Conserved Structural Mechanisms for Autoinhibition in IpaH Ubiquitin Ligases*

Background: The IpaH ubiquitin ligases are bacterial type III effectors that ubiquitinate and inactivate eukaryotic host proteins. Results: Two distinct modes of autoinhibition confer substrate-dependent activation of IpaH enzymes. Conclusion: Conserved intramolecular interactions regulate IpaH activity. Significance: IpaH autoregulatory mechanisms may be potential targets for therapeutic intervention. The IpaH family of novel E3 ligase (NEL) enzymes occur in a variety of pathogenic and commensal bacteria that interact with eukaryotic hosts. We demonstrate that the leucine-rich repeat (LRR) substrate recognition domains of different IpaH enzymes autoinhibit the enzymatic activity of the adjacent catalytic novel E3 ligase domain by two distinct but conserved structural mechanisms. Autoinhibition is required for the in vivo biological activity of two IpaH enzymes in a eukaryotic model system. Autoinhibition was retro-engineered into a constitutively active IpaH enzyme from Yersinia pestis by introduction of single site substitutions, thereby demonstrating the conservation of autoregulatory infrastructure across the IpaH enzyme family.

The IpaH family of novel E3 ligase (NEL) enzymes occur in a variety of pathogenic and commensal bacteria that interact with eukaryotic hosts. We demonstrate that the leucine-rich repeat (LRR) substrate recognition domains of different IpaH enzymes autoinhibit the enzymatic activity of the adjacent catalytic novel E3 ligase domain by two distinct but conserved structural mechanisms. Autoinhibition is required for the in vivo biological activity of two IpaH enzymes in a eukaryotic model system. Autoinhibition was retro-engineered into a constitutively active IpaH enzyme from Yersinia pestis by introduction of single site substitutions, thereby demonstrating the conservation of autoregulatory infrastructure across the IpaH enzyme family.
The ubiquitin system regulates diverse cellular processes in all eukaryotes through the covalent attachment of ubiquitin to acceptor lysine residues on target substrates, thereby dictating protein stability, activity, and localization. Ubiquitin is activated and conjugated to its substrates by a sequential enzymatic cascade, E1 3 E2 3 E3 (1). E3 enzymes, also known as ubiquitin ligases, specifically recognize their substrates via protein interaction domains. The eukaryotic E3 enzymes fall into two major classes. The HECT 4 domain class facilitates transfer to substrates via a catalytic cysteine residue, whereas the RING/ U-box domain class serves as a platform to bridge the E2 enzyme to the substrate.
The crucial role of the ubiquitin system in the regulation of host immune surveillance poises ubiquitin-dependent processes as a pervasive target for a wide range of pathogens (2). Several type III secretion virulence effectors in Gram-negative pathogenic bacteria possess ubiquitin ligase activity, including AvrPtoB from Pseudomonas syringae (3), SopA from Salmonella enterica typhimurium (4), NIeG from enterohemorrhagic Escherichia coli (5), and the IpaH family members from Shigella flexneri and Salmonella enterica (6). AvrPtoB and NIeG are structural mimics of RING/U-box domain ubiquitin ligases, whereas SopA is a structural mimic of HECT domain ubiquitin ligases (5,7,8). In contrast, IpaH enzymes possess a novel catalytic fold that does not resemble known eukaryotic E3 enzymes (9 -11). The substrates for most pathogen-encoded E3 enzymes remain to be identified but are likely to function in host defense systems. For example, AvrPtoB and IpaH9.8, respectively, target the Fen and NF-B essential modulator/IB kinase ␥ (NEMO/IKK␥) protein kinases, both of which are key mediators of the innate immune response (12,13).
The domain architecture of the IpaH family consists of a N-terminal leucine-rich repeat (LRR) domain and a C-terminal catalytic domain, termed the novel E3 ligase (NEL) domain (9). The NEL domain bears a cysteine residue that mediates transfer of ubiquitin to substrates through the formation of a thioester intermediate (10,11), analogous to the structurally unrelated HECT domain in eukaryotes. The NEL domain is virtually identical (Ͼ98% identity) across all IpaH family members in Shigella and is highly conserved in other bacterial species. In contrast, the LRR domain that recognizes substrates is variable between IpaH family members and across bacterial species, which presumably reflects the diversity of specific target proteins. For instance, the LRR domain of the NEL enzyme SspH1 in Salmonella selectively targets PKN1 (14), whereas the LRR domain of Shigella IpaH9.8 targets NF-B essential modulator/IB kinase ␥ (13).
In addition to recruiting substrates, the LRR domain regulates the catalytic activity of the adjacent NEL domain. Removal of the LRR domain of IpaH9.8 or Salmonella SspH2 increases autoubiquitination and potentiates free polyubiquitin chain synthesis (9,10,11). The autoinhibitory effect of the LRR domain appears to be relieved through substrate interactions. For example, in the presence of the yeast surrogate substrate Ste7, autoubiquitination activity of the NEL domain of IpaH9.8 is enhanced (6). These observations suggest that substrate recognition is coupled to catalytic activation.
The mechanism by which the LRR domain regulates NEL domain catalytic activity has been suggested by recent structural studies (9,10). To date, three crystal structures of NEL enzymes have been solved: the isolated NEL domain of Shigella IpaH1.4 (Protein Data Bank (PDB) number 3CKD) (10) and two near full-length structures encompassing the LRR and NEL domains of Shigella IpaH3 (S. flexneri strain 2a, PDB number 3CVR) (11) and Salmonella SspH2 (PDB number 3G06) (9). Notably, the higher order orientations adopted by the LRR and NEL domains of the IpaH3 and SspH2 enzymes are strikingly different. As the SspH2 structure corresponds to an autoinhibited state (9) and as autoinhibition was not detected for IpaH3 (11), the IpaH3 structure has been inferred to represent an active state (9). However, given that the close homologue IpaH9.8 (77% identity to IpaH3) does exhibit potent autoinhibition (10), we re-examined the autoinhibitory properties of a series of IpaH enzymes. We demonstrate that there are two distinct structural modes of NEL domain autoinhibition and that the inhibited versus non-inhibited states can be interconverted through facile substitution of key residues.
Protein Expression and Purification-Protein constructs were expressed in BL21 (DE3) CodonPlus and induced with 0.3 mM isopropyl-1-thio-␤-D-galactopyranoside at 16°C for overnight. Expressed proteins were purified using nickel-nitrilotriacetic acid affinity chromatography and stored in a buffer containing 20 mM HEPES (pH 7.5), 150 mM NaCl, 5% glycerol, and 1 mM DTT. Purification tags on UBE1, UBE2D3, and ubiquitin were removed by tobacco etch virus protease.
In Vitro Ubiquitination-Reactions were performed in a 20-l reaction mixture containing 25 mM Tris-HCl (pH 7.5), 50 mM NaCl, 10 mM MgCl 2 , 5 mM ATP, 0.25 mM DTT, 1 g of ubiquitin, 1 M UBE1, 5 M UBE2D3 (ϭUbcH5c), and 1 M of indicated IpaH proteins in the presence or absence of 2 M PKN1. Reactions were incubated at 25°C for 60 min and stopped by SDS sample buffer containing 100 mM DTT. Reaction mixtures were separated by SDS-PAGE, transferred onto PVDF membrane, and probed with specific antibodies as indicated. Presented experiments were representative of minimally 3 independent replicates.
Protein Stability in Yeast-Yeast cells were grown at 30°C in medium containing 2% raffinose to a density of 10 7 cells/ml followed by the addition of 2% galactose. After 1 h of galactose induction, cells were washed and resuspended in medium containing 2% glucose and 50 g/ml cycloheximide. Cells were collected, resuspended in 20% trichloroacetic acid, and flashfrozen at the indicated time points before cell lysis. The experiments presented in Fig. 4, A and B, were representative of minimally 2 and 3 independent replicates, respectively.
Yeast Viability Assays-Previous studies established a FUS1 pr -HIS3 yeast reporter strain that is sensitive to IpaH9.8 expression (6). Wild type and the indicated IpaH9.8 FLAG mutant constructs under control of the GAL1 promoter were transformed into a strain bearing a FUS1 pr -HIS3 reporter construct (SY2625 MATa bar1 ade2 his3::FUS1-HIS3 mfa2::FUS1-lacZ leu2 can1 ura3 trp1 (6)) and spotted as a dilution series on glucose-or galactose-containing synthetic medium lacking histidine. For wild type and mutant SspH1 FLAG constructs, the same procedure was used except that constructs were transformed into an S288C wild type yeast strain.

RESULTS
To investigate whether autoinhibition by the LRR domain is a general feature of NEL enzymes, we tested the previously uncharacterized Yersinia Y3400 enzyme for autoinhibition, along with selected Shigella and Salmonella enzymes. IpaH enzymes possess three specific ubiquitin transferase activities that can be monitored in vitro: (i) the polymerization of free ubiquitin chains; (ii) the ubiquitination of a substrate; and (iii) enzyme autoubiquitination. We first assayed the formation of free ubiquitin polymers as a measure of NEL domain enzymatic activity. As expected, deletion of the LRR domain liberated potent catalytic activity of the Shigella IpaH9.8 and Salmonella SspH1 NEL domains (Fig. 1A). Unlike a previous study (11), we also observed enhanced activation of the IpaH3 (S. flexneri strain 5a) catalytic domain. In contrast, we observed no effect of LRR domain deletion on the Yersinia Y3400 enzyme, which appeared constitutively active. Autoinhibition thus appears to be a feature of some but not all IpaH enzymes.
To understand the molecular basis for autoinhibition, we compared the three known IpaH structures. Consistent with the IpaH3 and SspH2 structures reflecting higher order offstates, their catalytic loop orientations differ markedly from that observed in the on-state structure of the isolated catalytic domain (Fig. 1B). However, the conformation of the catalytic loop also differed between the two off-state structures of fulllength enzymes. To investigate whether these differences reflect diversity in autoinhibitory mechanisms across the IpaH family, we set out to design mutants to more precisely dissect the relevance of both the previously identified autoinhibitory mode revealed by the SspH2 structure (denoted mode 1) and the second putative autoinhibitory mode revealed by the IpaH3 structure (denoted mode 2).
The interaction of the LRR and NEL domains in the SspH2 structure (mode 1) is bridged by the LRR-NEL linker; Ile-479, Phe-481, and Met-483 in the linker make hydrophobic contacts with Leu-465 in ␣4 of the LRR domain and with NEL domain residues Leu-638 in the ␣G-␣H linker and Phe-667 and Val-670 in the ␣H-␣I linker (Fig. 1C). Additionally, Asp-482 in the linker hydrogen bonds with Ser-579 in the catalytic loop (␣E-␣F) of the NEL domain. In contrast, the LRR-NEL interaction in the IpaH3 structure (mode 2) is mediated by direct contacts between the two domains. Notably, Phe-421 in the ␣G-␣H linker of the NEL domain interacts with Ile-222 at the center of a hydrophobic pocket on the LRR domain (Fig. 1D). Hydrogen bonds are also observed between main-chain amino groups of Phe-421 and His-420 in the ␣G-␣H linker with Glu-224 in ␣3, between Ser-454 in the ␣H-␣I linker with His-221 in the LRR-␣3 linker, and between Asp-321 in ␣C with Ser-244 in ␣4, within the NEL and LRR domains, respectively. Although the mode 1 versus mode 2 structures appear quite dissimilar, a common feature is the involvement of the NEL domain residue at the apex of the ␣G-␣H linker (Leu-638 in SspH2 and Phe-421 in IpaH3). With this in mind, we designed a set of mutants to selectively test the functional relevance and generality of the two intramolecular binding modes for autoinhibition.
We generated mutations in two representative IpaH homologues, Shigella IpaH9. 8 and Salmonella SspH1, and tested for effects on polyubiquitin chain synthesis in vitro. Consistent with previous findings for Salmonella SspH2 (9), an I391D substitution (corresponding to Ile-479 in SspH2) or the double F393A/M395A substitution (corresponding to Phe-481/Met-483 in SspH2) circumvented the mode 1 autoinhibition in SspH1 ( Fig. 2A). Introduction of the analogous substitutions in the more divergent Shigella IpaH9.8 enzyme also abrogated autoinhibition. Thus, the autoinhibitory mechanism represented in the SspH2 structure (mode 1) appears to operate in other IpaH family enzymes.
We next probed the predicted autoinhibitory mode 2 contacts in the two representative IpaH homologues, IpaH9.8 and SspH1. Single site substitutions of either of the domain contact residues, F395R within the NEL domain and the reciprocal substitution I196D in the LRR domain of Shigella IpaH9.8, effec- tively abrogated autoinhibition (Fig. 2B). This result suggested that the mode 2 conformation also reflects an autoinhibited state, rather than an active state as inferred previously (11). Surprisingly, introduction of the analogous mode 2 mutations in Salmonella SspH1 also relieved autoinhibition. The Phe residue at the apex of the ␣G-␣H linker of the IpaH3 structure is occupied by a conserved Leu residue in SspH1. A conservative L550F mutation had no effect on autoinhibition of SspH1, whereas a non-conservative L550S mutation overcame autoinhibition (Fig. 2B). Thus, autoinhibitory mode 2 revealed by the IpaH3 structure also applies to other IpaH enzymes.
To verify that mutations disrupting LRR-NEL interfaces function through relieving the LRR-mediated autoinhibition, the L550S and the F395R substitutions were introduced to the isolated SspH1 and IpaH9.8 NEL domain, respectively. Neither of the substitutions had any effect on basal activity of the isolated SspH1 or IpaH9.8 NEL domain, thereby demonstrating that the effect was due entirely to relief of autoinhibition (Fig.  3A).
The coupling of substrate recognition to catalytic activation of E3 ubiquitin ligases is thought to prevent promiscuous ubiquitination of unintended substrates and/or enzyme degrada-tion due to autoubiquitination (15). To determine whether this regulatory principle operates in IpaH enzymes, we examined SspH1 and IpaH9.8 interface mutants in autoubiquitination assays. Mutations that activated polyubiquitin chain synthesis also promoted autoubiquitination of both SspH1 and IpaH9.8 enzymes (Fig. 3B). Disruption of the LRR-NEL interface did not affect the ability of SspH1 to ubiquitinate its endogenous human substrate PKN1 (Fig. 3C). Therefore, autoinhibition of IpaH enzymes may serve in part to restrict catalytic activity toward specific physiological targets while preventing ubiquitination of the enzymes themselves.
We previously established a FUS1 pr -HIS3 strain of the budding yeast Saccharomyces cerevisiae as a surrogate biological model for IpaH9.8 enzyme function (6). To examine the relevance of autoinhibition for IpaH function in vivo, we conditionally expressed IpaH9.8 from the GAL1 promoter in this established FUS1 pr -HIS3 yeast strain (Fig. 4A). As reported previously (6), IpaH9.8 inhibited the mating pheromone pathway by eliminating the Ste7 MAPK kinase, thereby preventing expression of a FUS1 pr -HIS3 reporter construct and causing the consequent inability to grow on ϪHis medium. Interestingly, SspH1 caused a severe growth defect in a wild type yeast strain background (Fig. 4B), presumably by targeting one or more unidentified essential proteins for degradation. Mutations that disrupt either of the modes of autoinhibition in vitro abrogated toxicity of both IpaH enzymes in the respective surrogate genetic assays. Consistent with the possibility that autoinhibition prevents unintended autoubiquitination and destruction of the IpaH enzyme itself, the in vivo half-lives of constitutively active IpaH enzyme mutants were somewhat decreased ( Fig. 4C and supplemental Fig. S3). Autoinhibition by either mode 1 or mode 2 can therefore regulate IpaH enzyme function in vivo.
We then addressed the mechanism by which NEL enzymatic activity is perturbed by autoinhibition. The binding of IpaH9.8 to ubiquitin-charged E2 enzyme facilitates discharge of the ubiquitin thioester in a manner that depends on the NEL domain active site cysteine (6). Full-length SspH1 and the constitutively active isolated NEL domain, but not the C492A catalytic site mutant, were likewise able to stimulate discharge of the E2-ubiquitin thioester (Fig. 5). As SspH1-mediated discharge of ubiquitin from the E2 depended on the NEL domain catalytic cysteine residue, we infer that autoinhibition does not interfere with ubiquitin thioester formation on the NEL domain. Autoinhibition must instead perturb the transfer of ubiquitin from the NEL domain to an acceptor lysine residue, either on the E3 enzyme itself or on free ubiquitin chains.
If the autoregulatory features of the IpaH9.8 and SspH1 enzymes demonstrated above are generally applicable, it should be possible to explain the regulatory features of other IpaH enzymes in structural terms and to engineer artificial autoinhibitory mechanisms into enzymes that lack autoinhibition, such as the Y3400 enzyme of Y. pestis. In Y3400, the residue corresponding to the critical SspH1 Leu-550 is polar in nature and analogous to the activating L550S substitution in SspH1 (supplemental Fig. S1), thereby explaining the absence of autoinhibition in the Yersinia enzyme. However, Y3400 does possess a hydrophobic pocket in the LRR domain and the bridging Ile residue in the LRR-NEL linker, both of which are compo-  nents of the autoinhibitory mechanism in other enzymes. We retro-engineered an artificial autoinhibitory interaction between the LRR and NEL domains of Y3400 by introducing a S449F mutation in the apex of the ␣G-␣H linker. This mutation did not adversely affect the catalytic activity of the isolated Y3400 NEL domain but substantially reduced activity of the full-length Y3400 enzyme (Fig. 6A). Importantly, the engineered autoinhibition of Y3400 S449F was abolished by mutations that disrupt either of the two modes of autoinhibitory interactions in SspH2 and IpaH3 (Fig. 6, B and C). These results demonstrate that the LRR-NEL domain architecture of Y3400 retains the capacity for autoinhibition through mode 2 hydro-phobic pocket interactions, as in IpaH3, and/or through mode 1 bridging linker interactions, as in SspH2.

DISCUSSION
Eukaryotic ubiquitin ligases are regulated through diverse mechanisms. For example, the catalytic efficiency of the cullin-RING ligase enzymes is activated by conjugation of the ubiquitin-like modifier Nedd8, interactions with inhibitory factors, and dimerization (16). HECT domain E3 enzymes are similarly subject to multiple levels of control, including autoinhibition through higher order intramolecular interactions of the HECT domain with flanking interaction domains (17). For a number of E3 enzymes, autoubiquitination mediates proteasome-dependent degradation of the E3 itself, which intrinsically selflimits enzymatic activity (15). The overall weak effect on protein stability by disrupting LRR-NEL interaction in yeast models (supplemental Fig. S3) suggests that autoregulation may influence biological functions of IpaH enzymes through additional mechanisms.
Our results demonstrate that the two outwardly dissimilar higher order structures displayed by the IpaH3 and SspH2 enzymes reflect two different modes of autoinhibition. These modes differ in the relative positioning of the LRR domain with respect to the NEL domain. In mode 1, the concave surface of the LRR domain orients toward the NEL domain, whereas in mode 2, the concave surface is oriented opposite to the NEL domain. Despite this overt structural difference, in both modes, the LRR-NEL interaction impinges on the active site region of the NEL domain to down-regulate catalytic function. Based on our mutational analysis, both inhibitory modes appear to operate in the same enzyme, suggesting that the two autoinhibited states exist in equilibrium. The observation that mutations that disrupt mode 1 interactions are more potent activators than mutations that disrupt mode 2 interactions (supplemental Fig.  S2) suggests that mode 1 might be the predominant conformation that is sampled in solution. These distinct autoregulatory orientations may also allow plasticity in how the LRR domain engages substrates and/or how substrate interactions alleviate autoinhibition. Future structural work of reaction intermediates will be required to understand precisely how the LRR domain inhibits the catalytic transfer of ubiquitin from the NEL domain to substrates and how the autoinhibition of IpaH enzymes influences pathogen virulence.
Autoinhibition is a conserved structural and regulatory feature of IpaH enzymes, although it is latent in some enzymes, such as Yersinia Y3400. As noted above, autoinhibition may prevent the ubiquitination of unintended protein substrates and/or preclude autoubiquitination that might otherwise lead to destruction or inactivation of the enzyme itself. Recent studies suggest another advantage of autoinhibition, namely to allow pathogenic bacteria to escape the host surveillance system that detects untethered ubiquitin chains (18). These products of pathogen-encoded E3 enzymes can activate the TAK1 and IB kinase (IKK) kinase complexes, which in turn activate NF-B and the innate immune response (18). The elucidation of full-length LRR-NEL enzyme structures bound to different substrates should reveal how substrate engagement alleviates catalytic constraints of the autoinhibited structure. In turn, these interdomain interactions might be targeted to mitigate pathogen virulence.