Pathogenic bacteria are in a constant battle for survival with their host. In order to gain a competitive edge, they employ a variety of sophisticated strategies that allow them to modify conserved host cell processes in ways that favor bacterial survival and growth. Ubiquitylation, the covalent attachment of the small modifier ubiquitin to target proteins, is such a pathway. Ubiquitylation profoundly alters the fate of a myriad of cellular proteins by inducing changes in their stability or function, subcellular localization or interaction with other proteins. Given the importance of ubiquitylation in cell development, protein homeostasis and innate immunity, it is not surprising that this post-translational modification is exploited by a variety of effector proteins from microbial pathogens. Here, we highlight recent advances in our understanding of the many ways microbes take advantage of host ubiquitylation, along with some surprising deviations from the canonical theme. The lessons learned from the in-depth analyses of these host–pathogen interactions provide a fresh perspective on an ancient post-translational modification that we thought was well understood.
This article is part of a Minifocus on Ubiquitin Regulation and Function. For further reading, please see related articles: ‘Mechanisms of regulation and diversification of deubiquitylating enzyme function’ by Pawel Leznicki and Yogesh Kulathu (J. Cell Sci. 130, 1997–2006). ‘Cell scientist to watch – Mads Gyrd-Hansen' (J. Cell Sci. 130, 1981–1983).
During infection, bacterial pathogens encounter a variety of challenges such as the innate immune defense of the host and nutritional restriction. To counteract these limitations and establish conditions favorable for their survival and growth, infectious microbes often employ virulence strategies aimed at modifying key host signaling pathways. Ubiquitylation, the post-translational modification with the protein modifier ubiquitin, is such a process. Ubiquitin is a 76-amino-acid protein that is highly conserved in all eukaryotes. Its covalent attachment to target proteins typically occurs on lysine residues, although conjugation to other side chains such as threonine, serine, tyrosine or cysteine has also been reported (McDowell and Philpott, 2013; Komander, 2009; Kulathu and Komander, 2012). Importantly, any of the seven lysine residues in ubiquitin (positions 6, 11, 27, 29, 33, 48 and 63) can serve as recipient for additional cycles of ubiquitin attachment, resulting in the formation of branched or non-branched ubiquitin chains (polyubiquitylation). The type of ubiquitylation, whether mono-ubiquitylation, multi-mono-ubiquitylation, or poly-ubiquitylation, commonly referred to as the ‘ubiquitin code’, determines the fate of the substrate protein (Fig. 1; Komander and Rape, 2012).
Protein ubiquitylation is a cascade catalyzed by an E1 (ubiquitin-activating enzyme), an E2 (ubiquitin-conjugating enzyme) and an E3 (ubiquitin ligase) (Fig. 1; Dye and Schulman, 2007; Hershko and Ciechanover, 1998). Like most post-translational modifications, ubiquitylation is a reversible process, with ubiquitin removal being catalyzed by a specific class of proteases called deubiquitylating (or deubiquitinating) (DUBs) (D'Andrea and Pellman, 1998; Komander et al., 2009). In the following sections of this Commentary, we will outline strategies used by pathogens to exploit the host ubiquitylation pathway. Although this post-translational modification has also been shown to contribute to the virulence of viral pathogens (Wimmer and Schreiner, 2015), the current review will focus on pathogenic bacteria as well as their effector proteins.
Exploiting ubiquitylation through molecular mimicry
The ability of E3s to specifically interact with both substrate proteins and E2–ubiquitin complexes makes them central players in the ubiquitylation cascade. There are two major classes of E3s, HECT type, and RING or U-box (RING/U-box)-type E3s (Ardley and Robinson, 2005; Buetow and Huang, 2016; Pickart, 2001), and bacteria encode molecular mimics of both of them (Fig. 1; Table S1). In addition, F-box domains, the substrate recognition component in the Skp1–Cul1–F-box (SCF) complex that mediates ubiquitylation with the E3 ligase Rbx1 (Zheng et al., 2002), are also found in bacterial effectors (Fig. 2). In the following sections, we will review examples of bacterial effectors that are HECT-, RING/U-box-type E3 ligases, or have F-box domains.
HECT-type E3 ligases
HECT (for ‘homologous to the E6AP C-terminus’)-type E3 ubiquitin ligases form a reactive intermediate with ubiquitin before it is transferred to the substrate protein (Fig. 2) (Bernassola et al., 2008). The first bacterial effector protein identified as a HECT-type E3 ubiquitin ligase was SopA, a type-III secretion system (T3SS) effector in Salmonella (Zhang et al., 2006). SopA is required for the stimulation of transepithelial migration of polymorphonuclear leucocytes, as well as Salmonella-induced inflammatory responses in the intestinal epithelium (Zhang et al., 2006; Kamanova et al., 2016). Both processes are dependent on the E3 ligase activity of SopA towards TRIM65 and TRIM56, two host E3 ubiquitin ligases that modulate interferon (IFN)-β expression to stimulate the inflammatory response (Kamanova et al., 2016). A Salmonella variant encoding the catalytically inactive SopA(C753S) was attenuated in triggering cell migration and TRIM65-depedent inflammation (Zhang et al., 2006). Another bacterial HECT-type E3 ligase is NleL, a member of the non-LEE-encoded (Nle) effector family, from enterohemorrhagic Escherichia coli (EHEC) (Lin et al., 2011; Piscatelli et al., 2011). NleL downregulates Tir-mediated pedestal formation in host cells via its E3 ligase activity, and EHEC encoding ligase-deficient NleL(C753A) showed enhanced pedestal formation compared to that seen in the wild-type strain (Piscatelli et al., 2011).
Even though there is only little sequence similarity between SopA or NleL and their eukaryotic counterparts, they possess a similar bilobal architecture to that of eukaryotic HECT domains (Diao et al., 2008; Lin et al., 2012). The HECT domain of SopA and NleL is composed of two lobes (the N- and C-lobe), which are connected by a long flexible helix. The position of the E2-binding site in the N-lobe of the HECT domain of SopA and NleL differs from that found in eukaryotic HECT domains. However, the large movement of the C-lobe between the E2-binding site and the substrate protein binding site is reminiscent of that in eukaryotic HECT domains (Lin et al., 2012), indicating that the domain mobility in HECT E3s is universally important for the successful transfer of ubiquitin.
RING/U-box E3 ligases
Unlike HECT E3s, RING (for ‘really interesting new gene’) or U-box-type E3s mediate transfer of ubiquitin from the E2 directly to the substrate protein, without formation of an E3–ubiquitin intermediate (Fig. 2) (Aravind and Koonin, 2000; Deshaies and Joazeiro, 2009; Metzger et al., 2014). AvrPtoB, a T3SS effector protein of the plant pathogen Pseudomonas syringae pv. tomato, contains a U-box E3 ligase domain and functions in suppressing the resistance mechanism of the host (Abramovitch et al., 2006; Janjusevic et al., 2006). Upon binding of the N-terminal domain of AvrPtoB to host proteins, the E3 ligase activity within the C-terminal U-box domain is triggered in order to mediate ubiquitylation of several host kinases, including Pto and Fen (Ntoukakis et al., 2009; Rosebrock et al., 2007), the LysM receptor kinase CERK1 (Gimenez-Ibanez et al., 2009) and the pattern recognition receptor FSL2 (Göhre et al., 2008; Xiang et al., 2008). Ubiquitylation of these host proteins targets them for degradation, and leads to an inhibition of the pattern-triggered immunity by the plants.
NleG is the largest family of Nle effectors with over 20 representatives in different strains of pathogenic E. coli and Salmonella (Wu et al., 2010). Owing to the lack of sequence homology, the molecular function of this protein family remained unknown until recently when it was shown that the most conserved C-terminal portion of NleG effectors is structurally reminiscent to RING/U-box domains and exhibits E3 ligase activity, thus providing important insights into the biological role of this protein family (Wu et al., 2010).
Two U-box E3s have been identified among Legionella pneumophila type-IV secretion system (T4SS) effector proteins: LubX and GobX. LubX possesses two U-box domains (Quaile et al., 2015; Kubori et al., 2008), but only the N-terminal one interacts with E2s and exhibits E3 ligase activity. Interestingly, the ubiquitylation activity of LubX is not directed against a host factor, but against another L. pneumophila effector protein, SidH (Kubori et al., 2010). It appears that LubX accumulates in host cells at a later infection stage and ubiquitylates SidH to target it for proteasomal degradation. This feature makes LubX a so-called ‘metaeffector’, an effector that targets another effector. In vitro, LubX has been shown to interact with and polyubiquitylate the host Cdc2-like kinase 1 (Clk1) (Kubori et al., 2008). However, the level of Clk1 ubiquitylation within cells coexpressing lubX was moderate, and L. pneumophila with a lubX deletion mutant replicated normally in bone marrow-derived mouse macrophages (Kubori et al., 2008). Therefore, the exact role of Clk1 and whether LubX targets it for degradation during L. pneumophila infection remains to be elucidated.
The L. pneumophila effector GobX contains a central U-box domain with remote sequence similarity to other E3s (Lin et al., 2015). Its E2-binding interface appears to be conserved with eukaryotic U-box domains as mutation of either Ile58 or Trp87, residues found to be functionally important in other U-box E3s, attenuated GobX-catalyzed polyubiquitylation (Lin et al., 2015). Interestingly, upon production in transiently transfected cells, GobX was highly enriched in the Golgi region, a feature that relied on the covalent attachment of a hydrophobic palmitoyl group to the Cys175 residue of GobX. A C175S mutation rendered GobX cytosolic, demonstrating that GobX exploits two highly conserved eukaryotic post-translational modification machineries, ubiquitylation and S-palmitoylation, in order to fulfill its function within host cells (Lin et al., 2015).
F-box domain proteins
The F-box motif is a small 50-amino-acid domain that binds Skp1, which serves as an adapter for Cul1 (Fig. 2). Cul1 and the RING-box 1 (Rbx1) protein form the catalytic core that recruits the E2 and mediates ubiquitylation (Kipreos and Pagano, 2000; Zheng et al., 2002). F-box domain-containing proteins control substrate binding by the SCF complex, and bacterial effectors that mimic F-box domains have been found in plant pathogens Ralstonia solanacearum and Agrobacterium, as well as in L. pneumophila, as discussed below.
R. solanacearum encodes a family of T3SS effectors called GALA (named after a conserved GAxALA sequence) that are essential for virulence. They contain an F-box domain that interacts with the Arabidopsis Skp1-like proteins ASK1 and ASK2 (Angot et al., 2006). Similarly, VirF, a T4SS effector of Agrobacterium that mediates gene transfer to plant cells, also has an F-box domain that interacts with ASK1 and ASK2 (Schrammeijer et al., 2001). The association of VirF with the plant SCF complex directs the ubiquitin ligase activity not only against the plant protein VIP1 but also its own VirE2 effector (Tzfira et al., 2004), which makes VirF a metaeffector, similar to LubX. Proteasomal degradation of VIP1 and VirE2 ultimately leads to the unpacking of the transfer-DNA and its integration into the plant genome (Tzfira et al., 2004).
L. pneumophila strain Philadelphia-1 encodes at least five effector proteins with F-box domains (Ensminger and Isberg, 2010). Among them, LegU1 and LegAU13 (also known as AnkB or Ceg27) stably associate with Skp1 and Cul1 of the host SCF complex (Ensminger and Isberg, 2010). LegU1 interacts with and mediates the ubiquitylation of the host protein HLA-B-associated transcript 3 (BAT3; also known as BAG6), a ribosome-associated factor that is involved in a wide range of biological functions, including apoptosis, because it chaperones tail-anchored membrane proteins (Ensminger and Isberg, 2010; Mariappan et al., 2010). LegAU13 has an N-terminal F-box domain and a C-terminal ankyrin repeat region (Cazalet et al., 2004; de Felipe et al., 2005). The F-box domain of LegAU13 forms extensive contacts with human Skp1 through a hydrophobic interface (Wong et al., 2017), and mutation of key residues within this interface prevents LegAU13 from binding to Skp1 and from recruiting polyubiquitylated species to the Legionella vacuole (Price et al., 2009). The ankyrin repeat domain of LegAU13 has been proposed to contribute to the binding of a yet-to-be identified substrate (Wong et al., 2017). Notably, the biological role of LegAU13 varies in different L. pneumophila backgrounds. Whereas deletion of legAU13 from the strain Philadelphia-1 did not cause a detectable replication defect in either mouse macrophages or the protist Acanthamoeba castellanii (Ivanov and Roy, 2009; Ensminger and Isberg, 2010), its absence from strains AA100/130b (Wadsworth) (Al-Khodor et al., 2008; Price et al., 2010b) or Paris (Lomma et al., 2010) caused a reduction in both intracellular proliferation and accumulation of polyubiquitylated species around the Legionella-containing vacuole. A possible explanation for the discrepancy in phenotypes is that functionally redundant effectors that could compensate for the loss of legAU13 are present in Philadelphia-1, but not in Paris or Wadsworth. Alternatively, LegAU13 homologs from various strains or isolates could have a different set of host targets and play diverse roles during infection. Regardless of these differences, most LegAU13 homologs possess a CaaX box motif at their C-terminal end that allows them to be farnesylated within the host cell, which enhances their membrane association (Ivanov et al., 2010; Price et al., 2010a).
Exploiting ubiquitylation through unconventional E3s
The abundance of bacterial effectors that mimic host E3 ligases or F-box domains underscores the importance of exploiting ubiquitylation for microbial pathogenesis. These molecular mimics were most likely acquired by bacterial pathogens through horizontal gene transfer. Furthermore, with a less clear evolutionary origin, bacterial pathogens also encode novel E3s, namely, novel E3 ligases (NELs), XL-box-containing E3 ligases, and SidC-like E3s (Fig. 2), or effectors that mediate ubiquitylation through a unique E1- and E2-independent fashion (SidE) (Fig. 3). These extended strategies, which are unseen in eukaryotes, are outlined in the following sections.
Novel E3 ligases
NELs have been discovered in Salmonella (SspH1, SspH2 and SlrP) (Quezada et al., 2009), Shigella (IpaH family proteins) (Singer et al., 2008; Zhu et al., 2008; Rohde et al., 2007) and other bacteria (Xin et al., 2012; Soundararajan et al., 2010) (see also Table S1). Similar to what is found in HECT-type E3s, NELs contain a conserved cysteine residue that mediates ubiquitylation by forming a thioester intermediate with the ubiquitin molecule. NEL domains have a unique fold that is composed of two long finger-like helices that extend from a globular domain (Quezada et al., 2009; Singer et al., 2008; Zhu et al., 2008). The study of SspH2 showed that its NEL domain binds to the UbcH5–ubiquitin conjugate (UbcH5 is also known as UBE2D1) by interacting with both UbcH5 and the ubiquitin molecule, and the E2 surface recognized by SspH2 is substantially different from that in known eukaryotic E3 ligases (Levin et al., 2010). Notably, SspH2 is able to stimulate K48-linked polyubiquitin chain formation on UbcH5 in vitro (Levin et al., 2010), suggesting that, during infection, substrate degradation could be expedited by mediating the transfer of preformed polyubiquitin chains onto substrate proteins.
Substrate binding of NELs occurs through the leucine-rich repeat (LRR) domain adjacent to the NEL domain (Fig. 2). The LRR domain is connected to the NEL domain through a flexible linker and inhibits its ubiquitin ligase activity in the absence of substrate proteins (Chou et al., 2012; Quezada et al., 2009; Keszei and Sicheri, 2017). Removal of the LRR domain from SspH1, SspH2, SlrP or IpaH9.8 increases their E3 ligase activity (Quezada et al., 2009; Chou et al., 2012; Zouhir et al., 2014). A structural study of SspH1 showed that the binding of the host protein kinase N1 (PKN1) to the LRR domain releases its inhibitory effect on the NEL domain, which promotes the ubiquitylation and degradation of PKN1, and leads to the attenuation of the androgen response (Haraga and Miller, 2006; Keszei et al., 2014). Site-directed mutagenesis suggests that the LRR domain inhibits the activity of NELs not by preventing the catalytic cysteine from forming an intermediate with ubiquitin, but by selectively blocking the transfer of ubiquitin onto substrate proteins (Keszei and Sicheri, 2017).
During Shigella infection, chromosomally encoded IpaH proteins play a key role in dampening the host inflammatory response by mediating proteasomal degradation of NF-κB-related proteins (Ashida and Sasakawa, 2015). IpaH9.8 promotes the ubiquitylation and proteasome-dependent degradation of the NF-κB essential modulator (NEMO, also known as IKKγ), a component of the inhibitor of κB kinase (IKK) complex. Without NEMO, NF-κB does not translocate to the nucleus, which prevents the activation of the transcription of proinflammatory cytokines and antimicrobial peptide genes (Ashida et al., 2010; Takagi et al., 2016). IpaH0722 ubiquitylates the tumor-necrosis factor (TNF) receptor associated factor 2 (TRAF2) and targets it for proteasome-dependent degradation, thereby inhibiting the protein kinase C-dependent activation of NF-κB (Ashida et al., 2013). IpaH4.5 directly interacts with and mediates the ubiquitylation and degradation of the NF-κB protein subunit p65 (also known as RelA) (Wang et al., 2013). Moreover, IpaH4.5 has been shown to ubiquitylate TANK-binding kinase 1 (TBK1) to inhibit IFN regulatory factor 3 (IRF3)-mediated cytokine activation and the host antibacterial response (Zheng et al., 2016). IpaH1.4 and its homolog IpaH2.5, ubiquitylate HOIP (also known as RNF31), the catalytic component of the linear ubiquitin chain assembly complex (LUBAC) (de Jong et al., 2016). Degradation of HOIP decreases LUBAC-mediated linear ubiquitin chain formation, which suppresses the signal for NF-κB activation. Finally, instead of targeting the NF-κB pathway, the E3 ligase activity of IpaH7.8 activates the inflammasome to stimulate macrophage death (Suzuki et al., 2014).
XL-box-containing E3 ligase XopL
Xanthomonas outer protein L (XopL) is one of ∼30 T3SS effectors translocated to plant cells during Xanthomonas campestris pv. vesicatoria (Xcv) infection (Fig. 2) (Singer et al., 2013). Production of XopL in Nicotiana benthamiana induces plant cell death, a phenomenon that is dependent on the E3 ligase activity of XopL. The E3 ligase domain of XopL catalyzes K11-linked polyubiquitin chain formation with selective E2s. This domain, which was subsequently named the XL-box, assumes a unique fold not seen in other known structures (Singer et al., 2013). It is composed of two lobes that assemble into an ‘L’-shaped protein. The cavity between the two lobes contains residues important for the E3 ligase activity, but the lack of a catalytic cysteine residue suggests that XopL mediates ubiquitylation through a mechanism that is different from that of the HECT-type E3 ligases and NELs. However, similar to NELs, XopL contains an N-terminal LRR domain, which is proposed to be involved in substrate recognition and, together with the XL-box domain, is required for the induction of plant cell death (Singer et al., 2013).
L. pneumophila SidC
Another bacterial E3 ligase with a fold unlike any other eukaryotic E3 is the L. pneumophila T4SS effector SidC (Sid is a mnemonic for ‘substrate of Icm/Dot’) (Fig. 2). Upon translocation into host cells, SidC associates with the Legionella-containing vacuole membrane through its C-terminal phosphatidylinositol-4-phosphate [PI(4)P]-binding domain, where it promotes the recruitment of endoplasmic reticulum (ER) membranes (Horenkamp et al., 2014; Luo and Isberg, 2004; Ragaz et al., 2008). Although the crystal structure of the N-terminal domain of SidC has been independently solved by three groups (Gazdag et al., 2014; Horenkamp et al., 2014; Hsu et al., 2014), only one group noticed a canonical Cys-His-Asp catalytic triad located at the surface of this unique crescent-like structure. This domain was subsequently confirmed to possess E3 ligase activity that is dependent on these catalytic triad residues (Hsu et al., 2014). Moreover, the structure of full-length SidC revealed that the E3 ligase activity may be allosterically regulated through binding of the C-terminal domain to PI(4)P (Luo et al., 2015). The discovery that XopL and SidC are novel E3 ligases, but not NELs, raises the possibility that other pathogens, or even eukaryotic cells, may utilize proteins with similar folds to catalyze ubiquitylation.
Exploiting ubiquitylation via non-E3 ligase enzymes
All thus far described ubiquitylation processes require an E1 and E2, as well as one of the many structurally diverse E3 enzymes. A paradigm shift away from this fundamental principle came with the discovery of the SidE family of L. pneumophila effector proteins, which catalyzes ubiquitylation in an E1- and E2-independent manner (Qiu et al., 2016). SidE and its homologs SdeA, SdeB and SdeC contain a conserved mono-ADP-ribosyltransferase (mART) domain with a putative catalytic R-S-ExE motif (Fig. 3A) (Qiu et al., 2016). Conventional mART domain-containing enzymes covalently attach the ADP-ribose moiety of nicotinamide adenine dinucleotide (NAD) to substrate proteins (Ueda and Hayaishi, 1985). The mART domain of SidE family members was found to catalyze ADP-ribosylation of ubiquitin on its Arg42 residue to activate it for the ubiquitylation of several Rab proteins including Rab33 (Fig. 3B) (Qiu et al., 2016). L. pneumophila mutant strains lacking the SidE family are attenuated for both growth in the host Dictyostelium discoideum and recruitment of the ER marker GFP–HDEL to Legionella-containing vacuole membranes (Qiu et al., 2016). Both defects are complemented by plasmid-encoded SdeA but not by the catalytically inactive SdeA(E/A), suggesting that a functional mART domain is critical for the biological role of the SidE family (Qiu et al., 2016).
Two subsequent studies elucidated the exact molecular mechanism underlying SidE-mediated ubiquitylation (Bhogaraju et al., 2016; Kotewicz et al., 2017). The mART domain of SidE is flanked by a nucleotidase-phosphohydrolase (NP) domain that cleaves the phosphodiester bond in ADP-ribosylated ubiquitin (ADPr-Ub) to release adenosine monophosphate (AMP) and phospho-ribosylated ubiquitin (Pr-Ub), which is the ubiquitin moiety that is ultimately attached to substrate proteins (Bhogaraju et al., 2016; Kotewicz et al., 2017). The mART domain and NP domain are also required for the ubiquitylation of the ER protein reticulon 4 (Rtn4) and its recruitment to the Legionella-containing vacuole (Kotewicz et al., 2017). Replacement of the catalytic histidine residue in the NP domain renders it inactive, which leads to an accumulation of ADPr-Ub and a failure of substrate ubiquitylation (Bhogaraju et al., 2016; Kotewicz et al., 2017). Similar to what is observed for the mART domain, the NP domain activity is required for intracellular growth of L. pneumophila (Kotewicz et al., 2017). Moreover, unlike conventional ubiquitylation, which primarily occurs on lysine residues, Pr-Ub is preferentially attached to serine residues within substrate proteins via a phosphodiester linkage (Bhogaraju et al., 2016). Bhogaraju and colleagues also proposed that SidE-mediated accumulation of Pr-Ub in cells attenuates the conventional ubiquitin cascade, possibly by steric hindrance generated by the phospho-ribosyl group at the Arg42 residue in ubiquitin (Bhogaraju et al., 2016). Notably, this inhibitory effect has only been observed within the context of transiently transfected HEK293T cells (Bhogaraju et al., 2016), conditions where the levels of ectopically produced SidE are likely orders of magnitude higher than those during infection, and it remains to be examined whether this inhibitory role of SidE is indeed biologically relevant during infection.
Taken together, these discoveries suggest that targeting host protein ubiquitylation through a diverse set of ubiquitin ligases is indeed an important armory utilized by bacterial pathogens during host infection.
Effectors that reverse host protein ubiquitylation
Given that ubiquitylation is a reversible process, it is not surprising that pathogens exploit this post-translational modification by encoding effectors that mimic host DUBs to remove ubiquitin from substrate proteins (Fig. 1). DUB effectors have been found in bacterial pathogens including Salmonella, Chlamydia, Xanthomonas and Legionella, and their functions during infection are summarized below (see also Table S1).
The Salmonella T3SS effector SseL hydrolyzes mono- and poly-ubiquitin substrates in vitro (Rytkönen et al., 2007). Salmonella mutant strains lacking SseL show an enhancement of ubiquitylated aggregates near the Salmonella-containing vacuole within infected macrophages (Rytkönen et al., 2007; Mesquita et al., 2012) and an increase in the number of lipid droplets within infected gall bladder epithelial cells (Arena et al., 2011). One study proposed that the ubiquitylation of IκBα and subsequent NF-κB activation is inhibited in the presence of SseL (Le Negrate et al., 2008a), whereas another study found no evidence for such changes (Mesquita et al., 2013). Hence, further investigation is needed to clarify the role of SseL in controlling the NF-κB pathway. The Chlamydia trachomatis effectors ChlaDub1 and ChlaDub2 remove both ubiquitin and NEDD8, a ubiquitin-like modifier, from host proteins during infection (Misaghi et al., 2006). Studies in transiently transfected cells have revealed that ChlaDub1 prevents ubiquitylation of IκBα, and the impaired degradation of IκBα ultimately suppresses NF-κB signaling (Le Negrate et al., 2008b).
Unlike the DUB effectors from human pathogens, which preferentially hydrolyze and remove ubiquitin and ubiquitin chains, XopD, the T3SS effector from the plant pathogen Xanthomonas, hydrolyzes chains of the small ubiquitin-like modifier (SUMO) (Chosed et al., 2007; Hotson et al., 2003). XopD targets several transcription factors in plants (Canonne et al., 2011; Kim et al., 2013; Tan et al., 2015), and the hydrolysis of the SUMO modification on those transcription factors causes their destabilization, which leads to an inhibition of the anti-Xcv immunity in plants (Kim et al., 2013; Tan et al., 2015). How substrate specificity of DUBs and ubiquitin-like protein proteases is accomplished has recently been revealed through a comparative analysis of the bacterial DUBs SseL, ChlaDub1 and XopD, and two human ubiquitin-like protein proteases (Pruneda et al., 2016). Substrate recognition by these proteases of the cysteine endopeptide (CE) clan was found to occur via three variable regions that constitute the ubiquitin- or ubiquitin-like-protein-binding sites. Notably, these three regions not only determine the specificity of CE clan proteases for a certain type of modifier (ubiquitin versus ubiquitin-like proteins), but also allow them to distinguish different chain linkages (Pruneda et al., 2016).
Further studies of SidE, the L. pneumophila T4SS effector that catalyzes the E1- and E2-independent ubiquitylation described in the previous section, revealed yet another surprise, as this protein was found to also contain a DUB domain with a canonical Cys-His-Asp catalytic triad (Sheedlo et al., 2015) (Fig. 3A). The DUB activity of SdeA preferentially hydrolyzes K63-linked polyubiquitin chains. Intracellular L. pneumophila lacking all SidE family members showed enhanced association with ubiquitylated species, and this phenotype was reversed through exogenous production of wild-type SdeA, but not the deubiquitylation-defective SdeA(C118A) variant (Sheedlo et al., 2015). However, unlike the aforementioned mART-NP-mediated ubiquitylation activity (Qiu et al., 2016), the DUB activity of SidE family proteins was not required for proficient intracellular replication of L. pneumophila (Sheedlo et al., 2015). The presence of domains of opposite function within a single protein is unusual, yet it appears that the DUB activity of SdeC does not serve the purpose of reversing the ubiquitylation catalyzed by the mART and NP domains (Kotewicz et al., 2017). In fact, cleavage of K63-linked tetra-ubiquitin chains by SdeC is prohibited in the presence of NAD, suggesting that the mART domain interferes with the DUB activity of SidEs. One possible function of the DUB domain, as proposed by Isberg and colleagues (Kotewicz et al., 2017), is to prevent the accumulation of K63-linked ubiquitin chains around the Legionella-containing vacuole, thus increasing the availability of monomeric ubiquitin for the generation of ADPr-Ub and Pr-Ub. Whether L. pneumophila encodes effector proteins that can reverse the Pr-Ub modification of substrate proteins remains to be seen.
Post-translational modification of ubiquitin components
As described above, many bacterial effectors catalyze either ubiquitin attachment to (E3s) or removal from (DUBs) substrate proteins. While those may be more common approaches to directly interfere with host protein ubiquitylation, bacterial effectors with activities that indirectly regulate host protein ubiquitylation through post-translational modification have also been discovered.
NleE is a T3SS effector from EPEC with S-adenosyl-L-methionine-dependent methyltransferase activity (Fig. 4A). It methylates a Zn2+-coordinating cysteine residue in the ubiquitin-sensing proteins TAK1-binding proteins 2 and 3 (TAB2 and TAB3), which blocks their capacity to bind ubiquitin chains and thereby disrupts NF-κB signaling (Zhang et al., 2012; Yao et al., 2014). Another EPEC effector, NleB, harbors N-acetyl-D-glucosamine (O-GlcNAc) transferase activity (Fig. 4B). NleB mediates O-GlcNAcylation of glyceraldehyde 3-phosphate dehydrogenase (GAPDH), which disrupts its interaction with TRAF2 and blocks TRAF2 polyubiquitylation, ultimately suppressing NF-κB activation (Gao et al., 2013). Last but not least, the Shigella T3SS effector OspI selectively deamidates UBC13 (also known as UBE2N), the E2 enzyme required for TRAF6 self-ubiquitylation (Fig. 4C) (Deng et al., 2000; Fukushima et al., 2007; Lamothe et al., 2007). OspI converts the glutamine residue (Gln100) located near the interacting surface of UBC13 with TRAF6 into a glutamic acid residue. This alteration suppresses polyubiquitin chain formation on TRAF6 and decreases NF-κB activation (Sanada et al., 2012), possibly by blocking its association with UBC13 (Nishide et al., 2013). The OspI-binding site on UBC13 largely overlaps with the binding region through which UBC13 interacts with other E3 ligases, suggesting that the bacterial effector exploits a binding surface that is similar to that of its eukaryotic E3 competitors (Fu et al., 2013; Nishide et al., 2013).
Regulation of bacterial effector degradation through ubiquitin
In addition to regulating host proteins through ubiquitylation, bacterial pathogens also hijack the ubiquitin machinery to control the activity of their own effector proteins during infection. Bacterial effectors are targeted to the proteasome primarily by host protein-mediated ubiquitylation and, occasionally, by effector protein-mediated ubiquitylation, as discussed above for the L. pneumophila effectors SidH and LubX. Below, we will summarize how host-mediated ubiquitylation and proteasomal degradation of bacterial effectors from Salmonella, Yersinia and Pseudomonas is used to fine-tune virulence.
SopE and SptP are effector proteins that regulate the invasion of Salmonella into non-phagocytic cells. SopE is a guanine nucleotide exchange factor (GEF) that activates host Rho family GTPases, such as Cdc42 and Rac1 (Fig. 5A), which causes rearrangements of the actin cytoskeleton that result in local membrane ruffles and Salmonella uptake (Hardt et al., 1998; Stender et al., 2000; Zhou et al., 2001). In contrast, SptP functions as a GTPase-activating protein (GAP) towards Rho family GTPases and reverses the effect of SopE on cytoskeletal dynamics (Fu and Galan, 1999). Although both SopE and SptP are simultaneously translocated into host cells during the early stage of infection, the intracellular level of SopE decreases much more rapidly than that of SptP (Kubori and Galán, 2003). This difference in protein levels is due to the preferential ubiquitylation and subsequent proteasome-dependent degradation of SopE, which is determined by intrinsic properties of its N-terminal secretion and translocation domain (Kubori and Galán, 2003). The identity of the host E3 ligase(s) that preferentially ubiquitylate SopE over SptP, and the molecular details underlying the differential recognition of their secretion and translocation domains have remained unclear.
A similar yet more subtle strategy has been discovered for YopE, a T3SS effector of Yersinia enterocolitica with anti-phagocytic function (Fig. 5B). Similar to what is seen with SptP, YopE exhibits GAP activity towards Rho GTPases, which leads to the disruption of actin microfilament structures (Black and Bliska, 2000; Von Pawel-Rammingen et al., 2000). While YopE from Yersinia enterocolitica serogroup O8 is degraded within host cells (Ruckdeschel et al., 2006), the highly homologous YopE proteins from serogroup O3 or O9 evade polyubiquitylation and proteasomal destabilization owing to the lack of two lysine residues at position 62 and 75, the ubiquitylation sites in serogroup O8 YopE (Gaus et al., 2011; Hentschke et al., 2007). Accumulation of stabilized YopE is accompanied by increased cytotoxicity within infected host cells, and Yersinia mutant strains producing stabilized YopE show attenuated virulence during infection in mice and reduced dissemination into the liver and spleen (Gaus et al., 2011).
The T3SS effector ExoT from Pseudomonas also undergoes ubiquitylation and proteasomal degradation within host cells (Fig. 5C). However, unlike YopE, ubiquitylation of ExoT limits P. aeruginosa infection in C57BL/6 mice (Balachandran et al., 2007). The host RING-type E3 ubiquitin ligase Cbl-b is required for ExoT ubiquitylation and degradation. Overproduction of Cbl-b increases the rate of ExoT degradation within transiently transfected cells, and Cbl-b-knockout C57BL/6 mice (Cbl-b–/–) are significantly more susceptible to P. aeruginosa PA103ΔexoU infection, with a bacterial burden several orders of magnitude higher than that seen in the lungs of Cbl-b+/+ mice. This underscores the importance of Cbl-b-mediated ubiquitylation in limiting the dissemination of ExoT-producing bacteria (Balachandran et al., 2007).
Ubiquitin-dependent regulation of effector protein functions
In addition to its fundamental role in promoting proteasomal degradation, ubiquitin has also been found to critically contribute to the diversification of bacterial effector functions.
Salmonella SopB is a phosphoinositide phosphatase that promotes actin cytoskeletal rearrangements and enhances bacterial uptake (Zhou et al., 2001). Upon T3SS-mediated translocation into host cells, SopB localizes to the inner leaflet of the plasma membrane (Fig. 6A) (Zhou et al., 2001; Marcus et al., 2002; Galyov et al., 1997; Mallo et al., 2008; Norris et al., 1998). Soon after, however, SopB relocates to internal vesicular compartments and the Salmonella-containing vacuole, where it recruits Rab5 and Vps34 and promotes generation of phosphatidylinositol 3-phosphate [PI(3)P] (Mallo et al., 2008; Marcus et al., 2002; Hernandez et al., 2004). This spatiotemporal behavior is controlled by the mono-ubiquitylation of multiple lysine residues in SopB by the host E3 ligase TRAF6 (Ruan et al., 2014; Knodler et al., 2009; Patel et al., 2009). A ubiquitylation-resistant variant of SopB shows prolonged localization to the plasma membrane, which results in sustained actin cytoskeletal rearrangements and intensified activation of Akt proteins, which ultimately leads to defective intracellular replication of Salmonella in macrophages (Patel et al., 2009; Knodler et al., 2009).
The Pseudomonas aeruginosa protein ExoU is a T3SS effector with phospholipase A2 (PLA2) activity. ExoU is cytotoxic in both yeast cells and mammalian cells but not in E. coli, and recombinant ExoU showed catalytic activity in vitro only upon the addition of eukaryotic cell extract, suggesting that a host factor is needed for its function (Sato et al., 2003). It was revealed that the phospholipase activity of ExoU required the presence of ubiquitin (Fig. 6B) (Anderson et al., 2013, 2015), and while ubiquitin chains of various lengths (mono-, di- or poly-ubiquitin) and from different organisms (yeast or mammals) can trigger the phospholipase activity of ExoU, ubiquitin-like proteins such as SUMO-1, ISG15, FAT10 and NEDD8 had no stimulating effect (Anderson et al., 2015). Given that ExoU itself is ubiquitylated within host cells, one hypothesis is that the ubiquitylated ExoU may self-activate upon translocation into host cells (Stirling et al., 2006).
Shigella OspG is a T3SS effector whose kinase activity is needed to dampen the TNF-stimulated inflammatory response by preventing the degradation of phosphorylated IκBα (Kim et al., 2005). Shigella strains lacking OspG cause an enhanced inflammatory response in the rabbit ileal loop model of infection (Kim et al., 2005) and reduced virulence in a murine model of shigellosis (Pruneda et al., 2014). It appears that the kinase activity of OspG is enhanced by formation of complexes with various human E2–ubiquitin conjugates and, to a lesser extent, with the unconjugated form of either E2s (Grishin et al., 2014; Pruneda et al., 2014) or ubiquitin (Fig. 6C) (Zhou et al., 2013). The co-crystal structures of OspG and UbcH5c–ubiquitin (Pruneda et al., 2014) or UbcH7–ubiquitin (UbcH7 is also known as UBE2L3) (Grishin et al., 2014) revealed that binding of the E2–ubiquitin conjugate involves two non-overlapping surfaces in OspG; E2 binding occurs primarily through residues in the N-terminal lobe of OspG, whereas the ubiquitin subunit is bound via the C-terminal lobe. The simultaneous interaction of E2–ubiquitin with both lobes of OspG is believed to stabilize the kinase domain in a conformation that is either more active or less prone to proteolytic degradation within host cells, or both (Grishin et al., 2014).
The large number of bacterial effector proteins dedicated to the interference with or exploitation of host ubiquitylation described in this Commentary is evidence that this conserved machinery, in addition to being critical for eukaryotic cell development and homeostasis, also plays a major role for the outcome of many bacterial infections. The wide variety of molecular mechanisms used by these effectors to accomplish their goal is fascinating. Not only do the effectors modulate ubiquitylation by functioning as molecular mimics of host factors, such as E3 ligases or DUBs, but they also hijack host ubiquitylation in unconventional ways by catalyzing E1- and E2-independent ubiquitylation, by post-translationally modifying components of the ubiquitylation cascade or by coupling their own catalytic activity to the presence of components of the ubiquitylation machinery. The knowledge gained from studying the function of these effectors during ubiquitin exploitation revealed that this conserved post-translational modification machinery, which we think of as being reasonably well understood, appears to offer a much more diverse range of regulatory concepts and intervention points than initially anticipated.
The existence of bacterial effectors that function in ways that are noticeably distinct from molecular mimicry is particularly astonishing as it raises several important questions that need to be addressed. For example, have these effectors evolved from bacterial ancestor proteins, or were they acquired by the pathogens through horizontal gene transfer from a host? If the latter is the case, do functional homologs to these unique effectors still exist in eukaryotes and, if so, what is their biological role? Alternatively, have they been eliminated or replaced with alternative regulators over time and, if so, why? Regardless of what the answers might be, the existence of ubiquitin-regulatory effectors that are unique to bacterial pathogens may represent an Achilles heel for the development of novel therapeutics aimed at treating infectious diseases in the future. Moreover, the fact that ubiquitylation can occur in an E1- and E2-independent manner by simply pairing the activity of an ADP-ribosyltransferase with that of a nucleotidase raises the intriguing possibility that eukaryotic cells, including human cells, once did or still may encode proteins or protein complexes that catalyze a similar reaction. If this is the case, the spectrum of unconventional post-translational modifications, such as the addition of ADP-ribose-conjugated ubiquitin to substrate proteins, may be more widely distributed among organisms than is currently appreciated. Evidently, the resourcefulness of microbes and the functional diversity of their effectors described in this Commentary serve as evidence that much can be learned from intracellular bacteria about eukaryotic cell biology and the many roles of ubiquitylation required for human health.
We would like to thank members of the Machner laboratory for the critical reading and comments on the manuscript.
The work of our laboratory is support by the Intramural Research Program of the National Insitutes of Health (NIH). Deposited in PMC for release after 12 months.
The authors declare no competing or financial interests.