Ubiquitin‐targeted bacterial effectors: rule breakers of the ubiquitin system

Abstract

Regulation through post‐translational ubiquitin signaling underlies a large portion of eukaryotic biology. This has not gone unnoticed by invading pathogens, many of which have evolved mechanisms to manipulate or subvert the host ubiquitin system. Bacteria are particularly adept at this and rely heavily upon ubiquitin‐targeted virulence factors for invasion and replication. Despite lacking a conventional ubiquitin system of their own, many bacterial ubiquitin regulators loosely follow the structural and mechanistic rules established by eukaryotic ubiquitin machinery. Others completely break these rules and have evolved novel structural folds, exhibit distinct mechanisms of regulation, or catalyze foreign ubiquitin modifications. Studying these interactions can not only reveal important aspects of bacterial pathogenesis but also shed light on unexplored areas of ubiquitin signaling and regulation. In this review, we discuss the methods by which bacteria manipulate host ubiquitin and highlight aspects that follow or break the rules of ubiquitination.

Pruneda and colleagues discuss how bacteria manipulate host‐cell ubiquitination and what this can teach us about previously unexplored areas of ubiquitin signaling and regulation.

Introduction

The small protein ubiquitin (Ub) is a highly conserved post‐translational signaling molecule that is deeply integrated into many aspects of eukaryotic cell biology. Owing to the many complexities that together comprise the “ubiquitin code,” different forms of ubiquitination (i.e., post‐translational modification with/of ubiquitin) direct distinct cellular fates of the modified substrate, which range from degradative roles in proteasome or autophagy recruitment to non‐degradative roles that include protein trafficking or kinase activation. The vastly different outcomes following ubiquitination arise from a diversification of the signal itself. Ubiquitin can be added as a single modification (monoUb), or attached to itself to create at least eight different homotypic and innumerous heterotypic polymeric ubiquitin chains (polyUb), tagging substrates with a cellular fate dependent on the type of modification. To manage these signaling complexities, regulatory proteins write, read, and erase ubiquitin modifications with strict oversight. Humans have evolved hundreds of ubiquitin regulators—about 5% of the protein‐coding genome—to control the ubiquitination of thousands of proteins at tens of thousands of modification sites. The consequences of ubiquitin dysregulation can be severe and lead to human diseases such as neurodegeneration, autoimmunity, and cancer. As a result, the therapeutic potential of correcting or even redirecting ubiquitin signaling has attracted a lot of recent attention. Unfortunately, humans were not the first to recognize the power of manipulating the ubiquitin system; microbial pathogens appear to have been exploiting this potential for millennia.

Despite the fact that they do not otherwise encode a functional ubiquitin system, many bacterial and viral pathogens have evolved strategies to manipulate ubiquitin signaling in their host. Bacteria, in particular, employ secreted effector proteins that subvert the host ubiquitin system though numerous strategies. Following infection, the cellular “ubiquitinome” undergoes large‐scale and dynamic changes as both host and pathogen vie for control and overall survival. From the host perspective, ubiquitin signaling plays critical roles in both innate and adaptive immunity. Pattern recognition receptors (PRRs) instigate a Ub‐dependent inflammatory response following the detection of pathogen‐associated molecular patterns (PAMPs). Signaling from cell surface and cytosolic PRRs to the nucleus for transcriptional activation of inflammatory cytokines proceeds through MAPK and NF‐κB signaling pathways, both of which require Ub‐dependent processes. Antigen processing and presentation also require ubiquitin signaling, as do activation of B and T cells. Cell‐autonomous immunity also relies upon ubiquitin and related Ub‐like signaling molecules for pathogen recognition and activation of defense responses such as xenophagy. Thus, the underlying dependence of the host immune response on ubiquitin signaling makes it an attractive target for intervention by secreted bacterial effectors. In addition to modulating the immune response, targeting the host ubiquitin system also grants the pathogen access to critical systems such as protein degradation, cell morphology, and vesicular trafficking. Manipulation of host ubiquitin signaling can offer a competitive edge during infection and is emerging as a common strategy for bacterial virulence.

Given the numerous cellular processes that ubiquitin signaling controls, it is sensible that there would be an evolutionary advantage for bacteria that evolve methods to hijack it (Franklin & Pruneda, 2021). Although the field of ubiquitin research is expansive and continually advancing, the regulation of ubiquitin in eukaryotes tends to follow a set of commonly used “rules”. Barring some interesting exceptions that will be discussed in more detail below, the rules of the ubiquitin system establish how ubiquitin signals are typically written or erased, how the ubiquitin modification should look, and what signals these ubiquitin modifications should encode. Some bacterial regulators of ubiquitin follow the rules established by the eukaryotic system; for example, through employing protein folds and/or enzymatic mechanisms similar to their eukaryotic counterparts, or by otherwise utilizing ubiquitin signaling in a “conventional” manner. It is possible that bacteria acquired rule‐following effectors such as these through an ancient horizontal gene transfer event, after which the primary sequences diverged but the underlying function was retained. An alternative hypothesis would be that certain rules of the ubiquitin system are hard‐set, and even organisms that have converged upon ubiquitin regulation must adhere to them. Remarkably, however, the evolutionary pressure placed on human pathogens appears to have led to convergent evolution of many ubiquitin regulators that break the rules of the eukaryotic system. These bacterial rule breakers can use entirely distinct protein structures and mechanisms to manipulate ubiquitin signaling in familiar or sometimes completely foreign ways. By revealing which rules of ubiquitin signaling can be bent or broken, bacterial effector proteins have opened our eyes to new possibilities of ubiquitin regulation in and outside of bacterial infection. Here, we will briefly present the rules of the eukaryotic ubiquitin system, much of which has been reviewed extensively elsewhere, and discuss examples of Ub‐targeted bacterial effector proteins that follow or break these rules.

Ubiquitin ligation by RING‐type E3 ligases

The rule makers

The typical route to ubiquitin conjugation follows a three‐enzyme cascade that links the ubiquitin C‐terminus to the amino group of a substrate lysine side chain or N‐terminus, forming an isopeptide or peptide linkage, respectively. The C‐terminus of ubiquitin is first activated by an E1 ubiquitin‐activating enzyme in an ATP‐dependent manner. In humans, two E1 enzymes, UBA1 and UBA6, are responsible for activating the cellular pool of ubiquitin. Following an ATP‐consuming adenylation reaction, the ubiquitin C‐terminus is loaded onto the E1 active‐site cysteine to form a high‐energy thioester intermediate (Streich & Lima, 2014). This activated ubiquitin is then transferred onto the active‐site cysteine of an E2 ubiquitin‐conjugating enzyme, of which there are ~ 40 in humans. In some cases, this activated E2~Ub conjugate can independently catalyze substrate ubiquitination. Typically, however, 1 of ~ 600 E3 ubiquitin ligases facilitates this process by scaffolding the E2~Ub and substrate. Really interesting new gene (RING) E3 ligases form the largest family, separated from the others by structural and mechanistic differences (Deshaies & Joazeiro, 2009). A RING domain includes eight Cys, His, and/or Asp residues that coordinate two zinc ions in a cross‐brace arrangement, which orients two loops and a central helix as the primary E2‐binding site (Fig 1A). A related E3 family called U‐box ligases adopts a similar RING‐type protein fold but lack zinc coordination. Substrate specificity is achieved through additional protein interaction sites within or outside of the RING domain. In addition to their scaffolding role, RING domains also allosterically activate the E2~Ub thioester linkage to enhance direct ubiquitin transfer onto a substrate (Metzger et al, 2014). Activation of the E2~Ub conjugate is accomplished by promoting its “closed” conformation that orients the thioester linkage for nucleophilic attack. RING domains achieve this conformational rearrangement using a conserved “linchpin” Lys/Arg residue that, in some cases, works in concert with additional contacts directly between the E3 and ubiquitin (Fig 1A and D). Many RING E3 ligases exist in larger protein complexes that can serve important roles in E2~Ub activation as well as substrate recruitment (Baek et al, 2021). For example, some RING ligases form homo‐ or heterotypic dimers, whereas others function through a large, multi‐subunit cullin‐RING ligase (CRL) complex that uses exchangeable substrate adaptor molecules to recruit target proteins for ubiquitination (Fig 1E).

Figure 1. Cysteine‐independent ubiquitin ligases.

(A) Structure of the human U‐box E3 ligase UBE4B (PDB 3L1Z; Benirschke et al, 2010), with regions involved in E2 binding and E2 activation colored in blue and gold, respectively. (B) Structure of the Pseudomonas syringae U‐box E3 ligase AvrProB (PDB 2FD4; Janjusevic et al, 2006), highlighting conserved regions for E2 binding and activation. (C) Structure of the Xanthomonas campestris XL‐box E3 ligase (PDB 4FC9; Singer et al, 2013), highlighting its distinctive fold and E2 binding. (D) Mechanism of RING/U‐box‐type E3 activation through directing a dynamic E2~Ub conjugate toward a closed and reactive conformation (Pruneda et al, 2011, 2012; Dou et al, 2012; Plechanovová et al, 2012). Bacterial followers of this mechanism include U‐box effectors. (E) Mechanism of multicomponent SCF E3 ligase complexes, highlighting the versatility imparted by F‐box substrate adaptors (Baek et al, 2020). Bacterial followers of this mechanism include F‐box effectors. (F) Mechanism of SidE noncanonical ubiquitination, whereby an mART domain consumes NAD to ADP‐ribosylate Arg42 of ubiquitin (releasing NAM), which is then activated onto a PDE domain (releasing AMP) and subsequently transferred onto a substrate serine residue (Bhogaraju et al, 2016; Qiu et al, 2016).

The rule followers

Pseudomonas syringae AvrPtoB

The effector AvrPtoB from the plant pathogen Pseudomonas syringae pv. tomato was the first bacterially encoded E3 ligase to be identified (Abramovitch et al, 2006; Janjusevic et al, 2006). AvrPtoB was initially discovered as a potent inhibitor of plant resistance responses that are characterized by localized programmed cell death of the host (Abramovitch & Martin, 2005). The surprising finding that AvrPtoB encodes E3 ligase activity arose from two simultaneous studies. In one study, the AvrPtoB C‐terminal domain was found to interact with ubiquitin using a yeast two‐hybrid screen and subsequently shown to ligate ubiquitin when combined with human E1 and E2 enzymes (Abramovitch et al, 2006). At the same time, a crystal structure of the C‐terminal domain of AvrPtoB demonstrated both structural homology and mechanistic mimicry to eukaryotic RING‐type E3 ligases (Janjusevic et al, 2006). The AvrPtoB C‐terminal domain is homologous to eukaryotic U‐box domains and contains a highly conserved hydrophobic patch that superimposes with established E2 interaction sites of RING‐type ligases. Disruption of the E2‐binding site abrogates AvrPtoB ligase activity in vitro and, importantly, blocks the ability of AvrPtoB to inhibit programmed cell death in planta (Abramovitch et al, 2006; Janjusevic et al, 2006). Despite a remarkable divergence in primary sequence, so much so that homology to a RING‐type protein fold could not have been predicted without the structure, AvrPtoB closely follows the rules of RING‐type ligases, including a recognizable fold, E2‐binding interface, and even a conserved linchpin residue (Fig 1B).

Enterohemorrhagic Escherichia coli NleG

NleG (non‐LEE‐encoded G) proteins encoded in enteropathogenic Escherichia coli (EPEC), enterohemorrhagic E. coli (EHEC), and Citrobacter rodentium make up a significant portion (up to 14 members identified in EHEC O157:H7) of the arsenal of type III secreted effectors (Tobe et al, 2006). NleG effectors are ~ 200 amino acids in total size, with a ~ 100 amino acid variable N‐terminal domain and a highly conserved C‐terminal domain. While there is low sequence homology to established ubiquitin enzymes, a solution structure of NleG2‐3 was the first to reveal pronounced structural similarities (~ 2 Å RMSD) between the conserved NleG C‐terminal domain and eukaryotic RING‐type E3 ligases (Wu et al, 2010). Similar to eukaryotic U‐boxes, NleG proteins maintain a RING‐type fold by noncovalent interactions rather than zinc binding. Screening for in vitro ligase activity with a panel of 30 human E2 enzymes revealed that several NleG effectors preferred the UBE2D and UBE2E families of host E2s, similar to many human RING‐type ligases (Brzovic & Klevit, 2006; Wu et al, 2010). The UBE2D2/NleG2‐3 interface was characterized by NMR chemical shift perturbation analysis and mutagenesis to show that canonical E2:RING‐type E3 interfaces were important for NleG ligase activity. UBE2D1 interacts with a small groove on NleG2‐3, which corresponds to a common E2‐binding site of RING‐type domains. Residues within the E2‐binding site, as well as a linchpin arginine, are highly conserved among NleG family members.

Subsequent work revealed an important role for sequence plasticity among NleG N‐terminal domains in targeting distinct host proteins for ubiquitination (Valleau et al, 2018a). A crystal structure of full‐length NleG5‐1 from EHEC provided the first view of how the N‐terminal domain may contribute to ligase function. The structure showed a C‐terminal U‐box domain similar to that of NleG2‐3 and eukaryotic ligases, as well as an N‐terminal α/β sandwich fold connected via an unstructured linker. While the two domains are likely flexibly linked, the conformation observed in the crystal structure was compatible with previous RING‐type E3/E2~Ub complex structures. By comparing the surface residues of the NleG5‐1 N‐terminal domain to an analogous domain modeled for NleG2‐3, the authors were able to identify unique surface patches that explained their distinction for the host target proteins MED15 and Hexokinase 2, respectively (Valleau et al, 2018a). The combination of a RING‐type E3 ligase module with unique substrate‐targeting domains extends the reach of bacterial effector ligases to target multiple host proteins for ubiquitination, while still following the rules for RING‐type substrate specificity established in eukaryotes.

Salmonella enterica StoD

The effector StoD from typhoidal Salmonella enterica was first identified as a homolog of E. coli NleG effectors (McDowell et al, 2019). Sequence homology and mutational analyses identified a C‐terminal U‐box domain that was capable of auto‐ubiquitination activity with human E2s from the UBE2D and UBE2E families. NMR studies confirmed that the E2 interaction is confined to the C‐terminal U‐box domain of StoD and uses a conserved hydrophobic pocket. A crystal structure of the N‐terminal domain revealed a similar fold as observed in NleG effectors, despite low sequence homology, and the authors identified a similarity to Ub‐like folds. Interestingly, both the N‐ and C‐terminal domains were shown to bind ubiquitin directly with affinities of ~ 50 μM. After mapping the ubiquitin‐binding sites by NMR, neither was consistent with expected interactions with an E2~Ub conjugate. This, coupled with a higher affinity for diUb chains (5–15 μM) led the authors to propose that StoD may demonstrate “E4” ligase activity, that is, act to extend existing ubiquitin modifications (McDowell et al, 2019). Such an activity would be consistent with related eukaryotic U‐box ligases, and could allow StoD to alter existing ubiquitination events with new types of signals (Hoppe, 2005).

Legionella pneumophila LubX

Legionella pneumophila utilizes a Dot/Icm type IV secretion system to deliver a substantial (> 300) arsenal of effectors. Among these are a number of E3 ubiquitin ligases that work to assemble and maintain a “coat” of ubiquitin modifications on the Legionella‐containing vacuole (LCV). The first identified example, named LubX (Lpg2830), was identified by sequence‐based methods to contain a secretion signal and two consecutive U‐box domains (Kubori et al, 2008). LubX U‐box 1 was demonstrated to be critical for ubiquitin ligase activity in vitro and functions with host E2s from the UBE2D and UBE2E families, as well as UBE2W and UBE2L6, to form polyUb chains (Kubori et al, 2008; Quaile et al, 2015). A crystal structure of the LubX/UBE2D3 complex shows significant similarity to eukaryotic RING‐type E3 ligases and their interactions with E2s, but displays an expanded hydrogen bond network (Quaile et al, 2015). Interestingly, U‐box 1 retains all the critical hydrophobic residues for RING‐type E3/E2 interactions, including a linchpin residue, but U‐box 2 does not (Kubori et al, 2008; Quaile et al, 2015). Initially, U‐box 2 of LubX was proposed as a substrate interaction domain for the targeting of the host kinase Clk1 (Kubori et al, 2008). Subsequent work provided the first evidence that LubX functions as a “metaeffector” ligase that targets another Legionella effector, SidH (details of this mechanism are discussed below; Kubori et al, 2010; Quaile et al, 2015). LubX U‐box 2 was shown to also directly bind SidH, mediating the polyubiquitination and subsequent proteasomal degradation of SidH by U‐box 1. The timed translocation of LubX and subsequent proteasomal destruction of SidH advances the formation and maintenance of the LCV during infection (Kubori et al, 2010).

Legionella pneumophila GobX

Another effector from Legionella, GobX (lpg2455), hijacks both the host ubiquitin and prenylation systems. GobX exploits host S‐palmitoylation at Cys175 to localize to Golgi membranes. GobX also encodes a functional U‐box domain that is capable of auto‐ubiquitination with host E2s from the UBE2D and UBE2E family, although its cellular targets during infection are unknown (Lin et al, 2015). While its structure has yet to be resolved, the GobX sequence shows conservation of the common E2‐binding interface, as well as a linchpin residue, and thus likely follows the eukaryotic rules of RING‐type E3 ligases.

Legionella pneumophila RavN

As demonstrated by the structure‐guided discoveries of AvrPtoB and the NleG E3 ligases, primary sequences can often be too divergent to predict bacterially encoded E3 ligase domains. Despite no recognizable sequence homology, Legionella RavN (lpg1111) was identified to possess E3 ligase activity through an approach to identify host interaction partners of Legionella effectors (Lin et al, 2018). Proteomic studies following effector protein pull‐downs from HEK293T cell lysate revealed that RavN was monoubiquitinated under these conditions. Subsequent studies with recombinant RavN revealed that the N‐terminal domain is a bona fide E3 ligase in vitro and can function with the UBE2D and UBE2E families of host E2s. A crystal structure of the N‐terminal ligase domain revealed an unusual RING‐type domain most similar to E. coli NleGs, but lacking certain characteristic features such as the central helix (Fig 1A). Despite these differences, RavN maintains conserved E2‐interacting residues and an atypical histidine residue at the linchpin position (Lin et al, 2018). Mutation of these E2‐interacting residues, just as with other U‐box E3 ligases, diminishes ligase activity.

Legionella pneumophila F‐box substrate adaptors

Skp1‐Cul1‐F‐box (SCF) ubiquitin ligases, a subclass of the CRL family, utilize F‐box proteins as exchangeable substrate interaction modules (Nguyen & Busino, 2020). F‐box proteins bind to the CRL complex via a conserved F‐box motif, and recruit substrates through additional protein interaction domains such as a WD or leucine‐rich repeats. The F‐box sequence itself consists of approximately 50 amino acids with only a few invariant residues, making the prediction of F‐box motifs by primary sequence difficult (Kipreos & Pagano, 2000). Bacterial F‐box proteins were first discovered among plant pathogens, including Agrobacterium tumefaciens and Ralstonia solanacearum (Tzfira et al, 2004; Angot et al, 2006; Schulze et al, 2012).

Seven identified F‐box‐containing proteins have been described in Legionella pneumophila, including LegU1, LicA, Lpg1975/Lpp1959, AnkB/LegAU13, PpgA/Lpg2224, Lpg2525, and Lpp2486 (Price et al, 2009; Ensminger & Isberg, 2010; Qiu & Luo, 2017). Of these F‐box domain‐containing proteins, AnkB/LegAU13 and LegU1 have been shown to possess E3 ligase activity (Price et al, 2009; Ensminger & Isberg, 2010; Lomma et al, 2010). AnkB initially gained attention because it is a conserved effector across multiple Legionella pneumophila strains and is an essential effector for intracellular replication during infection (Habyarimana et al, 2008). Translocation of AnkB occurs within 5 min of bacterial attachment to macrophages, which results in the rapid accumulation of polyubiquitinated proteins at the LCV.

AnkB and LegU1 exhibit functional mimicry of host F‐box proteins by directly interacting with host Skp1 and redirecting this conserved ligase complex toward ubiquitination of new targets (Fig 1E; Price et al, 2009; Ensminger & Isberg, 2010; Lomma et al, 2010). In addition, AnkB and LegU1 follow the bipartite domain composition of eukaryotic F‐box proteins, with an N‐terminal F‐box motif and a C‐terminal substrate interaction domain. However, these bacterial F‐box proteins differ from their eukaryotic counterparts by using unconventional substrate recognition domains, such as an ankyrin repeat (Ensminger & Isberg, 2010; Lomma et al, 2010). The crystal structure of AnkB/LegAU13 in complex with Skp1 revealed two eukaryotic‐like domains: a Skp2‐like N‐terminal F‐box domain and a C‐terminal ankyrin repeat domain (Wong et al, 2017). Mutation of the highly conserved Skp1‐binding F‐box residues within AnkB inhibits SCF E3 ligase activity and subsequently impedes intracellular Legionella replication during infection (Price et al, 2009; Wong et al, 2017). Structural analysis of the AnkB ankyrin domain identified a set of residues that likely form the basis for substrate recruitment, as their mutation resulted in a loss of ubiquitin recruitment to the LCV as well as a defect in Legionella replication (Wong et al, 2017). AnkB targets several reported substrates for ubiquitination, and its function is believed to be the generation of amino acids necessary for Legionella growth through the proteasomal degradation of its targets (Ensminger & Isberg, 2010; Lomma et al, 2010; Price et al, 2011). Interestingly, AnkB has also been shown to be targeted for K11‐linked polyubiquitination by a host E3 ligase, although this modification does not affect AnkB stability and its function remains unknown (Bruckert & Abu Kwaik, 2016).

It is noteworthy that other F‐box‐containing bacterial effectors may serve nontraditional roles. For example, although LicA can bind to Skp1, it fails to integrate into the full SCF complex as would be required for ubiquitin ligase activity. Furthermore, despite encoding F‐box motifs, PpgA and Lpg2525 fail to interact with Skp1, raising questions as to whether they play some other role or require an additional layer of regulation. The extent to which other human pathogens utilize F‐box proteins for ubiquitin ligation or other functions is an open question. The pervasiveness of F‐box proteins among plant pathogens would suggest that their existence in human pathogens could be more widespread than currently appreciated.

The rule breakers

Xanthomonas campestris XopL

The presence of a substrate‐binding leucine‐rich repeat domain shared among another family of bacterial E3 ligases (NEL family, discussed below) led to the identification of novel ubiquitin ligase activity in the effector protein XopL from the hemibiotrophic plant pathogen Xanthomonas campestris (Singer et al, 2013). The ligase activity was isolated within a cysteine‐free C‐terminal domain, the structure of which revealed no relationship to eukaryotic RING‐type ligases. In fact, this domain showed no structural homology whatsoever and was coined as “XL‐box.” The lack of possible catalytic cysteine residues implies a RING‐type mechanism, but aside from the identification of several E2‐interacting residues, nothing is known about how the XopL XL‐box catalyzes ubiquitin conjugation (Fig 1C; Singer et al, 2013). Only some E2 residues that are required for RING‐type E3 ligases are also required for XopL. The E3 ligase activity of XopL has been linked to a striking phenotype in plant stromule formation, and has been reported to play a role in promoting proteasomal degradation of host autophagy machinery despite being targeted for selective autophagy‐mediated degradation itself (Erickson et al, 2018; Leong et al, 2022).

Legionella pneumophila SidE family E3 ligases

The roles of many of the Dot/Icm type IV secretion system effectors in L. pneumophila are not fully understood because, when deleted individually, these genes have no impact on bacterial replication inside the host cell. However, early secretion of the SidE family of effectors (SdeA, SdeB, SdeC, and SidE) is required for establishing the LCV and efficient intracellular replication (Luo & Isberg, 2004; Bardill et al, 2005). The function of SidE effectors remained a mystery until over a decade later when they were discovered to constitute noncanonical, ATP/E1/E2‐independent ubiquitin ligases (Qiu et al, 2016). SidE effectors were shown to catalyze ubiquitination in a manner that requires NAD⁺ and a conserved mono‐ADP‐ribosyltransferase (mART) domain. Subsequent work provided a full picture of this reaction, whereby Arg42 of ubiquitin is ADP ribosylated by the SidE mART domain, producing ADPr‐Ub that is then processed by the neighboring phosphodiesterase (PDE) domain to release AMP and produce phosphoribosyl‐Ub (Pr‐Ub), which can, in turn, be released or transferred directly onto a substrate serine residue (Fig 1F; Bhogaraju et al, 2016; Kotewicz et al, 2017). If Pr‐Ub is released, it becomes highly cytotoxic as it disrupts normal ubiquitination processes (Bhogaraju et al, 2016).

A series of follow‐up studies provided molecular detail to the reaction by capturing structural snapshots of nearly every reaction step (Akturk et al, 2018; Dong et al, 2018; Kalayil et al, 2018; Kim et al, 2018; Wang et al, 2018). A cleft within the mART domain forms a low‐affinity binding site for ubiquitin, making contacts primarily with Arg72 and Arg74 near the C‐terminus and positioning Arg42 near the NAD⁺‐binding site. Within the PDE domain, a catalytic histidine forms a covalent intermediate with Pr‐Ub that can then be hydrolyzed or transferred onto a serine residue contained within an unstructured region of the substrate. Mutations within this substrate‐binding site indicate that substrate ubiquitination plays a more relevant role during Legionella infection than release of Pr‐Ub (Kalayil et al, 2018). Recent data demonstrate that SdeC can also modify tyrosine residues, raising the possibility that SidE family members may target distinct substrates (Zhang et al, 2021). Surprisingly, the catalytic centers of the mART and PDE domains are quite distant from each other, suggesting they act independently and the ADPr‐Ub intermediate relies on diffusion instead of direct transfer. There are some discrepancies, however, as to whether the orientation of the two domains can be perturbed (either through mutation or truncation) without repercussions to activity (Akturk et al, 2018; Wang et al, 2018). Since SidE effectors are localized to the exterior of the LCV, local diffusion of reaction intermediates like the ADPr‐Ub is plausible (Bardill et al, 2005).

SidE ligases have been shown to ubiquitinate a variety of targets including ER‐associated Rab GTPases and reticulons, Golgi tethering proteins, and a range of other proteins involved in mitochondrial metabolism, autophagy, and proteasomal degradation (Qiu et al, 2016; Kotewicz et al, 2017; Wan et al, 2019a; Shin et al, 2020b). This links SidE activities to established roles in remodeling the ER (Shin et al, 2020b; preprint: Kim & Isberg, 2023), regulating autophagy (De Leon et al, 2017), and disrupting Golgi function (Liu et al, 2021). While SidE effectors also encode a CE‐clan DUB (deubiquitinase) domain at their N‐termini that modulates ubiquitin signals at the LCV (Sheedlo et al, 2015, 2021), products of SidE's noncanonical ubiquitination are resistant to these and other conventional DUBs (Puvar et al, 2017). Instead, SidE‐catalyzed ubiquitination is regulated by two specialized Legionella effectors that act as deubiquitinases for Pr ubiquitination (DUPs; Wan et al, 2019a; Shin et al, 2020b). DupA and DupB encode PDE domains related to those found in the SidE ligases, but display a higher affinity for ubiquitin that allows them to reverse the reaction, producing an unmodified serine residue and soluble Pr‐Ub.

When expressed on their own in yeast or human cell culture, SidE proteins are highly toxic (Havey & Roy, 2015; Jeong et al, 2015; Bhogaraju et al, 2016). Interestingly, it is specifically mART function and not DUB activity or PDE activity that inhibits yeast growth, suggesting ADPr‐Ub‐dependent stress on the ubiquitin system (Havey & Roy, 2015; Wan et al, 2019a). While neither DupA nor DupB can rescue yeast cells that co‐express SdeA, expression of another Legionella effector called SidJ can (Havey & Roy, 2015; Jeong et al, 2015; Wan et al, 2019a). SidJ is located near the SidE family genes in the Legionella genome, and while originally it was thought to reverse SidE‐dependent ubiquitination, it was later shown to inhibit SidE ligase activity (Qiu et al, 2017; Bhogaraju et al, 2019; Black et al, 2019; Sulpizio et al, 2019; Gan et al, 2019b). When activated by host calmodulin, the pseudokinase fold of SidJ catalyzes polyglutamylation on a conserved glutamate residue within the SidE mART domain that is required for ubiquitin ADP ribosylation. Structures of SidJ:SidE complexes reveal the mechanistic basis for this activity, whereby the target SidE glutamate is AMPylated in an activation step prior to nucleophilic attack by a free glutamate amino acid (Adams et al, 2021; Osinski et al, 2021). A related Legionella effector, SdjA, can perform the same function but exhibits target preference among the SidE family members (Osinski et al, 2021). During Legionella infection, concentrations of SidJ accumulate and inhibit SidE effectors within the first few hours post‐infection (Jeong et al, 2015). Thus, DupA, DupB, SidJ, and SdjA provide Legionella with finely tuned spatial and temporal control over noncanonical SidE ligase function.

In sum

The combination of sequence (e.g., LubX, GobX, and F‐boxes), structural (e.g., AvrPtoB and NleG), and functional (e.g., RavN) approaches that have been required to identify bacterial RING‐type E3 ligases nicely illustrates the divergence of these enzymes. Still, despite significant sequence diversity, bacterial RING‐type ligases follow the rules of eukaryotic examples by retaining a conserved E2‐binding site, functioning with a similar set of host E2s, and encoding a catalytic linchpin residue to activate the E2~Ub conjugate. The striking convergence of the distinct XopL XL‐box fold upon cysteine‐independent ubiquitin ligation and remarkable E1/E2/ATP‐independent activity of the SidE ligase family raise a series of new questions on the existence of additional examples, their functions, and how they have evolved mechanisms of ubiquitin ligation that break the rules established in eukaryotes.

Ubiquitin ligation by cysteine‐dependent E3 ligases

The rule makers

Unlike RING‐type E3 ligases that facilitate direct transfer from activated E2~Ub to substrate, cysteine‐dependent E3 ligases first receive ubiquitin from the E2 onto an internal active‐site cysteine, forming one final thioester intermediate before ultimately transferring ubiquitin onto a substrate. Humans encode ~ 40 known cysteine‐dependent E3 ligases, the bulk of which fall into the HECT (homologous to E6‐AP C‐terminus) and RBR (RING‐in‐between‐RING) families (Lorenz, 2018; Cotton & Lechtenberg, 2020), but also include the newly described RING‐Cys‐relay (RCR) and RNF213‐ZNFX1 (RZ) families (Pao et al, 2018; Otten et al, 2021). Although structurally each family is unique, a common theme among cysteine‐dependent E3 ligases is a multidomain architecture that allows flexibility between an E2‐binding domain and a catalytic domain that presents the active‐site cysteine. For example, HECT E3 ligases are characterized by an E2‐binding N‐lobe that is flexibly linked to a catalytic C‐lobe (Fig 2A). The movement between these domains is important for recruitment of an activated E2~Ub conjugate and subsequent transfer of ubiquitin onto the E3 active site. The resulting E3~Ub thioester intermediate can then carry out substrate ubiquitination and, in some cases, polyUb chain formation (Fig 2D). One key property allowed by HECTs and other cysteine‐dependent ligase mechanisms is that it imparts control over both substrate and polyUb chain specificity to the E3.

Figure 2. Cysteine‐dependent ubiquitin ligases.

(A) Structure of the human HECT E3 ligase NEDD4L (PDB 3JVZ; Kamadurai et al, 2009), with the E2‐binding region and active site colored in blue and gold, respectively. (B) Structure of the Salmonella Typhimurium HECT‐like E3 ligase SopA (PDB 2QYU; Diao et al, 2008), highlighting topologically analogous E2‐binding region and active site. (C) Structure of the Salmonella Typhimurium NEL SspH2 (PDB 3G06; Quezada et al, 2009), highlighting its distinctive tri‐lobed fold (N‐, M‐, and C‐lobe), E2‐binding region, and active site. (D) Mechanism of HECT E3 ligases, highlighting flexibly‐linked N‐ and C‐lobes that facilitate docking and transfer of ubiquitin from an E2, as well as activation of ubiquitin for substrate ubiquitination and/or polyUb chain formation (Kamadurai et al, 2009; Jäckl et al, 2018; preprint: Franklin et al, 2023). (E) Mechanism of bacterial HECT‐like E3 ligases, highlighting common themes including interlobe flexibility and E2 transthiolation (Diao et al, 2008; Lin et al, 2012). (F) Mechanism of bacterial SNL E3 ligases, highlighting distinctive bilobed topology and E2~Ub binding (Hsu et al, 2014; Wasilko et al, 2018).

The HECT ligase family consists of 28 members, each sharing a highly conserved HECT domain that contains the E2‐interacting N‐lobe, the active‐site cysteine‐containing C‐lobe, and a small, flexible linker domain that bridges the N‐ and C‐lobes (Lorenz, 2018). All of the HECT ligases encode their HECT domain at the C‐terminus and, consequently, feature their substrate‐binding regions upstream of the HECT domain. While the HECT domain is highly conserved, the type and number of domains upstream of the HECT domain vary greatly. The presence of certain accessory domains is the basis for categorizing the HECT subfamilies, such as the C2 and WW domains that characterize the NEDD4 family. Among the well‐characterized HECT ligases, there appears to be a predominant specificity for assembly of either K48‐ or K63‐linked polyUb chains.

The rule followers

Bacterial HECT‐like E3 ligases

The bacterial HECT‐like E3 ubiquitin ligases are presently composed of just two members: SopA from Salmonella typhimurium and NleL from EHEC. SopA, before being described as a HECT‐like ubiquitin ligase, was primarily appreciated for its role among several bacterial effectors in inducing inflammatory responses and enteritis during Salmonella infection (Wood et al, 2000; Zhang et al, 2002). Another study reported that SopA shared 29% amino acid homology to a protein of unknown function from EHEC, later to be renamed EspX7 and, finally, NleL (Layton et al, 2005; Tobe et al, 2006; Lin et al, 2011). A yeast two‐hybrid screen for host interacting proteins identified the RING‐type E3 ligase HsRMA1 (Zhang et al, 2005). During the characterization of SopA and HsRMA1, an unexplainable ubiquitin ligation activity led to the discovery that SopA is an E3 ligase with a cysteine active site 30 residues upstream of its C‐terminus, a signature shared by all eukaryotic HECT ligases (Zhang et al, 2006). Despite poor sequence and structural homology, crystal structures of SopA and NleL revealed topological mimicry of the HECT‐like domain to eukaryotic HECT domains, with two distinct lobes joined by a linker (Fig 2B; Diao et al, 2008; Lin et al, 2011). Based on sequences and structural features of NleL and SopA, the bacterial HECT‐like ligase family has recently been expanded to include examples from additional human and plant pathogens (preprint: Franklin et al, 2023).

Subsequent crystal structures bound to the E2 UBE2L3 revealed additional similarities to eukaryotic HECT E3s. Both SopA and NleL contact UBE2L3 within their N‐lobes, each making contact with Phe63 of UBE2L3, a residue known to be critical for interaction with eukaryotic HECT domains (Nuber & Scheffner, 1999; Kamadurai et al, 2009; Lin et al, 2012). One notable deviation from eukaryotic HECTs is that the linker domain bridging the E2‐binding N‐lobe and catalytic C‐lobe is considerably longer in NleL and SopA, which encode an ~ 23 residue linker relative to 3–4 amino acids in eukaryotic HECTs (Lorenz, 2018). Comparison of apo and E2‐bound structures showed that the NleL C‐lobe adopted an entirely different conformation with respect to the N‐lobe, suggesting flexibility within the linker domain (Lin et al, 2012). Coupled with structural work on catalytic intermediates (preprint: Franklin et al, 2023), these observations indicate that not only do NleL and SopA share the architecture of eukaryotic HECT domains but also some of the dynamic features associated with HECT‐type ligation (Fig 2E).

Like eukaryotic HECT E3s, NleL and SopA also encode their HECT domains as the very C‐terminal domain, with the substrate‐binding regions located upstream (Lorenz, 2018). While SopA appears to preferentially ligate K48‐linked polyUb chains (Fiskin et al, 2017), similar to many eukaryotic HECTs, NleL demonstrates the unusual ability to ligate K6‐linked polyUb with equal preference as K48 polyUb, a specificity that is not observed to such an extent in eukaryotic HECTs (Lin et al, 2011; Jäckl et al, 2018). Recent work found these two polyUb specificities to be separable by mutation, providing tools to address whether the two activities play distinct roles during EHEC infection (preprint: Franklin et al, 2023). Although SopA was first reported to target HsRMA1 (Zhang et al, 2005), two studies subsequently identified the host RING‐type ubiquitin ligases TRIM56 and TRIM65, components of interferon signaling in innate immune responses, as additional targets of ubiquitination (Kamanova et al, 2016; Fiskin et al, 2017). The first study found these modifications to be nondegradative and stimulate interferon responses, while the second study concluded they were degradative, consistent with SopA's specificity for the degradative K48 polyUb signal. Interestingly, the substrate‐binding region was mapped to SopA's beta‐barrel domain, located immediately upstream of the HECT domain (Fiskin et al, 2017). NleL, on the other hand, has been observed to target at least one of its substrates using the disordered region upstream of its beta barrel (Sheng et al, 2017). NleL has been studied for its contributions to the attaching and effacing mechanism of EHEC, but conflicting results make the exact function of NleL in this pathway unclear (Piscatelli et al, 2011; Sheng et al, 2017). Additionally, another recent study provided evidence that NleL may be targeting components of the NF‐κB immune signaling pathway (Sheng et al, 2020).

Interestingly, the effector protein TRP120 from Erlichia chaffennesis was recently reported to also have ubiquitin ligase activity dependent on a cysteine residue located ~ 30 amino acids upstream from its C‐terminus (Zhu et al, 2017). While TRP120 otherwise has no identifiable HECT‐like features and thus may represent an unresolved rule breaker, future work on TRP120 may hold interesting insights into its unexpected ubiquitin ligase activity. The discovery of TRP120, alongside NleL and SopA, indicates that more HECT‐like and cysteine‐dependent ligases may exist in the effector repertoires of other pathogenic bacteria.

The rule breakers

NEL E3 ligases

While the IpaH protein family in Shigella flexneri had already been established as type III secretion system effectors (Demers et al, 1998), it was not apparent that they constituted a “novel E3 ligase” (NEL) family until IpaH9.8 was shown to direct the proteasomal degradation of MAPKK Ste7 in a yeast surrogate model (Rohde et al, 2007). Sequence analysis showed no homology to known eukaryotic E3 ligases, but identified an active‐site cysteine residue that is perfectly conserved among all 12 S. flexneri IpaH family members as well as in another three related effectors in S. Typhimurium called SspH1, SspH2, and SlrP, and a series of other orthologs in bacteria that include several Pseudomonas species. These examples consist of leucine‐rich repeat (LRR) domains followed by the catalytic NEL domain. Additional members that only retain the NEL domain have since been discovered in plant pathogens, including Sinorhizobium fredii and Ralstonia solanacearum (Mukaihara et al, 2010; Xu et al, 2018). Other examples, such as YopM from Yersinia pestis and Yersinia pseudotuberculosis, have lost the catalytic NEL domain but are still proposed to act as E3 ligases through an alternative active site within the LRR domain (Wei et al, 2016).

The ~ 300‐residue catalytic domain is highly conserved among NEL effectors, and crystal structures have revealed an all‐helical fold that has no resemblance to eukaryotic or other bacterial E3 ligases (Fig 2C; Singer et al, 2008; Zhu et al, 2008; Quezada et al, 2009). The catalytic cysteine is surrounded by several conserved acidic residues that are important for activity, including an aspartate residue two positions C‐terminal from the active site that is required for polyUb chain formation (Singer et al, 2008; Zhu et al, 2008). Unlike the bilobed structure of HECT ligases, the NEL domain displays more subtle dynamics between three lobes. Recent work demonstrates that E2 binding occurs on a separate lobe from the catalytic cysteine, suggesting that interlobe flexibility is important for catalysis just as it is for HECT ligases (Keszei & Sicheri, 2017; Cook et al, 2019). The NEL domain interacts with E2s quite differently from other E3 ligases. Mutations that typically ablate activity with RING‐ and HECT‐type ligases have no effect on NEL function (Singer et al, 2008; Zhu et al, 2008), and NMR studies have revealed a distinct binding interface on the E2 as well as extensive interactions with ubiquitin in the context of the E2~Ub conjugate (Levin et al, 2010). NEL effectors are generally thought to encode K48‐linked polyUb chain specificity, although there are some reports of other products including K27‐linked and K63‐linked polyUb (Ashida et al, 2010; Wei et al, 2016; Wang et al, 2020b). How polyUb linkage specificity is achieved is not yet clear, but biochemical data suggest a “seesaw” mechanism for ubiquitin chain formation, whereby a growing polyUb chain is passed back and forth between E2 and E3 (Levin et al, 2010).

While the C‐terminal NEL domain is highly conserved, the N‐terminal LRR domains are more variable and, unsurprisingly, implicated in localization, regulation, and substrate targeting. In addition to encoding the secretion signal for translocation into the host cell, N‐termini and LRR domains influence the subcellular localization of the effector ligase activity (Quezada et al, 2009). SspH2, for example, encodes an N‐terminal palmitoylation site that localizes it to the host plasma membrane (Hicks et al, 2011). LRR domains also play an important role in regulating activity of the NEL domain through intramolecular contacts that limit polyUb chain formation (Singer et al, 2008; Quezada et al, 2009; Chou et al, 2012; Keszei & Sicheri, 2017). Substrate binding to the LRR opens the interdomain structure and relieves auto‐inhibition (Keszei et al, 2014; Zouhir et al, 2014). Without auto‐inhibition, the constitutive ligase activity leads to NEL auto‐ubiquitination, which, in cells, decreases the protein half‐life via proteasomal degradation (Chou et al, 2012). The versatility of the LRR domain leads to binding and regulation of a broad set of host targets, many of which are subject to proteasomal degradation following NEL ubiquitination (Rohde et al, 2007; Bernal‐Bayard & Ramos‐Morales, 2009; Ashida et al, 2010, 2013; Bhavsar et al, 2013; Wang et al, 2013; Suzuki et al, 2014; de Jong et al, 2016; Zheng et al, 2016; Li et al, 2017; Piro et al, 2017; Wandel et al, 2017; Otsubo et al, 2019; Hansen et al, 2021; Luchetti et al, 2021). Some of these LRR:substrate interactions have now been resolved in molecular detail (Keszei et al, 2014; Zouhir et al, 2014; Ji et al, 2019; Ye et al, 2020; Liu et al, 2022; Wang et al, 2023; Zhong et al, 2023). Cellular pathways and substrates targeted by other NEL effector E3 ligases remain to be determined, and mechanisms of Ub ligation by NELs, including determinants of polyUb chain specificity, remain unknown.

Legionella pneumophila SidC/SdcA/SdcB

A number of the Dot/Icm type IV‐secreted proteins contribute to the formation of the specialized LCV, which provides a suitable environment for Legionella intracellular replication. A prominent feature of the LCV is the enrichment of polyUb and endoplasmic reticulum (ER) markers around the vacuole. The effector SidC and its paralog SdcA localize to the cytosolic surface of the LCV by virtue of a highly specific interaction with phosphatidylinositol‐4‐phosphate (PI(4)P; Luo & Isberg, 2004; Weber et al, 2006). In the absence of SidC/SdcA, the LCV loses both the polyUb coat as well as ER markers (Ragaz et al, 2008; Hsu et al, 2014).

Three groups independently solved crystal structures of the SidC N‐terminal region consisting of a bilobed architecture formed by a small, all‐helical domain inserted into a larger protein fold (Gazdag et al, 2014; Horenkamp et al, 2014; Hsu et al, 2014). These structures revealed a novel fold with no known structural homology. Sequence conservation highlighted a putative Cys/His/Asp active site reminiscent of those observed in DUBs, but functional assays instead revealed a cryptic cysteine‐dependent E3 ligase activity (Hsu et al, 2014). As a result, this SidC N‐terminal ubiquitin ligase region was renamed the SNL domain. Full SNL activity required all three components of the active site, although only the cysteine was essential. Subsequent structural work on SidC and SdcA with E2 and E2~Ub bound highlighted a role for the insertion (INS) domain in binding E2 and, by virtue of a flexible linker, orienting it near the active site contained within the larger SNL domain (Fig 2F; Luo et al, 2015; Wasilko et al, 2018). This bilobed architecture separating E2 binding from a catalytic subdomain is analogous to HECT E3 ligases, although the relative sizes of these domains are swapped in the case of SNL enzymes. The insertion domain binds to E2s at the typical E3‐binding interface, and mutations such as R6A and F63A in UBE2L3 hinder ligase activity. Although highly similar, SidC and SdcA exhibit preferred function with distinct E2 enzymes. In the E2~Ub‐bound structure, the ubiquitin is sandwiched between the main and insertion domains, with extensive contacts to the C‐terminus that lock it into an extended conformation reminiscent of activated E2 and E3 enzymes in eukaryotes.

C‐terminal to the SidC and SdcA SNL domains lies the PI(4)P‐binding domain (P4C) responsible for localization to the LCV. A crystal structure of nearly full‐length SidC revealed an intramolecular interaction between P4C and the SNL domain that results in occlusion of the active site (Luo et al, 2015). In addition to subcellular localization, PI(4)P binding into the P4C relieves the auto‐inhibition and enhances E3 ligase activity. Interestingly, a third Legionella effector E3 ligase in this family, SdcB, was identified by sequence similarity to the SNL domain but encodes an ankyrin repeat instead of the P4C, suggesting it may exhibit distinct localization and regulatory properties (Lin et al, 2018). Regulation of SidC/SdcA through auto‐inhibition would suggest that the timing or locale is particularly important for Legionella biology. SidC/SdcA have been associated with ubiquitination of Rab1 and Rab10 following their recruitment to the LCV (Horenkamp et al, 2014; Jeng et al, 2019). While the product of SidC/SdcA in vitro is a combination of K11‐ and K33‐linked polyUb chains, Rab1 and Rab10 are primarily monoubiquitinated (Horenkamp et al, 2014; Hsu et al, 2014; Jeng et al, 2019; Liu et al, 2020). Direct ubiquitination of Rab10 has been reported in vitro, however, and this process can be reversed by the Legionella DUB LotC (Liu et al, 2020). Further work on targets and specificities of Legionella SidC, SdcA, and SdcB ligases will be important to understanding their pivotal role in maturation of the LCV and interplay with other effectors.

In sum

Cysteine‐dependent effector E3 ligases are highly potent weapons of virulence. Not only are they extremely active, but they also direct the type of polyUb signal that is appended onto the substrate. This gives them access to discrete areas of host cell biology, including proteasomal degradation (e.g., K48 polyUb specificity among the NEL family) as well as other pathways that are yet to be resolved (e.g., K6 polyUb activity of NleL). It is no wonder that bacteria like Salmonella, Shigella, and Legionella have evolved and expanded upon novel E3 ligase scaffolds. Whether other bacterial pathogens have convergently evolved their own effector E3 ligases, how they might function, and in what ways they might contribute to infection remain to be seen.

Ubiquitin hydrolysis by CA‐clan deubiquitinases

The rule makers

Ubiquitin signaling is reversed through the action of specialized proteases, termed deubiquitinases (DUBs), that hydrolyze the ubiquitin linkage and release it for recycling. Humans encode ~ 100 DUBs that fall into seven structurally distinct families. Aside from the one metalloprotease family (JAMM/MPN), the majority of DUB families (USP, UCH, OTU, MJD/Josephin, MINDY, and ZUP1) hydrolyze ubiquitin using a cysteine‐dependent mechanism and belong to the papain‐like CA clan of proteases. CA proteases are structurally related and, with few exceptions, encode a catalytic triad in the sequential order of Cys, His, and Asn/Asp (Rawlings & Barrett, 1993; Fig 3A). DUB activities can be directed toward specific cellular roles through regions within or outside of the catalytic domain that affect aspects such as localization, regulation, and DUB–target interactions (Mevissen & Komander, 2017). Through a variety of different mechanisms, DUBs can also demonstrate varying degrees of specificity for distinct polyUb chain types. Many of the polyUb linkage‐specific examples belong to the OTU family of DUBs (Mevissen et al, 2013), which typically directs chain specificity through Ub‐binding sites (termed the S1 and S1′ sites) on either side of the catalytic center that select a particular diUb arrangement (Fig 3D). Each DUB family uses characteristic structural modules that form a Ub‐interacting S1 site. Among OTU family members, for example, three common structural regions form the S1 site: the arm (VR1), the β‐sheet edge (VR2), and an extended β‐turn (VR3) that together recognize common interaction surfaces on ubiquitin (Schubert et al, 2020). S1′ sites, if encoded at all, are much more variable in their modes of orienting the ubiquitinated target (ubiquitin itself, in the case of polyUb specificity) for hydrolysis.

Figure 3. CA‐clan deubiquitinases.

(A) Structure of the human OTU deubiquitinase OTULIN (PDB 3ZNV; Keusekotten et al, 2013; Rivkin et al, 2013), with Ub‐binding sites and the catalytic triad colored in blue and gold, respectively. (B) Structure of the Legionella pneumophila OTU deubiquitinase LotC (PDB 7BU0; Liu et al, 2020; Shin et al, 2020a), with analogous Ub‐binding sites and catalytic triad colored in blue and gold, respectively. (C) Structure of the Legionella pneumophila deubiquitinase RavD (PDB 6NII; Wan et al, 2019b), highlighting its distinctive CA‐clan fold yet related Ub‐binding regions (blue) and catalytic triad (gold). (D) Mechanism of diUb proteolysis by eukaryotic OTU deubiquitinases (Mevissen & Komander, 2017). (E) Mechanism of diUb proteolysis by bacterial OTU deubiquitinases, highlighting a common route with distinctive features (Schubert et al, 2020). (F) Mechanism of UBE2N inactivation by Legionella pneumophila MavC, a CA‐clan effector that has evolved transglutaminase function. MavC forms a thioester intermediate with the ubiquitin Gln40 side chain before facilitating transfer onto a UBE2N lysine residue (Gan et al, 2019a). This process is reversed by the related Legionella pneumophila effector, MvcA (Gan et al, 2020).

The rule followers

Chlamydia pneumoniae OTU deubiquitinase

The first bacterial OTU was predicted in Chlamydia pneumoniae and other members of the Chlamydophila lineage based on primary sequence (Makarova et al, 2000). The Chlamydia OTU (ChlaOTU) was only later experimentally shown to hydrolyze polyUb chains in vitro and play a role in removing ubiquitin signals at Chlamydia entry sites (Furtado et al, 2013). Interestingly, this study observed DUB activity in full‐length ChlaOTU purified from transfected HeLa cells, whereas a more recent study failed to observe any activity against multiple substrates using a shorter construct purified from E. coli (Schubert et al, 2020). At the time, the remainder of the catalytic triad motif was not immediately apparent from sequence analysis (Schubert et al, 2020). Based on a recent AlphaFold2 model, however, ChlaOTU appears to encode an extremely large inserted domain within VR‐1, which displaces catalytic residues on either end of the protein, separated by ~ 750 amino acids (Varadi et al, 2022).

Legionella pneumophila OTU deubiquitinases

A family of related Legionella OTU (LOT) DUBs has recently been identified by sequence homology (Pruneda et al, 2023). The first example, LotA (lpg2248), contains two active OTU‐like domains, the first of which is capable of hydrolyzing K6‐linked polyUb, while the second appears to prefer longer K48‐linked and K63‐linked polyUb chains (Kubori et al, 2018; Luo et al, 2022; Takekawa et al, 2022; Kang et al, 2023; Warren et al, 2023). Interestingly, only the second LotA OTU domain appears to be important for Legionella growth in mouse bone marrow‐derived macrophages, although only in a background where the SidE family ligases are deleted. LotA was shown to localize to the outside of the LCV, where its dual DUB activities act to edit the coat of ubiquitin modifications and function to restrict localization of the AAA ATPase VCP (Kubori et al, 2018; Warren et al, 2023). The related LOT DUBs, LotB (lpg1621) and LotC (lpg2529), have also been shown to localize to the LCV in order to edit the ubiquitin coat (Liu et al, 2020; Ma et al, 2020; Shi et al, 2023a). LotB demonstrates a preference for K63‐linked polyUb and acts to remove this signal type from the LCV as part of establishing a replicative endoplasmic reticulum‐like niche (Hermanns et al, 2020; Kitao et al, 2020; Ma et al, 2020; Schubert et al, 2020; Shin et al, 2020a). LotC targets a wide range of polyUb signals, including K6‐, K11‐, K33‐, K48‐, and K63‐linked polyUb chains, and removes ubiquitin modifications catalyzed by the Legionella E3 ligase SidC (Liu et al, 2020; Shin et al, 2020a). A fourth member of the LOT DUB family, which we coin LotD (lpg0227), also shows a range of activities toward K6‐, K11‐, K33‐, K48‐, and K63‐linked polyUb, but its localization and function during Legionella infection have yet to be studied (Schubert et al, 2020).

Crystal structures have been determined for LotB, LotC, and both OTU domains of LotA (Liu et al, 2020; Ma et al, 2020; Shin et al, 2020a; Luo et al, 2022; Takekawa et al, 2022; Kang et al, 2023; Warren et al, 2023). These structures reveal a typical OTU fold that orients the linear arrangement of Cys, His, and Asn/Asp residues into a folded catalytic triad (Fig 3B). The first OTU domain of LotA is distinctive for its stringent specificity for K6‐linked polyUb, the molecular basis for which was shown to be substrate‐assisted catalysis (Warren et al, 2023). A common characteristic of the LOT DUB family is a large (~ 180 amino acid) insertion in VR‐1 of the S1 Ub‐binding site (Schubert et al, 2020). To date, only structures of LotC and the first OTU domain of LotA have been determined in complex with ubiquitin or diUb (Liu et al, 2020; Luo et al, 2022; Warren et al, 2023), but these structures as well as modeling efforts for LotB and LotA all indicate that the inserted helical domain is a critical component of the S1 Ub‐binding site (Ma et al, 2020; Shin et al, 2020a; Takekawa et al, 2022). Although the scale of this Ub‐binding helical insertion is thus far unique to LOT DUBs, this adaptation of VR‐1 within the OTU fold follows a general pattern for adaptability of the S1 site (Schubert et al, 2020). Interestingly, a structural relationship has recently been established between the LOT insertion domains and a related Ub‐binding domain from a DUB encoded by Orientia tsutsugamushi. This domain adapts Ub‐binding sites within a variable region of an otherwise common fold, and was therefore named an “adaptive Ub‐binding domain” (A‐UBD; Warren et al, 2023).

Other OTU deubiquitinases

Sequence‐based approaches have also enabled the identification of bacterial OTU DUBs outside of Chlamydia and Legionella species. Active OTU DUBs have been described in Burkholderia ambifaria (BurkOTU), Rickettsia massiliae (RickOTU), and Wolbachia pipientis (wMelOTU; Schubert et al, 2020). Interestingly, a related Wolbachia pipientis OTU domain (wPipOTU) exhibited no activity against a panel of ubiquitin and Ub‐like substrates in vitro, suggesting that it is either a pseudoenzyme, has altered target specificity, or requires a host cofactor for activity. Against simplified fluorescent substrates, all three bacterial OTUs showed a strong preference for targeting ubiquitin over Ub‐like modifiers. Similar to the substrate‐induced catalysis mechanism used by Legionella LotA and the human OTU DUB called OTULIN (Keusekotten et al, 2013), BurkOTU is only active against polyUb and not monoUb substrates. Among a panel of all eight amide‐linked polyUb chains, BurkOTU, RickOTU, and wMelOTU all cleaved a combination of K6‐, K11‐, K48‐, and K63‐linked polyUb chains, with some minor preferences among these.

Structural analysis of wMelOTU confirmed its relationship to the OTU family, including the canonical Cys, His, and Asn/Asp topology of the catalytic triad (Schubert et al, 2020). While an apo crystal structure of wMelOTU showed a misaligned active site and unstructured S1 Ub‐binding site, a Ub‐bound structure showed structural rearrangements in both sites that favor DUB activity (Fig 3E; Schubert et al, 2020). The S1 site is centered around beta‐hairpin insertions in VR‐1 and VR‐3 that become ordered and embrace the bound Ub. Nearer to the active site, wMelOTU exhibits many features that follow a typical OTU fold, including conserved aromatic residues that enclose the ubiquitin C‐terminus and a well‐structured oxyanion hole to support hydrolysis.

Burkholderia pseudomallei TssM

The obligate intracellular pathogen Burkholderia pseudomallei and related Burkholderia species encode the cysteine protease TssM, which resembles the USP (ubiquitin‐specific protease) family of eukaryotic DUBs (Shanks et al, 2009; Tan et al, 2010). Crystal structures of TssM revealed a catalytic domain related to eukaryotic USPs, with deviations primarily residing within the Ub‐binding S1 site (preprint: Szczesna et al, 2023; preprint: Hermanns et al, 2023). In vitro, TssM was shown to be an active DUB with specificity toward ubiquitin over a panel of related Ub‐like modifiers and the capacity to cleave (at least) K48‐ and K63‐linked polyUb chains (Shanks et al, 2009). During infection, TssM is secreted by the type II secretion system and has been proposed to suppress NF‐κB activation by reversing TRAF3, TRAF6, and IκBα ubiquitination (Tan et al, 2010; Burtnick et al, 2014). More recently, the DUB activity of TssM was found to counteract noncanonical, ester‐linked ubiquitination of bacterial lipopolysaccharide (LPS) directed by the host E3 ligase RNF213, which otherwise earmarks cytosolic bacteria for destruction by xenophagy (Otten et al, 2021; preprint: Szczesna et al, 2023). Enzymatic assays of TssM demonstrated ~ 30‐fold higher activity for ester‐linked ubiquitin over canonical isopeptide linkages, possibly explaining why TssM was unique among a panel of bacterial DUBs tested for their ability to reverse LPS ubiquitination (preprint: Szczesna et al, 2023).

The rule breakers

Escherichia albertii EschOTU

Among the more typical, rule‐following bacterial OTU domains predicted by Schubert et al (2020), an example from Escherichia albertii (EschOTU) was identified that, by sequence analysis of OTU motifs, appeared to display a permutated threading of the catalytic triad residues. In vitro, EschOTU was shown to be active and demonstrated a preference for K6‐, K11‐, K48‐, and K63‐linked polyUb. A crystal structure of EschOTU confirmed its altered sequence topology, which deviates from the Cys, His, and Asn/Asp arrangement that defines the CA protease clan and instead adopts a His, Asn, and Cys arrangement that is more similar to the CE protease clan. In three‐dimensional structure, EschOTU very closely resembles a typical OTU fold, but the permutated sequence shifts the N‐ and C‐termini to the opposite side of the domain, near VR‐1. Following this topology, an artificially permutated sequence of a viral OTU DUB was shown to retain proper folding and enzymatic activity, demonstrating a unique plasticity of this protein fold to altered sequence threading (Schubert et al, 2020). In fact, it has been proposed previously that the CA and CE protease clans diverged through an ancient sequence permutation event (Rawlings & Barrett, 1993; Fox et al, 2014). Thus, while EschOTU breaks the rules set by eukaryotic OTU domains, it may resemble an intermediary or ancestor to the CA and CE protease clans.

Bacterial–viral tegument‐like deubiquitinases

The large tegument proteins of α, β, and γ herpesviruses contain a CA‐clan DUB domain that is distinct in sequence and structure from any DUBs found in humans (Kattenhorn et al, 2005). These DUBs demonstrate a preference for K48‐linked polyUb, and structural characterization reveals an unusual Cys, Asp, and His arrangement of the catalytic triad (Schlieker et al, 2007). While initially thought to be unique to Herpesviridae, a recent bioinformatic analysis detected additional members from this viral tegument‐like DUB (VTD) family in animals, fungi, protists, and the bacterium Waddlia chondrophila, a Chlamydia‐like intracellular pathogen (Erven et al, 2022). The study found two Waddlia VTD genes, Wc‐VTD1 (wcw_1294) and Wc‐VTD2 (wcw_1327). Both Wc‐VTD proteins were found to be specific for ubiquitin over Ub‐like modifiers, but exhibited distinct preferences among polyUb chains: Wc‐VTD1 strongly preferred K6‐linked polyUb, while Wc‐VTD2 targeted K63‐linked polyUb instead. A structure of Wc‐VTD1 confirmed the distinctive VTD‐type DUB fold, revealed structural determinants for ubiquitin binding into the S1 site, and indicated a substrate‐assisted catalysis mechanism underlying the specificity for K6‐linked polyUb (Erven et al, 2022). Interestingly, the specificity of Wc‐VTD1 for K6‐linked polyUb matched the specificity of a VTD family member identified in protists—the natural host of Waddlia—as well as Legionella, which encodes the K6‐linked polyUb‐specific DUB LotA (Kubori et al, 2018; Luo et al, 2022; Kang et al, 2023; Warren et al, 2023). Recently, an additional K6‐linked polyUb‐specific VTD was identified among the ~ 12 DUBs in the Chlamydia‐like bacterium Simkania negevensis, which also lives inside protist hosts (preprint: Boll et al, 2023). What K6‐linked polyUb signals might be regulating in these contexts is an open area of study. The evolution of VTD DUBs is also fascinating, as current evidence suggests branching from a general protease family, coincident with developing a reorganized active‐site topology and ubiquitin specificity. From there, VTD DUBs are believed to have spread via transposons, consistent with their somewhat sporadic occurrence across kingdoms of life (Erven et al, 2022).

Legionella pneumophila RavD

In a functional screen of over 40 bacterial cell lysates, Legionella pneumophila was shown to uniquely possess DUB activity capable of cleaving M1‐linked polyUb chains (Wan et al, 2019b). After cloning and expressing nearly 150 of the known Legionella effector proteins, RavD was identified as the responsible DUB and subsequently shown to exclusively target M1‐linked polyUb over any other chain type (Wan et al, 2019b). Through interactions with phosphatidylinositol‐3‐phosphate (PI(3)P), RavD was shown to localize to the outside of the LCV, where it edits M1‐linked polyUb to downregulate NF‐κB signaling and block endolysosomal maturation (Pike et al, 2019; Wan et al, 2019b). A RavD crystal structure revealed a classic CA‐clan protease fold with a Cys‐His‐Ser catalytic triad, but otherwise little structural homology to any eukaryotic DUBs (Fig 3C; Wan et al, 2019b). Thus, RavD represents an entirely new family of DUBs within the CA protease clan. A co‐crystal structure with M1‐linked diUb shows contacts to both proximal and distal ubiquitin moieties that place the peptide bond at the RavD catalytic center. Interestingly, despite a distinct protease fold, RavD binds diUb in a similar manner as the human M1‐specific DUB, OTULIN (Keusekotten et al, 2013; Rivkin et al, 2013). While OTULIN uses substrate‐assisted catalysis as a mechanism of achieving polyUb specificity, however, it is less clear whether RavD follows this trend. The RavD co‐crystal structure shows no structural rearrangements that suggest polyUb‐dependent activation, but recent molecular dynamics simulations indicate reorganization of the catalytic triad upon polyUb binding (Wan et al, 2019b; Schulze‐Niemand et al, 2022).

In sum

A strong foundation of mechanistic knowledge on eukaryotic CA‐clan DUBs has greatly facilitated the study of related examples in bacterial pathogens. Although there is still much to learn, many of the bacterial CA DUBs identified thus far appear to follow the rules established by the eukaryotic system, such as the arrangement of catalytic triad residues and the use of modular Ub recognition elements. However, because most of these examples in bacteria have been identified by sequence homology to eukaryotic DUB families, it remains to be seen whether “rule breakers” such as EschOTU, VTDs, and RavD are the exceptions or the norm in bacterial CA‐clan DUBs.

Ub‐like hydrolysis by CE‐clan proteases

The rule makers

The CE‐clan proteases, in contrast to the CA clan, are marked by a permutated active‐site architecture (His‐Asn/Asp‐Cys; Barrett & Rawlings, 2001). In humans, the CE protease clan is comprised of the sentrin protease (SENP) family, of which six target Ub‐like SUMO modifications and one targets the Ub‐like modifier NEDD8. Structurally, SENPs resemble DUBs from the CA clan with key exceptions in the catalytic center and S1 substrate‐binding site (Fig 4A). The difference in arrangement of catalytic residues within the primary sequence shifts the location of the His and Asn/Asp residues onto neighboring β‐strands for SENPs, compared to their typical location on the same β‐strand for CA‐clan DUBs. Exquisite specificity for the SUMO and NEDD8 Ub‐like modifiers over ubiquitin is controlled by interactions at the S1 site which, similar to CA‐clan DUBs, is modular in nature and composed of three variable regions (Pruneda et al, 2016). Further specificity among the SUMO1, SUMO2, and SUMO3 modifications are also dictated by the S1 site, whereas an S1’ site can influence specificity for the modified substrate (Fig 4D; Kunz et al, 2018). Additional layers of functional regulation are encoded by accessory domains outside of the catalytic CE fold, which affect subcellular localization and protein–protein interactions of SENPs (Kunz et al, 2018).

Figure 4. CE‐clan Ub/Ubl proteases.

(A) Structure of the human NEDD8 protease NEDP1 (PDB 2BKR; Reverter et al, 2005; Shen et al, 2005), with S1 site variable regions (VR1‐3) and catalytic triad colored in blue and gold, respectively. (B) Structure of the Legionella pneumophila CE‐clan deubiquitinase SdeA (PDB 5CRB; Sheedlo et al, 2015), highlighting an analogous S1 site and catalytic triad in blue and gold, respectively. (C) Structure of the Chlamydia trachomatis CE‐clan deubiquitinase/acetyltransferase ChlaDUB1 (PDB 6GZT; Pruneda et al, 2018), highlighting an analogous Ub‐binding S1 site (blue) but distinctive coenzyme A‐binding site (teal) that endows moonlighting functionalities. (D) Mechanism of NEDD8 proteolysis by Ub‐like proteases (Reverter et al, 2005; Shen et al, 2005). (E) Mechanism of diUb proteolysis by bacterial CE‐clan deubiquitinases, highlighting a common route with distinctive substrate‐binding sites (Sheedlo et al, 2015; Pruneda et al, 2016). (F) Mechanism of acetylation by bacterial CE‐clan acetyltransferases, typified by Chlamydia trachomatis ChlaDUB1 (Pruneda et al, 2018). Acetylation proceeds via a “ping‐pong” mechanism wherein an acetyl group is transferred onto the active site cysteine before modifying a target serine, threonine, or lysine residue (Mukherjee et al, 2006).

The rule followers

Chlamydia trachomatis ChlaDUB1 and ChlaDUB2

The development of ubiquitin activity‐based probes that can covalently capture DUB active sites provided a powerful tool for the identification of novel DUBs (Borodovsky et al, 2002). Using an HA‐tagged ubiquitin probe with a reactive “vinylmethylester” warhead installed at the C‐terminus, a novel probe‐reactive protein present in cells infected with the obligate intracellular pathogen Chlamydia trachomatis was identified (Misaghi et al, 2006). Mass spectrometry analysis identified the captured protein as CT868, later renamed ChlaDUB1 or Cdu1. Sequence analysis of ChlaDUB1 showed its relationship to the CE clan of proteases, identified a related paralog that was coined ChlaDUB2/Cdu2, and revealed conservation among all pathogenic Chlamydia species except for C. pneumoniae. Subsequent work using a modified approach was also able to capture ChlaDUB2 from infected cells using a ubiquitin activity‐based probe (Claessen et al, 2013). In contrast to the Ub‐like modifier specificities of eukaryotic CE‐clan proteases, both ChlaDUB1 and ChlaDUB2 were shown to cleave ubiquitin and, to a lesser extent, NEDD8 modifications (Misaghi et al, 2006; Pruneda et al, 2016). Among a panel of polyUb chains, ChlaDUB1 preferentially cleaved K63‐linked polyUb, indicating the possible existence of an S1’ Ub‐binding site.

Crystal structures of ChlaDUB1 and ChlaDUB2 from C. trachomatis and a related ortholog from C. abortus reveal a similar CE protease core surrounded by a more variable surface structure (Pruneda et al, 2016, 2018; Fischer et al, 2017; Ramirez et al, 2018; Hausman et al, 2020). The catalytic triads of ChlaDUB1 and ChlaDUB2 are arranged in a spatially conserved fashion, despite several structural insertions between members of the triad. Ub‐bound structures of ChlaDUB1 and ChlaDUB2 clarify the roles of these insertions as modular variable regions that contribute to the S1 Ub‐binding site (Fig 4E; Pruneda et al, 2018; Ramirez et al, 2018; Hausman et al, 2020). One such insertion at VR‐3 encodes a short α‐helix that appears to clamp down over the bound ubiquitin substrate as its C‐terminus threads underneath into the active site. Structural information on polyUb recognition outside of the ChlaDUB1 or ChlaDUB2 S1 site is limited, although biochemical data indicate the existence of an S1′ site that dictates K63‐linked polyUb preference and, at least in the case of ChlaDUB2, possibly other Ub‐binding sites that determine preferences in polyUb chain length (Pruneda et al, 2016; Hausman et al, 2020).

Salmonella Typhimurium SseL, Escherichia coli ElaD, and Shigella flexneri ShiCE

The Salmonella Typhimurium effector protein SseL was first identified to encode a CE‐clan protease domain by sequence similarity and shown to possess DUB activity important for regulation of autophagy, maintenance of the Salmonella‐containing vacuole, and macrophage killing (Rytkönen et al, 2007; Mesquita et al, 2012; Kolodziejek et al, 2019). Closely related genes were subsequently identified in diverse Escherichia coli strains and rarely among Shigella flexneri (Catic et al, 2007; Pruneda et al, 2016). All three examples were shown to be specific for ubiquitin over related Ub‐like modifiers and preferred to cleave K63‐linked polyUb (Catic et al, 2007; Rytkönen et al, 2007; Pruneda et al, 2016). A crystal structure of SseL confirmed its relationship to the CE protease clan and highlighted unique variable regions that make up the S1 Ub‐binding site (Pruneda et al, 2016). The same crystal structure also resolved a domain N‐terminal to the catalytic domain that resembled a VHS domain used in eukaryotes for protein recognition and vesicular trafficking. Indeed, the SseL VHS domain was shown to form an additional Ub‐binding site that was important for localization of DUB activity to the cytosolic surface of the Salmonella‐containing vacuole (Pruneda et al, 2016). Thus, like examples found in eukaryotes, the biological function of CE protease activities is regulated through accessory domains. Whether/how the related VHS domains present in ElaD and ShiCE regulate their functions remains to be determined.

Legionella pneumophila SdeA

SdeA and its paralogs SdeB, SdeC, and SidE, discussed previously as rule‐breaking ligases, also harbor CE‐clan protease domains at their N‐termini (Sheedlo et al, 2015). The CE domains exhibit DUB activity with a preference for Lys63‐linked polyUb and lesser ability to cleave K48‐ and K11‐linked polyUb chains. Furthermore, SdeA is capable of deNEDDylation activity in vitro. Notably, the protease activity at the N‐terminus does not impact the noncanonical ligase activity of the C‐terminal domains (Puvar et al, 2017). A crystal structure of a covalent complex between the SdeA DUB domain and a Ub‐VME (vinylmethylester) activity‐based probe revealed three distinct Ub‐interacting regions positioned at the active‐site cleft, around SdeA Tyr33, and a third site that engages Leu8 of ubiquitin with Ser29 of SdeA (Fig 4B). There are no contacts to ubiquitin Arg72, which would otherwise impart ubiquitin‐versus‐NEDD8 specificity, nor with the ubiquitin Ile44 patch, which may in part explain SdeA's reactivity with ISG15‐VS (vinyl sulfone) and NEDD8‐VS activity‐based probes. Unlike the ligase domain of SidE effectors, the DUB domain is not required for Legionella replication in amoebae, or its toxicity in yeast. It does, however, play a significant role in editing Lys63‐linked polyUb at the LCV (Sheedlo et al, 2015).

Legionella pneumophila LupA

Following the discovery of LubX (Kubori et al, 2010), an E3 ligase that targets another effector from L. pneumophila, a growing number of “metaeffectors” have been found. A systemic analysis of functional interactions between Legionella effectors in a yeast toxicity model identified nine novel metaeffectors that directly suppress the activity of other Legionella effectors (Urbanus et al, 2016). Among them was a CE‐clan DUB, termed lpg1148 or LupA (Legionella Ub‐specific protease A), that was found to abrogate the toxic effect of the Q‐SNARE‐like effector LegC3 when overexpressed in yeast. LupA possesses a canonical protease catalytic triad and has specificity for ubiquitin over Ub‐like substrates NEDD8 and SUMO (Urbanus et al, 2016; Hermanns et al, 2020). LupA was found to remove Ub from LegC3 when co‐expressed in HEK293 cells (Urbanus et al, 2016); how this affects LegC3 function remains an unanswered question.

Rickettsia bellii RickCE

The tick‐borne bacterium Rickettsia bellii also encodes a CE‐clan DUB called RickCE. The CE fold of RickCE was originally identified by sequence homology (Catic et al, 2007), but a large stretch of sequence with unknown function is inserted into the catalytic domain. It was shown to catalyze cleavage of both ubiquitin and NEDD8, and possess strong preference for Lys63‐linked polyUb like other bacterial CE proteases (Pruneda et al, 2016). Further work is needed to elucidate the biological activity of RickCE.

Wolbachia pipientis (wPip) CidB

Wolbachia is an obligate intracellular bacterium that infects arthropods and causes a condition known as cytoplasmic incompatibility, leading to a reproductive advantage among infected females. Recently, cytoplasmic incompatibility in Wolbachia pipientis has been linked to CidB, which encodes a CE‐clan DUB module, and the associated CidA (Beckmann et al, 2017). In this pair, CidB and CidA act as the toxin and antitoxin, respectively. In the absence of CidA, CidB DUB activity is toxic in yeast and linked to male insect sterility. Like other CE‐clan DUBs, CidB demonstrates a preference for K63‐linked polyUb. Interestingly, CidA does not affect CidB DUB activity, but more likely alters its localization and access to substrates (Beckmann et al, 2017; Oladipupo et al, 2023; Terretaz et al, 2023).

Orientia tsutsugamushi OtDUB

Orientia tsutsugamushi, another obligate intracellular pathogen, also encodes a CE‐clan DUB, termed OtDUB, that was identified by sequence homology (Pruneda et al, 2016; Berk et al, 2020). OtDUB follows the trend of other CE‐clan DUBs in preferring K63‐ and K48‐linked polyUb, but in this case also demonstrates a requirement for triUb chains or longer. This specificity is directed by an S2 Ub‐binding site that selects for a K63 linkage with respect to the S1 site. At the level of diUb, OtDUB becomes inhibited through binding at the S1‐S2 sites, rather than across the active site. Interestingly, OtDUB also encodes a third Ub‐binding site that demonstrates unprecedented (5 nM) affinity for ubiquitin. While this high‐affinity ubiquitin interaction in OtDUB does impact both DUB activity as well as a DUB‐independent role in membrane trafficking, a physiological function during Orientia infection remains an open question (Berk et al, 2020, 2022). The unusually high affinity, however, does make the OtDUB Ub‐binding domain a valuable research tool for ubiquitin detection and enrichment (Zhang et al, 2022).

Xanthomonas campestris XopD

XopD, from the tomato pathogen Xanthomonas campestris, encodes a CE domain that was initially identified by sequence similarity to mammalian SUMO proteases. XopD indeed possesses deSUMOylase activity and was later shown to localize to subnuclear foci, where it regulates host transcription factors (Hotson et al, 2003; Kim et al, 2008). Structural characterization of the CE domain identified mechanisms of specificity for tomato and closely related plant SUMOs over any mammalian orthologs (Chosed et al, 2007). In a later cross‐kingdom comparison of CE‐clan proteases, it was revealed that XopD also exhibits DUB activity, with preferences for Lys48‐ and Lys11‐linked polyUb (Pruneda et al, 2016). Interestingly, the unique SUMO/polyUb specificity of XopD likely reflects the evolutionary pressures of its plant host, as opposed to CE DUBs from human pathogens that predominantly prefer K63‐linked polyUb. Structures of XopD bound to tomato SUMO or ubiquitin show that the flexibility of substrate recognition is accommodated by an unstructured VR‐1 region within the S1 site that adjusts to either binding partner. XopD VR‐1 represents the first linear Ub‐binding peptide, termed the low‐complexity Ub‐binding region (LC‐UBR; Pruneda et al, 2016). Whether such an interaction with ubiquitin exists in other contexts remains to be seen. The dual activity between ubiquitin and SUMO is also unique, given their low sequence homology, and describing roles for the DUB activity in plant pathogenesis will be an interesting area of future work.

The rule breakers

CE‐clan acetyltransferases

The CE‐clan effector from Yersinia pestis, YopJ, was initially considered a SUMO protease and a DUB capable of cleaving Lys48‐ and Lys63‐linked polyUb, but later was found to be a highly active Ser/Thr acetyltransferase (Orth et al, 2000; Zhou et al, 2005; Mittal et al, 2006; Mukherjee et al, 2006; Pruneda et al, 2016). Similar results were found for two other CE‐clan effectors from Legionella pneumophila and Salmonella Typhimurium, LegCE and AvrA, which both show no proteolytic, but rather acetylation activity in vitro (Jones et al, 2008; Pruneda et al, 2016). YopJ and AvrA also lack a highly conserved tryptophan in the CE‐clan fold that is required for Ub/Ub‐like protease function, a sequence divergence that is consistent with their roles as dedicated acetyltransferases. Interestingly, both YopJ and AvrA target serine and threonine residues for acetylation, as opposed to canonical lysine residues, and require activation by the host cofactor IP6 (Mittal et al, 2006, 2010; Mukherjee et al, 2006; Jones et al, 2008).

A broader study of bacterial CE‐clan proteases revealed that Ub/Ub‐like protease and acetyltransferase activities are not mutually exclusive, and that the Chlamydia trachomatis effector ChlaDUB1 can perform both functions through the same catalytic center (Fig 4F; Pruneda et al, 2018). Unlike YopJ and AvrA, the acetyltransferase activity of ChlaDUB1 is targeted toward lysine residues. Interestingly, ChlaDUB1 orthologs demonstrate varying degrees of the two activities: the C. abortus ChlaDUB ortholog is a dedicated acetyltransferase, while the closely related C. trachomatis paralog ChlaDUB2 is a dedicated DUB. Crystal structures of intermediate stages from both reactions showed spatially distinct binding sites for ubiquitin and coenzyme A on opposite faces of a unique VR‐3 helix, allowing the two activities to be decoupled by mutation (Fig 4C). The DUB function of ChlaDUB1 has been linked to several biological processes, including inhibition of NF‐κB signaling, Mcl‐1 stabilization, and Golgi fragmentation/recruitment to the Chlamydia‐containing inclusion (Le Negrate et al, 2008; Fischer et al, 2017; Pruneda et al, 2018; Auer et al, 2020). Recently, a role for ChlaDUB1 acetyltransferase function has been described in protecting other Chlamydia effectors from Ub‐mediated degradation, thereby supporting bacterial egress (preprint: Bastidas et al, 2023). Whether ChlaDUB1 is unique in its ability to catalyze both proteolysis and acetylation remains to be seen.

Legionella pneumophila RavZ

The autophagy pathway is used by eukaryotes to sequester invading cytosolic bacteria into the autophagosome, which fuses with the lysosome to degrade the cargo. There is limited evidence that pathogenic bacteria inhibit autophagy through direct manipulation of regulatory Ub‐like modifiers such as Atg8/LC3, Atg12, and FAT10. However, a genetic screen identified Legionella pneumophila RavZ for its ability to cleave the autophagy marker LC3 from phosphatidylethanolamine (PE) at the autophagosomal membrane (Choy et al, 2012). Remarkably, unlike canonical Ub/Ub‐like proteases, RavZ cleaves LC3 one amino acid short of the C‐terminus, thereby leaving a Gly‐PE remnant and preventing LC3 from recycling through the conjugation pathway (Choy et al, 2012). RavZ binds LC3 via an N‐terminal LC3‐interacting region (LIR) motif, which facilitates extraction of LC3 from the membrane and removal of the PE moiety (Yang et al, 2017a). Recent studies have also implicated RavZ in clearing ubiquitin from the bacteria‐containing vacuole (Kubori et al, 2017; Shi et al, 2023b), although direct DUB activity remains to be shown. The strategy RavZ uses to simultaneously remove and inactivate LC3 is akin to how the foot‐and‐mouth disease viral protease Lb^pro cleaves ISG15 (Swatek et al, 2018), but whether this strategy has been adopted by other pathogen‐encoded Ub/Ub‐like proteases is an open question.

Coxiella burnetii EmcB

In the process of delivering effector proteins into the host cytosol, bacterial secretion systems can also activate host immune sensors. L. pneumophila, for instance, can trigger a type I interferon response through the immune sensor RIG‐I, which presumably detects leakage of its ligand, double‐stranded RNA, through the Dot/Icm type IV secretion system (Monroe et al, 2009). Interestingly, the obligate intracellular pathogen Coxiella burnetii encodes a related type IV secretion system but does not elicit a type I interferon response, suggesting that it may encode an effector that counters immune signaling. A recent genetic screen identified EmcB (cbu2013) as one of two effectors capable of suppressing an interferon response (Duncan‐Lowey et al, 2023). EmcB was found to act as a DUB that reverses RIG‐I ubiquitination, a necessary signaling step for downstream immune activation. Although it lacks sequence or predicted structural homology to any known family of DUB, biochemical studies did identify putative catalytic His and Cys residues that fit the topology of a CE‐clan protease. Similar to many bacterial CE‐clan DUBs, EmcB preferentially cleaves long K48‐ and K63‐linked polyUb chains, with a lower activity toward NEDD8 modifications (Duncan‐Lowey et al, 2023). Activity toward long K63‐linked polyUb is most consistent with its role in countering RIG‐I ubiquitination. Further work on EmcB and its related Coxiella orthologs will be necessary to understand its relationship, if any, to established CE‐clan bacterial DUBs.

In sum

With examples across all kingdoms of life, the CE clan of proteases exhibits a range of diverse activities including general peptidase, Ub/Ub‐like‐specific protease, or even acetyltransferase functions. Some defining features of each activity have been identified, but in general, the catalytic scaffold appears to be highly malleable to evolution of ubiquitin, Ub‐like, or acetyl CoA‐binding sites that influence function. At least in the case of ChlaDUB1, it is even possible to adopt multiple enzymatic functions within the same CE‐clan scaffold. The evolutionary origins of this fold and what the ancestral activity might have been will be an interesting question moving forward, the answer to which may help identify additional discriminating features among CE‐clan enzymes. Another interesting feature is the prevailing preference for K63‐linked polyUb among bacterial CE‐clan DUBs from human pathogens described thus far, as this substrate specificity is unique from examples found in plant pathogens (e.g., XopD) and has arisen through discrete mechanisms (e.g., OtDUB). How pervasive this trend is and whether bacterial CE‐clan DUBs subvert a common element of host defense remains to be seen.

Regulation via ubiquitination

The rule makers

The ubiquitin–proteasome system (UPS) is an energy‐dependent protein quality control system that specifically degrades ubiquitinated substrates (Fig 5D). Protein ubiquitination was once thought to be a simple marker for protein degradation by the proteasome, but has since been found to regulate nearly all aspects of eukaryotic biology. This diversity in signaling comes from a capacity to be further modified to form complex polymeric chains at eight canonical sites (M1, K6, K11, K27, K29, K33, K48, and K63), although recent work has expanded this signal diversity even further (Komander & Rape, 2012; Swatek & Komander, 2016; McCrory et al, 2022). Specific polyUb chain types signal distinct cellular outcomes. For example, K48‐linked polyUb chains target proteins for degradation by the proteasome, while other linkage types can play non‐degradative roles in protein trafficking and signal amplification. The signaling rules for many forms of ubiquitin modifications, including the so‐called “atypical” chains, are still being actively studied.

Figure 5. Regulation of and by ubiquitin.

(A) Structure of the insect Ub kinase PINK1 (green) bound to Ub (red; PDB 6EQI; Schubert et al, 2017), highlighting proximity of Ub Ser65 to the PINK1 active site. (B) Structure of the Legionella pneumophila SdeA mART domain (PDB 5YIJ; Akturk et al, 2018; Dong et al, 2018; Kalayil et al, 2018; Kim et al, 2018; Wang et al, 2018), highlighting proximity of Ub Arg42 to the mART active site. (C) Structure of the Burkholderia pseudomallei Ub/NEDD8 deamidase CHBP (PDB 4HCN; Yao et al, 2012), highlighting proximity of Ub Gln40 to the CHBP active site. (D) Mechanism of targeted protein degradation via the ubiquitin–proteasome system. (E) Mechanism of hijacking the host ubiquitin–proteasome system by bacterial metaeffector E3 ligases (Kubori et al, 2010). (F) Mechanism of dual ubiquitination/phosphorylation regulation by the Shigella flexneri effector kinase OspG, wherein OspG stabilizes E2 ~ Ub conjugates in an inactive conformation while simultaneously stabilizing an active OspG conformation for phosphorylation (Pruneda et al, 2014).

The rule followers

Hijacking targeted protein degradation

Numerous bacterial effectors modulate the UPS during infection. By encoding E3 ligase domains with K48‐linked polyUb specificity, bacterial effectors can stimulate the degradation of select host factors. The NEL family of bacterial ligases is particularly well characterized for this role in targeted degradation through modular use of catalytic and substrate‐binding domains, which may explain its expansion among the effector repertoire in S. flexneri (Tanner et al, 2015). Members of this family have been shown to target host cytoskeletal and immune defense factors for Ub‐dependent degradation (Rohde et al, 2007; Ashida et al, 2010, 2013; Suzuki et al, 2014; de Jong et al, 2016; Zheng et al, 2016; Li et al, 2017; Wandel et al, 2017; Otsubo et al, 2019; Hansen et al, 2021; Luchetti et al, 2021). Most recently, NEL family E3 ligases IpaH 9.8, IpaH 1.4/2.5, and IpaH 7.8 from S. flexneri have been shown to direct the degradation of guanylate‐binding proteins (thereby blocking cytosolic restriction), components of the linear ubiquitin chain assembly complex/LUBAC (thereby blocking inflammatory signaling), and gasdermin pores (thereby blocking pyroptosis and bacterial lysis), respectively, illustrating the versatility of this family to shunt innate immune responses (de Jong et al, 2016; Li et al, 2017; Wandel et al, 2017; Hansen et al, 2021; Luchetti et al, 2021). Numerous other bacterial ligases have also been reported to ubiquitinate host factors for targeted degradation (Rosebrock et al, 2007; Göhre et al, 2008; Gimenez‐Ibanez et al, 2009; Chen et al, 2017; Fiskin et al, 2017; Alix et al, 2020). Interestingly, the host NLRP1 inflammasome can act as a molecular tripwire to detect bacterial ubiquitin ligase activities, and through a mechanism termed “functional degradation” activate pyroptotic cell death (Sandstrom et al, 2019). In contrast, some bacterial DUBs protect host factors from degradation. Chlamydia ChlaDUB1 has been reported to protect IκBα and Mcl‐1 from ubiquitination, thereby inhibiting NF‐κB activation as well as cell death (Le Negrate et al, 2008; Fischer et al, 2017). Similar inhibition of NF‐κB signaling through protein stabilization has been shown for the Burkholderia DUB TssM (Tan et al, 2010). Many other bacterial DUBs have a demonstrated capability to hydrolyze K48‐linked polyUb signals, but whether they specifically stabilize target proteins from proteasomal degradation remains to be seen.

The Ub‐dependent proteasomal degradation of effectors themselves has been shown to both negatively and positively impact the progress of bacterial infection. Salmonella effectors SptP and SopE are ubiquitinated with K48‐linked polyUb and subsequently degraded, which regulates Salmonella infection. The degradation of these effectors is temporally regulated, and they display distinct susceptibility to host proteasomal degradation, the determinants for which are encoded in the effectors themselves (Kubori & Galán, 2003). The cytotoxic effect of the ADP‐ribosyltransferase, ExoT, from Pseudomonas aeruginosa is regulated by the host E3 ligase Cbl‐b. ExoT is targeted for proteasomal degradation, limiting the dissemination of ExoT‐producing Pseudomonas (Balachandran et al, 2007). In contrast, the proteasomal degradation of the Yersinia effector YopE contributes to dissemination of the bacteria (Hentschke et al, 2007; Gaus et al, 2011).

The U‐box E3 ligase “metaeffector” LubX from L. pneumophila targets another bacterial effector protein, SidH, for proteasomal degradation during infection (Fig 5E; Kubori et al, 2010). The U‐box‐2 region of LubX binds directly to SidH, while U‐box‐1 mediates polyubiquitination. Interestingly, the translocation of SidH and LubX is distinctly regulated. Although a relatively small amount is needed to efficiently degrade SidH in the host cell, LubX concentration continues to increase over the first 10 hours post‐infection, suggesting there are other roles at later stages. In support of this, LubX has also been shown to target the host kinase Clk1 for proteasomal degradation (Kubori et al, 2008).

The rule breakers

Ubiquitin as a stimulatory cofactor

ExoU from P. aeruginosa possesses phospholipase activity that is detectable only in the presence of ubiquitin as an activating cofactor (Anderson et al, 2011). ExoU binds ubiquitinated host proteins prior to activation and has a 3.5‐fold greater activity in the presence of K48‐ and K63‐linked (tetraUb or longer) polyUb chains compared to monoUb in vitro. MonoUb is sufficient for ExoU activity in cells. Furthermore, Ub‐like proteins such as ISG15, FAT10, SUMO‐1, and NEDD8 do not activate ExoU, indicating a high specificity for ubiquitin. In addition to using ubiquitin as an activator, ExoU is ubiquitinated itself at K178 with a K63‐linked diUb modification. The host enzymes responsible and the role of this modification during infection remain unclear. An ExoU K178R mutant retained full toxicity, however, implying that ubiquitination of ExoU is not required for phospholipase activity. A unique Ub‐binding domain localized within the ExoU‐bridging domain is required for full binding affinity and activation of the effector in vitro (Tessmer et al, 2018). Two other Ub‐activated phospholipase orthologs of ExoU are present in Rickettsia and Achromobacter species (Tessmer et al, 2019). Interestingly, other Ub‐activated phospholipases have been recently described in Mycobacterium tuberculosis and L. pneumophila (Chai et al, 2022; Li et al, 2022). Whether all of these examples are linked by a common mechanism or biological context remains to be investigated.

The effector kinase OspG encoded by S. flexneri binds several host E2~Ub conjugates to attenuate the host immune response during infection. The connection between OspG and ubiquitin was first discovered using a yeast two‐hybrid screen (Kim et al, 2005). OspG was shown to inhibit IκBα ubiquitination during Shigella infection, thereby stabilizing it and suppressing NF‐κB signaling. It was later demonstrated that ubiquitin binding enhanced the kinase activity of purified OspG, but not to the same extent as E2~Ub conjugates (Zhou et al, 2013; Grishin et al, 2014; Pruneda et al, 2014). Co‐crystal structures of OspG bound to either UBE2D3~Ub or UBE2L3~Ub revealed that complex formation stabilizes the minimal kinase domain of OspG, thereby enhancing kinase activity (Grishin et al, 2014, 2018; Pruneda et al, 2014). While related OspG orthologs can be found in Yersinia and EHEC, only Yersinia YspK would be predicted by sequence conservation to employ a similar mechanism of kinase activation (Pruneda et al, 2014). Meanwhile, OspG interaction competes for binding sites on E2 and Ub that are required for E3 ligase interactions and diminishes the reactivity of UBE2D3~Ub through maintaining a stable extended conformation that is less active for ubiquitin transfer (Grishin et al, 2014; Pruneda et al, 2014). In this way, the OspG interaction with host E2~Ub intermediates results in simultaneous kinase activation and E2 inhibition (Fig 5F).

In contrast to the kinase activity of OspG, the M. tuberculosis effector PtpA is protein tyrosine phosphatase that is activated by ubiquitin binding. PtpA is required for mycobacterial survival in macrophages and has been shown to play roles in blocking phagosome acidification as well as suppression of innate immune signaling (Bach et al, 2008; Wong et al, 2011; Wang et al, 2015). Ubiquitin was found to interact with PtpA in a yeast two‐hybrid screen, and subsequently shown to activate PtpA phosphatase activity toward substrates including phosphorylated JNK and p38 (Wang et al, 2015). Structural analysis identified a novel Ub‐interacting motif‐like (UIML) region in PtpA that was required for Ub‐dependent activation. An interaction between PtpA and the immune signaling adaptor TAB3 has also been discovered by yeast two‐hybrid, the result of which is to compete with TAB3's ability to recognize K63‐linked polyUb and trigger NF‐κB signaling (Wang et al, 2015). Thus, like OspG, PtpA plays a dual role in subverting host immune signaling.

Molecular decoys of ubiquitination

Host‐directed ubiquitination of the M. tuberculosis effector Rv0222 has been shown to play a role in immune suppression during infection (Wang et al, 2020a). Following secretion, Rv0222 is modified with K11‐linked polyUb by the catalytic E3 ligase subunit of the anaphase promoting complex, ANAPC2. Through an unknown mechanism, ubiquitinated Rv0222 stimulates recruitment of the protein tyrosine phosphatase SHP1 to TRAF6, an E3 ligase that controls a key stage of the NF‐κB pathway. SHP1 binding inhibits the ligase activity of TRAF6, thereby suppressing inflammatory signaling. The mechanisms by which Rv0222 acts as a decoy for ubiquitination, avoids proteasomal degradation, and induces SHP1 localization will be an interesting area of future research.

In sum

It is remarkable to imagine that bacteria have evolved secreted effectors that not only regulate but are also regulated by host ubiquitin. The temporal control endowed by the ubiquitin–proteasome system allows bacterial effectors to act in sequence, while ubiquitin itself can act as a molecular cue for a eukaryotic environment. Both host and pathogen appear to have evolved decoys to detect opposing E3 ligase activity, and it will be important to determine whether this strategy is employed across additional host–pathogen interfaces.

Regulation through modifying ubiquitin

The rule makers

In addition to modifications with ubiquitin or Ub‐like proteins, ubiquitin can be subject to additional post‐translational modifications, thereby expanding the complexity of the ubiquitin code (Swatek & Komander, 2016; Lacoursiere et al, 2022). Ubiquitin acetylation has been shown to impact activation by the E1 enzyme as well as the types of polyUb linkages produced, but how these modifications are regulated remains an open question (Ohtake et al, 2015; Lacoursiere & Shaw, 2021). Phosphorylation of ubiquitin can occur at multiple sites, but biological roles have only been ascribed to a few. For example, phosphorylation of ubiquitin Thr12 has been implicated in the DNA damage response, while the mitochondrial kinase PINK1 catalyzes phosphorylation of ubiquitin Ser65 (Fig 5A), which in turn activates the E3 ligase Parkin to promote generation of autophagy signals leading to removal of damaged mitochondria (Kane et al, 2014; Kazlauskaite et al, 2014; Koyano et al, 2014; Ordureau et al, 2014; Wauer et al, 2015; Walser et al, 2020). ADP ribosylation of the ubiquitin C‐terminus has also been reported as a mechanism of short‐term inactivation catalyzed by a specialized family of E3 ligases called the DELTEX family (Yang et al, 2017b; Chatrin et al, 2020), although more recent work proposes that DELTEX ligases instead ubiquitinate ADP‐ribosyl modifications (Zhu et al, 2022). How the ubiquitin signal can be fine‐tuned through these and additional forms of post‐translational regulation is an active area of research.

The rule followers

As part of an effort to identify mechanisms of NF‐κB suppression, a type III secreted effector protein from Chromobacterium violaceum called CteC was found to ADP‐ribosylate ubiquitin at Thr66 (Yan et al, 2020). This covalent modification was found to prevent the cycling of monoUb through the E1, E2, and E3 transfer cascade, as well as the receptor recognition of existing polyUb signals that were modified during infection. CteC is part of a family of related enzymes that includes members within the Burkholderia cepacia complex. The catalytic domain is reminiscent of C3‐like ADP ribosyltransferases but has a more minimal set of catalytic residues. The precise mechanism of ubiquitin recognition and catalysis awaits additional structural studies.

As the first step toward noncanonical ubiquitination, the L. pneumophila SidE ligase family ADP‐ribosylates Arg42 of ubiquitin (Fig 5B; Bhogaraju et al, 2016; Qiu et al, 2016). In the absence of suitable substrate, the SidE PDE domain will hydrolyze ADPr‐Ub to Pr‐Ub. Reversal of SidE‐catalyzed ubiquitination by DupA and DupB also produces Pr‐Ub as a byproduct (Wan et al, 2019a; Shin et al, 2020b). Pr‐Ub exhibits a defect in passing through conventional E1/E2/E3 conjugation enzymes, and if levels accumulate becomes toxic in cells (Bhogaraju et al, 2016). Whether a host‐ or Legionella‐derived mechanism exists to prevent this toxicity by clearing Pr‐Ub remains unknown.

The rule breakers

Cycle‐inhibiting factors (Cifs) are a family of type III secreted effectors conserved in Gram‐negative bacterial pathogens such as E. coli, B. pseudomallei, and Y. pseudotuberculosis. Cif effectors are related to CA‐clan proteases, but instead catalyze deamidation of Gln40 in ubiquitin or NEDD8, with most demonstrating a preference for NEDD8 (Fig 5C). Deamidation is an extremely small change, effectively altering a glutamine to a glutamate, but has large implications on local charge and hydrogen‐bonding capability. The irreversible modification of ubiquitin impairs its ability to transfer through the E1‐E2‐E3 conjugation cascade, while modification of NEDD8 within the context of CRLs disrupts conformational changes required for ubiquitin ligation (Cui et al, 2010; Jubelin et al, 2010; Morikawa et al, 2010; Boh et al, 2011; Toro et al, 2013; Yu et al, 2015). Structural studies have provided mechanistic detail for the deamidation reaction, which takes place within a papain‐like enzyme scaffold (Crow et al, 2012; Yao et al, 2012).

While structurally homologous to Cif deamidases and CA‐clan proteases, the L. pneumophila effectors MavC and MvcA exhibit several key differences. First, despite a similar structure, MavC and MvcA are specific to ubiquitin and will not deamidate NEDD8 (Valleau et al, 2018b). Second, by virtue of a unique insertion domain, MavC specifically inhibits the host E2 enzyme UBE2N, which results in a loss of K63‐linked polyUb signals required for NF‐κB activation (Valleau et al, 2018b; Gan et al, 2019a). The mechanism by which MavC specifically inhibits UBE2N is that it does not deamidate ubiquitin Gln40, but actually performs a transglutaminase reaction that crosslinks the ubiquitin Gln40 side chain to a nearby lysine residue of UBE2N, forming an isopeptide bond (Fig 3F; Gan et al, 2019a). Structures of the MavC:UBE2N:Ub ternary complex reveal the mechanism of this noncanonical ubiquitin ligase activity (Guan et al, 2020; Mu et al, 2020; Puvar et al, 2020). Remarkably, the preferred target of MavC is actually the activated UBE2N~Ub intermediate, which is inactivated by MavC‐dependent transglutamination. While this crosslinked UBE2N‐Ub product is resistant to eukaryotic DUBs, the closely related Legionella enzyme MvcA actually reverses the reaction and releases UBE2N and deamidated ubiquitin (Gan et al, 2020). In fact, both MavC and MvcA can enzymatically perform similar functions but are only biased in either direction by slight differences in sequence (Valleau et al, 2018b; Gan et al, 2020; Guan et al, 2020; Mu et al, 2020). The bi‐directionality of these reactions is reminiscent of the dual hydrolase/transferase activity of Chlamydia ChlaDUB1, and even eukaryotic DUBs can be reprogrammed to perform the opposite transamidation reaction (Chang & Strieter, 2018; Pruneda et al, 2018).

In sum

Post‐translational modification of host ubiquitin is a powerful yet dangerous mechanism of virulence. While producing very potent effects on host signaling, without some form of focusing this activity on specific processes, the overall effects can be detrimental to cell viability. It is intriguing how little overlap has been described thus far between eukaryotic ubiquitin modifications (phosphorylation and acetylation) and those catalyzed by bacterial effectors (ADP ribosylation, deamidation, and transglutamination). Whether this could be due to an incomplete picture of ubiquitin regulation, or possibly that orthogonal ubiquitin modifications are better suited for bacterial virulence, remains to be seen.

Concluding remarks

The tendency of bacterial effectors to break the rules of ubiquitin regulation creates an interesting challenge for identifying more examples. Activity‐based approaches or harnessing the power of artificial intelligence breakthroughs may be the way forward for characterizing this host–pathogen interface. At the same time, although it establishes a nice framework for understanding the similarities of some bacterial Ub‐targeted effectors and the contrasting properties of others, the “rules” of the eukaryotic ubiquitin system are continually evolving. As our appreciation for the distinctive properties of some ubiquitin regulators grows, we must continue to expand the categories by which we describe them (Abdul Rehman et al, 2016; Haahr et al, 2018; Hermanns et al, 2018; Hewings et al, 2018; Kwasna et al, 2018; Pao et al, 2018; Otten et al, 2021). Unveiling new areas of biology regulated by ubiquitin and the interplay it has with other forms of post‐translational regulation continues to challenge how we think of ubiquitination and the tools we have at hand to study it (Otten et al, 2021; Kelsall et al, 2022; Sakamaki et al, 2022; Zhu et al, 2022). Just as it has for other areas of biology (Welch, 2015), studying ubiquitin signaling through the lens of bacterial intervention will undoubtedly provide new tools and perspectives for future research on this fascinating post‐translational modification.

Author contributions

Cameron G Roberts: Conceptualization; writing – original draft; writing – review and editing. Tyler G Franklin: Conceptualization; writing – original draft; writing – review and editing. Jonathan N Pruneda: Conceptualization; supervision; funding acquisition; writing – original draft; writing – review and editing.

Disclosure and competing interests statement

The authors declare that they have no conflict of interest.

The EMBO Journal (2023) 42: e114318