Host manipulation by bacterial type III and type IV secretion system effector proteases

Proteases are powerful enzymes, which cleave peptide bonds, leading most of the time to irreversible fragmentation or degradation of their substrates. Therefore they control many critical cell fate decisions in eukaryotes. Bacterial pathogens exploit this power and deliver protease effectors through specialised secretion systems into host cells. Research over the past years revealed that the functions of protease effectors during infection are diverse, reflecting the lifestyles and adaptations to specific hosts; however, only a small number of peptidase families seem to have given rise to most of these protease virulence factors by the evolution of different substrate‐binding specificities, intracellular activation and subcellular targeting mechanisms. Here, we review our current knowledge about the enzymology and function of protease effectors, which Gram‐negative bacterial pathogens translocate via type III and IV secretion systems to irreversibly manipulate host processes. We highlight emerging concepts such as signalling by protease cleavage products and effector‐triggered immunity, which host cells employ to detect and defend themselves against a protease attack.


| INTRODUCTION
Gram-negative bacteria evolved virulence factors that enable them to compete with environmental predators, such as protozoa, or to exploit unicellular and higher organisms such as plants and animals as hosts. Protein secretion systems, particularly Type III and IV secretion systems (T3SS and T4SS), play a fundamental role in these interactions (Costa et al., 2015). T3SS and T4SS are highly optimised multi-protein machineries that use distinct operation modes to translocate so-called effector proteins from the bacteria directly into the host cell cytoplasm, enabling effective manipulation of host processes (Costa et al., 2021;Galan & Waksman, 2018). Diverse combinations of T3SSs and T4SSs are found in numerous bacteria, and they are key virulence factors for many clinically and economically important pathogens, including Salmonella, Legionella and Pseudomonas species (Abby et al., 2016;Buttner, 2012). Variable effector repertoires shape host and tissue tropism, virulence and pathology.
Effectors are structurally diverse, combining one or more activity domains with an essential, shared determinant, a translocation signal, encoded at the N-terminus for T3SS or at the C-terminus for T4SS effectors (Galan & Waksman, 2018). Activity domains can exert their effect on cellular processes by molecular mimicry of host proteins, for example, eukaryotic GTPase-activating proteins (GAPs) to control the state of small GTPases (Elde & Malik, 2009) and/or have enzymatic activity. Many enzymatic activity domains manipulate proteins by introducing or removing post-translational modifications (PTMs), e.g., phosphorylation or ubiquitination Tahir, Rashid, & Afzal, 2019). This overrides host control of protein activity, but as many PTMs are reversible, the duration of the manipulation depends on the availability and ability of cognate host proteins to counteract the effectors and restore normal protein states.
To irreversibly manipulate host processes, several pathogens deploy T3SS or T4SS protease effectors, which cleave a peptide bond in the polypeptide backbone of target proteins. An exception are deubiquitinase (DUB) effectors that cleave the isopeptide bonds between proteins and ubiquitin or ubiquitin-like modifiers releasing the intact proteins (recently reviewed by (Kitao, Nagai, & Kubori, 2020;Kubori, Kitao, & Nagai, 2019). Here, we review the enzymology and functions of effector proteases of Gram-negative bacterial pathogens, which irreversibly cleave the polypeptide backbone of target proteins (Table 1).

| ENZYMOLOGY AND CLASSIFICATION OF PROTEASES
Proteases are believed to have evolved in primaeval organisms as destructive enzymes mediating protein catabolism and production of amino acids (Neurath, 1984), but their vital roles in cell signalling and physiology are now recognised (Lopez-Otin & Bond, 2008).
They are classified based on sequence and structural homology, cleavage specificity and catalytic mechanism (Rawlings, 2013), information which is captured in the MEROPS database (https://www. ebi.ac.uk/merops/, Rawlings et al., 2018]). Key terminology to describe the cleavage specificity is illustrated in Figure 1a. Proteolysis requires the catalytic activation of the target peptide bond and/or an incoming water molecule to enable hydrolysis. Proteases have evolved different substrate binding pockets and catalytic mechanisms to facilitate this. Six major classes are distinguished: aspartic, glutamic, threonine, metallo-, serine and cysteine proteases (Rao, Tanksale, Ghatge, & Deshpande, 1998;Rawlings, 2013). All identified effector proteases, but one mechanistically not fully characterised threonine protease belong to the last three classes.
Within these classes, a few peptidase families seem to have given rise to the vast majority of effectors. The expansion of these families might have been promoted by horizontal gene transfer and subsequent functional diversification, an important driver for the evolution of virulence (Diard & Hardt, 2017). Adaptation of existing effectors for new purposes seems favourable compared to the evolution of new effectors from non-effector genes, which would require, for example, the acquisition of secretion signals and integration into the regulatory circuits controlling virulence. In addition, unknown intrinsic features might make these peptidases in particular suitable for their use as effectors. However, as we only start to reveal the functions of effectors from environmental pathogens such as Legionella pneumophila, which have much larger effector arsenals than host-restricted pathogens and a high propensity to convert host cell genes to effectors, additional effector peptidase families are likely to be discovered. proteases, which rely on a catalytic triad consisting of a serine or cysteine, a histidine and a third residue, often aspartate (Rao et al., 1998;Rawlings, 2013), and metalloproteases, which depend on one or two divalent metal ion co-factors for catalysis (Cerda- Costa & Gomis-Ruth, 2014). Variations of this catalytic core occur; for example, some cysteine proteases only require cysteine and histidine residues. Mutation of the catalytic triad or metal-coordinating residues renders these proteases usually inactive. The substrate specificity is determined by the active site binding pocket, but often also by exosites, which recruit the substrates and can also control the subcellular localisation of proteases ( Figure 1c). Production of inactive pro-enzymes, which only get activated in the host, and secretion of finite amounts of effector during a limited time window are additional mechanisms employed by pathogens to exert control of protease activity. Despite their destructive potential, most effector proteases perform selective manipulations of the host ( Figure 1d).

| EFFECTOR PROTEASES OF BACTERIAL ENTEROPATHOGENS
Enteropathogenic and enterohemorrhagic Escherichia coli (EPEC & EHEC), Salmonella, Shigella, Yersinia and Vibrio spp. cause a wide spectrum of diarrheal diseases and are a global health problem (Collaborators, 2018). They all rely on T3SSs as key virulence factors to deliver distinct repertoires of effectors to shape their pathogenesis.
However, due to shared evolutionary origins and horizontal gene transfer, homologue effectors are often found in several enteropathogens (Pinaud, Sansonetti, & Phalipon, 2018

Rac1
Disruption of cytoskeletal dynamics, inhibition of cell migration and phagocytosis Abbreviations: M ID, MEROPS ID; n.a., not assigned; PT, protease type; SST, secretion system type.
a Most common name, which was used in this review, synonyms might exist. b ID of representative protein in Uniprot database https://www.uniprot.org/.
c Organism/Group of organisms, in which the prototype protease was described. F I G U R E 1 (a) Scheme illustrating the terminology used to describe protease cleavage specificity. Endopeptidases catalyse the hydrolysis of a peptide bond in the central region, exopeptidases close to the N-or Cterminus of a substrate and are also termed aminopeptidases or carboxypeptidases, respectively. To describe the cleavage site sequence specificity, the residues N-terminal to the cleaved bond are designated P1, P2, etc. and the ones C-terminal P1 0 , P2 0 , etc., with numbers increasing with distance to the scissile bond.
(b) Scheme of the active sites of serine/ cysteine-and metalloproteases. Serine and cysteine proteases rely on a catalytic triad, consisting of a serine or cysteine, a histidine and a third residue, aspartate, asparagine or glutamate. The residues are found in a spatially conserved constellation, facilitating deprotonation of the hydroxyl (serine) or thiol (cysteine) groups via the histidine, activating them for a nucleophilic attack on the carbonyl carbon of the peptide bond. The resulting covalent enzyme-substrate intermediate is hydrolysed by a water molecule releasing two peptide fragments and the protease. Metalloproteases rely on one or two divalent metal ion co-factors, for example, Zn 2+ , which polarise a water molecule, priming it for a nucleophilic attack, and the substrate carbonyl group, making its carbon more electrophilic and susceptible to the water molecule. The attack leads to a non-covalent intermediate which resolves  (Creuzburg et al., 2017). JNK cleavage is associated with reduced pro-apoptotic signalling (Baruch et al., 2011).
NleC is the prototype of the second important zinc metalloprotease effector family conserved in many A/E pathogens. Each S.
enterica isolate encodes at least one of three related proteases, PipA, GtgA and GogA (Sun, Kamanova, Lara-Tejero, & Galan, 2016), which are encoded in SPI-5 (PipA) (Wood et al., 1998)  Salmonella enterica also uses the SPI-2 effector SpvD to inhibit NF-κB-driven pro-inflammatory signalling, contributing to full virulence in a murine infection model Rolhion et al., 2016). SpvD is structurally related to CA clan papain-like proteases (Barrett & Rawlings, 1996), the largest family of cysteine proteases, which have evolved to target a huge diversity of substrates.
Its activity in vitro and vivo depends on its catalytic triad . In vitro, SpvD variants from different Salmonella isolates hydrolyse an RLRGG peptide-based small-molecule substrate indicative of DUB-like activity, but only one variant shows activity towards a full-length ubiquitin-probe . SpvD targets but does not cleave cytoplasmic Exportin-2 (Xpo2), derailing nuclear transport and shuttling of p65 into the nucleus Rolhion et al., 2016). Hence, the exact mechanism of action of SpvD during infection remains elusive.

| Inhibition of necroptosis by EspL-type cysteine proteases
The EPEC effector EspL also belongs to the papain-like CA clan proteases (

| DISRUPTION OF SMALL GTPASE FUNCTIONS BY PROTEASE EFFECTORS
The multitudinous Ras-family of small GTPases controls numerous essential cellular processes, including cytoskeletal dynamics and membrane trafficking (Wennerberg, Rossman, & Der, 2005) and members are therefore prime targets for pathogens. Yersinia outer protein T (YopT) is a T3SS cysteine protease effector produced by many pathogenic Yersinia species, for example, Y. pestis, Y. pseudotuberculosis and Y. enterocolitica (Iriarte & Cornelis, 1998;Palace, Proulx, Szabady, & Goguen, 2018). YopT cleaves directly before the prenylated C-terminal cysteine of RhoGTPases RhoA, Rac, Cdc42 and RhoG independently of the activation state, releasing them from the membrane (Mohammadi & Isberg, 2009;Shao, Vacratsis, et al., 2003;Shao, Merritt, Bao, Innes, & Dixon, 2002). Prenylation and a polybasic region in the Rho GTPases are key for recognition and processing (Fueller & Schmidt, 2008;Shao, Vacratsis, et al., 2003), and small differences in these elements lead to different turnover rates, making RhoA the preferred substrate (Fueller & Schmidt, 2008); however, during infection substrate selection might be influenced by cell-type-specific expression levels of the GTPases and other effectors, in particular YopE, which is a Rho GAP (Von Pawel- Rammingen et al., 2000). RhoA, Rac and Cdc42 have important functions in coupling cytoskeletal and membrane dynamics and microinjection of YopT causes cell rounding and death (Iriarte & Cornelis, 1998;Sorg, Burns, Goehring, Aktories, & Schmidt, 2001). In cellular infection models, YopT contributes to the inhibition of uptake into nonphagocytic and phagocytic cells and interferes with podosomal adhesion structures required for macrophage migration (Adkins et al., 2007;Aepfelbacher et al., 2003;Grosdent, Maridonneau-Parini, Sory, & Cornelis, 2002). While YopT obstructs functions of the GTPases at the membrane, it generates distinct cytosolic and nuclear pools of the delipidated but active proteins, for example, Rac1, which could have biological functions (Wong & Isberg, 2005). In macrophages, enzymatically active YopT augments TNFα production in response to T3SS activity (Auerbuch, Golenbock, & Isberg, 2009) and stimulates a pyrin-dependent innate immune surveillance pathway, which senses the inactivation of RhoA and drives inflammasome activation leading to proinflammatory cytokine release and pyroptosis (Medici, Rashid, & Bliska, 2019). This effector-triggered immunity (ETI) (Lopes Fischer, Naseer, Shin, & Brodsky, 2020) might explain why YopT has only a minor role (Palace et al., 2018;Viboud, Mejia, & Bliska, 2006)  YopT is the prototype of the peptidase C58 family (Shao et al., 2002 pneumophila T4SS effectors (Schroeder et al., 2010) and large toxins such as PaTox and MARTX (Bogdanovic et al., 2019), but their physiological activities still need to be defined. Interestingly, MARTX from Vibrio cholerae contains an additional cysteine protease domain (CPD), which is required for auto-processing upon binding of the host factor inositol hexakisphosphate (IP6) (Egerer & Satchell, 2010). The T3SS effector VPA1380 of Vibrio parahaemolyticus, a marine and foodborne human pathogen, which causes gastroenteritis (Broberg, Calder, & Orth, 2011), shares similarity with CPDs and inhibits the growth of yeast in a catalytic cysteine and IP6 dependent manner, but actual proteolytic activity of VPA1380 still needs to be validated (Calder et al., 2014). The presence of these related protease domains in effectors and secreted toxins suggests that they are common building blocks of new virulence factors.

| T4SS EFFECTOR PROTEASES IN HOST SUBVERSION BY LEGIONELLA PNEUMOPHILA
Legionella pneumophila, the causative agent of the potentially fatal pneumonia Legionnaires' disease, translocates more than 330 effectors through its Defect in organelle trafficking/Intracellular multiplication (Dot/Icm) T4SS to establish a replication-permissive niche, the Legionella containing vacuole (LCV), in macrophages and protozoa (So, Mattheis, Tate, Frankel, & Schroeder, 2015). Key for the intracellular lifestyle is the evasion of phagolysosomal degradation and innate defence mechanisms, such as autophagy, which can trap intracellular bacteria in membrane-bound compartments that mature into the degradative auto-phagolysosomes (Huang & Brumell, 2014).
L. pneumophila uses effector proteases to interfere with several host processes.
After membrane nucleation autophagosome maturation proceeds with recruitment of Microtubule-associated protein light chain 3 (LC3) to PI3P-containing early, pre-autophagosomal membranes and modification of LC3 with a phosphatidylethanolamine (PE) lipid anchor (Dikic & Elazar, 2018). LC3 then drives cargo sequestration and autophagosome expansion. The Ulp family DUB-like cysteine protease effector RavZ (Lpg1683) blocks this step (Choy et al., 2012;Horenkamp et al., 2015). RavZ localises to these membranes using a PI3P-binding domain and sensing their unique curvature, interacts with LC3 via three eukaryotic-like LC3-interacting region (LIR) motifs, extracts the PE anchor from the membrane and cleaves it together with the C-terminal glycine to which it is conjugated (Choy et al., 2012;Horenkamp et al., 2015;Kwon, Kim, Jung, et al., 2017;Yang, Pantoom, & Wu, 2017). Mammalian orthologues of LC3, such as GABARAPL1, which fulfil similar but non-redundant functions, are also cleaved (Choy et al., 2012;Schaaf, Keulers, Vooijs, & Rouschop, 2016). The mechanism of action of RavZ was studied in great detail, showing that it requires RavZ-membrane, RavZ-PE lipid and RavZ-LC3 interactions that are enhanced by a unique conformation of the LIR motifs generating higher binding affinity than classical LIR motifs and that are distinct from the binding mode of the cognate host protease ATG4, leading to irreversible rather than reversible delipidation (Kauffman et al., 2018;Park et al., 2021;Park et al., 2019;Yang, Pantoom, & Wu, 2020). While this an effective mechanism, a L. pneumophila Philadelphia ΔravZ strain still prevents sequestration of the LCV by autophagy, suggesting that this strain as well as L. pneumophila isolates, which do not encode RavZ, most likely deploy additional effectors to disrupt the process (Omotade, Roy, & Brodsky, 2020).
The host cytoskeleton is targeted by the L. pneumophila zincdependent metalloprotease effector RavK (Lpg0969), which cleaves actin, rendering it unusable for polymerisation (Liu et al., 2017). The cleavage of actin is observed upon infection with L. pneumophila overexpressing RavK but not with wild-type bacteria and the effector is dispensable for intracellular growth, leaving the role of its activity during infection uncertain.
Ectopic expression of the effector RavJ (Lpg0944) also perturbs the cytoskeleton, manifesting as the accumulation of F-actin on the plasma membrane (Liu et al., 2017). Its overexpression is toxic for yeast and this can be controlled by the meta-effector LegL1, which binds and seals the active site of RavJ (Urbanus et al., 2016). RavJ has two domains, an N-terminal domain with papain protease fold and a Cterminal domain that interacts with several septins in vitro (Urbanus et al., 2016). The catalytic residues of RavJ are essential for toxicity in yeast; however, Cys-His-Asp catalytic triads are also employed by other enzymes for different reactions, for example, the ubiquitin transfer by the ubiquitin E3 ligase effector SidC (Hsu et al., 2014), and actual protease activity and substrates of RavJ remain to be confirmed.

| EFFECTOR PROTEASES OF PHYTOPATHOGENS
Plants rely on two main cell autonomous innate immune mechanisms to recognise and fight invading pathogens, as they lack circulating immune cells and adaptive immunity (Jones & Dangl, 2006). Plant pattern recognition receptors (PRRs) are activated by PAMPs such as flagella, initiating PAMP-triggered immunity (PTI). Pathogens disable PTI using effectors, but resistance (R) proteins guard this and activate ETI (Yuan, Ngou, Ding, & Xin, 2021 Pseudomonas syringae AvrPphB, a YopT-like, peptidase C58 effector (Zhu, Shao, Innes, Dixon, & Xu, 2004), is produced as a proprotease that upon translocation undergoes autoproteolysis, allowing modification of unmasked myristoylation and palmitoylation sites with lipid moieties that tether the effector to membranes (Dowen, Engel, Shao, Ecker, & Dixon, 2009;Nimchuk et al., 2000;Puri et al., 1997;Shao et al., 2002). AvrPphB then cleaves the cytoplasmic kinase PBS1 and other PBS1-like kinases including BIK1, PBL1 and PBL2 (Shao, Golstein, et al., 2003;Zhang et al., 2010) These examples illustrate that phytopathogens encode many protease effectors, which have often homologues in pathogens with different host ranges, and, thus, lessons learned about their enzymology and recognition by the host can be exemplary beyond plant-pathogen interactions. It will be interesting to see if this promise holds true for P. syringae effector HopB1, which binds the plant immune receptor FLS2 to cleave the co-receptor BAK1 and disrupt flagellin-induced PTI using an unusual catalytic triad formed by a threonine, a histidine and two aspartates, making it the first threonine protease effector (L. Li et al., 2016).

| CONCLUSION
Proteases are powerful enzymes, which irreversibly cleave proteins,

ACKNOWLEDGMENTS
Protein cartoons were created with the program "Illustrate" (Goodsell, Autin, & Olson, 2019). This work was supported by the Medical Research Council UK grant MR/R010552/1.

CONFLICT OF INTEREST
The authors declare no potential conflict of interest.

DATA AVAILABILITY STATEMENT
Not applicable -no new data was generated for this literature review.