a tumour-suppressive role, most evidence supports its role as an oncogene115. Indeed, in a transgenic mouse model with constitutively activated UCHL1, sporadic tumours developed in many tissues116. Moreover, in vitro tumorigenesis studies showed that UCHL1 expression stimulated oncogenesis and an invasive phenotype117–119, whereas UCHL1 depletion had antitumour effects and blocked cell migration in a lung cancer cell line117. The precise mechanism by which UCHL1 contributes to tumorigenesis remains unclear, although reports suggest that it contributes to cell survival signalling, cell cycle regulation, DNA repair and regulating pools of free ubiquitin in ways that affect protein degradation and function115. UCHL1 inhibitors have been described, the most potent being isatin O-acyl oximes (such as LDN-57444; Table 2) with some selectivity over UCHL3 (Ref. 120). In addition, a series of pyridinones have been identified as moderate UCHL1 inhibitors121. Enzyme kinetic studies revealed that these compounds are uncompetitive inhibitors and are selective for UCHL1, exhibiting no inhibition of other cysteine hydrolases tested. Using X-ray crystallography, a weak tripeptide fluoromethyl ketone inhibitor was subsequently shown to bind within the UCHL1 active site, irreversibly modifying the active-site cysteine122. Mission Therapeutics has also developed several series of potent and selective UCHL1 inhibitors123,124. Although no UCHL1 inhibitors have demonstrated antitumour activity in vivo, inducible depletion of UCHL1 has been shown to cause disease regression in an orthotopic multiple myeloma mouse model125.

USP22. Another DUB associated with oncogenesis is USP22, the catalytic subunit of a deubiquitylase module in the SAGA (Spt-Ada-Gcn5-acetyltransferase) complex. The best-characterized substrates for the SAGA complex include several acetylation sites in histone H3 and a ubiquitylation site in histone H2B, post-translational modification of which regulates gene expression29. USP22 has strong links to oncogenesis29, having been identified in microarray screens as part of an 11-gene 'death-from-cancer' signature for highly aggressive, therapy-resistant tumours. USP22 was later shown to act as an oncogene product, regulating cell cycle progression, proliferation and apoptosis126. Increased expression of USP22 has been connected with poor prognosis in several cancers, including liver127, colorectal127 and breast cancers128 as well as oesophageal squamous cell carcinoma129 and oral squamous cell carcinoma130. If USP22 DUB activity can be linked to survival and progression of these cancers, then inhibitors may provide attractive prospects for new therapies.

Cancer immunotherapy

Given the role of ubiquitin modifications and DUBs in many inflammatory processes (see below), as well as the renewed interest in targeting the immune system to fight cancer, the antineoplastic potential of therapeutically inhibiting DUBs involved in the immune system is being investigated. Among these is USP7, which positively regulates the stability of FOXP3, a crucial transcription factor controlling the differentiation of regulatory T cells (Treg cells)34. In a search for DUBs that contribute to GATA3 stabilization in FOXP3-expressing cells, both USP7 and USP21 were shown to upregulate GATA3-mediated activity using a reporter assay131. Furthermore, depletion of USP21 in Treg cells resulted in downregulation of FOXP3, compromised expression of Treg signature genes and impaired their suppressive activity132. As Treg cells restrict antitumour immune responses and promote tumour survival133, these results suggest that depletion of FOXP3 in Treg cells by targeting USP7 and USP21 offer promise for anticancer immunotherapies. In this regard, Mission Therapeutics is investigating USP7 as an immuno-oncology target and has developed USP7 inhibitors (See Mission Therapeutics pipeline in Further information).

DUBs in neurodegenerative disease

Identification of ubiquitin in protein aggregates associated with neurodegenerative pathologies — such as neurofibrillary tangles in Alzheimer disease, neuronal inclusion bodies (also known as Lewy bodies) in Parkinson disease or intranuclear inclusions in hereditary polyglutamine expansion disorders — has prompted much interest in understanding how ubiquitylation and deubiquitylation affect such aggregates134. DUB function in the central nervous system has been described in detail elsewhere135,136, therefore, below we focus on a selected number of DUBs connected to neurodegenerative disease.

Mitochondrial quality control

Mitochondrial dysfunction and UPS impairment have been described as hallmarks of ageing137 and have been implicated in the aetiopathogenesis of many age-related diseases, particularly neurodegenerative disorders such as Alzheimer disease and Parkinson disease. In accord with this connection, ubiquitylation has close links to mitochondrial function, with the UPS maintaining mitochondrial homeostasis by regulating organelle dynamics, the mitochondrial proteome and mitophagy138. Conversely, mitochondrial dysfunction can impair cellular protein homeostasis by generating oxidative damage. Notably, mutations in the ubiquitin E3 ligase Parkin are causally associated with certain cases of familial Parkinson disease139. As Parkin ubiquitylates mitochondrial components, thus promoting turnover of mitochondria by lysosome-mediated mitophagy, defective mitophagy and accumulation of defective mitochondria that cause enhanced oxidative stress could be an underlying cause of Parkinson disease140,141. A corollary of this is that Parkin activation — or inhibition of factors counteracting Parkin — could provide opportunities for disease alleviation.

A screen for DUBs that oppose Parkin function identified the mitochondrion-associated DUB USP30 as an antagonist of Parkin-mediated mitophagy142,143, with USP30 depletion significantly decreasing mitochondrial numbers in cells, a phenotype that was rescued by wild-type but not catalytically inactive USP30. Furthermore, USP30 depletion in vivo provided stress protection in Drosophila melanogaster models of Parkinson disease. In line with such findings, USP30 depletion in human HeLa cells led to elongated and interconnected mitochondria144, suggesting a role for USP30 in regulating mitochondrial fusion and fission. Current models invoke USP30 functioning under normal physiological conditions to prevent inappropriate mitophagy. However, in response to stresses such as membrane depolarization, Parkin is recruited to mitochondria to promote mitophagy145. Accordingly, under conditions of mitochondrial dysfunction — for example, caused by defects in Parkin (or its positive regulator PINK1) — USP30 is thought to counteract clearance of damaged mitochondria, leading to a build-up of metabolically and energetically deficient cells142. It is thus hypothesized that, in the context of certain mitochondrial dysfunctions, USP30 inhibition would have therapeutic benefits. So far, only one chemical inhibitor of USP30 has been described, 15-oxospiramilactone (Table 2), which induced mitochondrial elongation in Mfn1-knockout mouse fibroblasts, with no effect on cell viability146. Mission Therapeutics is exploring USP30 inhibition for the treatment of Parkinson disease and other mitochondrial disorders, and has published several patent applications describing USP30 inhibitors124,147.

Two other DUBs connected to mitophagy are USP8 and USP15. Notably, USP8 depletion was found to delay Parkin translocation onto depolarized mitochondria, as well as mitochondrial clearance, and USP8 displayed an ability to remove K6-linked ubiquitin chains from Parkin in vitro148. In addition, USP8 has been shown to remove K63-linked ubiquitin chains from α-synuclein149, a protein known to aggregate — often in a ubiquitylated form — in Lewy bodies associated with neurodegenerative diseases such as Parkinson disease. Depletion of USP8 in either human SH-SY5Y cells or D. melanogaster resulted in increased lysosomal degradation of α-synuclein149. Meanwhile, USP15 was identified as a Parkin-interacting protein that colocalizes with mitochondria150. In HeLa cells overexpressing Parkin, overexpression of wild-type but not catalytically dead USP15 strongly inhibited mitophagy143,150. Furthermore, depleting endogenous USP15 enhanced mitophagy in HeLa cells, in a human dopaminergic neuronal cell line and in primary fibroblasts from human patients150. USP15 does not deubiquitylate Parkin under basal conditions or when cells are treated with mitochondrial depolarizing agents. USP15 depletion also does not seem to affect Parkin translocation to mitochondria150, although it can oppose Parkin-mediated mitochondrial ubiquitylation. Finally, loss of USP15 in D. melanogaster was found to rescue both locomotor defects and accumulation of dysfunctional mitochondria in flight muscles of parkin-knockout flies150. Collectively, these findings highlight the potential for USP8 and USP15 inhibitors in Parkinson disease and perhaps other diseases associated with mitochondrial dysfunction.

Further highlighting connections between Parkinson disease and DUBs, ATXN3 has been shown to interact with Parkin in a manner that counteracts Parkin autoubiquitylation151. In addition, USP7 was recently shown to remove K63-linked ubiquitin chains from α-synuclein149, a protein that aggregates and accumulates in Lewy bodies, which are hallmarks of Parkinson disease.

USP14

As described above, USP14 removes ubiquitin from certain substrates targeted to the proteasome, thus rescuing such substrates from degradation and maintaining a pool of free ubiquitin54,152. IU1 (Table 2), a reversible small-molecule USP14 inhibitor, was shown to target the USP14 catalytic site and promote degradation of several overexpressed proteins, such as Tau, TAR DNA-binding protein 43 (TDP43) and ATXN3, whose accumulation is linked to neurodegenerative diseases46. Notably, IU1 only promoted degradation in Usp14+/+ murine embryonic fibroblasts46 but not in Usp14−/− cells, suggesting that this compound functions specifically through USP14. Furthermore, IU1 reduced accumulation of menadione-induced oxidized proteins and ameliorated menadione- or hydrogen peroxide-induced cell death in human HEK293 cells46. Proteostasis Therapeutics (in collaboration with Biogen) is developing USP14 inhibitors for the clearance of aggregation-prone proteins, including α-synuclein in Parkinson disease and Tau in Alzheimer disease (see Proteostasis pipeline in Further information) and has published several patent applications describing USP14 inhibitors153,154,155.

Despite the growing interest in USP14 as a therapeutic target in cancer and neurodegeneration, the fact that its loss causes severe morbidity and postnatal lethality requires further investigation, especially in regard to its role in neuromuscular junctions: the neuromuscular phenotype of USP14-deficient ax J mice is rescued by neuronal-specific expression of USP14 (Ref. 156). Furthermore, the extent to which USP14 contributes to the clearance of proteins involved in neurodegeneration in vivo remains controversial157. The development and use of USP14 inhibitors in disease-relevant models may shed further light on such issues and hopefully will define potential therapeutic windows for USP14 inhibition in disease settings.

USP16

Down syndrome is a congenital disorder driven by triplication of human chromosome 21, on which the USP16 gene resides. USP16 has been reported to regulate cell cycle progression and gene expression through deubiquitylation of histone H2A158. Defects in haematopoietic stem cell self-renewal in a mouse model of Down syndrome were rescued by reducing USP16 expression to levels similar to those in control mice159. In addition, USP16 overexpression in normal human fibroblasts and neural progenitors led to reduced cell expansion159, which was similar to the strong proliferation defects observed in fibroblasts from individuals with Down syndrome160. Thus, USP16 is a key regulator that controls stem cell self-renewal and senescence in Down syndrome, suggesting that inhibitors of USP16 might provide therapeutic benefits to such individuals.

DUBs in immunity and inflammation

Pathogens are recognized by several families of pattern recognition receptors and activate various signal transduction cascades via the RIG-I-like receptors (RLRs), the nucleotide-binding oligomerization domain-like receptors (NLRs) and the toll-like receptors (TLRs)161. These signalling events mediate induction of inflammation that is important for recruiting immune cells to sites of infection. Ubiquitylation is a critical post-translational modification in this process161. Non-degradative K63-linked and M1-linked ubiquitin chains mediate the key upstream event of recruiting the TGFβ-activated kinase 1 (TAK1) and the IκB kinase (IKK) complexes, respectively162. K63-linked polyubiquitylation activates the TAK1 complex, which phosphorylates the subunit IKKβ at key serine residues in the activation loop, resulting in IKK activation and transcriptional activation of target genes that include mediators of immune and inflammatory responses as well as feedback inhibitors of the NF-κB pathway163. Negative regulators include DUBs that cleave K63-linked and linear ubiquitin chains such as A20 (also known as TNFAIP3), CYLD and OTU domain-containing deubiquitinase with linear linkage specificity (OTULIN; also known as FAM105B and Gumby)161,164,165.

A20 is probably the best-characterized DUB linked to inflammation166. This DUB plays a key part in restricting TLR signalling and maintaining immune homeostasis through deubiquitylation of NF-κB signalling factors such as NF-κB essential modulator (NEMO), receptor-interacting serine/threonine-protein kinase 1 (RIPK1) and tumour necrosis factor (TNF) receptor-associated factor 6 (TRAF6)167. In addition, A20 can bind to polyubiquitin chains through its zinc finger domain, allowing for interaction with ubiquitylated NEMO. This ubiquitin-induced recruitment of A20 to NEMO is sufficient to block IKK phosphorylation by its upstream kinase TAK1, preventing NF-κB activation168. Thus, A20 deficiency promotes local or systemic inflammation in vivo, underscoring why inactivating TNFAIP3 mutations have connections with both inflammatory and autoimmune syndromes169.

CYLD is another DUB known to negatively regulate ubiquitylation of RIG1 (one of the major RLRs) and RIG1-mediated interferon (IFN)-regulated gene induction170,171. CYLD binds to RIG1 and thereby inhibits ubiquitylation and signalling functions of RIG1. CYLD also inhibits the ubiquitylation of TBK1 and IKKɛ, which contributes to the negative regulation of IFN responses171. Consistent with this, CYLD deficiency causes constitutive activation of TBK1 and IKKɛ in dendritic cells. Despite enhanced RIG1 signalling, CYLD-deficient cells and mice are more susceptible to vesicular stomatitis virus infection due to attenuated signalling and antiviral gene expression induced by IFNβ, suggesting a positive role for CYLD in the regulation of type I IFN receptor function161.

M1-linked ubiquitin chains are generated by the linear ubiquitin chain assembly complex (LUBAC), which consists of haem-oxidized IRP2 ubiquitin ligase 1 (HOIL1; also known as RBCK1), HOIL1-interacting protein (HOIP; also known as RNF31) and SHANK-associated RH domain-interacting protein (SHARPIN). LUBAC is recruited to many immune receptors and ubiquitylates target proteins, including RIPK1, RIPK2, MYD88, interleukin-1 receptor-associated kinases (IRAKs) and NEMO172,173. Genetic loss of LUBAC components leads to immunodeficiency174 and inflammatory phenotypes in mice175,176,177,178, and mutations in LUBAC components also cause inflammatory conditions in humans179,180. Hence, loss of M1-linked ubiquitin chains imbalances immune signalling. OTULIN is the only DUB known to specifically cleave M1 linkages181,182. Accordingly, a homozygous hypomorphic mutation in human OTULIN has recently been shown to cause a potentially fatal auto-inflammatory condition termed OTULIN-related autoinflammatory syndrome183. Similar to ubiquitin, the UBL IFN-stimulated gene 15 (ISG15) has a key role in cellular signalling in response to pathogens. Conjugation of ISG15 to various cellular substrates is reversed by the IFN-inducible isopeptidase USP18. USP18 is upregulated after viral infection, type I and type III IFNs, lipopolysaccharide, TNFα or genotoxic stress. In addition to its isopeptidase activity, USP18 negatively regulates type I and type III IFN signalling by blocking the IFNAR2 subunit of the interferon receptor184.

Inflammatory and autoimmune disorders

Debilitating autoimmune diseases range from those with genetic components — such as Crohn's disease, diabetes mellitus type 1, Graves disease and rheumatoid arthritis185 — to sporadic conditions, including coeliac disease, inflammatory bowel disease, multiple sclerosis, psoriasis and systemic lupus erythematosus. In addition, chronic inflammatory diseases are characterized by a prolonged and persistent pro-inflammatory state, and include autoimmune disease as well as metabolic syndrome, neurodegenerative disease, chronic obstructive pulmonary disease and cardiovascular disease.

Following pattern recognition receptor stimulation, dendritic cells secrete various cytokines that regulate the differentiation of CD4+ T cells to different subsets of T helper cells (TH cells), including inducible Treg cells, T follicular helper cells, and TH1, TH2, TH9 and TH17 cells186. TH17 cells mediate pro-inflammatory functions through the secretion of pro-inflammatory cytokines, including interleukin-17A (IL-17A), IL-17F, and IL-22 (Ref. 187). Moreover, TH17 cells have been implicated in the development of autoimmune diseases such as multiple sclerosis, rheumatoid arthritis and systemic lupus erythematosus188.

USP4 has been shown to stabilize the nuclear receptor retinoid-related orphan receptor-γt (RORγt) in TH17-activated T cells and has been proposed as a possible therapeutic target for rheumatoid arthritis189. One report showed that USP4 is highly expressed in TH17 cells, and its depletion resulted in decreased RORγt as well as IL-17A expression189. In addition, use of the reported USP4 inhibitor vialinin A (Table 2) also diminished RORγt and IL-17A expression190. Furthermore, expression of USP4, IL-17A and IL-17F mRNA have been shown to be significantly elevated in CD4+ T cells from patients with rheumatoid arthritis compared with healthy controls189, providing further evidence for a role of USP4 in rheumatoid arthritis.

TRABID (also known as ZRANB1) is required for TLR-mediated expression of the inflammatory cytokines IL-12 and IL-23 in dendritic cells191. It has been proposed that TRABID deubiquitylates and stabilizes the histone demethylase Jumonji domain-containing protein 2D (JMJD2D; also known as KDM4D), which regulates histone modification at the Il12 and Il23 promoters to facilitate recruitment of the NF-κB family member REL191. Conditional deletion of Zranb1 in dendritic cells impairs IL-12 and IL-23 production and the generation of TH1 and TH17 subsets of inflammatory T cells, rendering mice refractory to the induction of experimental autoimmune encephalomyelitis (EAE)191.

Another DUB associated with the activity of TH17 cells is USP18. Although this DUB has been extensively studied in the context of viral infection, USP18 regulates the TAK1–TAB interaction, which is required for TH17 cell differentiation and the autoimmune response192. Consistent with this, USP18-deficient mice were resistant to EAE192.

T cell receptor signalling has been shown to be facilitated by the DUB cellular zinc finger anti-NF-κB protein (CEZANNE; also known as OTUD7B), which binds to and deubiquitylates ζ-chain-associated protein (ZAP70), thus preventing the interaction of ZAP70 with negative regulatory phosphatases193. ZAP70 is a cytoplasmic protein tyrosine kinase that plays a crucial part in T cell signalling and is recruited to phosphorylated sites on the T cell receptor, where it is subsequently phosphorylated by the SRC kinase LCK. Phosphorylation of ZAP70 is required for full activation and downstream phosphorylation of adaptor proteins, which facilitate T cell signalling194. In addition, CEZANNE-deficient mice exhibited attenuated T cell responses to bacterial infection and were refractory to EAE193. While young Otud7b-knockout mice had similar naive and memory-like T cells compared with wild-type mice, older mice deficient for CEZANNE had reduced IFNγ-producing TH1 cell subsets193.

Similar to TH17 cells, TH1 cells have the capacity to cause inflammation and autoimmune disease. The development, differentiation and function of TH1 cells is driven by the T-box transcriptional factor T-bet (also known as TBX21), which promotes TH1-mediated immune response primarily through promoting expression of the cytokine IFNγ195. The DUB USP10 has been shown to deubiquitylate and stabilize T-bet, resulting in enhanced secretion of IFNγ196. In addition, USP10 mRNA expression was found to be elevated in peripheral blood mononuclear cells from patients with asthma compared with healthy donors196.

Although it is currently unclear why so many DUBs are involved in the regulation of immune responses, it is possible that different DUBs function in distinct cell types. Many published studies are based on cell lines and overexpression systems, and the expression of endogenous DUBs in various immune cells will be an important area for future investigation. Similarly, the generation of genetic models and the development of inhibitors for CEZANNE, TRABID, USP4, USP10 and USP18 will help determine their therapeutic potential.

DUBs in infectious diseases

As described below, there is growing interest in DUBs as potential therapeutic targets for various infectious diseases of man and other animals. Such potential is being explored both by developing compounds that inhibit the activity of pathogen-encoded DUB-like proteins, or target host cell DUBs that control the pathogen life cycle or infectivity.

Viral infections

Ubiquitylation is important for modulation of protein–protein interactions, including the activation of innate immune signalling pathways, so perhaps it is not surprising that various viruses have genes for DUBs, which are used as a strategy to inhibit ubiquitin- and ISG15-dependent antiviral pathways197. Severe acute respiratory syndrome coronavirus (SARS-CoV) and Middle East respiratory syndrome coronavirus (MERS-CoV) are two of the six known human coronaviruses. Both are highly pathogenic, with the potential for human-to-human transmission, and contain papain-like cysteine proteases termed SARS-CoV PLpro and MERS-CoV PLpro, respectively. In addition to processing viral polyprotein, these proteases remove ubiquitin and ISG15 from host cell factors, resulting in antagonism of the host antiviral immune response198. Hence, both SARS-CoV PLpro and MERS-CoV PLpro have been proposed as important antiviral targets. The X-ray structures of both proteases have shown structural similarity to the USP family of DUBs199,200,201.

OTU domain-containing proteases from diverse RNA viruses, including the nairoviruses Crimean-Congo haemorrhagic fever virus and Dugbe virus, the papain-like protease (PLP2) domain of the arterivirus equine arteritis virus, and the protease (PRO) domain of the tymovirus turnip yellow mosaic virus can hydrolyse ubiquitin and ISG15 from cellular target proteins197,202. Many positive-strand RNA viruses, including arteriviruses and tymoviruses, encode polyproteins that are post-translationally cleaved by internal protease domains. In accord with this, both arterivirus PLP2- and tymovirus PRO-domain-containing proteases are crucially required for viral replication due to their primary role in polyprotein maturation197. Thus, viral OTU domain-containing proteases may represent promising therapeutic targets.

Bacterial infections

Bacteria use a repertoire of effector proteins that target the eukaryotic ubiquitin system to promote bacterial pathogenicity. Protease activities from human bacterial pathogens, including Salmonella enterica serovar Typhimurium (SseL), Escherichia coli (ElaD), Shigella flexneri (ShiCE), Chlamydia trachomatis (ChlaDUB1), Rickettsia bellii (RickCE) and Legionella pneumophila subsp. pneumophila (LegCE), have recently been characterized203. LegCE showed no proteolytic activity; SseL, ElaD, and ShiCE demonstrated ubiquitin-specific protease activity; whereas ChlaDUB1 and RickCE cleaved both ubiquitin and, to a lesser extent, NEDD8-modified peptides. Interestingly, these DUBs encoded by human pathogens showed strong preference for K63-linked ubiquitin chains, only targeting K48-linked and K11-linked chains at later time points or higher enzyme concentrations. Therefore, bacterial DUBs are potential therapeutic targets.

Parasitic infections

In addition to expressing DUBs that target host functions, similar to viruses and bacteria, eukaryotic parasites also possess UBL pathways of their own. The use of ubiquitin-based activity probes to identify DUBs in Plasmodium falciparum led to the identification of PfUCH54, which was shown to have deubiquitylating activity and also an ability to remove adducts of the UBL NEDD8 (Ref. 204). Further investigation of the parasite Toxoplasma gondii using a similar strategy identified four DUBs, one of which was orthologous to mammalian UCHL3 (Ref. 205). Structural studies on PfUCHL3 explained the dual specificity of the enzyme, and PfUCHL3 was found to be required for parasite survival206. Distinct differences in the ubiquitin-binding site between PfUCHL3 and its human counterpart suggest that this parasitic DUB can be selectively targeted by inhibitors. Based on the above findings, it will be of great interest to further explore anti-infective opportunities for DUB inhibitors.

Challenges and emerging technologies

Despite the significant and growing attractiveness of DUBs as drug targets, DUB-focused drug discovery has been challenging, with researchers in this arena facing various obstacles. First, although DUBs have clear catalytic pockets that a priori seem suitable for drug development, a key challenge has been to identify potent compounds that show selectivity among related DUBs and that have properties commensurate with their development for clinical use. Second, ubiquitylation and deubiquitylation are intracellular processes that, at least at present, are only amenable to classical small-molecule chemical approaches. Third, because most DUBs execute the transfer of ubiquitin molecules via a reactive thiol group, most standard assays used to identify inhibitors are prone to non-selective redox or alkylating false positives207. Fourth, the mechanisms of action of DUB enzymes are often complex, involving regulation of enzymatic activity through allosteric effects and/or substrate-mediated catalysis, and many DUBs alternate between active and non-active conformations (see below)208,209. This makes it challenging both to design predictive biochemical assays and develop drug-like compounds. Finally, DUBs often display specificity for ubiquitin chains as well as the target proteins. Hence, to optimize the likelihood of identifying genuine inhibitors, it is prudent to develop bespoke primary screening and secondary assays that recapitulate the most physiological substrate and ubiquitin-linkage setting for each DUB.

Despite the above issues, DUBs are fundamentally catalytically driven proteins with known enzymatic functions and, as such, present researchers with the opportunity to identify small-molecule inhibitors either within the active site or at adjacent allosteric pockets. Indeed, over the past few years there has been an increasing rate of progress in successfully screening for and evolving small-molecule DUB inhibitors, with the most developed ones now moving towards or into clinical evaluation (for examples, see Table 2).

Understanding DUB–substrate interactions

Understanding the mechanism of action of individual DUBs is important when initiating any screening and subsequent drug discovery campaign. DUBs are generally isopeptidases that, in most cases, catalyse a proteolytic reaction between a lysine ɛ-side chain and a carboxyl group corresponding to the ubiquitin C terminus209. The last two C-terminal amino acid residues are glycines (Gly75–Gly76) that lack side chains, resulting in a narrow linker on either side of the isopeptide bond, which is mirrored in a long and narrow DUB catalytic cleft209. Moreover, cysteinyl protease DUB catalytic activity tends to rely on two or three crucial residues comprising a catalytic diad or triad, generally containing a His side chain that — by lowering the pKa of the catalytic Cys — leads to a nucleophilic attack on the ubiquitin– substrate isopeptide linkage12. Collectively, these properties bring complexity to the identification of selective small-molecule inhibitors that target DUB catalytic sites and are likely to restrict the breadth of series that are suitable for developing potent and selective DUB inhibitors.

The Proteostasis thiophene pyrimidine core-derived USP14 inhibitors are known to bind in the ubiquitin pocket and prevent the ubiquitylated substrate from binding210. However, the majority of historical and current DUB drug discovery programmes have focused on chemical series that include the provision of an active 'warhead' that forms a reversible or irreversible covalent adduct with the DUB catalytic cysteine. The high reactivity of some of these warheads, which include oxidative, alkylating and arylating moieties210, is likely to limit drug selectivity, may hamper the development of acceptable pharmacokinetic and pharmacodynamic parameters, and may also pose risks of idiosyncratic toxicities in patients.

For this reason, less-reactive warheads are being explored that are closely related to warheads utilized by non-DUB cysteine protease inhibitors in the clinic. For example, the USP8 inhibitor identified from a library of amidomethyl methyl acrylates (Compound 6)211 contains a Michael acceptor group also found in rupintrivir, an inhibitor of rhinovirus 3C protease and a GSK cathepsin C inhibitor210. In addition, USP9X inhibitors WP1130 and EOA1342143 (Ref. 212) contain a Michael acceptor group similar to that found in a Principia Biopharma's Bruton tyrosine kinase (BTK) inhibitor210. However, these examples are few in number, and the compounds are weak DUB inhibitors. Mission Therapeutics has discovered covalent active-site series that are 'drug-like', unrelated to any previously described DUB inhibitor, and that achieve submicromolar cell-based potencies and exhibit good oral bioavailability123,124,147.

Allosteric regulation: implications

Most peptidases, including many cysteine proteases, recognize a small linear-polypeptide motif and cleave either before or after the peptide bond213. DUBs, however, are more complex. Most DUBs cleave an isopeptide linkage between the side chain of a lysine residue and the C-terminal glycine of ubiquitin, with the isopeptide linkage providing specificity and flexibility to the mechanism of proteolysis214. Also, DUBs need to accommodate substantial globular post-translational modifications (ubiquitin, UBL, or ubiquitin–UBL chains) into their catalytic site215. Furthermore, unlike most other cysteine peptidases, the catalytic triad of cysteinyl peptidase DUBs is not usually in a 'functional' configuration, with allosteric regulation being required to render DUBs fully functional and processive. Such allosteric regulation can be substrate-mediated (for example, OTULIN)181, triggered by intramolecular reorganization (for example, USP7)216 or induced by key cofactors (as for USP1)74. In addition, several DUBs are associated with multi-protein complexes such as the proteasome217, valosin-containing protein (VCP; also known as p97)218, or the COP9 signalosome219. These associations can allosterically regulate the affinity of DUBs for their substrates208,220, and in some instances, DUBs coexist in the same complex as the ubiquitylation machinery221. The above issues must therefore be carefully considered when establishing screening and compound evaluation assays for a DUB. Some DUB inhibitors have been suggested to target allosteric sites, such as the USP1 inhibitor ML323 (Ref. 78).

Screening technologies

Approximately twenty years ago, a general assay was established for measuring DUB enzymatic activity based on the substrate ubiquitin C-terminal 7-amido-4- methylcoumarin (Ub-AMC). This substrate is efficiently cleaved or hydrolysed by various DUBs, releasing a highly fluorescent AMC moiety. While this assay has been used in various DUB inhibitor screens, for example to identify USP1 (Ref. 222) and USP7 (Refs 207,223,224) inhibitors, one significant drawback is that it is prone to fluorescence interference exhibited by many small molecules225. Moreover, AMC and alternative tags such as rhodamine and tetramethylrhodamine (TAMRA), which have been used because they are less prone to fluorescence artefacts, contain a peptide linkage and thus differ quite substantially from most natural DUB substrates. Processing of such substrates thus requires the DUB to function in a non-physiological manner, thereby potentially diminishing prospects for identifying compounds that will operate in cellular or therapeutic settings.

A further challenge for development of DUB inhibitor screening assays is oxidative hydrolysis of the active-site cysteinyl residue of purified DUBs in biochemical buffers. This sensitivity requires use of protective reducing agents such as dithiothreitol (DTT), usually in millimolar concentrations, to maintain DUB enzymatic activity. Altering the concentration or type of reducing agent (for example, 2-mercaptoethanol, cysteine, glutathione or tris(2-carboxyethyl)phosphine (TCEP)) can considerably affect inhibition obtained for hit compounds207. Following a high-throughput screen to identify USP7 inhibitors, the ability of compounds to inhibit USP7 in the presence of different reductants was evaluated207. Many compounds showed the greatest inhibition in the absence of any reductant, being less potent in the presence of cysteine or glutathione, and least potent in the presence of DTT or TCEP. A further subset of molecules showed an alternative profile, only demonstrating inhibition in the presence of DTT or TCEP. A final set of molecules only inhibited USP7 when no additional reductant was added. Together, these data demonstrate the critical nature of the reducing environment on DUB activity and inhibition. Thus, most screens based on high concentrations of reducing agents and using first-generation fluorescent substrates generate high false-positive rates, an issue that has likely been the most significant challenge in identifying genuine and selective DUB inhibitors.

Indeed, the non-selective nature of some DUB inhibitors is highlighted in biochemical selectivity-profiling assays, with relatively few DUB inhibitors reported in the literature showing promise in such studies76. Matrix-assisted laser desorption/ionization time-of-flight (MALDI-TOF) mass spectrometry was used to screen for DUB activity and specificity by systematically assessing the specificities of 42 recombinant human DUBs against di-ubiquitin isomers with all possible chain linkages (M1- (linear), K6-, K11-, K27-, K29-, K33-, K48- and K63-linked)76. Subsequently, a panel of 32 DUBs was screened against 9 reported DUB inhibitors. Their findings demonstrated that none of the compounds displayed strong selectivity towards a single DUB and that many inhibited most DUBs on the panel.

Novel technologies based on chemically synthesized DUB substrates containing isopeptide linkages, ubiquitin chains and/or assay technologies that are less prone to false positives — such as luminescence, time-resolved fluorescence or mass spectrometry — are advancing screening campaigns and are therefore now being exploited76,226,227,228. For example, a ubiquitin–aminoluciferin substrate was used with a variety of DUBs to demonstrate a suitable assay window for high-throughput screening207,229. Subsequently, USP2 was used as a representative DUB to demonstrate statistical robustness of this reagent in a screening campaign for inhibitors. We believe that such developments are crucial to optimize the prospects for identifying and developing DUB inhibitors for ultimate clinical use.

Monitoring DUB activity or inhibition

A key issue when studying DUBs and their modulation in cells is understanding substrate specificity. Some DUBs have preferences for monoubiquitylated substrates, whereas others favour specific ubiquitin chain types, chains bearing mixed linkages, or mixed chains containing ubiquitin and UBLs230,231. Furthermore, many DUBs have some specificity for the substrate protein itself, with this being mediated through mechanisms often involving regions of the DUB distinct from its catalytic site. DUB substrates can be determined by biochemistry, yeast two-hybrid interactions, proteomic profiling and genetics232, but this is often challenging and time-consuming. Clearly, the ability to directly monitor DUB activity within a native biological system is essential to understanding the physiological and pathological role of individual DUBs as well as the effects of DUB inhibition233.

DUB activity in cells can be monitored by chemical probes that generate readily detectable covalent complexes with the DUB catalytic site (recently reviewed in Ref. 234). Activity probes label DUBs based on their catalytic-site thiol group235, with DUB reactivity towards such probes depending on the type of electrophilic warhead fused to ubiquitin. In addition to profiling DUB levels, activity and catalytic inhibition, activity probes have also been used to identify DUBs by affinity purification combined with mass spectrometry236. More recently, activity-based probes (ABPs) bearing a fluorescent reporter tag have been generated to replace the initial tags (for example, the haemagglutinin (HA) epitope) to allow fluorescent imaging instead of detection by immunoblotting226,227. Although production of ubiquitin ABPs was historically based on a trypsin-catalysed transpeptidation to modify ubiquitin at its C terminus with a vinyl sulfone group, recent approaches involve the full chemical synthesis of ubiquitin ABPs226,237. This advance allows incorporation of modified amino acid residues at any position in the ABPs, whether natural or not. Mass spectrometry has become an important tool to monitor ubiquitin adducts as well as changes in ubiquitin levels232,238. Indeed, combining ABPs with immunoblotting or mass spectrometry can generate powerful tools for monitoring DUB activity and inhibition by small molecules98,239 as well as assessing drug–enzyme target engagement in cells or tissues. For example, one study used ABPs to demonstrate the selectivity of P22077 for USP7 in cells, in contrast to PR-619, which inhibited a broad range of DUBs239. In addition, another study demonstrated the cellular selectivity of HBX19818 for USP7 against a panel of DUBs using ABPs and immunoblotting98.

Activity-based proteomic probes have facilitated the development of pharmacologically active enzyme inhibitors. This approach represents a cell-based assay in which treatment with the inhibitor is carried out in intact cells, allowing for a range of cellular enzymes to be assessed simultaneously239. Competition assays between an inhibitor and the ABP lead to a reduced labelling profile for the ABP, with loss of signal for ABP-labelled target enzymes allowing for assessment of the specificity of inhibition. The limitation to this approach, however, is the number of enzymes successfully labelled by the ABP and the representation of active enzymes in the cellular proteome. ABPs were used to characterize the DUB inhibitors PR-619 and P22077 by immunoprecipitation, combined with identification and label-free quantification by mass spectrometry-based proteomics239. Using this approach, quantitative data for 25 cellular DUBs was obtained. PR-619 was confirmed as a broad DUB inhibitor, whereas P22077 was found to be a selective inhibitor of USP7 and USP47 that may therefore provide the basis for exploring therapeutic opportunities in oncology (see preceding sections and Table 2).

Concluding remarks

During the past decade, we have witnessed dramatic advances in our understanding of DUB functions, mechanisms of action, regulation and disease linkages. In parallel, there have been major improvements in DUB biochemical assays and screening technologies, leading to the development of increasing numbers of small-molecule DUB inhibitors whose selectivity is now being explored and, where possible, refined. Such inhibitors are providing the basis for drug-like molecules suitable for clinical evaluation and are also providing versatile tools to further investigate DUB cell biology, regulation and biochemical mechanisms, as well as to test therapeutic hypotheses in disease models. Although it is still too early to predict the extent of the broad therapeutic potential of DUBs, the next few years certainly seem set to produce further exciting developments in the arenas of DUB biology and drug discovery.

Publisher's note

Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.