Structural mechanism of ubiquitin and NEDD8 deamidation catalyzed by bacterial effectors that induce macrophage-specific apoptosis

Abstract

Targeting eukaryotic proteins for deamidation modification is increasingly appreciated as a general bacterial virulence mechanism. Here, we present an atomic view of how a bacterial deamidase effector, cycle-inhibiting factor homolog in Burkholderia pseudomallei (CHBP), recognizes its host targets, ubiquitin (Ub) and Ub-like neural precursor cell expressed, developmentally down-regulated 8 (NEDD8), and catalyzes site-specific deamidation. Crystal structures of CHBP–Ub/NEDD8 complexes show that Ub and NEDD8 are similarly cradled by a large cleft in CHBP with four contacting surfaces. The pattern of Ub/NEDD8 recognition by CHBP resembles that by the E1 activation enzyme, which critically involves the Lys-11 surface in Ub/NEDD8. Close examination of the papain-like catalytic center reveals structural determinants of CHBP being an obligate glutamine deamidase. Molecular-dynamics simulation identifies Gln-31/Glu-31 of Ub/NEDD8 as one key determinant of CHBP substrate preference for NEDD8. Inspired by the idea of using the unique bacterial activity as a tool, we further discover that CHBP-catalyzed NEDD8 deamidation triggers macrophage-specific apoptosis, which predicts a previously unknown macrophage-specific proapoptotic signal that is negatively regulated by neddylation-mediated protein ubiquitination/degradation.

Protein deamidation refers to hydrolysis of the side-chain carboxamide of a glutamine or rarely an asparagine. The related transamidation involves sequential deamidation and incorporation of a lysine ε-amine, resulting in cross-linking of two polypeptides (1). Humans have eight structurally related transglutaminases (TG1-7 and factor XIII) that regulate many biological processes. TGs can catalyze deamidation in the absence of a proper amine. Other promiscuous or dedicated enzymes capable of performing deamidation are also present in humans (2).

Deamidation is more widely used in microorganisms. In bacterial chemosensory system, CheB from Escherichia coli/Salmonella enterica (3) and CheD from most nonenteric chemotactic bacteria (4) catalyze site-specific deamidation of chemotaxis receptors. Many secreted bacterial toxins also have the deamidase activity. Cytotoxic necrotizing factor 1/2 (CNF1/2) from certain virulent E. coli strains deamidate Gln-63/61 in Rho GTPases, rendering constitutively active GTPases and altered actin cytoskeleton (5, 6). Pasteurella multocida toxin (PMT) activates heterotrimeric G proteins by deamidating a conserved Gln in Gα (7). Cycle-inhibiting factor (Cif) from enteropathogenic E. coli (EPEC) and Cif homolog in Burkholderia pseudomallei (CHBP), both delivered into host cells through the type III secretion system (TTSS), induce a cytopathic effect of cell cycle arrest and actin stress fiber formation on epithelial cells. The Cif/CHBP family deamidates a conserved Gln-40 in host ubiquitin (Ub) and Ub-like protein (UBL) neural precursor cell expressed, developmentally down-regulated 8 (NEDD8) (8). CHBP targets both Ub and NEDD8 with a preference for NEDD8 in vitro and during infection (8–10). Deamidated Ub is impaired in supporting Ub chain synthesis. NEDD8 is monoconjugated (neddylation) to Cullins that mediate the assembly of a large repertoire of Cullin-RING Ub ligases (CRLs) (11). Neddylation stimulates CRL Ub ligase activity, but this effect is reversed when deamidated NEDD8 is conjugated onto Cullins. NEDD8 deamidation and its inhibition of CRL activity are responsible for Cif/CHBP-induced cytopathic effect (8).

Deamidases and transglutaminases fall into two structural categories (12–20). CNFs and CheD feature a central β-sandwich surrounded by helices and loops; PMT, the CHBP family, and transglutaminases all bear a papain-like core structure (one α-helix and three to four antiparallel β-strands) and a Cys-His-Asp/Asn/Glu/Gln catalytic triad, with Cys being the nucleophile. All available deamidase/transglutaminase structures are substrate-free; the mechanism for substrate recognition, site-specific deamidation, and determination of a deamidation versus transglutamination reaction is largely unknown. Here, we determine crystal structures of CHBP–Ub/NEDD8 complexes and show that CHBP recognizes Ub/NEDD8 in a manner resembling Ub/NEDD8 recognition by their E1 activation enzymes. The structures also establish the mechanisms for site-specific deamidation and the deamidation-only property of CHBP. Molecular-dynamics simulation identifies electrostatic interactions mediated by Glu-31 in NEDD8 as the determining factor for CHBP substrate preference for NEDD8. We further explore the idea of using CHBP deamidase as a cell biology probe and discover that NEDD8 deamidation induces massive macrophage-specific apoptosis due to inactivation of the CRL pathway.

Results

Overall Structure of CHBP–Ub Complex.

To understand the mechanism of Ub deamidation by the CHBP family, we determined a 2.6-Å crystal structure of CHBP-N78 C156A (residues 78–328) in complex with Ub (Fig. 1A, Fig. S1A, and Table S1). The structure contains one CHBP and one Ub in an asymmetric unit, forming a 1:1 enzyme–substrate complex. Ub-bound CHBP is structurally identical to apo-CHBP (17) (Fig. S1B). Ub is well situated in a cleft between the CG and the HE lobe (Fig. S1C), which renders the to-be-deamidated Gln-40 facing the catalytic center in the CG lobe (Fig. S1A). The structure of Ub in the CHBP complex is also nearly identical to that of free Ub (Cα rmsd, 0.77 Å), except for the highly flexible C-terminal tail (Fig. S1D). The C-terminal tail of Ub passes through the bottom of the cleft in CHBP with its extreme terminus extending out of the complex. This implicates that CHBP could potentially modify Ub in the context of Ub chains, and cellular Ub recycled from Ub chains might also exist in the deamidated form when CHBP is present.

Fig. 1.

Structure and functional analyses of CHBP–Ub complex. (A) Overall structure of CHBP (orange)–Ub (cyan) complex in cartoons. HE and CG refer to the helical extension and core globular lobe of CHBP, respectively. (B) The four contacting interfaces between CHBP (orange surface) and Ub (cyan cartoon) are indicated by red dashed circles. (C–F) Detailed views of contact A (C), contact B (D), contact C (E), and the C-terminal contact (CTC) (F). Key residues involved in CHBP (orange) and Ub (cyan) interaction are labeled and shown as sticks. The black dashed lines are hydrogen bonds, and the green dots are water molecules. (G and H) Mutation analyses of CHBP–Ub interaction using in vitro deamidation assay. Intact and deamidated Ub were separated on a native gel following the deamidation reaction. The percentage of Ub deamidation was quantified and plotted against CHBP concentration. Ub del-C in H means deletion of the C-terminal six residues.

K11 Surface in Ub Is Most Critical for CHBP Interaction and Deamidation.

CHBP binding of Ub buries 2,779-Ų surface area, corresponding to 11 and 29% of the total surface area of CHBP and Ub, respectively. The interface has four contacting areas, namely contact A, B, and C, and the C-terminal contact (CTC) (Fig. 1B). Contact A in Ub overlaps with the canonical Leu-8/Ile-44/Val-70 hydrophobic patch recognized by most Ub-binding proteins. The corresponding interface on CHBP, composed of Phe-133, Tyr-139, and Leu-142, surrounds His-68 of Ub (Fig.1 B and C). Contact B, involving the loop of the N-terminal β-hairpin in Ub, features extensive hydrophilic interactions; Thr-9 of Ub forms a side-chain hydrogen bond with Asn-107 of CHBP; Thr-7 side-chain hydroxyl and Lys-11 side-chain amine in Ub are hydrogen bonded to Asp-108 of CHBP (Fig. 1D). These interactions require side-chain rotation of Thr-9 and Lys-11 by ∼165 and 140°, respectively. The backbone carbonyl oxygen of Thr-9 forms a water-mediated hydrogen bond with Ile-122 of CHBP. Contact C involves Ub C-terminal α-helix and contains three hydrogen bonds between backbone carbonyl oxygens of Glu-34, Lys-33, and Gly-35 of Ub and side chains of Asn-161, Thr-177, and Lys-212 of CHBP, respectively (Fig. 1E). Contact C, adjacent to the site of deamidation, likely facilitates presentation of Gln-40 to the catalytic center of CHBP. The CTC contains four pairs of hydrogen bonds: the main-chain carbonyl oxygen of Ub Leu-71 with the amide nitrogen of CHBP Val-155, the guanidyl of Ub Arg-72 with the main-chain carbonyl oxygen of CHBP Asp-147, the amide nitrogen of Ub Leu-73 with the main-chain carbonyl oxygen of CHBP Thr-153, and the backbone carbonyl oxygen of Ub Leu-73 with the amide nitrogen of CHBP Thr-153 through a water molecule (Fig. 1F). The CTC reduces the flexibility of the C-terminal tail of Ub, which has a well-defined structure in our CHBP complex (Fig. S2A).

CHBP F133A/Y139A/L142A and Ub H68A mutations, designed to disrupt contact A interaction, decreased the deamidation efficiency by 10- to 30-fold (Fig. 1 G and H). The activity of contact C-deficient mutant of CHBP (N161A/T177A) was decreased more severely by 30- to 100-fold. Most notably, CHBP N107A/D108A was almost completely inactive, and the corresponding contact B mutants of Ub (T7A/T9A and K11A) were largely resistant to deamidation by CHBP due to the loss of their physical interactions with CHBP (see Fig. S5 A–C). Among all CHBP-contacting residues in Ub, only T7/T9/K11 are conserved in NEDD8, and this motif is indeed only present in NEDD8 among all UBLs (Fig. S2B). Thus, contact B is most critical for CHBP recognition and deamidation of Ub and NEDD8. Last, deletion of Ub C-terminal six residues drastically decreased the efficiency of deamidation by CHBP (Fig. 1H). Given that Gln-40 is spatially close to the C terminus of Ub, the CTC may function in orienting Gln-40 during catalysis.

Ub Recognition Mode of CHBP Resembles That of Ubiquitin-Activating Enzyme.

Ub-binding domain (UBD)-containing proteins, recognizing a single Ub or poly-Ub chains (21), and Ub enzymes like deubiquitinases (DUBs) and Ub-activating enzyme (E1) are two major groups of Ub-interacting proteins. UBDs contact less than 10% of total Ub surface area, which often relies on the Ile-44 patch. Differently, CHBP recognizes a much larger surface region in Ub (29% of total Ub surface); the related contact A primarily binds His-68 rather than Ile-44 of Ub (Fig. 1C). Thus, CHBP and UBDs adopt completely different structural modes of Ub recognition.

DUBs can be divided into five families (USP, UCH, OTU, Josephin, and JAMM) by their primary sequences; the JAMM family is metalloprotease (22), and all others are papain-like hydrolases. The USP family contacts the entire surface of the β-sheet of Ub (22, 23), which is significantly larger than that by CHBP. UCH, OTU, Josephin, and JAMM families use a similar Ub-binding mode that mainly involves contact A and B (seen in the CHBP–Ub complex) as well as Ub C-terminal tail (Fig. 2 and Fig. S3). Different from CHBP, DUBs barely bind Ub through contact C. This is because DUBs target the very C-terminal carboxyl group of Ub for hydrolysis whereas CHBP deamidates Gln-40, which is nearby contact C.

Fig. 2.

CHBP recognition surfaces in Ub/NEDD8 resemble those of E1 enzymes. Shown are the surface presentation of Ub (the left three columns) and NEDD8 (the right two columns), in which the areas contacted by CHBP, yeast Ub E1 enzyme Uba1p (PDB ID code 3CMM), deubiquitinating enzyme OTU1 (PDB ID code 3BY4), and human NEDD8 E1 enzyme APPBP1-UBA3 (PDB ID code 1R4N) are colored. The C-terminal contact is in orange; Ile-44 is in cyan; other contacting surfaces are in red. Contacts A, B, and C are marked on CHBP-bound Ub/NEDD8 by dashed circles.

According to the Ub and yeast E1 (Uba1p) complex structure (24), interaction with Uba1 involves 33% of Ub surface area, comparable to the 29% observed with CHBP–Ub interaction. Remarkably, Uba1p contacts almost the same Ub surfaces as CHBP, including the His-68/Ile-44 surface (contact A), the loop of N-terminal β-hairpin (contact B), C terminus of the long α-helix (contact C), and the C-terminal tail (Fig. 2). Lys-11 of Ub, critically contacted by CHBP, is also hydrogen bonded to Uba1p. These analyses reveal a striking similarity between CHBP and the E1 enzyme in their Ub recognition modes despite that there are no structural similarities between the two enzymes.

Catalytic Center and Structural Basis for the Deamidation-Only Property.

In CHBP–Ub complex, the small helical turn bearing Gln-40 is bracketed by contact C and the CTC, which positions the side chain of Gln-40 proximal to CHBP catalytic center (Fig. 3A). Asp-39 of Ub is hydrogen bonded to His-300 of CHBP, likely contributing to deamidation site selection. A similar situation is noted with bacterial receptor deamidase CheD (13). Asp-39, together with a steric hindrance provided by the side chain of CHBP Val-155, forms a clamp to precisely fix the side-chain carboxamide of Gln-40 at a position only 2.5 Å away from the catalytic Cys-156 in CHBP (Fig. 3A). Consistently, activity of CHBP H300A mutant was decreased by ∼10-fold and CHBP V155G was completely inactive (Fig. S4A).

Fig. 3.

Catalytic center and mechanism of site-specific deamidation. (A) The catalytic center of CHBP (orange)–Ub (cyan) complex. Catalytic triad residues (C156-H211-Q231), V155 and H300 of CHBP, as well as Q40 and E39 of Ub are in sticks. Hydrogen bonds are shown as black dashed lines. The loop (residues 233–241) wrapping around the open side of the catalytic center is in magenta. Contact C and CTC are marked with red dashed circles. (B) The loop prevents any amine acceptor from assessing the catalytic center. CHBP is shown as cartoon with partially transparent surface (gray). The short helical turn bearing Q40 in Ub is in cyan. The loop colored in magenta in the Upper panel is omitted in the Lower panel to visualize the catalytic triad residues in CHBP and the site of deamidation in Ub.

No deamidases/transglutaminases have a protein substrate-bound structure. When the CHBP–Ub structure was superimposed onto that of TG2 inhibitor [6-diazo-5-oxo-l-norleucine (DON)] complex (25), Gln-40 of Ub was found to occupy the same position as DON, which is covalently linked to the catalytic cysteine of TG2 (Fig. S4B), suggesting that our CHBP–Ub model captures a catalytically competent state. Different from TG2 and CNF-like enzymes, CHBP could not catalyze transamidation even in the presence of high-concentration ethylenediamine or putrescine. This deamidation-only property can be well explained by a unique loop in CHBP (residues 233–241) that wraps around the open side of the catalytic center and prevents any amine acceptor from assessing the thioester bond in the acyl-enzyme intermediate (Fig. 3B).

Structure of CHBP–NEDD8 Complex.

We also solved a 2.5-Å crystal structure of CHBP (C156A)–NEDD8 complex (Table S1), which is nearly identical to that of CHBP–Ub complex (Cα rmsd, 0.32 Å) (Fig. 4A). The NEDD8 complex preserves all contacting interfaces observed in the Ub complex. In contact A, the Phe-133/Tyr-139/Leu-142 patch of CHBP mediates hydrophobic interactions with the His-68 surface of NEDD8; CHBP Tyr-139 forms an additional hydrogen bond with NEDD8 Lys-6 (Fig. 4B). In contact B, Asn-107 and Asp-108 of CHBP form multiple hydrogen bonds with Thr-7/Thr-9/Lys-11 of NEDD8. In contact C, Lys-33, Glu-34, and Gly-35 of NEDD8 are hydrogen bonded to Thr-177, Asn-161, and Lys-212 of CHBP, respectively. Despite missing of NEDD8 C-terminal last four residues due to protein degradation during crystallization, Leu-71 remains hydrogen bonded to CHBP Val-155, indicating a C-terminal interaction similar to that observed with CHBP and Ub.

Fig. 4.

Structure and functional analyses of CHBP–NEDD8 complex. (A) Overall structure of CHBP–NEDD8 complex in comparison with that of CHBP–Ub complex. (B) Interfaces and catalytic center of CHBP (green)–NEDD8 (yellow) complex. Shown are close-up views of contacts A, B, and C, and the catalytic center. Key residues involved in binding and catalysis are in sticks. Corresponding structures in CHBP–Ub complex (gray) are superimposed onto those in CHBP–NEDD8 complex for comparison. (C) Filamentous actin (F-actin) staining of HeLa cells transduced with CHBP. WT or mutant CHBP proteins were delivered into HeLa cells using the LFn-PA system (see SI Materials and Methods for details). F-actin (red) and nuclei (blue) were stained with rhodamine-labeled phalloidin and DAPI (4′-6-diamidino-2-phenylindole), respectively.

Each of H68A, K11A, and T7A/T9A mutations in NEDD8 diminished or severely attenuated the CHBP–NEDD8 interaction (Fig. S5 A–C). Also similarly to that noted with Ub, NEDD8 H68A showed a modestly decreased deamidation, whereas NEDD8 K11A and T7A/T9A mutants were hardly modified by CHBP (Fig. S5D). Conversely, mutations in CHBP on contact A (F133A/Y139A/L142A), B (N107A/D108A), and C (N161A/T177A) affected NEDD8 deamidation to the same extent as that on Ub (Fig. S5E). All three mutants failed to induce actin bundle assembly (Fig. 4C) and cell cycle arrest (Fig. S5F).

NEDD8 surface area contacted by CHBP shows an analogous pattern to that by NEDD8 E1 enzyme APPBP1/UBA3 despite a slightly smaller contact B and a more extensive contact C surface (Fig. 2). Thus, CHBP also mimics the E1 enzyme to recognize NEDD8. NEDD8 Gln-39 is hydrogen bonded to CHBP His-300, which, together with CHBP Val-155, fixes NEDD8 Gln-40 in a position ready for hydrolysis by CHBP, just as that observed in the CHBP–Ub complex (Fig. 4B).

Mechanism of CHBP Substrate Preference for NEDD8.

CHBP-catalyzed NEDD8 deamidation is 10-fold more efficient than Ub deamidation (8). This is intriguing given that CHBP contacts the two substrates in a nearly identical manner. To gain insights into this, unbiased molecular-dynamics (MD) simulation of CHBP–Ub/NEDD8 complexes was performed. Remarkably, Ub in the CHBP complex showed a pronounced motion around its crystallographic position throughout the simulation (Fig. 5A and Movie S1), evident from a maximal and an average rmsd change of 4.5 and 2.5 Å, respectively (Fig. 5B). In contrast, NEDD8 exhibited limited motions with a maximal and an average rmsd of only 2.2 and 1.7 Å, respectively. The internal motions of Ub and NEDD8 were insignificant (Cα rmsd, ∼1 Å). The higher rmsd change of Ub indicates a relatively unstable complex with CHBP that is prone to be catalytically nonproductive. In fact, the dissociation constant (K_D) between Ub and CHBP (∼106.7 μM), measured by isothermal titration calorimetry, was higher by one order of magnitude than that of the CHBP–NEDD8 complex (∼9.4 μM) (Fig. S6).

Fig. 5.

Molecular-dynamics simulation reveals residue 31 of Ub/NEDD8 as the determinant of substrate selectivity of CHBP. (A) Positional changes of NEDD8 and Ub after 45-ns molecular-dynamics simulation (for additional information, see also Movie S1 and SI Materials and Methods). Presimulation and postsimulation structures of Ub and NEDD8 (N8) were aligned by using CHBP structure as the reference. (B) Cα rmsd changes of Ub/NEDD8 during simulation. (C) Electrostatic potential near residue 31 of Ub/NEDD8 in the CHBP complex. The surfaces are colored by relative electrostatic potential with red and blue denoting negative and positive, respectively. (Upper) CHBP (green)–NEDD8 (yellow) complex. (Lower) CHBP (orange)–Ub (cyan) complex. (Left) CHBP is shown as surface, whereas Ub/NEDD8 is shown as cartoons. (Right) CHBP is shown as cartoons, whereas Ub/NEDD8 is shown as surface. (Center) Magnified view of the surfaces around residue 31. E31 of NEDD8, Q31 of Ub, and the K212/K304/R306 patch on CHBP are marked.

We further analyzed the simulation trajectories and identified residue 31 of Ub and NEDD8 as one possible key determinant. In NEDD8, the negatively charged Glu-31 can potentially interact with a cluster of positively charged residues in CHBP (Lys-212, Lys-304, and Arg-306) (Fig. 5C). However, such electrostatic interactions are unlikely to occur in the CHBP–Ub complex; the neutral Gln-31 in Ub is not only significantly more distant to Lys-212/Lys-304/Arg-306 of CHBP during and after simulation (Fig. S7 A and B, and Movie S2) but also associated with higher interaction energy (Fig. S7C). Swapping this single residue completely reversed the distance between residue 31 in Ub/NEDD8 and the Lys-212/Lys-304/Arg-306 patch on CHBP as well as the corresponding interaction energy (Fig. S7 B and C). E31Q mutation in NEDD8 resulted in an increased rmsd value and enhanced overall fluctuation during simulation, and the same adverse effect also occurred with Ub Q31E mutation (Fig. 5B). Experimental analyses confirmed that Ub Q31E bound more tightly to CHBP and was deamidated nearly as efficiently as NEDD8 (Fig. S8 A and B). Conversely, NEDD8 E31Q interacted with CHBP less prominently and its deamidation by CHBP was drastically decreased. Thus, NEDD8 Glu-31-mediated electrostatic interaction is a critical determinant of CHBP substrate preference for NEDD8 over Ub.

NEDD8 Deamidation Triggers Macrophage-Specific Apoptosis.

Having revealed the structural basis for CHBP deamidation of Ub/NEDD8, we explored the idea of harnessing the unique activity of CHBP to discover new biological functions of ubiquitination. Interestingly, when recombinant CHBP was delivered into the cytosol of terminally differentiated primary bone marrow-derived macrophages (BMMs), massive cell death rather than cell cycle arrest occurred (Fig. 6A and Fig. S9A). Profiling a panel of additional cells showed that macrophages all died to a similar extent upon CHBP delivery, whereas nonmacrophage cells remained completely viable (Fig. S9 A and B). CHBP-triggered macrophage-specific death was not associated with rapid release of cytosolic contents (Fig. 6B), indicating preservation of an intact plasma membrane. Neither Nec-1, a specific inhibitor of programmed necrosis, nor deficiency of caspase-1 and Asc, which mediate macrophage pyroptosis, could prevent CHBP-induced macrophage death (Fig. S10 A–C). CHBP-transduced macrophage shrank severely, and evident membrane blebbing occurred (Fig. 6A). Additional typical apoptotic features, including caspase-3 activation and fragmentation of genomic DNA, were observed (Fig. 6 C and D, and Fig. S9C). Thus, CHBP-induced macrophage-specific death is apoptosis. Moreover, Toll-like receptor 4 (TLR4) and TNR receptor (TNFR)-deficient macrophages died to the same extent as wild-type macrophages (Fig. S10 D and E), suggesting that LPS contaminants and autocrined TNFα play little roles in CHBP induction of macrophage-specific apoptosis.

Fig. 6.

CHBP-catalyzed NEDD8 deamidation induces macrophage-specific apoptosis. (A) Intracellular delivery of recombinant CHBP by using the LFn-PA system triggers massive cell death in primary bone marrow-derived macrophage (pri-BMM). (Left and Center) Phase contrast images of untreated (Left) and CHBP-transduced (Center) cells. (Right) Cell viability determined by measuring amounts of cytosolic ATP. WT and C/A refer to wild type and the catalytic C156A mutant of CHBP, respectively. (B) Comparison of CHBP with anthrax lethal toxin (LT) on cytosol release and viability of RAW264.7 cells. Cell death (viability based) and cytosol release were determined by the ATP and lactate dehydrogenase assays, respectively. (C) DNA fragmentation of RAW264.7 cells delivered with wild type (CHBP^WT) or the C156S mutant (CHBP^C/S) CHBP. Staurosporine (STP) is included as a positive control. (D) Caspase-3 cleavage in CHBP-delivered RAW264.7 cells. (E) Effects of NEDD8 E1 inhibition on immortalized TLR4-deficient BMM [TLR4-KO i-BMM (Right)] or HeLa cells (Left). Cells were treated with 5 μM NEDD8 E1 inhibitor MLN4924 for 24 h. Macrophage viability was determined by the ATP assay and shown on the Right. Shown on the Left are phase-contrast images of HeLa cell morphology. (F) Effects of Ub E1 inhibition on TLR4-KO i-BMM. Macrophages were treated with Ub E1 inhibitor PYR-41 (50 μM) for 29 h. Cell viability was determined by the ATP assay.

CHBP C156A(S) mutant did not kill macrophage (Fig. 6A and Fig. S9B) despite that a much higher level of the mutant protein, compared with wild-type CHBP, was detected in macrophage cytosol (Fig. S11A). CHBP prefers NEDD8 over Ub for deamidation, as having been clearly demonstrated by our structural analyses. Consistently, nearly all cellular NEDD8 was found to be deamidated (Fig. S11B) and CRL substrate such as Nrf2 was accumulated in CHBP-delivered macrophages (Fig. S11C). Although Nrf2 accumulation was to a similar extent as observed in MG132-treated cells, degradation of non-CRL substrate (MOAP1) was not affected and levels of the total polyubiquitin conjugates showed little increase upon CHBP delivery (Fig. S11 C and D). Notably, a NEDD8 E1 inhibitor, MLN4924, not only recapitulated CHBP-triggered cell enlargement in HeLa cells but also induced massive macrophage death (Fig. 6E). In contrast, administration of a Ub E1 inhibitor PYR-41 triggered little macrophage death (Fig. 6F). The dose of PYR-41 used (50 μM) is sufficient to abolish Ub-E1 thioester formation in cells (26) and was already toxic to HeLa cells (Fig. S11E). Further dose titration demonstrated that macrophage death was much more sensitive to NEDD8 E1 inhibitor than that to Ub E1 inhibitor (Fig. S11F). An extremely high concentration (150 μM) of PYR-41 did kill macrophages, but the morphology differed from that of CHBP-induced apoptosis. These data suggest that NEDD8 deamidation and the resulting dysfunction of the Cullin pathway are mainly responsible for CHBP-induced apoptosis, further predicting a macrophage proapoptotic signal that is negatively regulated by neddylation and Cullin-mediated degradation.

Discussion

Although transglutaminase activity is abundant in higher eukaryotes, deamidation is more prevalent in prokaryotes and often used to modify host proteins. Following early discovery of Rho-deamidating CNF1 (5, 6), recent studies identify several important bacterial deamidase toxins including P. multocida PMT, which modifies and activates Gα (7), B. pseudomallei BPSL1549, which deamidates eIF4A and inhibits host protein synthesis (19), and Shigella flexneri TTSS effector OspI, which modifies Ub E2 enzyme Ubc13 for dampening host inflammatory response (27). All of these deamidase toxins use a catalytic cysteine with a CNF- or papain-like fold and target a key host protein(s) with high specificity. Thus, glutamine deamidation represents a previously underappreciated virulence mechanism for bacterial pathogens. Future studies will likely identify more bacterial toxins/effectors endowed with the deamidase activity.

Killing of macrophages by CHBP deamidase serves as a potential virulence mechanism for Burkholderia to counteract macrophage-mediated host defense. Several known effectors are associated with macrophage killing. Yersinia YopJ/YopP sensitizes macrophages to apoptosis by inhibiting MAPK and NF-κB pathways following TLR4 activation (28). AIP56 from Photobacterium damselae induces macrophages and neutrophils apoptosis (29), and its N terminus is homologous to EPEC effector NleC, which cleaves p65/RelA in the NF-κB pathway (30). VopS from Vibrio parahaemolyticus induces macrophage apoptosis also through NF-κB inhibition (31). However, apoptotic induction by CHBP does not involve TLR4 and TNFR-mediated NF-κB signaling. CHBP further differs from these effectors in that it alone is sufficient to kill macrophage, which is through blocking CRL-mediated ubiquitination and degradation.

Neddylation-stimulated Cullin pathway is critical for cell proliferation and cancer progression (32). The neddylation inhibitor MLN4924 is currently under clinical development to treat cancers such as lymphoma and multiple myeloma. The very similar cellular effects induced by CHBP and MLN4924 on both epithelial and macrophage cells suggest a therapeutic potential of the CHBP-family effectors. CHBP induction of apoptosis further indicates that defective neddylation can trigger macrophage apoptosis due to impaired degradation of a CRL substrate(s) likely of proapoptotic nature. Identification of the substrate may lead to discovery of a new macrophage-specific apoptotic signaling axis. Thus, the diverse and unique biochemical activities of bacterial effectors can be used to assist cell biological studies in a way similar to compounds-mediated chemical genetics (33).

Materials and Methods

Purified CHBP N78 C156A (20 mg/mL) and Ub (40 mg/mL) were mixed in a buffer containing 10 mM Hepes (pH 7.2) and 100 mM NaCl. Crystals of CHBP–Ub complex were grown in 12.5% (vol/vol) PEG 3350 and 125 mM magnesium formate at 20 °C for 1 wk using the vapor diffusion method. The crystal belongs to the C2 space group and native diffraction data were collected on Rigaku MicroMaxTM-007 X-ray generator with image plate detector. To crystallize CHBP–NEDD8 complex, CHBP N78 C156A (10 mg/mL) and NEDD8 (10 mg/mL) were mixed together in the buffer containing 10 mM Hepes (pH 7.2) and 100 mM NaCl, and crystals appeared in 1.5 M ammonium sulfate and 100 mM citrate (pH 5.5) several months after setting up the vapor diffusion drop. The CHBP–NEDD8 complex crystal also belongs to the C2 group and native diffraction data were collected at Beamline 17U of Shanghai Synchrony Radiation Facility with Quantum 315 CCD detector (ADSC). All animal experiments were conducted following the Ministry of Health national guidelines for housing and care of laboratory animals and performed in accordance with the regulations of the Institutional Animal Care and Use Committee at National Institute of Biological Sciences. The rest of the information about reagents, structure determination, in vitro deamidation and pull-down assays, MD simulation, and cell death assays is presented in SI Materials and Methods.