Active-Site Conformational Fluctuations Promote the Enzymatic Activity of NDM-1

β-Lactam antibiotics are the mainstay for the treatment of bacterial infections. However, elevated resistance to these antibiotics mediated by metallo-β-lactamases (MBLs) has become a global concern.

ABSTRACT

β-Lactam antibiotics are the mainstay for the treatment of bacterial infections. However, elevated resistance to these antibiotics mediated by metallo-β-lactamases (MBLs) has become a global concern. New Delhi metallo-β-lactamase-1 (NDM-1), a newly added member of the MBL family that can hydrolyze almost all β-lactam antibiotics, has rapidly spread all over the world and poses serious clinical threats. Broad-spectrum and mechanism-based inhibitors against all MBLs are highly desired, but the differential mechanisms of MBLs toward different antibiotics pose a great challenge. To facilitate the design of mechanism-based inhibitors, we investigated the active-site conformational changes of NDM-1 through the determination of a series of 15 high-resolution crystal structures in native form and in complex with products and by using biochemical and biophysical studies, site-directed mutagenesis, and molecular dynamics computation. The structural studies reveal the consistency of the active-site conformations in NDM-1/product complexes and the fluctuation in native NDM-1 structures. The enzymatic measurements indicate a correlation between enzymatic activity and the active-site fluctuation, with more fluctuation favoring higher activity. This correlation is further validated by structural and enzymatic studies of the Q123G mutant. Our combinational studies suggest that active-site conformational fluctuation promotes the enzymatic activity of NDM-1, which may guide further mechanism studies and inhibitor design.

INTRODUCTION

β-Lactam antibiotics, including penicillins, cephalosporins, and carbapenems, constitute more than half of the antibiotics prescribed in clinical settings (1). However, the efficiency of these antibiotics is being continuously challenged by the emergence of resistant pathogenic bacteria, among which resistance to carbapenems is of extreme concern since carbapenems are considered the last resort for multidrug-resistant infections. β-Lactamases, which inactivate β-lactam antibiotics by hydrolyzing the β-lactam bonds (2), play a major role in antibiotic resistance (3, 4). Based on sequence similarity, β-lactamases can be divided into four classes (A, B, C, and D) (5). Enzymes in classes A, C, and D are serine-β-lactamases which harbor a serine residue at the active site to facilitate catalysis. And members of class B are metallo-β-lactamases (MBLs) that retain one or two zinc ions at the active site to facilitate β-lactam cleavage. MBLs are broad-spectrum β-lactamases, active against almost all β-lactams, including carbapenems and even the clinically used inhibitors of serine-β-lactamases (6). Furthermore, most of the circulating MBLs are encoded in a transferable plasmid which may be rapidly disseminated in different strains and even different species of bacteria through horizontal gene transfer (7). For example, New Delhi metallo-β-lactamase-1 (NDM-1) is a newly added member of the MBL B1 subclass, and since its first identification in 2009 (8), NDM-1-positive bacteria have been identified on all continents except Antarctica (9).

NDM-1 shows broad activity toward all β-lactam antibiotics except monobactams (8). Because of the rapid dissemination of NDM-1-positive bacteria not only in clinical settings but also in community environments (10, 11), they pose a serious threat to our health care system. Since the identification of NDM-1 in 2009, much effort has been put into its biochemical characterization (12–15), structural and mechanical studies (16–30), and inhibitor screening (31–35). It is highly desirable to find some potent specific inhibitors for NDM-1 or even broad-spectrum inhibitors for all MBLs. However, because of the structural diversity of MBL active sites, the continuous evolving of these enzymes, and the fact that very few residues are tightly constrained by function, the design of broad-spectrum MBL inhibitors is very challenging (6). Therefore, the design of MBL inhibitors may be best achieved by mimicking the focused nature of interactions observed in both substrate recognition and the catalytic mechanism (6) as in the case of serine β-lactamases. Unfortunately, the catalytic mechanism of NDM-1 is subtle, and some detailed roles of active-site residues are still under debate, such as substrate coordination (23, 26, 36), the individual roles of the two zinc ions, the identity and source of hydroxide for nucleophilic attack (16, 23, 26, 27), the protonation of β-lactam nitrogen after C-N bond cleavage (16, 23, 26), the source of the next hydroxide for enzymatic turnover (23, 26), etc. Recent spectroscopic and structural investigations made significant progress in the study of the mechanisms of MBLs but did not reach a consistent conclusion on the hydrolysis of carbapenems by NDM-1 (29, 37), suggesting the need for further study toward a mechanism-based broad-spectrum inhibitor design.

In this work, we investigated the conformational changes of the active site through the combination of crystallography, site-directed mutagenesis, enzymatic measurements, and in silico simulation. Our results indicate that a dynamic active site and its conformational fluctuation promote the enzymatic activity of NDM-1, findings which may guide future inhibitor design.

RESULTS

NDM-1 in complex with hydrolyzed antibiotics.

In our previous study (16), we solved the first NDM-1 structure in complex with a hydrolyzed ampicillin and noticed that the distance between the two zinc ions (Zn1-Zn2 distance) at the active site was 4.6 Å, which was much longer than the distances in other MBLs. We proposed that this longer distance might correlate with the relatively lower enzymatic activity of NDM-1 than other B1 MBLs (16). Hence, the longer distance might be an intrinsic property of NDM-1 in the product binding state. However, there is also a possibility that the longer distance is a crystallographic artifact. To clarify whether the longer distance is an intrinsic product binding property or crystallographic artifact, we determined a series of NDM-1/Amp structures under different crystallization conditions. As shown in Table 1, we recrystallized NDM-1 in complex with ampicillin under the same conditions as we previously reported and got the highest-resolution NDM-1/product structures produced until now (1.00 Å). While the improvement in resolution revealed more accurate details, we noticed that the new structures were almost identical to a previously solved structure at the active site (Fig. 1A). We further compared the structures determined from crystals under different conditions, such as at different pHs and with different precipitants. As shown by the first four structures in Table 1, although the structures were crystallized at different pHs and even in high salt, they are almost identical and can be superimposed, with root mean square deviation (RMSD) values ranging from 0.1 to 0.26 Å for all visible Cα atoms (see Fig. S1A for overall structure superposition). In addition to the overall structures, the active-site confirmations of these structures are also very consistent with those of a previously solved structure, with the Zn1-Zn2 distances ranging from 4.57 to 4.61 Å in eight NDM-1 molecules while the Zn1 to OH (Zn1-OH) and OH-Zn2 distances averaged 2.00 ± 0.05 Å and 3.01 ± 0.04 Å (Fig. 1B), respectively.

TABLE 1.

NDM-1 in complex with hydrolyzed products

PDB accession no.	Complex	Resolution (Å)	M1-M2 distance(s) (Å)a	M1-OH distance(s) (Å)a	OH-M2 distance(s) (Å)a	Crystallization conditionsb	Reference and/or comment
5ZGE	Ampicillin	1.00	4.60, 4.61	2.08, 2.09	2.98, 2.94	0.1 M Bis-Tris, pH 5.5, 0.2 M Li2SO4, 25% PEG 3350	This work
5ZGP	Ampicillin	1.15	4.59, 4.59	1.97, 1.94	2.99, 3.04	0.1 M Bis-Tris pH 6.2, 0.2 M Li2SO4, 15% PEG 3350, 20 mg/ml ampicillin	This work
5ZGR	Ampicillin	1.15	4.59, 4.59	1.98, 1.99	3.01, 3.04	0.1 M HEPES, pH 7.3, 20% PEG 3350, 20 mg/ml ampicillin	This work
5ZGQ	Ampicillin	1.50	4.58, 4.57	1.99, 1.97	3.04, 3.04	0.1 M Tris-HCl, pH 7.5, 25% PEG 4000, 0.7 M (NH4)2SO4, 20mg/ml ampicillin	This work
4EY2	Methicillin	1.17	4.57, 4.58	1.95, 1.97	2.99, 3.00	0.2 M MgCl2, 25% PEG 3350, 0.1 M Bis-Tris, pH 5.5	22
4EYB	Oxacillin	1.16	4.54, 4.55	1.98, 1.96	2.94, 2.98	0.2 M MgCl2, 25% PEG 3350, 0.1 M Bis-Tris, pH 5.5	22
4EYF	Benzylpenicillin	1.8	4.63, 4.61	1.98, 1.93	3.11, 3.08	0.2 M MgCl2, 25% PEG 3350, 0.1 M Bis-Tris, pH 5.5	22
4RL2	Cefalexin	2.01	4.48, 4.55	1.83, 2.00	2.99, 2.78	28% (wt/vol) PEG 3350, 0.1 M Bis-Tris, pH 5.8, 0.2 M (NH4)2SO4	27
4H0D	Ampicillin	1.5	4.48, 4.49	2.08, 2.09	2.79, 2.79	0.2 M NaCl, 0.1 M HEPES, pH 7.5, 25 % (wt/vol) PEG 3350, 10 mM MnCl2	23
4HL2	Ampicillin	1.05	4.60, 4.60	1.94, 1.96	3.04, 3.04	0.2 M (NH4)2SO4, 0.1 M Bis-Tris, pH 5.5, 25% PEG 3350, 100 mM ampicillin	23
4RAW	Ampicillin	1.3	4.59, 4.60	2.05, 2.07	2.98, 2.97	0.2 M NaCl, 0.1 M Tris, pH 7.0, 30% PEG 3000, 5 mM CdCl2, 200mg/ml ampicillin	Cd1-Cd2 at active site M67V
4RL0	Cefuroxime	1.3	3.81, 3.83	2.01, 2.03	2.16, 2.17	30% PEG 3350, 0.1 M Bis-Tris, pH 6.0, 0.2 M Li2SO4	27; C8 Coo− did not coordinate to zn1
4EYL	Meropenem	1.9	4.05, 3.88			1 M trisodium cacodylate, 0.1 M sodium cacodylate, pH 6.5	22; C7 Coo− intercalates between two Zn ions
5N0H	Meropenem	1.9	4.01, 3.83	2.04, 2.08	2.66, 2.15	1 M trisodium cacodylate, 0.1 M sodium cacodylate, pH 6.5	Intercalates between two Zn ions
4RBS	Meropenem	2.4	4.00, 4.00			2.8 M sodium acetate, pH 7.0	C3 Coo− did not coordinate to Zn2
5YPK	Imipenem	2.0	3.87–4.15			28% (wt/vol) PEG 3350, 0.1 M Bis-Tris, pH 5.8, 0.2 M ammonium sulfate	29; EI2
5YPI	Imipenem	2.3	4.03–4.25			28% (wt/vol) PEG 3350, 0.1 M Bis-Tris, pH 5.8, 0.2 M ammonium sulfate	29; EI1
5YPL	Imipenem	1.8	3.89, 3.86			28% (wt/vol) PEG 3350, 0.1 M Bis-Tris, pH 5.8, 0.2 M ammonium sulfate	29; EP, intercalation
5YPM	Meropenem	2.15	3.93–4.30			28% (wt/vol) PEG 3350, 0.1 M Bis-Tris, pH 5.8, 0.2 M ammonium sulfate	29; EI1
5YPN	Meropenem	2.12	4.08, 4.31			28% (wt/vol) PEG 3350, 0.1 M Bis-Tris, pH 5.8, 0.2 M ammonium sulfate	29; EI2

^aDifferent values represent different monomers. M1 and M2 are metal ions 1 and 2, respectively.

^bPEG, polyethylene glycol.

FIG 1.

Active-site conformation of NDM-1 in complex with hydrolyzed ampicillin crystallized at pH 5.5 (A) and superposition of the active sites of NDM-1 crystallized under different conditions (B). Active-site residues and hydrolyzed ampicillin are shown as ball-and-stick models. Carbons in active residues are shown in gray, and those in hydrolyzed ampicillin molecules are shown in green and cyan in panels A and B, respectively. Zinc ions and hydroxide are shown as gray and red balls, respectively. Coordination bonds between zinc ions and active-site residues are shown as yellow dashed lines. The Zn1-OH, OH-Zn2, and Zn1-Zn2 distances are 2.1, 3.0, and 4.6Å (shown as red dashed lines), respectively.

After our first report of the NDM-1/ampicillin (Amp) structure, several structures of NDM-1 in complex with other hydrolyzed antibiotics were also reported, as listed in Table 1. For NDM-1 in complex with methicillin (PDB accession number 4EY2 [22]), oxacillin (4EYB [22]), benzylpenicillin (4EYF [22]), cephalexin (4RL2 [27]), and ampicillin (4H0D [23], 4HL2 [23], and 4RAW), the overall structures are very similar, with RMSD values ranging from 0.10 to 0.36 Å for all the superimposed Cα atoms (Fig. S1B). For the active sites of these structures, Zn1-Zn2 distances ranged from 4.48 to 4.63 Å (average value of 4.56 ± 0.05 Å) while the distances of Zn1-OH and OH-Zn2 were 1.98 ± 0.07 Å and 2.96 ± 0.10 Å (Fig. S2A), respectively. The active-site conformations observed in these structures are almost identical to those determined in our studies.

However, for structures of NDM-1 in complex with hydrolyzed meropenem (PDB accession number 4EYL [22], 5N0H, and 4RBS), the Zn1-Zn2 distances were a little bit shorter (ranging from 3.83 to 4.05 Å with an average of 3.96 ± 0.09 Å). Inspection of these structures revealed that the hydrolyzed meropenem did not bind the active site in the same way as other hydrolyzed antibiotics (Fig. S2A), with either the newly generated carboxylate group intercalating between the two zinc ions (PDB accession numbers 4EYL and 5N0H) or the side ring carboxylate group not coordinated to Zn2 (PDB accession numbers 4EYL and 4RBS) (Fig. S2B). The intercalated conformation might be a rare binding state during product release, which was also suggested by the quantum mechanics/molecular mechanics (QM/MM) calculation (38) as a rare inhibition state.

The structure of NDM-1 in complex with a reaction intermediate of cefuroxime (PDB accession number 4RL0 [27]) showed shorter Zn1-Zn2 distances (3.81 and 3.83 Å for two monomers) and OH-Zn2 distances (2.16 and 2.17 Å for two monomers). Active-site inspection revealed that the newly generated carboxylate group did not coordinate to Zn1 as other products did. The shorter distance observed in this structure might be attributed to the reaction intermediate, which needs further investigation.

Overall, in all NDM-1 structures with the hydrolyzed antibiotics coordinated to the active site in a product binding state, the Zn1-Zn2, Zn1-OH, and OH-Zn2 distances are very consistently found to be 4.6, 2.0, and 3.0 Å, respectively. This conformation should be an intrinsic NDM-1/product binding state and not a crystallographic artifact.

NDM-1 in native form.

Since the identification of NDM-1 in 2009, many structures of NDM-1 in native form and in complex with antibiotics have been solved. Some of the reported structures showed no metal, monometal (19), or no properly coordinated residue-metal interactions at the active site (39). NDM-1 with a Zn-Cd substitution (20) (PDB accession number 3ZR9) was obtained at pH 7.5 with a Zn-Cd distance of 3.64 Å, while another structure (18) (PDB accession number 3SPU) obtained at pH 8.5 reported 5 molecules in the asymmetric unit and Zn1-Zn2 distances ranging from 3.56 to 3.97 Å. It was then proposed that the Zn1-Zn2 distance was affected by different pHs (23). To reveal a clear and more accurate active site of native NDM-1 and what kind of factors can affect the active site conformation, we determined a series of native NDM-1 structures under different conditions, including crystal packing, buffer components, and pH variations.

Effect of crystal packing.

In our search for native NDM-1 crystallization conditions, two crystal forms were obtained, with form 1 retaining one molecule in the asymmetric unit and diffracted up to 0.95-Å resolution while form 2 harbored two molecules and diffracted between 1.05 and 1.55 Å, depending on the crystallization precipitants (Table 2). For crystals of form 2, although they were obtained at different pHs and with different buffer components (HEPES at pH 7.0 and pH 7.5, imidazole at pH 7.5, and Tris at pH 8.0), the active-site conformations were strikingly identical (Fig. 2A and S3) with Zn1-Zn2, Zn1-OH, and OH-Zn2 distances of 3.39 to 3.43 Å, 1.93 to 2.11 Å, and 1.95 to 2.13 Å, respectively (Table 2). The Zn1-Zn2 distance is also consistent with the one measured by an extended X-ray absorption fine-structure study (3.38Å) (21). However, when we compared the structures of NDM-1 in forms 1 and 2 obtained in the same buffer and at the same pH (imidazole, pH 7.5; PDB accession numbers 5ZGZ and 5ZH1), we noticed that the Zn1-Zn2 distance in form 1 was 3.62 Å while the distances in form 2 for two different monomers were 3.41 and 3.42 Å. The only difference between these two crystals is crystal packing, and so the Zn1-Zn2 distance is clearly affected by crystal packing.

TABLE 2.

Active-site conformations of native NDM-1 structures

PDB accession no.	Comment	Resolution (Å)	M1-M2 distance (Å)a	M1-OH distance (Å)a	OH-M2 distance (Å)a	Coordinate error (Å)	Reference
5XP6	Native, form 1, succinate, pH 5.5	0.95	3.95	1.92	2.55	0.009	This work
5ZGI	Native, form 1, succinate, pH 6.5	0.98	3.78	1.93	2.35	0.011	This work
5ZGX	Native, form 1, succinate, pH 7.5	0.95	3.59	1.96	2.09	0.010	This work
5ZGW	Native, form 1, succinate, pH 7.5 (back soaked from pH 5.5)	0.95	3.53	1.97	2.02	0.011	This work
5ZGF	Q123G, form 1, succinate, pH 7.5	1.20	3.84	1.90	2.43	0.029	This work
5ZGZ	Native, form 1, imidazole, pH 7.5	0.95	3.62	1.98	2.04	0.011	This work
5ZGY	Native, form 1, Bis-Tris, pH 7.5	0.95	3.49	1.96	2.00	0.011	This work
5ZGU	Native, form 2, HEPES, pH 7.0	1.55	3.43, 3.42	2.11, 2.11	2.13, 2.12	0.049	This work
5ZGT	Native, form 2, HEPES, pH 7.5	1.20	3.42, 3.39	2.05, 2.04	2.07, 2.05	0.020	This work
5ZH1	Native, form 2, Imidazole, pH 7.5	1.05	3.41, 3.42	1.93, 1.95	1.96, 1.95	0.022	This work
5ZGV	Native, form 2, Tris, pH 8.0	1.15	3.43, 3.40	2.05, 2.06	2.06, 2.07	0.021	This work
3SPU	Ammonium sulfate, pH 8.5	2.1	3.84, 3.88, 3.97, 3.84, 3.56	2.54, 1.93, 2.11, 2.08, 1.98	2.46, 2.60, 2.20, 2.37, 2.09		18
3ZR9	Zn-Cd, pH 7.5	1.91	3.64	1.93	2.38		20
5JQJ	Mutant, MgSO4, pH 6.75	1.67	3.61	1.89	2.21
4TZE	NDM-5, pH 7.5	1.57	3.49, 3.56	2.05, 2.05	2.07, 2.07
4TZF	NDM-8, Bis-Tris, pH 5.5	1.22	3.70	2.00	2.13
4TYF	NDM-4, pH 7	1.10	3.54	1.89	2.03
4RM5	D124N, Tris, pH 7.5	2.1	4.01, 4.10, 4.10, 4.11	2.61, 2.53, 2.85, 2.58	2.60, 2.72, 2.72, 2.57		27

^aDifferent values represent different monomers. M1 and M2 are metal ions 1 and 2, respectively. Significant values for the Q123G mutant are in boldface.

FIG 2.

Active-site conformations of native NDM-1. Active-site residues are shown as ball-and-stick models while zinc ions and hydroxide at the active site are shown as spheres. Coordination bonds are shown as yellow dashed lines. (A) Superposition of the active sites of 8 native NDM-1 molecules crystallized in form 2. The active-site conformations are strikingly identical among structures from different crystallization conditions, especially the Zn1-OH, OH-Zn2, and Zn1-Zn2 conformations and distances. (B) Superposition of active sites of native NDM-1 crystallized in form 1 at pH 5.5, 6.5, and 7.5 in succinate buffers. The zinc ions, hydroxide ion, and carbon atoms in residue H250 are shown in green, cyan, red, and magenta for crystals at pH 5.5, 6.5, 7.5, and 7.5 backsoak (crystals of pH 5.5 soaked at pH 7.5), respectively. Residue H250 pushes Zn2 to the active center along with a pH increase. Residue D124 adjusts its side chain carboxylate group accordingly to coordinate to Zn2.

Effect of buffer components.

Next, we compared the structures of form 1 obtained at pH 7.5 but prepared with different buffer components (imidazole, Bis-Tris, and succinate). Despite being crystallized at the same pHs, these structures showed different Zn1-Zn2 distances of 3.62 Å in imidazole, 3.59 and 3.53 Å for two structures solved individually in succinate, and 3.49 Å in Bis-tris (Table 2). Although the differences are subtle, they can be repeatedly observed, indicating that different buffer components do affect the active site conformations.

Effect of different pHs.

We then examined the pH effect on the active site of NDM-1. To avoid the buffer component effect, all the NDM-1 structures were determined in the same buffer components but at different pHs. The crystals could be easily obtained by seeding in Bis-Tris buffer at different pHs but with unidentifiable electron densities at the active sites at a lower pH (probably Bis-Tris, but the density could not be clearly modeled). To avoid potential perturbation of Bis-Tris to the active site, we then optimized the crystallization in succinate buffer and got crystals at pH 5.5, 6.5, and 7.5. Although succinate was observed at the active site in crystals at pH 5.5 and 6.5, it can be clearly modeled with one hydroxyl group coordinated to Zn2 and substituting an apical water. This kind of interaction did not affect the other coordination bonds at the active site (Fig. S4). Comparison of structures in succinate at pH 5.5, 6.5, and 7.5 (Fig. 2B) revealed a prominent decrease of the Zn1-Zn2 distances from 3.95 and 3.78 to 3.59 Å (Table 2). To further confirm this pH effect, crystals obtained at pH 5.5 were soaked in a cryo-protectant at pH 7.5, which yielded a structure with the Zn1-Zn2 distance changed from 3.95 to 3.53 Å, indicating a consistent pH effect on the active site. In addition to the clear tendency of Zn1-Zn2 distances changing according to different pHs, we also observed a decrease in the OH-Zn2 distance along with the pH increase, which is consistent with the Zn1-Zn2 changes (Table 2). However, Zn1-OH distances remained almost identical in all crystals that we obtained and that were reported by others (1.99 ± 0.07 Å). Although in some structures this hydroxide was modeled as a water molecule, the shorter distance with Zn1 in all of the structures indicates that it should be a hydroxide. Furthermore, QM/MM study showed that if a water molecule coordinates between the two zinc ions, the distance of Zn1-Zn2 will be 5.6 Å (23), which is much longer than the distances of all currently observed active sites. The observed shorter Zn1-OH distance is also consistent with the mostly accepted idea that this hydroxide is mainly activated by Zn1 and acts as a nucleophile during the C-N bond break of β-lactam hydrolysis.

MD simulations.

The distances of Zn1-Zn2 observed in different NDM-1 structures were also assessed by in silico calculations. Molecular dynamics (MD) simulations starting from the enzyme/hydrolyzed ampicillin structures (PDB accession numbers 5ZGE, 5ZGP, and 5ZGR) were run to compare with experimental results. The energy-minimized structures under different pH states presented similar Zn1-Zn2 distances of 4.7 Å (pH 5.5), 4.7 Å (pH 6.2), and 4.8 Å (pH 7.3). These results are very close to the experimentally determined distance of 4.6 Å and do not show pH dependence. The differences in Zn1-Zn2 distances observed in native NDM-1 at different pHs were also supported by our molecular dynamics analyses. Energy minimization showed different Zn1-Zn2 distances of 3.9 Å (pH 5.5), 3.7 Å (pH 6.5), and 3.5 Å (pH 7.5). Despite subtle differences, the tendency of Zn1-Zn2 distances to change is consistent with the experimentally observed values. Previous QM/MM studies also calculated Zn1-Zn2 distances of 3.58 Å (12), 3.60 Å (23), and 3.62 Å (40) in native NDM-1 molecules although pH values were not reported in these studies. We also modeled an intact ampicillin molecule into the active site of native NDM-1 and performed MD simulations at different pHs, which yielded Zn1-Zn2 distances of 4.8Å (pH 5.5), 3.8Å (pH 6.5), and 3.4Å (pH 7.5). The consistent decrease in Zn1-Zn2 distances in NDM-1/ampicillin models along with a pH increase is similar to that observed for native unbound NDM-1 and correlates well with the enzymatic activity of NDM-1 at different pH values (discussed later).

Overall, our results reveal that the longer distance in enzyme/hydrolyzed penam or cephalosporin complexes is an intrinsic feature of NDM-1 and that Zn1-Zn2 distance in the native enzyme is affected by crystal packing, pH, and buffer components. Next, we examined if the Zn1-Zn2 distance variation affects the enzymatic activity of NDM-1.

Enzymatic study of NDM-1.

We first measured the enzymatic activity of wild-type NDM-1 at different pHs (5.5, 6.5 and 7.5) which were prepared with succinate in all cases to avoid the influence of buffer components. Imipenem was used as the substrate, and the k_cat values of NDM-1 at pH 5.5, 6.5, and 7.5 were 1,283, 2,060, and 3,960 s⁻¹, respectively, representing roughly a 2-fold change for each pH increase (K_m and k_cat/K_m parameters are listed in Table 3). The same pattern of increase was also observed using a buffer prepared with Bis-Tris in our assays and in other reports (23) although the specific values are different depending on substrates and assay conditions. It is interesting that along with the pH increase, the enzymatic activity increases while Zn1-Zn2 distance decreases. In general, buffer pH may affect enzyme activity through charge changes on residues at the active site, which may influence all the steps involved in enzyme catalysis, including substrate binding, intermediate formation, and product release. The facts that almost identical zinc coordination bonds, in particular the Zn1-OH distances (1.99 ± 0.07 Å), and geometries are kept in all native structures obtained in different pHs (Table S3) and that identical product-enzyme interactions are also observed in structures at different pHs (Fig. 1) indicate that the charge effect on the active-site residues might be very subtle or indirect. The only obvious conformational change observed in these native structures at different pHs is the Zn1-Zn2 distance variation. Through careful structural comparison, we noticed that the Zn2 coordination residue H250 progressively pushes Zn2 toward the active center as the pH changes from low to high (Fig. 2B). Residue H250 locates at the tip of loop 12, which is relatively mobile among the structures at different pHs and has several charged residues at the other end. Through this long-range interaction, we speculate that pH may fine-tune Zn2 position and hence affect the enzymatic activity of NDM-1.

TABLE 3.

Enzymatic characterization of wild-type and mutant NDM-1using imipenem as substrate

Protein (buffer)	pH	kcat (s−1)	Km (μM)	kcat/Km (M−1 s−1)
WT (succinate)	5.5	1,283 ± 85	337 ± 44	3.82 × 106
	6.5	2,060 ± 245	523 ± 103	3.94 × 106
	7.5	3,960 ± 160	1415 ± 72	2.80 × 106
WT (HEPES)	7.5	1,815 ± 104	560 ± 52	3.24 × 106
A121F (HEPES)	7.5	221 ± 5.6	55 ± 4.6	4.02 × 106
Q123G (HEPES)	7.5	686 ± 38	201 ± 28	3.41 × 106

Mutagenesis studies to modulate the enzymatic activity and active-site conformation.

To further validate this speculation and test if modulation of the Zn1-Zn2 distance may affect enzyme activity, site-directed mutants of NDM-1 were generated, and their enzymatic activities were measured. To avoid possible perturbation of the active-site coordination and substrate binding, only second-shell residues of the active site were mutated. Through careful structural inspection, residues A121 and Q123 were selected for mutagenesis studies. While both A121F and Q123G showed lower enzymatic activities, only Q123G was successfully crystallized. As shown in Table 3, the k_cat value for the Q123G mutant was 686 s⁻¹, much lower than that of wild-type NDM-1 under the same assay condition (1,814 s⁻¹).

The active-site flexibility of the Q123G mutant was evaluated experimentally by measuring the Zn1-Zn2 distance in high-resolution crystal structures. We successfully determined the structure of the Q123G mutant at pH 7.5 in succinate buffer at a resolution of 1.2Å. The overall structure of Q123G superimposed well on that of wild-type NDM-1, with an RMSD value of 0.44 Å for all Cα atoms, and only minor flexibility in L3 loop was observed (Fig. 3A). The Q123G substitution did not perturb the active-site conformations or the interactions around Q123 in wild-type NDM-1 (Fig. 3B). The Zn1-Zn2 distance in the Q123G mutant was measured to be 3.84 Å, a value longer than that of wild-type NDM-1 at pH 7.5 (3.59 and 3.53Å for two structures solved individually under the same conditions) and near that of wild-type NDM-1 at pH 6.5 (3.78Å). The enzymatic activity of the Q123G mutant is much lower than that of wild-type NDM-1 at pH 7.5 but near the proportional value of NDM-1 at pH 6.5. In our enzymatic assay, imipenem was used as the substrate. Measuring from the recently published structure of NDM-1 in complex with imipenem (29), the side chain of residue Q123 makes no contacts with imipenem and therefore would not affect its binding with the active site or the release of hydrolyzed product. The activity decrease of the Q123G mutant should be solely attributed to the increase of Zn1-Zn2 distance, validating our notion that modulation of the Zn1-Zn2 distance will affect the enzymatic activity of NDM-1.

FIG 3.

Structural comparison between wild-type and Q123G mutant NDM-1 molecules. Q123G mutant NDM-1 was superimposed onto wild-type NDM-1 (with succinate, pH 7.5) (PDB accession number 5ZGW). Active-site residues are shown as ball-and-stick models while zinc ions and hydroxide at the active site are shown as spheres. Coordination bonds are shown as yellow dashed lines. (A) Overall structural superimposition. L3 loops show flexibility between these two structures. (B) Active-site superimposition. The zinc ions, hydroxide ion, and carbon atoms in residue H250 are shown in cyan and magenta for Q123G and wild-type NDM-1, respectively. The Q123G substitution did not obviously perturb the active-site conformation. The Zn1-Zn2 distance in Q123G is longer than that in wild-type NDM-1.

DISCUSSION

As revealed in the crystallographic structures, the active site of NDM-1, especially that of Zn1-Zn2, is in a dynamic or oscillation state, depending on environmental factors such as crystal packing, pH, and buffer components. This was also observed in a single crystal (18) (PDB accession number 3SPU) where there are five molecules in the asymmetric unit with five different Zn1-Zn2 distances ranging from 3.56 to 3.97 Å. Our enzymatic assays clearly showed that the Zn1-Zn2 distance correlates to the enzymatic activity of NDM-1, with the shorter Zn1-Zn2 distance in native NDM-1 associated with higher enzymatic activity. Interestingly, the Zn1-Zn2 distances in enzyme/product complexes are all at the same value of 4.6 Å. Thus, the relation of distance to enzymatic activity can be stated as follows: a bigger fluctuation (shorter Zn1-Zn2 distance) will favor higher enzymatic activity, or, in other words, the fluctuation of Zn1-Zn2 promotes the enzymatic activity of NDM-1. The oscillation of Zn1-Zn2 distance during the enzymatic turnover was observed in mechanism studies of the subclass B3 MBL L1 (41) and QM/MM calculations of NDM-1 (24, 36), but here we determined for the first time that such oscillations affect enzymatic turnover.

In a previous mutagenesis and structural study of another B1 MBL, β-lactamase II from Bacillus cereus, BCII, Vila and coworkers also noticed that the active-site conformational changes correlated with the enzymatic activity of BCII and proposed that fine-tuning of the Zn2 position was responsible for the change in BCII activity (42–44). Comparison among the wild-type and mutant BCII structures shows that the Zn1-Zn2 distances in activity-impaired BCII (BCII/HD, a double mutant; PDB accession number 2NYP), wild-type BCII (1BCII; 4C09), and activity-enhanced BCII (M5; 3FCZ) are 4.73, 3.50 and 3.85 (for different monomers), and 3.32 Å, respectively. The relationship between enzymatic activity and Zn1-Zn2 distance in BCII well supports our notion that in NDM-1 a shorter Zn1-Zn2 distance favors higher activity. Although fine-tuning of the Zn2 position can also explain the structure-activity relationship, Zn1-Zn2 distance variation is more specific. Considering that Zn1 also contributes to the conformational changes, albeit less than Zn2 (Fig. 2B), Zn1-Zn2 distance variation seems more accurate to explain current observations.

It is well known that conformational fluctuations are essential for substrate binding and product release and can even be a rate-limiting step in enzymatic reactions. However, the role of conformational fluctuation along the reaction pathway was not well recognized. In this study, through series of crystallographic, biochemical, and biophysical studies and simulation analyses, we presented a dynamic active site of NDM-1 which correlates well with its enzymatic activity and proved that conformational fluctuation of the active site facilitates the enzymatic reaction. This phenomenon was also observed in dihydrofolate reductase, where conformational fluctuation of an active-site loop enhances its activity (45). In contrast to the covalently bonded active-site residues in dihydrofolate reductase, the conformational fluctuation of the active site in NDM-1 is mainly presented by two noncovalently bonded metal ions. The active-site fluctuation may facilitate the enzyme's sampling of high-energy conformational substrates that are conducive to forming the transition state and promoting the enzymatic reaction (45). Our combinational studies reveal that modulation of the active-site fluctuation affects the enzymatic activity of NDM-1, which may provide clues for future inhibitor design. For example, Zn1-Zn2 fluctuation may be restricted to a certain state by coordinating chemicals such as substrate analogues, as shown in a recent study (32, 46).

In summary, through the combination of crystallographic determination, biochemical and biophysical measurements, site-directed mutagenesis, and in silico simulation, our results reveal the following: (i) that the longer distance of Zn1-Zn2 (4.6 Å) in an enzyme/product complex is an intrinsic feature of NDM-1; (ii) that the Zn1-Zn2 distance in native NDM-1 varies depending on external factors such as crystal packing, pH, and buffer components; (iii) that the variation of Zn1-Zn2 distance affects the enzymatic activity of NDM-1, with a shorter distance favoring higher enzymatic activity.

MATERIALS AND METHODS

Protein expression, purification, and site-directed mutagenesis.

Wild-type NDM-1 (G29-R270) was cloned into a His-maltose binding protein (MBP) vector to facilitate protein folding and purification. The fusion construct was expressed in Escherichia coli BL21(DE3), and the cells were allowed to grow to an optical density at 600 nm (OD₆₀₀) of 0.6, followed by induction with 0.5 mM isopropyl-β-d-thiogalactopyranoside (IPTG) overnight at 16°C. After centrifugation, the cell pellet was suspended in lysis buffer consisting of 20 mM HEPES, pH 7.4, and 0.5 M sodium chloride for sonication. The supernatant was loaded onto a nickel-nitrilotriacetic acid (NTA) chromatography column (GE Health Care), and NDM-1 was purified by an imidazole gradient. Effluents from the nickel column were further purified by an MBP column. The fusion tag was cleaved with tobacco etch virus (TEV) protease and separated from NDM-1 by passage through a Ni-NTA column again. NDM-1 was in the flowthrough fraction and was further purified by a Q-Sepharose ion exchange column (GE Health Care) at pH 7.0 with a sodium chloride gradient from 0 to 0.3 M. The purified protein was buffer exchanged into 50 mM NaCl and concentrated to 100 mg/ml (measured at an OD₂₈₀ using an extinction coefficient of 27,960) for later use.

For site-directed mutagenesis, wild-type NDM-1 was used as the template, and the mutations were introduced using the quick-change method. The mutants were confirmed by sequencing, and the expression and purification of the mutant proteins were performed in the same manner as that for wild-type NDM-1.

Crystallization, diffraction data collection, and structure refinement.

Wild-type NDM-1 crystals were screened by the sitting-drop vapor diffusion method. NDM-1 in complex with ampicillin crystals was observed under several conditions and further optimized by the hanging-drop method. Crystals for native NDM-1 were obtained in two forms, with form 1 having one molecule per asymmetric unit and form 2 having two molecules per asymmetric unit. The detailed crystallization and cryo-protection conditions are listed in Table S1 in the supplemental material. All crystals were cryo-protected and then flash frozen in liquid nitrogen. The diffraction data were collected at 100 K on station BL17U1 or BL19U1 at the Shanghai Synchrotron Radiation Facility and processed with HKL2000 software (47). The structures were solved by the molecular replacement method using the previously solved NDM-1/Amp complex (16) (PDB accession number 3Q6X) as the search model. The models were refined with Refmac (48, 49) in the CCP4 suite (50) and then cycled with rebuilding in Coot (51). TLS (translation-libration-screw) refinement (52) was incorporated into the later stages of the refinement process. Solvents were added automatically in Coot and then manually inspected and modified. The final models were analyzed with MolProbity (53) and showed that almost all amino acid residues were in favored regions of the Ramachandran plot while residue D90 was in a forbidden region, as noticed previously (16). Data collection and model refinement statistics for all data sets are summarized in Table S2.

Enzymatic activity measurement.

Hydrolysis of antibiotics by wild-type and mutant NDM-1 was monitored by detecting a reduction in the absorbance that resulted from the opening of the β-lactam ring using imipenem as the substrate (39). Activity for the pH dependence of wild-type NDM-1 was measured in a reaction buffer containing 50 mM succinate (pH 5.5, 6.5, or 7.5), 100 mM NaCl, 50 μM ZnCl₂, 10 μg/ml bovine serum albumin (BSA), 50 nM NDM-1, and various concentrations of imipenem. The comparison of enzymatic activities between wild-type and mutant NDM-1 molecules was assayed in a reaction buffer containing 50 mM HEPES (pH 7.5), 100 mM NaCl, 100 μM ZnCl₂, 10 μg/ml BSA, 50 nM enzymes, and various concentrations of imipenem. All the experiments were performed at 25°C, and the A₃₀₀ value was immediately measured using a Perkin Elmer EnSpire spectrophotometer. Kinetic parameters were determined by plotting the initial velocities against substrate concentrations and curve fitting with GraphPad Prism, version 5, software. All measurements were repeated three times, and the average values are presented.

Simulation.

The native models were taken from the crystal structures solved in succinate buffer at pH 5.5, 6.5, and 7.5. The enzyme/product models were taken from structures solved at pH 5.5 (PDB accession number 5ZGE), pH 6.2 (5ZGP), and pH 7.3 (5ZGR). The structure of the NDM-1 in complex with hydrolyzed ampicillin was superimposed onto the native models. The coordinates of hydrolyzed ampicillin were kept, and the chemical structure was modified to generate the substrate ampicillin. Then we determined the initial models of three enzyme-substrate systems at different pHs. All the water, solvent, and succinate molecules in the crystal structures were removed from the initial models, and the shared coordination hydroxide was kept. Polar hydrogens were added to the systems using the H++ web server (54).

MD simulations were carried out using Amber 12 (55). The script MCPB.py (56) was applied to generate parameters of the above models for further MD simulations. The ff14SB force field was used for the protein systems (57). Each model was then solvated in a periodic box surrounded by TIP3P water molecules at a distance of no less than 10 Å (58). Counter ions were added to maintain neutral charge of systems. To remove possible poor contacts between the solute and solvent, energy minimizations using a combination of the steepest descent and conjugated gradient method were performed. For models with ampicillin, each system was subjected to a slow heating process for 500 ps from 0 to 300 K and then equilibrated for 500 ps under an NPT (fixed number of atoms, pressure, and temperature) ensemble at a constant temperature of 300 K. Finally, production MD simulations were conducted for 20 ns. The Sander program was used to conduct the MD simulation at constant temperature (300 K) and pressure (1.0 atm) with a time step of 1 fs. The nonbonded cutoff was set to 12.0 Å, and electrostatic interactions were calculated using the particle-mesh Ewald method (59). The SHAKE algorithm (60) was used to constrain bonds involving hydrogen atoms.

Accession number(s).

The coordinates and structure factors were deposited in the Protein Data Bank under the following accession numbers: 5ZGE, 5ZGP, 5ZGQ, 5ZGR, 5ZGU, 5ZGT, 5ZH1, 5ZGV, 5ZGZ, 5ZGY, 5XP6, 5ZGI, 5ZGX, 5ZGW, and 5ZGF.