The molecular basis of ubiquitin-like protein NEDD8 deamidation by the bacterial effector protein Cif

Abstract

The cycle inhibiting factors (Cifs) are a family of translocated effector proteins, found in diverse pathogenic bacteria, that interfere with the host cell cycle by catalyzing the deamidation of a specific glutamine residue (Gln40) in NEDD8 and the related protein ubiquitin. This modification prevents recycling of neddylated cullin-RING ligases, leading to stabilization of various cullin-RING ligase targets, and also prevents polyubiquitin chain formation. Here, we report the crystal structures of two Cif/NEDD8 complexes, revealing a conserved molecular interface that defines enzyme/substrate recognition. Mutation of residues forming the interface suggests that shape complementarity, rather than specific individual interactions, is a critical feature for complex formation. We show that Cifs from diverse bacteria bind NEDD8 in vitro and conclude that they will all interact with their substrates in the same way. The “occluding loop” in Cif gates access to Gln40 by forcing a conformational change in the C terminus of NEDD8. We used native PAGE to follow the activity of Cif from the human pathogen Yersinia pseudotuberculosis and selected variants, and the position of Gln40 in the active site has allowed us to propose a catalytic mechanism for these enzymes.

Many pathogenic Gram-negative bacteria use a type III secretion (T3S) system to translocate effector proteins into target cells (1). Once inside the host cell, effectors act to subvert and/or otherwise manipulate vital cellular systems and represent a key virulence strategy for these pathogens (2). Type III secreted effectors (T3SEs) can encode a wide range of different enzymatic activities (3). Because of the generally low amino acid sequence conservation of effectors to proteins of known function, these activities are often only identified through structural studies.

In the past decade, progression of the host cell cycle has emerged as one cellular system targeted by multiple T3SEs from diverse pathogens (4–6). The cycle inhibiting factors (Cifs) comprise a family of T3SEs from animal pathogens and insect symbionts (7) that induce a cytopathic phenotype in host cells that includes cell-cycle arrest at the G₂/M or G₁/S transition (6, 8–12). It has been suggested that during the infection process, restriction of the host cell cycle might delay epithelial cell renewal and favor gut colonization (7). Recently, regulation of ubiquitin-mediated proteolysis has been implicated in the mechanism of Cif-induced cell-cycle arrest (7). Analysis of host cell proteins regulating cell-cycle checkpoints revealed accumulation of cyclin-dependent kinase inhibitors p21^Waf1/Cip1 and p27^Kip2 in response to Cifs (10, 11); these proteins are usually degraded by ubiquitin-mediated proteolysis. Further, ubiquitin-mediated proteolysis of GFP reporters expressed in HeLa cells was blocked following delivery of Cif from Burkholderia pseudomallei (Cif_Bp) (9).

One mechanism for managing eukaryotic cell-cycle progression is timed degradation of key regulators through ubiquitinylation and targeting to the 26S proteasome (13). In this system, ubiquitin molecules are covalently attached to proteins destined for destruction by the concerted action of an E1-E2-E3 enzyme cascade, with substrate specificity defined by E3 ligases (14). The largest family of E3 ligases is the cullin RING E3 ubiquitin ligases (CRLs) (15). As befitting their critical role in many cellular processes, the activities of CRLs are tightly regulated. One mechanism for CRL activation is through conjugation of the ubiquitin-like molecule NEDD8 (neural precursor cell expressed, developmentally down-regulated 8) to the cullin subunit (neddylation) (16, 17), stimulating substrate ubiquitination. Importantly, cycling of CRLs between neddylated and deneddylated forms is required for full activity (18, 19).

Cif from enteropathogenic Escherichia coli (Cif_Ec) interacts with NEDD8 in both yeast two-hybrid assays (20) and ProtoArray analysis (21). Cif_Ec also colocalizes with NEDD8 in the nuclei of HeLa cells (20) and specifically binds to neddylated CRLs, but not to the unmodified proteins, in immunoprecipitation assays (20). Significantly, Cif_Ec was shown to inhibit the E3 ligase activity of neddylated cullins (7, 9, 20, 21). Cif activity is correlated with accumulation of CRLs in their neddylated forms (22), preventing the neddylation/deneddylation cycle and locking CRLs in a neddylated but inactive form. This leads to stabilization of numerous CRL targets in cells, which presumably triggers the downstream cytopathic phenotype.

Structural studies of Cif_Ec, Cif_Bp, and Cif from Photorhabdus luminescens (Cif_Pl) (12, 23, 24) revealed a common fold, despite sharing low overall sequence identity. The proteins comprise a head-and-tail domain structure reminiscent of a comma or apostrophe. The C-terminal head domain comprises a cysteine protease-like fold and contains a conserved Cys-His-Gln catalytic triad. Regions of the N-terminal tail domain are important for Cif function, and it has been hypothesized that they contribute to substrate recognition (12, 20). A fundamental advance in understanding the mechanism by which Cifs inhibit CRL activity emerged when these effectors were shown to possess a specific deamidase activity (9). Cif_Bp and Cif_Ec both catalyze the deamidation of Gln40 in NEDD8 (converting this residue to a Glu). Cif_Bp also efficiently deamidates Gln40 of ubiquitin. Ectopic expression of a NEDD8(Gln40Glu) mutant in HeLa cells led to stabilization of CRL substrates and generated an equivalent effect to that exhibited by Enteropathogenic E. coli (EPEC) infection (9). This provides strong evidence that Cif deamidase activity toward NEDD8-Gln40 is necessary and sufficient for the Cif-mediated cytopathic phenotype.

The purpose of this study was to investigate the interaction between Cifs and NEDD8, both biochemically and structurally, and also to probe the mechanism of deamidation. Here, we report the crystal structures of two Cif/NEDD8 complexes, one including Cif from the human pathogen Yersinia pseudotuberculosis (Cif_Yp) and the second from the insect symbiont P. luminescens. The structure of Cif_Yp has not been determined before. The two complexes share a common mode of binding with interactions arising from both the head and tail domains. The Gln40 substrate residue of NEDD8 extends into the catalytic site of the Cifs. We also show that other members of the Cif family bind to NEDD8 and suggest that our structures are a model for all Cif/NEDD8 complexes. Using site-directed mutagenesis of Cif_Yp, we have probed the enzyme/substrate-binding interface and hypothesize that overall shape complementarity rather than any specific individual interaction is the driving force behind complex formation. Finally, we have also investigated the catalysis of NEDD8 and ubiquitin deamidation by Cif_Yp. This work has allowed us to propose a mechanism for the activity of Cifs that ultimately leads to inhibition of the host ubiquitin-dependent proteasomal degradation pathway and the cytopathic phenotype.

Results

Cifs Crystallize in a 1:1 Complex with NEDD8.

To define how the Cif family of effectors (sequence alignment is shown in Fig. 1) engages their substrate and catalyzes the deamidation of NEDD8, we attempted to crystallize a variety of Cif/NEDD8 complexes (details of gene cloning and protein expression are provided in Materials and Methods). We obtained diffraction-quality crystals of catalytic site mutants Cif_Pl(Cys123Ser) and Cif_Yp(Cys117Ala) in complex with NEDD8 following copurification of the enzyme and substrate (Materials and Methods, Fig. 2, Fig. S1, and Table S1). Both structures were solved by molecular replacement using uncomplexed Cifs and NEDD8 as search models. X-ray data collection, refinement, and validation statistics are given in Table 1. The Cif_Pl(Cys123Ser)/NEDD8 and Cif_Yp(Cys117Ala)/NEDD8 crystal structures comprise one and two 1:1 complexes of Cif and NEDD8 in the asymmetrical unit, respectively. Each of these complexes shows very similar overall arrangements with rmsds (25) of 0.59 Å between the two complexes in the Cif_Yp(Cys117Ala)/NEDD8 crystal and 1.41/1.47 Å between the Cif_Pl(Cys123Ser)/NEDD8 and the two Cif_Yp(Cys117Ala)/NEDD8 complexes (320, 279, and 284 equivalent C_α atoms considered). These structures most likely represent a substrate-binding mode that is conserved across the Cif family. A cartoon representation of the Cif_Yp(Cys117Ala)/NEDD8 complex (henceforth Cif_Yp/NEDD8) is shown in Fig. 3A.

Fig. 1.

Numbered sequence alignment, including four members of the Cif family. Highlighted in blue are the residues of the catalytic triad. Highlighted in red (tail domain) and green (head domain) are residues in Cif_Yp that were mutated to investigate either NEDD8 binding and/or catalysis. The residue that may be involved in NEDD8/ubiquitin selectivity is shown in yellow. Asterisks denote positions of identical residues.

Fig. 2.

Gel filtration enables purification of a 1:1 complex of Cif_Yp and NEDD8. (A) Gel filtration traces derived from purifications of Cif_Yp, NEDD8, and the complex. Details of the elution volumes are given in Table S1. mAU, milli-absorbance units. (B) SDS/PAGE analysis of fractions collected across the elution peaks corresponding to the Cif_Yp/NEDD8 complex and excess NEDD8 (from the same gel filtration experiment).

Table 1.

X-ray data collection, refinement, and validation statistics

	CifPl/NEDD8 (in-house)	CifPl/NEDD8 (DLS-I02)	CifYp/NEDD8 (DLS-I24)
Data collection
Space group	P21	P21	P6322
Cell dimensions
a, b, c; Å	40.7, 56.0, 67.6	40.7, 56.1, 67.6	125.4, 125.4, 169.9
α, β, γ; °	90.0, 104.0, 90.0	90.0, 104.0, 90.0	90.0, 90.0, 120.0
Resolution, Å	19.10–2.10 (2.21–2.10)	42.60–1.60 (1.69–1.60)	66.90–1.95 (2.06–1.95)
Rmerge	0.039 (0.123)	0.052 (0.288)	0.079 (0.463)
I/σ(I)	27.7 (12.7)	11.5 (3.0)	18.5 (4.2)
Completeness, %	95.0 (93.3)	96.4 (96.2)	100.0 (100.0)
Multiplicity	6.1 (5.9)	4.4 (4.3)	11.1 (9.2)
Refinement
Resolution, Å	—	38.20–1.60 (1.64–1.60)	66.90–1.95 (2.00–1.95)
Rwork, %	—	16.5 (26.3)	18.5 (22.1)
Rfree, %	—	23.6 (31.5)	23.5 (26.9)
No. of atoms
Cif	—	2,135	2,094, 2,078
NEDD8	—	634	635, 604
Water	—	376	318
Others	—	1	16
B-factors, Å2
Cif	—	22	27, 24
NEDD8	—	23	28, 36
Water	—	34	32
Others		25	33
rmsd
Bond lengths, Å	—	0.02	0.02
Bond angles, °	—	2.05	1.95
ESU (ML), Å	—	0.08	0.10
Ramachandran favored, %	—	98.4	98.7
Ramachandran outliers, %	—	0	0

Values in parentheses correspond to the highest resolution bin. DLS, Diamond Light Source; ESU (ML), Estimated Standard Uncertainty (Maxmium Likelihood).

Fig. 3.

Crystal structure of the Cif_Yp/NEDD8 complex. (A) Cartoon representation of the Cif_Yp/NEDD8 complex. Cif_Yp is shown in green, with NEDD8 shown in gray-blue. The residues that comprise the interface between the two proteins are colored copper (for Cif_Yp) and light blue (for NEDD8). (B) Space-filling surface representation of the complex. (C) Space-filling surface representation of Cif_Yp and NEDD8, with NEDD8 rotated ∼180° and translated to reveal the interface. Residues of Cif_Yp that have been mutated in this study are shown in yellow and are labeled with a single-letter amino acid code. (D) Close-up view of the Cif_Yp (surface)/NEDD8 (cartoon) interface. In NEDD8, the substrate Gln40 is shown and the reoriented C terminus is colored gold. In Cif, residues of the occluding loop are colored white. (E) Comparison of the free (Upper) and Cif-bound (Lower) conformations of NEDD8. Gln40 is shown, and the C-terminal residues are colored as in D. (F) Interactions in the Cif_Yp active site (green atoms in the cartoon) in complex with NEDD8 (gray-blue atoms in the cartoon). Hydrogen bonds (with distances) to the substrate Gln40 residue are shown as dashed lines.

Crystal Structure of Cif_Yp.

The crystal structure of free Cif_Yp has not been reported previously. Although in complex with NEDD8, the structure of Cif_Yp is very similar to other Cifs. It overlays on Cif_Bp (23), Cif_Pl (23), and the truncated structure of Cif_Ec (24) with rmsds (25) of 1.69 Å, 1.72 Å, and 1.20 Å (231, 229, and 167 equivalent C_α atoms considered). The catalytic triad residues Cys117, His173, and Gln193 (Cif_Yp numbering) occupy essentially equivalent positions in all structures.

An Extensive Binding Interface Is Formed Between Cif_Yp/Pl and NEDD8.

Structures of uncomplexed Cifs have been described as comprising a head-and-tail domain structure reminiscent of a comma or apostrophe (23). The structures of Cif_Yp/NEDD8 and Cif_Pl(Cys123Ser)/NEDD8 independently show that both the head domain [which comprises residues Val116/122 (Cif_Yp/Cif_Pl) to the C terminus] and tail domain [which comprises residues from the N terminus to Pro115/121 (Cif_Yp/Cif_Pl)] make significant contributions to the NEDD8 binding interface (Fig. 3 A–D). This interface buries 1508.5 Ų of the NEDD8 solvent-accessible surface area in the Cif_Yp/NEDD8 complex, equivalent to 31.1% of the total [1,376.1 Ų (29.2% of the total) is buried in the Cif_Pl(Cys123Ser)/NEDD8 complex]. Both the Cif_Yp/NEDD8 and Cif_Pl(Cys123Ser)/NEDD8 interfaces have a complexation significance score of 1.00 as defined by Protein Interfaces, Surfaces and Assemblies (PISA) (26). NEDD8 residues that form the interface with Cif reside on the loop between β₁ and β₂ (Lys6–Lys11), α₁ and β₃ (Glu31–Arg42, which contains the substrate Gln40 residue), β₃ and β₄ (Ile44–Gly47), and the C-terminal β-strand (Val66–Arg74). These four regions form interactions with the tail, head, tail, and tail-and-head domains of Cif, respectively. Full details of the residues contributing to the interfaces and the interactions they form are given in Tables S2–S5.

Changes in Conformation on Complex Formation.

The availability of the uncomplexed Cif_Pl (23) and NEDD8 (27) structures, alongside the Cif_Pl(Cys123Ser)/NEDD8 complex (henceforth Cif_Pl/NEDD8) described here has allowed analysis of the conformational changes occurring during complex formation. Overall, the structures of Cif_Pl and NEDD8 are very similar in their unbound and bound forms (rmsds of 0.86 Å and 0.62 Å, using 247 and 71 equivalent C_α atoms, respectively) (25) (Fig. 4). The only notable change in Cif_Pl on binding NEDD8 is a slight reorientation at the tip of the tail domain (residues Ile94–Tyr116) that bends toward the substrate protein (maximum C_α shift of 3.4 Å for Glu100). In NEDD8, the C-terminal five residues (Val70–Arg74) undergo a significant shift on binding Cif_Pl (9.7 Å for the C_α of Arg74) that breaks β-sheet hydrogen-bonding interactions and displaces the C-terminal residues away from the α₁/β₃ loop (Fig. 4). The α₁/β₃ loop contains the substrate NEDD8:Gln40 residue, and the displacement of the C terminus appears critical for positioning this residue in the Cif active site (Fig. 3 D–F). Cif_Pl residues Val122 and Leu203 are the key residues that enable the displacement of this region. The same displacement of the C terminus of NEDD8 is observed in the Cif_Yp/NEDD8 complex with Val116 and Leu196 substituting for Val122/Leu203 (Fig. 3 D–F). This Val/Leu pair is conserved in all Cifs except Cif_Ec, where the equivalent Val is a Ser residue.

Fig. 4.

Comparison of the overall structure of Cif_Pl and NEDD8 in their uncomplexed states and in the Cif_Pl/NEDD8 complex. The structures of Cif_Pl and NEDD8 as found in the complex are shown in wheat and gray-blue. Those of the overlaid uncomplexed proteins are shown in cyan and magenta.

Cif Proteins Bind NEDD8 in Solution.

Using gel filtration, we showed that Cif_Yp, Cif_Ec, Cif_Bp, and Cif_Pl (all containing mutations in the active site Cys) were able to bind NEDD8, with the increases in apparent molecular mass on complex formation consistent with 1:1 binding (Fig. 2, Fig. S1, and Table S1). We then focused on the Cif_Yp/NEDD8 interaction as a model system to probe the properties of Cif/NEDD8 binding attributable to the availability of the crystal structure and this pathogen’s relevance for human disease. We quantified the interaction between Cif_Yp(Cys117Ala) and NEDD8 using isothermal titration calorimetry (ITC). This experiment showed the affinity was in the submicromolar range (Fig. 5 and Table S6). The ITC data independently confirmed the 1:1 binding stoichiometry previously observed by gel filtration, further supporting the validity of the interaction observed in the crystals. The interaction between Cif_Yp(Cys117Ala) and NEDD8 was also analyzed using intrinsic tryptophan fluorescence (SI Text and Figs. S2 and S3).

Fig. 5.

Cif_Yp and NEDD8 binding monitored using ITC. (A) Example of binding isotherm and associated fit for the Cif_Yp(Cys117Ala)/NEDD8 interaction. (B) Bar chart representation of the dissociation constant between Cif_Yp(Cys117Ala)/NEDD8, and additional variants as labeled, derived from fits to the ITC curves (Table S6).

Targeted Mutagenesis of the Cif/NEDD8 Binding Interface Disrupts the Interaction.

Having quantified the interaction between Cif_Yp(Cys117Ala) and NEDD8, we investigated the importance of interfacing residues identified in the Cif_Yp/NEDD8 structure. First, we constructed five independent alanine substitution mutants in residues that we hypothesized would form important interactions with NEDD8 (Cif_Yp tail domain residues Asp66 and Asp67 and Cif_Yp head domain residues Asn122, Asn167, and Leu196; Fig. 3 B and C and Tables S2–S5). All but Cif_Yp(Asn122Ala) were expressed and purified, and displayed comparable chemical unfolding/refolding profiles to WT protein; therefore, these mutations do not significantly affect stability (Fig. S4 and Table S7). Somewhat surprisingly, each of these mutants, which included examples of removing intermolecular hydrogen bonds and hydrophobic interactions, gave only marginal decreases in the affinities as measured by ITC (Fig. 5B and Table S6); similar results were obtained using fluorescence-based assays (Fig. S2 and Table S6).

We then made six additional mutants in the Cif_Yp(Cys117Ala) background to test the effects on complex formation of introducing steric clashes and/or charged residues at interface positions. Two of these mutations [Asp67Arg and Val104Glu (tail domain)] were designed as controls and targeted residues that did not closely associate with NEDD8 in the structure, and four [Asp66Arg and Leu106Glu (tail domain) and Val116Asp and Gly118Thr (head domain)] were predicted to compromise Cif_Yp/NEDD8 interaction (Fig. 3C). All variant proteins except Asp67Arg were expressed and purified, and displayed comparable chemical unfolding/refolding profiles to WT protein (Table S7). Although the affinity of the interaction between Cif_Yp(Val104Glu) and NEDD8 was equivalent to WT, Cif_Yp(Leu106Glu), Cif_Yp(Val116Asp), and Cif_Yp(Gly118Thr) were each severely compromised in NEDD8 binding as measured by ITC (Fig. 5B and Table S6); similar results were obtained using fluorescence (Fig. S2 and Table S6). Cif_Yp(Asp66Arg) showed an intermediate effect but was still significantly impaired in binding to NEDD8.

NEDD8 Residue Glutamine 40 Occupies the Cif Active Site.

The structures of Cif_Yp/NEDD8 and Cif_Pl/NEDD8 revealed that the substrate NEDD8 residue Gln40 projects from the loop between α₁ and β₃ into the Cif active site (Fig. 3F). In the Cif_Pl/NEDD8 complex, the N^ε2 atom of NEDD8:Gln40 forms a hydrogen bond to the OH group of Ser123 (the mutated catalytic residue); it is unlikely that this represents a catalytically competent orientation for this side chain. In the Cif_Yp/NEDD8 complex, the Cys117Ala mutation allows the side chain of NEDD8:Gln40 to lie directly over what would be the catalytic center. We produced a model of the WT Cif_Yp/NEDD8 complex by mutating (in silico) Ala117 back to a Cys (Fig. 3F). In this model, the thiol group of Cys117 is well-positioned for nucleophilic attack of the NEDD8:Gln40C^δ atom, which would initiate the deamidation reaction.

From the structures of the complexes, we also identified additional Cif residues that may be relevant for the catalytic activity. Foremost among these is Asp195 (Cif_Yp numbering), a residue conserved in all Cifs. The Asp195-O^δ1 atom forms a hydrogen bond with NEDD8:Gln40N^ε2 (2.81 Å) in the Cif_Yp/NEDD8 complex (Fig. 3F). Other interactions in the Cif_Yp/NEDD8 active site include hydrogen bonds between NEDD8:Gln40O^δ1 with the backbone amides of Cys117 (2.77 Å) and Leu196 (2.94 Å) and NEDD8:Gln40N^ε2 with the backbone carbonyl of Gly172 (3.13 Å). The roles of these residues in orienting the NEDD8(Gln40) side chain are discussed below.

NEDD8 and Ubiquitin Deamidation by Cif_Yp.

Using tryptic digest liquid chromatography MS, we confirmed that the only modification to NEDD8 from exposure to Cif_Yp was the conversion of Gln40 to a Glu. We then used native PAGE (nPAGE) to investigate the deamidase activity of Cif_Yp (9, 28). This assay enables screening of selected structure-informed mutants for effects on activity. Incubation of NEDD8 with Cif_Yp results in a shift in electrophoretic mobility equivalent to NEDD8(Gln40Glu); no shift occurs for NEDD8(Gln40Ala) (Fig. 6A). Further, this shift is dependent on the catalytic residue Cys117 [incubation with Cif_Yp(Cys117Ala) does not result in a shift] (Fig. 6A).

Fig. 6.

Substrate deamidation by Cif_Yp (and selected variants) monitored using nPAGE. (A) nPAGE analysis of NEDD8 deamidation by Cif_Yp. C/A, Cif_Yp(Cys117Ala); D/N, Cif_Yp(Asp195Asn); L/E, Cif_Yp(Leu106Glu). (B) nPAGE analysis of ubiquitin deamidation by Cif_Yp. (C) Quantification of Cif_Yp deamidase activity against NEDD8 and ubiquitin from enzyme-titration experiments (Fig. S5). Cif_Yp (□), Cif_Yp(Asp195Asn) (◇), Cif_Yp(Leu106Glu) (△), and Cif_Yp(Cys117Ala) (○), all with a solid line and with NEDD8 as the substrate, are shown. Cif_Yp with ubiquitin (■), with a dashed line, is also shown.

We also tested the activity of the Cif_Yp(Leu106Glu) and Cif_Yp(Asp195Asn) variants using the nPAGE assay. The Cif_Yp(Leu106Glu) mutation compromises the interaction of Cif_Yp(Cys117Ala) with NEDD8, despite its position near the tip of the tail domain, distant from the active site. Consistent with this, catalytic activity for the Cif_Yp(Leu106Glu) variant is significantly reduced compared with WT (Fig. 6A). The structures of Cif_Yp/NEDD8 and Cif_Pl/NEDD8 suggested a putative role for Asp195 (Cif_Yp numbering) in catalysis. We found that this variant still retained a significant level of activity, suggesting that an Asp at this position is not critical for function and is unlikely to act as a general acid/base in catalysis (Fig. 6A).

We also tested the ability of Cif_Yp, Cif_Yp(Cys117Ala), and Cif_Yp(Leu106Glu) to deamidate ubiquitin (Fig. 6B). Similar to NEDD8, we observed an electrophoretic shift of ubiquitin in the presence of the WT enzyme but not with Cif_Yp(Cys117Ala) or Cif_Yp(Leu106Glu).

To obtain further details of Cif_Yp’s deamidase activity, we performed enzyme titration experiments (9, 28) with NEDD8 and ubiquitin (Fig. 6C and Fig. S5). These assays show that NEDD8 is a better substrate for Cif_Yp than ubiquitin (∼0.025 pmol of Cif_Yp required for complete conversion of 350 pmol of NEDD8 in the assay compared with ∼0.25 pmol needed for ubiquitin). This is consistent with previous studies of Cif_Bp and Cif_Ec that also show a preference for NEDD8 (9). To support these results, we also performed enzyme titration experiments with NEDD8 and the Cif_Yp(Cys117Ala), Cif_Yp(Leu106Glu), and Cif_Yp(Asp195Asn) variants (Fig. 6C and Fig. S5).

p21 Accumulates in HeLa Cell Culture in the Presence Cif_Yp.

Delivery of purified Cif proteins to HeLa cells (Cif_Ec, Cif_Bp, and Cif from Photorhabdus species) results in the accumulation of cell-cycle regulators, including p21 and p27, with this activity dependent on the active site cysteine (10). We have shown that WT Cif_Yp also stabilizes p21, although the effect was weaker than that observed for Cif_Ec (Fig. 7). Both Cif_Yp(Cys117Ala), a catalytic site mutation, and Cif_Yp(Leu106Glu), a tail domain mutation distant from the active site (that severely compromises Cif_Yp/NEDD8 interaction and catalysis in vitro), also prevent p21 accumulation in HeLa cells (Fig. 7).

Fig. 7.

Effects of Cif_Yp and variants on the cell-cycle arrest marker p21. (A) p21 accumulates in the presence of WT (wt) Cif_Ec and Cif_Yp but not in the active site Cys mutants (C/A) or the Cif_Yp(Leu106Glu) mutant (L/E). p21, actin, and His-tagged proteins were probed with appropriate antibodies. (B) Chemiluminescence signals of p21 and actin shown in A from six [or 3 for Cif_Yp(Leu106Glu)] independent experiments were quantified. The averages of p21 level are represented and expressed as ±SEM. Statistical differences between experimental groups were determined using one-way ANOVA with the Bonferroni multiple comparison posttest. **P < 0.01; ***P < 0.001.

Discussion

Pathogenic bacteria of both animals and plants have evolved mechanisms for directly modifying host cell targets through the delivery of enzymes by the T3S system. These enzymes encode a variety of activities (3). Hydrolytic activity is emerging as a key mechanism used by T3SEs, with many examples of proteases (29–33) and phosphatases (34, 35) acting in host cells.

Structural studies of Cifs have shown they belong to a clan of related enzymes that includes cysteine proteases, acetyltransferases, transglutaminases, and putative deamidases (36). Cifs are the first members of this clan with deamidase activity to have their structures determined. The Cif/NEDD8 complexes described here allow us to explore how these enzymes have adapted a cysteine protease-like fold to perform a deamidation reaction and evolved to recognize their substrate.

Both deamidation and proteolysis result in the hydrolysis of an amide bond. Therefore, it is likely that the reaction catalyzed by Cif will be fundamentally similar to the proteolysis reaction catalyzed by cysteine proteases (37). In Cif_Yp, residues that comprise the catalytic triad are Cys117, His173, and Gln193, with the main-chain amides of Val116 and Leu196 contributing to an oxyanion hole. The adaptation of the cysteine protease-like fold to catalyze substrate-specific side-chain deamidation is unique to these effectors and is most likely an example of divergent evolution from a common ancestor. However, what are the key features of Cif that deliver this unique adapted activity? These can be explored by considering the two primary enzyme/substrate interfaces formed by the Cif head and tail domains in the Cif_Yp/NEDD8 and Cif_Pl/NEDD8 complexes and their roles in productive substrate binding.

The head domain of Cifs supports the position of catalytic residues within an active site cleft. The equivalent region in cysteine proteases defines substrate specificity, recognizing residues to both the N and C termini of the scissile bond, with the substrate presented in an extended conformation. In the Cif_Yp/NEDD8 and Cif_Pl/NEDD8 complexes, a prominent cleft interacts with residues on the α₁/β₃ loop of NEDD8 to the N terminus of Gln-40. Recognition of residues to the C terminus of Gln-40 is blocked by the occluding loop (23, 24), resulting in a very different substrate-binding interface (Fig. 3D). A residue within this loop, Leu196 (Cif_Yp numbering), along with Val116, displaces the C-terminal β-strand of NEDD8 from its native conformation, helping to position Gln40 in the active site (Fig. 3 D and E). Introduction of a charged residue in place of Val116 [Cif_Yp(Val116Asp)] prevented Cif/NEDD8 interaction. Further, the O^δ1 atom of Asp195 forms a hydrogen bond with the N^ε2 atom of NEDD8:Gln40. This interaction, along with the backbone amides of Cys117 and Leu196 and the backbone carbonyl of Gly172, forms a pattern of hydrogen bond donor/acceptor residues that ensures the N^ε2 atom, rather than the O^δ1 atom, of Gln40 is oriented toward His173 (Fig. 3F). This role is consistent with the relatively minor impact of the Cif_Yp(Asp195Asn) mutation on catalysis in vitro. In contrast, the equivalent mutation in Cif_Ec (Asp187Asn) did not trigger G₂/M cell-cycle arrest in infected HeLa cells (12), suggesting this subtle mutation may have functional relevance in the context of a cellular environment.

The tail domains of Cifs bear no resemblance to the structurally equivalent region in cysteine proteases and appear to be an adaptation evolved to interact with NEDD8/ubiquitin substrates. Mutation of a Leu106 in Cif_Yp (a residue that directly contacts NEDD8) to a Glu severely impairs formation of the Cif_Yp(Cys117Ala)/NEDD8 complex, catalysis in vitro, and activity in vivo. The Cif_Yp(Val104Glu) mutation, in a residue that does not contact NEDD8, still binds NEDD8 to WT levels. At the base of the tail domain, Asp66 (Cif_Yp) forms hydrogen-bonding interactions with both Thr7 and Thr9 of NEDD8; the adjacent Asp67 residue is oriented away from the substrate. Neither the Cif_Yp(Asp66Ala) nor Cif_Yp(Asp67Ala) mutation revealed any significant effect on NEDD8 binding in vitro. Binding of NEDD8 to Cif_Yp(Asp66Arg) was still observed but was significantly lower than WT. Inspection of the Cif_Yp/NEDD8 structure suggests an Arg side chain could be accommodated at this position in the complex. A double mutant of Cif_Ec (Asp58Ala/Asp59Ala) lost the ability to deamidate NEDD8, as assayed by nPAGE (9). Because structural data for this region of Cif_Ec are not currently available, the potential impact of this double mutation on the structure and how this might affect function (and/or enzyme stability) are difficult to assess.

No single alanine substitution variant we made in Cif_Yp significantly affected binding to NEDD8 in vitro, suggesting that none of the interactions probed are, in their own right, critical to binding. However, the introduction of bulky/charged substitutions in key interface amino acids all resulted in disruption of the interaction. In the case of Cif_Yp(Leu106Glu), we also show that this mutation affects catalysis in vitro and activity in vivo, and predict that the other bulky/charged substitutions will result in the same effects, although this remains to be tested. Given these observations, as well as the extensive buried surface formed between Cif and NEDD8, we conclude that shape complementarity is a key feature governing the interaction between these proteins.

In this study, we have mainly focused on the Cif/NEDD8 interaction. However, Cifs can also catalyze the deamidation of Gln40 in ubiquitin. In vitro, we observe a preference of Cif_Yp for NEDD8 over ubiquitin, similar to that seen for Cif_Bp (9). Modeling ubiquitin on the structure of NEDD8 in the Cif_Yp/NEDD8 complex revealed only three positions at the interface that are different between these proteins: 31 (NEDD8:Glu/ubquitin:Gln), 39 (Gln/Asp), and 72 (Ala/Arg). The residue at position 72 of NEDD8/ubiquitin is known to be a key determinant of specificity in both deneddylation (38) and E1-specificity (39) pathways. However, as supported by the structure and activity work presented here, all these changes can be accommodated in the Cif_Yp/NEDD8 complex. Similarly, docking the structure of Cif_Bp on the Cif_Yp/NEDD8 complex does not suggest any strong selective pressure for one or another substrate. Unlike Cif_Yp and Cif_Bp, Cif_Ec is reported to have an ∼1,000-fold preference for NEDD8 over ubiquitin (9). One hypothesis that could explain this differential activity is the lack of conservation in amino acids of Cif (Cif_Yp:Ile259, Cif_Bp:His234, Cif_Ec:Gln251, and Cif_Pl:Lys269; Fig. 1) that contributes to the recognition of the residue preceding Gln40 in the substrate (Gln39 in NEDD8 and Asp39 in ubiquitin). Second, Cif_Ec has a Ser residue at the position of an otherwise conserved Val (Val116 in Cif_Yp), which is implicated in the reorientation of the C-terminal β-strand of NEDD8, although how this could contribute to substrate selectivity is not clear. Although acknowledging that the published structure of Cif_Ec does not include the tail domain, these are two differences that may contribute to Cif_Ec’s NEDD8 selectivity.

Despite recent advances in defining the activity of Cifs, many questions remain. First, does Cif deamidate “free” NEDD8 or NEDD8 already conjugated to CRLs? An overlay of NEDD8 in the Cif/NEDD8 complexes with neddylated Cul5^ctd-Rbx1 (16) shows it is unlikely that Cif can interact with neddylated cullins in a catalytically competent orientation (because of a significant clash), suggesting that Cifs deamidate NEDD8 before conjugation. Further, when Cif_Bp is overexpressed in HEK293T cells with Cullin1, only very weak interaction is observed (22). Second, what is the precise mechanism by which NEDD8:Gln40 deamidation leads to stabilization of neddylated CRLs? A recent study concluded that Cif-mediated NEDD8 deamidation inhibits CRL deneddylation by the COP9 signalosome (CSN) in vivo (22). This study proposed two mechanisms by which this could be achieved: (i) Deamidation of Gln40 prevents NEDD8-induced conformational changes in CRLs that are important for recognition by CSN, or (ii) deamidation of Gln40 directly inhibits recruitment of CSN to neddylated CRLs. The latter supports hypotheses that deamidated ubiquitin impairs ubiquitination pathways, because the Gln40 side chain is involved in interaction with E3 ligases (9, 40). It is difficult to see how Gln40 deamidation could alter the structure of neddylated CRLs; this residue is presented to bulk solvent in the neddylated Cul5^ctd-Rbx1 structure [Protein Data Bank (PDB) ID code 3DQV] (16). Importantly, however, Gln40 is positioned adjacent to the thioester link between the C terminus of NEDD8 and Lys724 of Cul5^ctd (separated by only 6.1 Å). Therefore, it is more likely that Cif-mediated deamidation of Gln40 directly interferes with CSN binding to neddylated CRLs, although this remains to be verified experimentally. Finally, what is the benefit to the pathogen of interfering with CRL activity? Perhaps the resulting inhibition of the cell cycle delays epithelial cell renewal at the site of infection, favoring colonization, or reduced CRL activity augments the activity of other codelivered effectors, enhancing the virulence of the pathogen. By targeting CRLs, Cifs could interfere with the stability of hundreds of substrates, suggesting that this effector may modulate many critical host cell functions, including cell proliferation, apoptosis, differentiation, and immune responses (7).

Cifs are part of a growing number of microbial products that interfere with ubiquitination pathways in eukaryotic cells. Future studies will undoubtedly use Cifs not only to address the role of these pathways in host/pathogen interactions but as tools to dissect the role of such pathways in other aspects of cell biology.

Materials and Methods

Cloning of Cifs and NEDD8 for Protein Production.

The cloning of Cif_Yp, Cif_Bp, Cif_Ec, and Cif_Pl has been described previously (10, 41). In this study, we generated unique constructs for expression of Cif_Bp and Cif_Yp using existing vectors as templates. For Cif_Bp, we used primers 5′-AAGTTCTGTTTCAGGGCCCGatgataacgccgatcatttcatcg-3′ (Forward) and 5′-ATGGTCTAGAAAGCTTTAgccaaggccggcgacgtattgtgc-3′ (Reverse) to amplify the region encoding the full-length protein (23). These primers included DNA sequences (capitalized letters) that enabled recombination-based cloning into the pOPIN-F expression vector using published procedures (42).

To produce recombinant Cif_Yp in a soluble form, we generated an N-terminally truncated construct of this protein using primers 5′-AAGTTCTGTTTCAGGGCCCGgtttcacattccataaataacccttcg-3′ (Forward) and 5′-ATGGTCTAGAAAGCTTTAattacagtgagttttaatgattgacatattg-3′ (Reverse) that amplified residues 33–290 of the native sequence. The resulting PCR product was cloned into pOPIN-F as above. We also generated the Cif_Yp(Cys117Ala) mutant in pOPIN-F using the same primers and a template DNA that already included this mutation.

DNA encoding full-length NEDD8 was amplified from a synthetic template obtained from Geneart and PCR primers 5′-AGGAGATATACCATGctaattaaagtgaagacgctgaccggaaag-3′ (Forward) and 5′-GTGATGGTGATGTTTctgcctaagaccacctcctcctctcagagc-3′ (Reverse). The resulting product was cloned into pOPIN-E (42), generating a C-terminal 6× His-tag.

The DNA sequence of all constructs was verified by sequencing.

Mutagenesis of Cif_Yp and NEDD8.

All the additional mutants established in the Cif_Yp(Cys117Ala) background were produced by Genscript, using the pOPIN-F:Cif_Yp(Cys117Ala) plasmid as a template. The Gln40Glu and Gln40Ala mutants of NEDD8 were also produced by Genscript, using the pOPIN-E:NEDD8 plasmid as a template.

Expression and Purification of Cif Proteins.

Cif_Yp and Cif_Bp were expressed from pOPIN-F, and Cif_Ec and Cif_Pl were expressed from pET28a-based plasmids. Soluble Cif_Yp was obtained using E. coli SoluBL21 (Gelantis); expression of all other Cifs used E. coli BL21(DE3). Bacterial cultures were grown in LB broth at 37 °C, supplemented with carbenicillin (pOPIN-F, 100 mg/L) or kanamycin (pET28a, 30 mg/L), before induction with 1 mM isopropyl-β-d-thiogalactopyranoside and overnight growth at 16 °C. Cells were pelleted at 5,000 × g, frozen, and stored at −80 °C until ready for purification. Thawed cell pellets were resuspended in 50 mM Hepes (pH 7.8), 300 mM NaCl, and 25 mM imidazole before being lysed by sonication. Unbroken cells and debris were cleared by centrifugation at 30,000 × g for 25 min. Cleared lysates were loaded onto preequilibrated Ni²⁺-immobilized metal ion affinity chromatography (IMAC) columns; washed extensively in load buffer; and step-eluted in 50 mM Hepes (pH 7.8), 300 mM NaCl, and 250 mM imidazole. The eluate was directly loaded onto a Hi-Load 26/60 Superdex 75 gel filtration column (GE Healthcare); equilibrated; and run in 20 mM Hepes (pH 7.5), 150 mM NaCl, and 1 mM DTT (DTT was excluded from the buffer for purification of proteins used in ITC experiments). Proteins were concentrated by ultrafiltration to between 350 μM and 950 μM before aliquoting and flash-freezing in liquid nitrogen. All protein aliquots were stored frozen until required for experiments.

Preparation of NEDD8 and Ubiquitin.

NEDD8 was expressed from pOPIN-E using E. coli SoluBL21. Purification proceeded as described for Cifs, and DTT was not included in the gel filtration buffer.

Ubiquitin was purchased from Sigma (bovine, identical sequence to human ubiquitin) and redissolved in 20 mM Hepes (pH 7.5) and 150 mM NaCl for use.

Purification and Crystallization of Cif/NEDD8 Complexes.

To generate Cif/NEDD8 complexes for purification and subsequent crystallization, we adopted a “cosplitting” approach, where pelleted bacterial cells previously induced to express individual Cif proteins were mixed and resuspended with pelleted cells containing expressed NEDD8. The mixture of resuspended cells was then sonicated, and proteins were purified as above, retaining elution peaks that contain complexes as judged by SDS/PAGE. We attempted to saturate Cif with NEDD8 by adding excess culture expressing this second protein. Saturation was evident from the presence of a large peak corresponding to free NEDD8 in the gel filtration profile at a high retention volume and the absence of a peak or shoulder at elution volumes that would correspond to free Cifs. For Cif_Bp and Cif_Yp, we also produced samples in which the N-terminal His-tag was removed using 3C protease (samples were concentrated to ∼700 μM and treated with 0.01 mg/mL 3C protease overnight at room temperature).

Crystallization of Cif/NEDD8 Complexes.

We produced Cif/NEDD8 complexes using Cif_Yp, Cif_Bp, Cif_Ec, and Cif_Pl with various catalytic site mutants. Extensive screening of crystallization conditions for these complexes using robotic setups identified crystals for Cif_Pl(Cys123Ser)/NEDD8 and Cif_Yp(Cys117Ala)/NEDD8. The latter was from a sample treated with 3C protease. Cif_Pl(Cys123Ser)/NEDD8 crystals were confirmed to comprise both proteins using MS. Diffraction quality crystals of Cif_Pl(Cys123Ser)/NEDD8 were obtained from 20% (vol/vol) PEG 4000, 200 mM sodium acetate, and 100 mM Mes (pH 6.7) using a 96-well sitting drop plate and a drop composed of 0.35 μL of protein solution [∼600 μM complex in 20 mM Hepes (pH 7.5), 150 mM NaCl] and 0.65 μL of the crystallization reagent. Diffraction quality crystals of Cif_Yp(Cys117Ala)/NEDD8 crystals were grown from 2.2 M sodium malonate, 44 mM Bis-Tris propane (pH 7), 66 mM Bis-Tris propane (pH 8) using a 1-μL sitting drop composed of 0.35 μL of the protein sample [∼500 μM complex, 20 mM Hepes (pH 7.5), 150 mM NaCl], and 0.65 μL of the crystallization reagent.

X-Ray Data Collection, Structure Solution, Refinement, and Validation.

Crystals of Cif_Pl(Cys123Ser)/NEDD8 and Cif_Yp(Cys117Ala)/NEDD8 were frozen in a cryoprotectant solution composed of 75% of the mother liquor (taken directly from the crystallization plate) and 25% ethylene glycol. All X-ray data were processed with iMosflm (43) and scaled with Scala (44), as implemented in the CCP4 suite (45). X-ray data collection statistics are given in Table 1.

The structure of Cif_Pl(Cys123Ser)/NEDD8 was solved by molecular replacement with data collected on a Rigaku RU-H3RHB generator/Mar345 detector and MolRep (45), using search models derived from Cif_Pl and NEDD8 [PDB ID codes 3GQY (Cif_Pl) and 1NDD (NEDD8)]. Iterative cycles of refinement with Refmac5 (45) and manual rebuilding with Coot (46) generated a model that was fitted to high-resolution data (from the same crystal) obtained at beamline I02 of the Diamond Light Source (Oxford, United Kingdom). Refinement/rebuilding cycles as above generated the final model. Anisotropic B-factors and alternate conformations were added toward the end of refinement.

The structure of Cif_Yp(Cys117Ala)/NEDD8 was solved by molecular replacement with data collected at beamline I24 of the Diamond Light Source and MolRep. An edited version of the Cif_Pl(Cys123Ser)/NEDD8 was used as a model (essentially polyalanine traces were generated). The resulting phases were modified by Parrot (47); Buccaneer (47) was used to rebuild the structure with the correct sequence. Iterative cycles of refinement with Refmac5 and manual rebuilding with Coot generated the final model.

Structure validation used Coot and Molprobity (48). A selection of refinement and validation statistics is given in Table 1. Structure-based overlays were calculated using secondary-structure matching algorithms as implemented in Superpose from the CCP4 suite (25, 45). The coordinates and structure factors for the Cif_Yp/NEDD8 and Cif_Pl/NEDD8 complexes have been submitted to the PDB with ID codes 4F8C (Cif_Yp/NEDD8) and 4FBJ (Cif_Pl/NEDD8).

Protein/Protein Interaction Studies (ITC).

Cifs and NEDD8 were both prepared in 20 mM Hepes (pH 7.5) and 150 mM NaCl, and were diluted consistently into the same buffer. All ITC experiments were performed with a MicroCal 205 calorimeter in high gain mode. In a typical experiment, the calorimetry cell was filled with 205 μL of Cif at ∼100 μM before making sequential injections of ∼1,200 μM NEDD8 from the syringe up to a final cumulative injection volume of 38 μL. Experiments were conducted at 25 °C, with typical injections of between 1.0 μL and 2.0 μL at 180-s intervals. Two control experiments were performed to ensure that the heat of dilution of NEDD8 or Cif would not be problematic under these experimental conditions; these involved direct injections of NEDD8 into buffer and injections of buffer into a solution of Cif. Binding isotherms were fitted to the integrated calorimetric data using Origin (Microcal). Each binding experiment was performed at least three times (except for the Cys117Ala/Asp66Arg variant, which was performed twice), with the mean and SD calculated for each variant.

nPAGE to Monitor Enzyme Activity.

A total of 700 pmol of NEDD8, NEDD8(Q40A), and NEDD8(Q40E) was incubated with 0.1 pmol of Cif_Yp (or variants) for 30 min at 30 °C. Total reaction volume was 10 μL. The reaction mixture was chilled on ice, diluted with 1:1 loading buffer [50 mM glycine (pH 10.0), 20% (vol/vol) glycerol], and 10 μL was loaded onto 12.5% nPAGE gels buffered with 50 mM glycine at pH 10.0 (gels were run with the same buffer). Therefore, 350 pmol of substrate was loaded on the gel. Proteins were visualized with InstantBlue staining, and the bands quantified using ImageJ (National Institutes of Health). The same protocol was used for the ubiquitin assays, except 1.0 pmol of Cif_Yp (or variants) was incubated with 700 pmol of substrate and gels were stained with Coomassie Blue.

For the titration experiments, the range of enzyme quantities used is shown in Fig. 6C and Fig. S5.

Lipofection Experiments.

Lipofection assays of purified Cif proteins were conducted essentially as previously described (10). HeLa cells (CCL-2; American Type Culture Collection) were cultured in DMEM (Invitrogen) supplemented with 10% FCS (Eurobio) and 80 mg/L gentamicin at 37 °C in a 5% CO₂ atmosphere. For lipofection assays, 80 μL of purified Cif_Yp, Cif_Ec (500 μg/mL), or PBS as a negative control was added to one Bio-PORTER tube (Genlantis) and resuspended in 420 μL of DMEM. The samples were added to the HeLa cells in six-well plates and incubated for 4 h before being replaced by fresh growth medium and further incubated for 20 h.

For Western blot analyses, ∼6 × 10⁵ cells were lysed in 80 μL of SDS/PAGE sample buffer, sonicated for 2 s to shear DNA, and then boiled for 5 min. Protein samples were resolved on either 10% or 4–12% NuPage gradient gels (Invitrogen) and blotted on PVDF membranes. Membranes were blocked in 10 mM Tris (pH 7.8), 150 mM NaCl, 0.1% Tween 20, and 5% nonfat dry milk, and then probed with primary antibody (0.5 mg/mL⁻¹) in the same buffer. Primary antibodies used were anti-actin (ICN), anti-p21 (Cell Signaling Technology), and anti-histidine (Qiagen). Bound antibodies were visualized with HRP-conjugated secondary antibody. Acquisitions were performed with a Molecular Imager ChemiDoc XRS system (Bio-Rad). Protein levels were quantified with Quantity One software (Bio-Rad) and normalized with the actin level.