The ubiquitin ligase SspH1 from Salmonella uses a modular and dynamic E3 domain to catalyze substrate ubiquitylation

SspH/IpaH bacterial effector E3 ubiquitin (Ub) ligases, unrelated in sequence or structure to eukaryotic E3s, are utilized by a wide variety of Gram-negative bacteria during pathogenesis. These E3s function in a eukaryotic environment, utilize host cell E2 ubiquitin-conjugating enzymes of the Ube2D family, and target host proteins for ubiquitylation. Despite several crystal structures, details of Ube2D∼Ub binding and the mechanism of ubiquitin transfer are poorly understood. Here, we show that the catalytic E3 ligase domain of SspH1 can be divided into two subdomains: an N-terminal subdomain that harbors the active-site cysteine and a C-terminal subdomain containing the Ube2D∼Ub–binding site. SspH1 mutations designed to restrict subdomain motions show rapid formation of an E3∼Ub intermediate, but impaired Ub transfer to substrate. NMR experiments using paramagnetic spin labels reveal how SspH1 binds Ube2D∼Ub and targets the E2∼Ub active site. Unexpectedly, hydrogen/deuterium exchange MS shows that the E2∼Ub–binding region is dynamic but stabilized in the E3∼Ub intermediate. Our results support a model in which both subunits of an Ube2D∼Ub clamp onto a dynamic region of SspH1, promoting an E3 conformation poised for transthiolation. A conformational change is then required for Ub transfer from E3∼Ub to substrate.

SspH/IpaH bacterial effector E3 ubiquitin (Ub) ligases, unrelated in sequence or structure to eukaryotic E3s, are utilized by a wide variety of Gram-negative bacteria during pathogenesis. These E3s function in a eukaryotic environment, utilize host cell E2 ubiquitin-conjugating enzymes of the Ube2D family, and target host proteins for ubiquitylation. Despite several crystal structures, details of Ube2DϳUb binding and the mechanism of ubiquitin transfer are poorly understood. Here, we show that the catalytic E3 ligase domain of SspH1 can be divided into two subdomains: an N-terminal subdomain that harbors the activesite cysteine and a C-terminal subdomain containing the Ube2DϳUb-binding site. SspH1 mutations designed to restrict subdomain motions show rapid formation of an E3ϳUb intermediate, but impaired Ub transfer to substrate. NMR experiments using paramagnetic spin labels reveal how SspH1 binds Ube2DϳUb and targets the E2ϳUb active site. Unexpectedly, hydrogen/deuterium exchange MS shows that the E2ϳUbbinding region is dynamic but stabilized in the E3ϳUb intermediate. Our results support a model in which both subunits of an Ube2DϳUb clamp onto a dynamic region of SspH1, promoting an E3 conformation poised for transthiolation. A conformational change is then required for Ub transfer from E3ϳUb to substrate.
To promote infectivity and survival within a eukaryotic cell, numerous species of pathogenic bacteria use macromolecular secretion systems to translocate an array of effector proteins directly into the host cell cytoplasm. These effectors perform various functions, including alteration of the cytoskeleton, alteration of vesicle maturation, transport, and fusion, and can target pathways that regulate host innate immune responses (1). Ubiquitin (Ub) 3 signaling pathways, found exclusively in eukaryotes, are often common targets for bacterial effector proteins (2,3). Protein ubiquitylation is used to regulate numerous aspects of protein function, including activity, localization, macromolecular interactions, and lifetime (4). To exploit this regulatory system, one strategy used by both viral and bacterial pathogens is to introduce effectors that target specific host proteins for ubiquitylation (5,6).
In eukaryotes, protein ubiquitylation requires the successive action of three types of enzymes: an E1 Ub-activating enzyme, an E2 Ub-conjugating enzyme, and an E3 Ub ligase (7)(8)(9). E3s are typically divided into two mechanistic classes. RING/U-box type E3s serve an activation and scaffolding role to facilitate the direct transfer of Ub from the E2ϳUb conjugate to the target (9). In contrast, HECT-type or RBR-type E3s catalyze Ub transfer in two stages. In stage I, the E3 binds an activated E2ϳUb, and Ub is transferred via a trans-thiolation reaction from the E2 to E3, forming an obligate E3ϳUb thioester intermediate. In stage II, Ub is transferred from an E3ϳUb conjugate usually to a lysine residue of a target protein or another Ub subunit, forming a covalent isopeptide bond (7)(8)(9).
Another general feature shared by HECT-type E3s, both eukaryotic and prokaryotic, is their modular construction. A key question arises as to how individual domains work in con- cro ARTICLE cert to facilitate Ub transfer. As part of our effort to understand the molecular basis of Ub transfer, we focused on the bacterial E3 SspH1 from Salmonella. SspH1 targets a eukaryotic kinase, PKN1, involved in the regulation of a number of pathways, including NF-B and androgen receptor signaling, both of which are involved in the cellular response to bacterial invasion (17)(18)(19). First identified as interacting with SspH1 in two-hybrid screens (18), PKN1 has been confirmed as a substrate through in vitro ubiquitylation and in vivo infection assays (10,(17)(18)(19). The minimal construct of PKN1 sufficient for binding and ubiquitylation is homology region b (HRb) (Fig. S1C), which functions in the regulation of PKN1 activity. A structure of the HRb region of PKN1 bound to the SspH1 LRR domain has been determined (19). Although the structure of the SspH1 E3 domain has not yet been solved, those of several E3s, including the closely related SspH2 from Salmonella (77.6% identical in the E3 domain), are available. In addition, several studies have shown that SspH/IpaH E3s use E2ϳUbs from the Ube2D (UbcH5) family as a source of activated Ub (12)(13)(14)(15)(16)20). Thus, SspH1 is a structurally tractable system for investigating protein interactions and domain motions that occur during substrate ubiquitylation.
Here, we show that mobility of structural elements within the SspH1 E3 domain is critical for Ub transfer. The SspH1 E3 domain can be divided into separate subdomains, an N-terminal subdomain (NSD) harboring the catalytic cysteine, and a C-terminal subdomain (CSD) responsible for binding Ube2DϳUb. Hydrogen/deuterium exchange MS (HDX-MS) reveals that, in contrast to eukaryotic ubiquitin E3s, the SspH1 Ube2DϳUb-binding site is dynamic in the absence of an E2ϳUb conjugate. Unexpectedly, the E2ϳUb-binding site is stabilized in the E3ϳUb intermediate. Site-specific incorporation of paramagnetic spin labels reveals where and how Ube2DϳUb binds SspH1 and how the SspH1 catalytic residue approaches the E2ϳUb active site. The data presented support a mechanistic model in which both subunits of Ube2DϳUb grasp onto a binding site in the CSD, positioning the E2ϳUb for reaction with the E3. In stage II of the transfer reaction, the NSD, which now holds the activated Ub, must move relative to the CSD to deliver Ub to the target substrate.

SspH1 LRR-E3 is monomeric
A previous study suggested that SspH/IpaH constructs containing both LRR and E3 domains form dimers in solution (15). Such high-molecular weight complexes could negatively impact NMR-based experiments and complicate data analysis. To determine the quaternary structure of SspH1 in solution, we used analytical size-exclusion chromatography (SEC) to examine the behavior of SspH1 LRR-E3 (residues 162-700, 61.8 kDa) in the presence and absence of the PKN1 HRb motif (residues 108 -201). The constructs and mutants described in this work are cataloged in Table S1.
On SEC, both SspH1 and the SspH1/PKN1 complex elute as single peaks at volumes consistent with monomeric SspH1 or a 1:1 SspH1/PKN1 complex (Fig. S2A). The isolated SspH1 E3 domain, which does not bind PKN1, also behaves as a mono-meric species on SEC. Analysis of the SspH1 LRR-E3 elution peak by multiangle light scattering yielded a molecular mass of ϳ63 kDa, in good agreement with the calculated molecular mass of this construct (Fig. S2B). Thus, under our experimental conditions, in both the absence and presence of substrate, purified SspH1 E3 and LRR-E3 constructs behave as monomeric proteins.

SspH/IpaH E3 domains can be divided into two distinct subdomains
As originally described by Singer et al. (13), the structure of SspH/IpaH E3 catalytic domains consists of three regions: an N-terminal lobe, a middle region, and a C-terminal lobe or "thumb" segment that protrudes away from the E3 domain (Fig.  1B). Subsequent structures all exhibit the same overall ␣-helical fold (12)(13)(14)21). Despite this similarity, close inspection reveals conformational heterogeneity within the E3 domain. Superposition of structurally conserved C␣ backbone atoms from IpaH3 and SspH2 E3 domains yields an RMSD of ϳ3.8 Å (Fig.  S3A). However, division of each E3 domain into an NSD (helices 1-7 based on SspH2) and a CSD (helices 8 -14) reveals that these individual regions have much higher structural similarity and superpose with backbone RMSDs of less than ϳ1 Å ( Fig. 1C and Fig. S3B). Likewise, the corresponding subdomains of SlrP, IpaH1.4, IpaH4, and IpaH9.8 superpose with similar backbone RMSDs. In addition, some level of structural variation is present in the CSD thumb region (Fig. S3, A and D).
Alignment of the IpaH3 and SspH2 CSDs reveals an ϳ20°r otation of the relative orientation of NSD helix 7 and, therefore, the entire NSD (Fig. 1D). The catalytic cysteine is located near the N-terminal end of helix 7, and this change in orientation alters the relative position of the catalytic residue by more than 21 Å (C␣-C␣ distance). Intriguingly, comparison of IpaH3 and SspH2 with the available crystal structures of IpaH1.4, IpaH4, and SlrP shows that each of these E3 domains adopts one or the other of these two conformational states (Fig.  S3C).
A short loop segment (Fig. 1B, magenta loop) connects E3 helices 7 and 8 and links the NSD (cyan) to the CSD (gray). By introducing a stop codon into the loop segment, we were able to express and purify the NSD, which is well-behaved and highly soluble. In generating a CSD construct, we found that inclusion of some helix 7 residues was needed to produce soluble protein. This suggests that at least a portion of helix 7 interacts with the CSD and these contacts aid in the stability and solubility of this subdomain. Purified NSD (residues 403-516) and CSD (residues 489 -700) constructs yield dispersed 1 H, 15 N TROSY NMR spectra (Fig. 2, A and B). For each construct, the peak intensities are uniform, and most of the expected number resonances are observed (ϳ100 of the predicted 120 peaks for the NSD; ϳ200 of the predicted 210 peaks for the CSD). By comparison, only ϳ150 of the predicted 320 resonances are clearly defined in the spectrum of the full SspH1 E3 domain, and variations in peak intensities are evident (Fig. 2C). Due to the size and ␣-helical nature of the E3 domain, significant overlap is observed in the spectrum (Fig. 2D). Nonetheless, dispersed resonances for the E3 domain overlay well with NSD and CSD resonances (Fig.  2D), indicating that the isolated subdomains adopt the same

SspH1 uses a modular and dynamic E3 domain for Ub transfer
overall tertiary structure as in the full E3 domain. As we have no evidence for nonspecific association or dimerization of the E3 domain (see above), we propose that the variable intensities observed in the spectrum of the full E3 domain arise from conformational exchange involving relative motions of the NSD and CSD, presumably between states observed in available crystal structures.

Identification of an E2ϳUb-binding site in the CSD of the E3 domain
Although neither subdomain exhibits E3 ligase activity on its own, the ability to purify separate E3 subdomains provided an opportunity to explore important E3 functional roles, particularly the interaction with activated E2ϳUb. An active-site Ube2D3 C85S mutation was used to generate a stable oxyesterlinked E2-Ub conjugate (Ube2D3-O-Ub) for NMR experiments. A second Ube2D3 mutation, S22R, was introduced into the ␤-sheet of the E2 to prevent additional noncovalent interactions of Ub with the E2 that could complicate analysis and interpretation of NMR spectra (22). S22R Ube2D3 still functions with SspH1 to ubiquitylate PKN1 (Fig. S4A). As reported for SspH2 (20), no significant chemical shift perturbations are observed in the 1 H, 15 N TROSY spectrum of free Ube2D3 or Ub upon the addition of SspH1 E3 domain (Fig. S4B). However, like SspH2, selective peak broadening and intensity loss is observed when SspH1 is added to 15 N-labeled Ube2D3-O-Ub conjugate ( Fig. 3A and Fig. S5A). These results show that both subunits of the E2ϳUb conjugate engage in binding. Furthermore, both SspH1 and SspH2 recognize a surface of Ube2D that differs from the canonical surface shown to bind eukaryotic E3s (20). These results are consistent with the observed selectivity of SspH/IpaH E3s for Ube2D family E2s.
Despite harboring the active site cysteine, the addition of the NSD had no effect on the Ube2D3-O-Ub spectrum. In contrast, the addition of the CSD produces significant spectral changes ( Fig. 3 (B and C) and Fig. S5 (B and C)). Mapping the most perturbed resonances onto a surface representation of Ube2D3ϳUb reveals that both the full E3 domain and the CSD affect the same Ube2D-Ub residues (Fig. 3D). Thus, our NMR data suggest that Ube2DϳUb binding is primarily restricted to the CSD. Keszei and Sicheri (16) demonstrated that mutation of residues in the thumb region of the IpaH9.8 E3 domain, far removed from the catalytic site, reduced binding of free Ube2D2 and decreased the ability of IpaH9.8 to form an E3ϳUb Figure 1. Domain architecture and topology of SspH1. A, a type 3 secretion system (T3SS) targeting sequence is present at the N terminus of SspH1, followed by the substrate-binding LRR domain and catalytic ubiquitin ligase E3 domain. The positions of the catalytic cysteine, Cys-492, the NSD (cyan), and the CSD (gray) are depicted. B, the topology of SspH/IpaH E3 domains, based on the structure of SspH2 (PDB code 3G06), can be divided into an NSD (cyan) and a CSD (gray), which are connected by a short linker (magenta). C, the NSDs and CSDs from E3 domains of SspH2 (gray) and IpaH3 (red; PDB code 3CVR) were separately superposed. The individual subdomains align with backbone RMSDs of structurally conserved residues of Ͻ1 Å. D, alignment of the SspH2 and IpaH3 CSDs reveals a 20°rotation in the position of NSD helix 7 relative to the CSD in the two structures.

SspH1 uses a modular and dynamic E3 domain for Ub transfer
intermediate. We used mutational analysis to further define the E2ϳUb-binding site. Single point mutations within the SspH1 thumb (L655A, A661Q, and G666A) each abrogate the ability of the SspH1 E3 domain to synthesize unanchored polyubiquitin chains (Fig. 4, A and B). Surprisingly, even substitution of Ala for Gly-666, a residue with Ͼ98% conservation within SspH/ IpaH E3s, significantly reduced the ability of the SspH1 E3 domain to bind Ube2D3-O-Ub (Fig. 4C). The homologous mutation in SspH2, G754A, also abolished E3 ligase activity (Fig. S4C). Overall, mutagenesis results show that the integrity of the tip of the thumb region is critical for binding Ube2D3ϳUb and E3 activity.

Orientation of bound Ube2D3ϳUb
To further characterize binding of Ube2D3ϳUb to SspH1, we incorporated TEMPO paramagnetic spin labels at specific locations in the SspH1 E3 domain (Fig. 5A). The unpaired paramagnetic electron in TEMPO selectively broadens the NMR resonances of residues in proximity to the spin label. In weakly interacting systems, the spin label effect can be observed in residues up to 15 Å away (23). TEMPO-SspH1 was added to 15 N-labeled Ube2D3-O-Ub, and 1 H, 15 N TROSY spectra were collected before (SL-active) and after (SL-quenched) the addition of ascorbate, which eliminates the paramagnetic effect (Fig. 5B). Resonance perturbations specifically induced by TEMPO were quantified by analyzing peak intensities in the SL-active and SL-quenched spectra (Fig. S6), and the corresponding residues were mapped onto a surface representation of the Ube2D3ϳUb conjugate (Fig. 5C). It should be noted that the intensities of Ube2D3-O-Ub resonances located at the Ube2D3ϳUb/SspH1 binding interface (Fig. 3A) are significantly reduced due to exchange broadening. Therefore, weak resonances that fall below one S.D. from the average intensity and resonances that could not be resolved were omitted from analysis (Fig. S6).
TEMPO spin labels were incorporated at three sites in the SspH1 thumb region. The E3 domain has only a single cysteine residue, the active site residue Cys-492. In combination with a C492A mutation, Glu-639, Ser-653, or Ser-672 was individually changed to cysteine to allow for single-site labeling (Fig. 5A). Consistent with mutagenesis results, SL653, located at the tip of the thumb, effectively abrogated binding of Ube2D3-O-Ub, presumably due to steric interference with the E2ϳUb conjugate (Fig. S7A). In contrast, variants in which a bulky spin label was incorporated at positions SL639 and SL672, adjacent to the

SspH1 uses a modular and dynamic E3 domain for Ub transfer
(B and C)). The results are mapped onto the surface of the Ube2D3ϳUb conjugate in green and red, respectively (Fig. 5C).
The pattern of Ube2D3-O-Ub chemical shift perturbations observed for binding to reduced SL639 and SL672 (SLquenched) is similar to that observed for the WT SspH1 E3 domain (Fig. S7B). These data indicate that Ube2D3-O-Ub interacts with SL639 and SL672 in much the same way as with WT SspH1. To verify that SL639 and SL672 form productive complexes with Ube2D3ϳUb, we developed a Ub transfer assay that is not dependent on Cys-492 at the SspH1 active site, which must be mutated to allow for selective spin labeling at other positions. We found that interaction of Ube2D3ϳUb with the SspH1 E3 domain enhances the intrinsic lysine reactivity of the E2ϳUb conjugate relative to Ube2D3ϳUb alone (Fig. S7C). By substituting lysine at the SspH1 active site, C492K, SspH1 can transfer Ub to the active site position to generate an E3ϳUb mimic in which Ub is tethered to the E3 via a covalent isopeptide bond. Analysis by reducing SDS-PAGE shows that C492K SspH1 E3 rapidly generates a stable E3-Ub species, whereas SspH1 C492A does not (Fig. S7D). Spin labels selectively introduced at positions 639 and 672 in a C492K background also readily transfer Ub to the E3 active site (Fig. 5D). Therefore, stage I of the Ub transfer reaction, from E2ϳUb to the E3 active-site residue, is still functional in the presence of these SspH1 modifications. The finding that spin label modifications adjacent to the tip of the thumb do not disrupt E2ϳUb binding, whereas mutations within this segment eliminate interaction and/or activity, strongly argues that E2ϳUb binding is localized to the top segment of the thumb region.
As the E3 and E2ϳUb active sites must interact during catalysis, a TEMPO spin label was also introduced at SspH1 Cys-492 (SL492). As shown in Fig. 5, SL492 has a highly localized effect on Ube2D3-O-Ub resonances in the E3/E2-Ub complex. Ube2D3 residues 81, 83, 89, 90, 92, 117, and 120 -125 are affected (Fig. 5B (cyan) and Fig. S6). These residues surround an entrance to the E2ϳUb active site that is opposite the E2 surface used to bind to the CSD (compare Fig. 3D with Fig. 5C) and reveal the direction from which the E3 catalytic cysteine in the NSD approaches the E2ϳUb active site (Fig. 5C, cyan surface). Together, the NMR results paint a consistent picture in which Ube2DϳUb binding is primarily a function of the SspH1 thumb region of the CSD. The E2 and Ub grasp opposite sides of the thumb positioning the E2 active site for attack by the E3 catalytic residue (Fig. 5E).

The SspH1 E2ϳUb-binding site is dynamic
Available SspH/IpaH E3 structures exhibit some conformational variation in the E2ϳUb-binding site defined above (Fig.  S3D). SspH2 contains two well-ordered anti-parallel helices. Although the IpaH1.4, IpaH4, IpaH9.8, and SlrP structures are also helical, there is a pronounced kink in the penultimate helix of the thumb region. Intriguingly, the electron density for this segment is absent in the structure of IpaH3 (Fig. S3). Is the observed diversity reflective of different E2ϳUb-binding modes among these E3s? Or is the observed heterogeneity indicative of dynamics in this region? To characterize the behavior of the SspH1 E3 domain, we used HDX-MS, a technique that provides information about local protein conformation and dynamics (24).
HDX reactions were carried out for varying amounts of time by incubating protein in buffered D 2 O followed by quenching in pH 2.5 ice-cold buffer. Quenched solutions were processed using standard protocols (see "Experimental procedures"), and masses of peptides derived from the E3 domain were obtained by MS and confirmed by MS-MS fragmentation. Fig. 6A plots SspH1 E3 domain peptide fractional deuteration centered on the middle residue of each peptide analyzed at different time points. The plot focuses on early time points as these provided the most insight into the dynamics and changes observed upon

SspH1 uses a modular and dynamic E3 domain for Ub transfer
Ub conjugation. When the extent of exchange at the earliest time point is depicted on an SspH1 homology model, regions that undergo rapid exchange are revealed: residues 394 -441 in the N terminus of the E3 domain, residues 571-591 located between CSD helices 9 and 11, and nearly complete exchange for residues 651-668 in the Ube2DϳUb-binding site (Fig. 6C  and Fig. S8).
These results are compared with the same analysis performed using the stable E3-Ub mimic generated from SspH1 C492K (Fig. 6, B and D). HDX-MS of SspH1-Ub shows that most of the E3 domain behaves similarly to free SspH1. However, modest changes in stability, indicated by reduction in fractional deuteration of identical peptides at identical times, are observed for residues 507-532, which encompasses the linker between the NSD and CSD, and CSD residues 551-556, located at the NSD/CSD interface. Unexpectedly, the largest increase in stability is found in the SspH1 thumb (residues 628 -668), even though no E2ϳUb was present in the sample.
The data are consistent with a structural model in which the SspH1 Ube2DϳUb-binding site lacks highly ordered second-ary structure, allowing facile exchange of backbone amide protons. This behavior makes it difficult to directly model the SspH1/Ube2DϳUb complex, as it is not clear what conformation is poised to bind E2ϳUb and how that is oriented relative to the rest of the E3 domain. The dynamic behavior of the SspH1 E2ϳUb-binding region stands in stark contrast to the well-structured E2-binding regions that have been characterized for eukaryotic RING, RBR, and HECT E3s (25)(26)(27)(28)(29). Formation of an obligate E3ϳUb intermediate leads to stabilization of a region of SspH1 that links the NSD and CSD and, in particular, a large segment of the SspH1 thumb region.

Mutations designed to block NSD/CSD conformational changes inhibit Ub transfer
The different E3 domain conformations observed in SspH/ IpaH crystal structures coupled with our finding that the SspH1 E3 domain can be partitioned into separate N-terminal and C-terminal subdomains suggests that movements of the two domains may be required for coordinated Ub transfer. ConSurf analysis (30, 31) of the SspH/IpaH family revealed a highly con-

SspH1 uses a modular and dynamic E3 domain for Ub transfer
served glycine residue (SspH1 Gly-515) in the short NSD-CSD linker far from the E3 active site and E2ϳUb-binding region (Fig. 7A). Substitution of Gly-515 with a conformationally constrained proline (G515P) does not significantly alter binding of Ube2DϳUb to the CSD thumb region (Fig. S9D) but has a large effect on the ability of SspH1 to ubiquitylate PKN1 HRab (Fig.  7B). Although not as efficient as WT SspH1, G515P catalyzes stage I of the reaction, as shown by the breakdown of purified Ube2D3ϳUb to free E2 and Ub (Fig. 7C). This reaction depends on Cys-492 and, therefore, formation of an E3ϳUb intermediate ( Fig. 7C and Fig. S9A). Thus, Cys-492 in the NSD reacts with CSD-bound E2ϳUb conjugate, but stage II of the Ub transfer reaction to substrate is impaired. This suggests that flexibility in the NSD-CSD linker is important for catalysis.
To further define the role for interdomain movement, we compared residues in the NSD/CSD interface in existing structures to identify residues likely to be solvent-exposed in one conformation but buried in the other. The goal was to bias SspH1 toward one of the two E3 conformations. Two SspH1 homology models based on the structures of IpaH3 and SspH2 were generated (Fig. S9B). The IpaH3-based model yields an SspH1 conformation in which the Cys-492 is closer to the Ube2D3ϳUb-binding site in the CSD. We refer to this state as the "proximal" conformation (Fig. S3C). In the SspH2-based "distal" conformation, the active-site cysteine is further from the E2ϳUb-binding site. Comparison of these models identified His-498 in helix 7 of the NSD, 10 Å from Cys-492, as a residue predicted to be in different environments in the two E3 conformations, solvent-exposed in the proximal model and buried in the distal model. His-498 is not highly conserved within the E3 family, as Leu, His, and Phe residues are all found at this site (Fig. 7A). Therefore, any observed differences in activity of His-498 mutants are unlikely to be caused by mutation of a catalytic residue.
Lysine was substituted for His-498 (H498K) to disfavor the buried (distal) conformation and tested for its ability to ubiquitylate PKN1-HRab. As predicted from our model, H498K greatly inhibits PKN1 ubiquitylation. Under the assay conditions, the WT-catalyzed reaction was complete after ϳ5 min with depletion of nearly all of the free Ub in the reaction mixture (Fig. 7B). In contrast, the H498K reaction shows synthesis of much shorter poly-Ub chains, slower formation of ubiquitylated PKN1, and slower depletion of free Ub (Fig. 7B). A variety of other mutations at this position were also tested to confirm that residues more similar to the WT show increasing ability to ubiquitylate substrate (Fig. S9C).
To determine what step(s) in the Ub transfer reaction are affected by H498K, we analyzed reaction mixtures by nonreducing SDS-PAGE to enable visualization of labile E2ϳUb and E3ϳUb species. After ϳ2 min, an E3ϳUb intermediate is observed for both the WT-and H498K-catalyzed reactions, although the steady-state level is much higher for the H498K mutant and persists much longer in the reaction mixture (Fig.  7D). Thus, H498K can efficiently carry out transfer of Ub from E2ϳUb to the E3 but is disabled in Ub transfer to substrate. NMR experiments confirm that H498K does not disrupt Ube2D3-Ub binding (Fig. S9D). Thus, the H498K mutation, designed to bias the E3 domain toward a proximal conformation, created an E3 that readily catalyzes stage I of the Ub transfer reaction to form an E3ϳUb intermediate but is impaired in its ability to catalyze stage II, where Ub is transferred from E3ϳUb to substrate.

Discussion
In eukaryotic cells, bacterial SspH/IpaH E3s compete with host E3s and other proteins for E2ϳUb conjugates to generate specific Ub signals advantageous to the invading pathogen. To fully understand bacterial E3 function, it is necessary to determine how they recognize and bind host proteins, how they assemble multiple proteins into a competent E3 Ub ligase com-

SspH1 uses a modular and dynamic E3 domain for Ub transfer
plex, and the mechanism of Ub transfer. As described herein, whereas SspH/IpaH E3s share some characteristics with their eukaryotic counterparts, they have distinct mechanisms for interacting with the host ubiquitination machinery that presumably confer a competitive advantage.
Much of the recent literature on SspH/IpaH E3s has focused on the large reorientation of the LRR and E3 domains observed in various crystal structures (16,19,21,32). Our further analysis of the same structures revealed that the E3 domains themselves can also adopt multiple conformations. Two types of conformational variation are observed among available structures: 1) local variation in the thumb region, especially near its tip (Fig.  S3) and 2) different relative orientations of the two E3 subdomains. The former may imply local flexibility, and the latter may imply en bloc subdomain motions.
Here, we show that the Ube2DϳUb-binding site is localized to the tip of the thumb region and that this region is indeed flexible in solution. But does the flexibility matter? The E2ϳUb-binding site resides in a segment bounded by conserved residues Tyr-642 in helix 13 and Gly-666 in helix 14 of SspH1 (Fig. 4A). In the crystal structures of IpaH1.4, IpaH4, IpaH9.8, and the highly homologous SspH2, the structurally conserved glycine (Ͼ98%) resides in the middle of the long terminal helix of the thumb. Yet for SspH1 and SspH2, substitution of the glycine for Ala, a residue with higher helix propensity not predicted to introduce unfavorable steric clashes, has profound effects on Ube2D3ϳUb binding and E3 activity (Fig. 4C  and Fig. S4C). The strict requirement for a glycine residue at this position strongly suggests that local flexibility in the E2ϳUb-binding site is a key feature of the mechanism by which this class of bacterial E3s recruits host Ube2DϳUb to generate reactive E3ϳUb intermediates (stage I of the overall reaction).
The dynamic behavior of the Ube2DϳUb-binding site is in stark contrast to the well-structured canonical binding sites used by eukaryotic RING, HECT, and RBR E3s (25)(26)(27)(28)(29). Furthermore, whereas eukaryotic E3s bind both free E2 and E2ϳUb species, we do not detect binding of free E2 to SspH1 or SspH2, even at high protein concentrations, although weak binding has been reported for Ube2D2 and IpaH9.8 (K d ϳ185 M) (16). In contrast, SspH1 and SspH2 bind E2ϳUb much more strongly. One explanation is that flexibility within the binding site lowers the E2 affinity so that complex formation requires both Ube2D and Ub in the form of E2ϳUb conjugate. This strategy could allow the bacterial E3s to specifically recruit the reactive species of an E2 and compete more effectively with host E3s. Together, the observations provide a compelling argument that local flexibility within the E2ϳUb-binding site of SspH1 does matter and that this property is likely shared among the bacterial E3 family.
A second type of dynamic behavior is suggested by the different relative orientations of the NSD and CSD in available bacterial E3 structures that are highly suggestive of subdomain movements important for catalysis (Fig. 1). Reminiscent of eukaryotic HECT-type E3s, the SspH1 E3 domain can be separated into semi-independent N-and C-terminal subdomains with different functions. The NSD contains the active-site cysteine required to generate an activated SspH1ϳUb intermedi-ate, whereas the CSD binds to the E2ϳUb in a reactive conformation. Importantly, mutations in the NSD/CSD linker (G515P) or in the subdomain interface (H498K), both far removed from the SspH1 active and E2ϳUb-binding sites, still support formation of the E3ϳUb intermediate (stage I) but disrupt Ub transfer to substrate (stage II). Thus, the ability of the two E3 subdomains to reorient after formation of the E3ϳUb intermediate appears to be required for stage II of the reaction. The high degree of conservation of the glycine corresponding to SspH1 Gly-515 and the detrimental effect of a proline mutation identify this as the likely pivot position for the motion between the NSD and CSD. In sum, our analyses reveal that local dynamics at the E2ϳUb-binding site in the CSD are critical for stage I, and large-scale subdomain motions are critical for stage II.
Although HECT/RBR-like in requiring an obligate E3ϳUb intermediate, the binding mode of SspH1 for Ube2DϳUb differs from eukaryotic E3s in important ways. HECT/RBR E3s are known to bind their cognate E2ϳUbs in an open conformation that has low reactivity toward Lys, while retaining reactivity toward cysteine (26,27,33,34). This limits unwanted side reactions with nearby lysine residues, an important consideration for RBR E3s whose active sites are occluded and unavailable for immediate reaction with a bound E2ϳUb. Despite their dependence on transthiolation to generate the E3ϳUb intermediate, SspH/IpaH E3s do not engage the open E2ϳUb state, but one in which the Ube2D and Ub subunits grasp onto opposite sides of the SspH1 thumb, resulting in a "semi-closed" conformation ( Fig. 5). Although the SspH1-bound Ube2DϳUb is reactive toward lysine (Fig. S7C), the proximity of the E3 active site on the NSD guarantees rapid and high-fidelity transfer between the two enzyme active sites. This point is brought home strikingly by the ability to stoichiometrically generate isopeptidelinked E3-Ub using SspH1 C492K. Notably, how SspH1 attacks the E2ϳUb active site in solution (Fig. 5B) is similar to that captured in the crystal structure of the HECT E3 NEDD4L/ Ube2D3-O-Ub complex (26). Intriguingly, this direction of approach has also been observed for the substrate lysine of Ran-GAP1 modified with the ubiquitin-like protein SUMO in the Ubc9 active site (35,36). It appears that a similar approach into the E2ϳUb active site is exploited by both prokaryotic and eukaryotic E3 ligases and can support both transthiolation and aminolysis reactions.
Overall, our results support a model in which Ube2DϳUb clamps onto the top of the thumb of the CSD and is positioned to react with the E3 active-site cysteine (Fig. 8B). Local flexibility in the thumb region helps to favor binding of the productive E2ϳUb species over the individual (nonreactive) components. We propose that the proximal conformation of the E3 domain places the active site cysteine closer to the thumb region to react with the E2ϳUb conjugate, and flexibility in the NSD/ CSD linker centered around a highly conserved glycine enables exchange between proximal and distal conformations. Ube2D is released from the enzyme upon formation of the SspH1ϳUb activated intermediate (Fig. 8C). Subsequently, the E3 domain must undergo a conformational change to the distal conformation to bring the E3-tethered activated Ub into position to react with substrate (Fig. 8D). We note that a structure of SspH2

SspH1 uses a modular and dynamic E3 domain for Ub transfer
LRR-E3 shows the E3 domain in a distal conformation with the substrate-binding site-containing LRR domain close to the E3 active site (14); this may represent a conformation poised to carry out stage II of Ub transfer reaction. It is likely that movements of the LRR domain and the E3 proximal-to-distal conformational change are closely coupled, although the details of such coupling will require further investigation.

Protein expression, purification, and paramagnetic spin-label conjugation
DNA sequences for SspH1, SspH2, and PKN1 constructs were cloned into pET28 vectors in frame with an N-terminal hexahistidine tag followed by a thrombin cleavage sequence; mutations were generated by the QuikChange protocol. His tags were not removed unless otherwise indicated. The DNA sequences for WT Ube2D3 and C85S/S22R were cloned into pET28 vectors out of frame of tags. Unlabeled proteins were expressed in Escherichia coli BL21(DE3) cells in lysogeny broth medium. 15 N-Labeled proteins were grown in minimal MOPS medium supplemented with 15 NH 4 Cl. His-tagged proteins were purified by Ni 2ϩ affinity purification; Ube2D3 variants without affinity tags were purified by strong cation-exchange chromatography on an SP column (GE Healthcare) via a 0 -1 M salt gradient in 25 mM MES, pH 6, at 4°C (21). The final step of purification for all proteins was size-exclusion chromatography over Superdex75 resin equilibrated in 25 mM sodium phosphate, 150 mM NaCl, pH 7, at 4°C or room temperature. Cloning and purification of human and wheat E1, ubiquitin, and generation of purified Ubce2D3-O-Ub were performed as described previously (21,22). Single-cysteine SspH1 mutants were modified with TEMPO spin labels by incubation in 25 mM sodium phosphate, 25 mM sodium chloride, 1-2 mM TCEP, pH 7, overnight at 4°C or for 30 min at 37°C in the presence of a 10ϫ molar excess of iodoacetamide-TEMPO (Sigma-Aldrich) dissolved in DMSO. Excess spin label was removed by buffer exchange via centrifugal filtration.

SEC-MALS and analytical SEC
For SEC-MALS, purified 50 M SspH1 was separated by sizeexclusion chromatography on a 1.4-ml Superdex200 column equilibrated in 25 mM sodium phosphate, 150 mM sodium chloride, 1 mM DTT, pH 7, at a flow rate of 0.1 ml/min. In-line MALS was performed using a Wyatt mini-Dawn Treos, and the differential refractive index was detected using a Wyatt Optilab T-rEX. Molecular weight was determined using Wyatt Astra version 6.1 software. For analytical SEC, 50 M SspH1 E3 domain, 50 M LRR-E3 domain, or 50 M LRR-E3 domain in complex with 75 M PKN1 108 -201 (HRb region) was eluted on a 24-ml GE Superose6 10/300 GL column equilibrated in 25 mM sodium phosphate, 150 mM sodium chloride, pH 7, at room temperature. Molecular mass standards were 1 mg/ml alcohol dehydrogenase (Sigma-Aldrich; 145 kDa), 1 mg/ml BSA (New England Biolabs; 66 kDa), and 100 M Ube2D3 (17 kDa).

Ubiquitylation assays
Ubiquitylation assays were performed in 25 mM sodium phosphate, 150 mM sodium chloride, pH 7, buffer at 37°C using 1 M human or wheat E1 (UBA1), 5 M E2 Ube2D3, 2 M E3, 50 or 100 M Ub, 5 M PKN1 1-201 (HRab region) or 108 -201 (HRb region), 5 mM MgCl 2 , and 5 mM ATP. Reactions were initiated by the addition of ATP. Sampled time points were taken by the addition of 5 l of reaction mixture to 5 l of reducing or nonreducing SDS-PAGE load dye followed by boiling (for reduced samples) or incubation on ice (for nonreduced samples) prior to electrophoresis, and 4 -20% gradient SDSpolyacrylamide gels were loaded with 7.5 l of sample (Bio-Rad). Intrinsic lysine reactivity assays were performed with purified 40 M Ube2D3ϳUb, 20 M SspH1, and 50 mM lysine, pH 7, 25°C. Gels were visualized by Coomassie staining and imaged on a LI-COR Odyssey CLx or Bio-Rad Gel Doc.

NMR spectroscopy
NMR data were collected at 25°C on a 500-MHz Bruker AVANCE III or 600-MHz AVANCE II fitted with a cryoprobe. Data were processed with NMRPipe (37) and analyzed using NMRViewJ (38). NMR samples were prepared in standard buffer with the addition of 10% D 2 O. Protein concentrations varied from 50 to 200 M. Paramagnetic spin labels were reduced by the addition of 10 mM sodium ascorbate in 25 mM sodium phosphate, 150 mM sodium chloride, pH 7, buffer. Surface structure representations of resonance perturbations were plotted using PyMOL (39) on molecular structures deposited in the Protein Data Bank (40). SspH1 homology models based on the IpaH3 (PDB code 3CVR) and SspH2 (PDB code 3G06) structures were generated using the SWISS-MODEL web server (41).

Hydrogen-deuterium exchange mass spectrometry
Hydrogen-deuterium exchange for SspH1 E3 domain (residues 389 -700) was initiated by incubating 1 mg/ml protein in Figure 8. Proposed mechanism and domain motions for SspH/IpaH-catalyzed Ub transfer. A, SspH1 must bind both substrate (PKN1 HRb domain) and an activated E2ϳUb conjugate (Ube2D3ϳUb). The dynamics of the SspH1-binding site may ensure selectivity for the E2ϳUb conjugate. B, both subunits of the Ube2D3ϳUb conjugate grasp onto the thumb region of the CSD, adopting a semi-closed conformation. With the NSD in the proximal conformation, the E3 active site is positioned to rapidly react with Ube2DϳUb. The trans-thiolation reaction could also be facilitated by the flexibility of the E2ϳUb-binding region. C, after Ub is transferred to the E3 active site, Ube2D3 dissociates from the complex. D, the E3 domain undergoes a conformational change to the distal conformation, with the NSD pivoting toward the substrate bound to the LRR domain. This motion is aided by the flexibility of the linker (magenta) connecting the NSD and CSD. This motion may be coupled to a large conformational change in the relative position of the LRR and E3 domains. The result is that Ub is brought into proximity of the substrate, where it is transferred from the E3 active site to a substrate lysine.

SspH1 uses a modular and dynamic E3 domain for Ub transfer
deuterated 25 mM sodium phosphate, 150 mM NaCl, pD 7, buffer for 3 s, 1 min, 30 min, or 20 h at room temperature. Exchange was stopped by the addition of ice-cold quench buffer (0.2% formic acid, 100 mM TCEP, 2 M guanidinium chloride, pH 2.5). Pepsin digestion was performed during quenching by the addition of 3 mg/ml pepsin and incubation on ice for 5 min. Undeuterated samples were prepared by incubation in nondeuterated 25 mM sodium phosphate, 150 mM NaCl, pH 7, buffer. Fully deuterated samples were prepared by first making a stock of unfolded protein in 4 M guanidinium chloride and boiling for 30 min prior to the initiation of exchange, followed by exchange in deuterated 25 mM sodium phosphate, 150 mM NaCl, pH 7, buffer for 30 min while boiling. Following exchange, quenching, and digestion, samples were flash-frozen in liquid nitrogen and stored at Ϫ80º C. Prior to MS, samples were thawed on ice, and peptides were separated by HPLC using a C18 column. Mass spectrometry was performed on a Waters Synapt G1. Mass spectra were analyzed by MassLynx software (Waters Corp.) and HX-Express2 (42,43). Peptides were identified using Protein Prospector (44).