The role of a conserved Asp-Asp pair in the structure, function, and inhibition of CTX-M Class A β-lactamase

Abstract

The Asp233-Asp246 pair is highly conserved in Class A β-lactamases, which hydrolyze β-lactam antibiotics. Here, we characterize its function using CTX-M-14 β-lactamase. The D233N mutant displayed decreased activity that is substrate-dependent, ranging from k_cat/K_m reduction of 20% for nitrocefin, to 6-fold for cefotaxime. In comparison, the mutation reduced the binding of a known reversible inhibitor by 10-fold. The mutant structures showed movement of the 213-219 loop and the loss of the Thr216-Thr235 hydrogen bond, which was restored by inhibitor binding. Mutagenesis of Thr216 further highlighted its contribution to CTX-M activity. These results demonstrate the importance of the aspartate pair to CTX-M hydrolysis of substrates with bulky side chains, while suggesting increased protein flexibility as a means to evolve drug resistance.

Introduction

Pairs of hydrogen-bonding carboxylate groups, especially between aspartate residues, are commonly found in protein structures[1-3]. While such pairs do not form favorable interactions when solvent-exposed and at physiological pH (due to electrostatic repulsion), side chain carboxylates with perturbed pK_a values inside the protein microenvironment can share a proton and establish strong hydrogen bonding contacts, including some very short (~2.5 Å) hydrogen bonds (HBs). Hydrogen-mediated bicarboxylates have been studied for their role in enzyme catalysis, including the reaction catalyzed by HIV protease[2]. However, a possible structural function of such bicarboxylate pairs has been much less investigated.

Pairs of highly conserved aspartate residues have been observed across many Class A β-lactamases; bacterial enzymes catalyzing the hydrolysis of β-lactam antibiotics, e.g., ampicillin (Figure 1)[4-6]. One such pair consists of Asp233 and Asp246, two residues located outside, but in close proximity to, the active site. In many of these enzymes, the Asp233-Asp246 HB length has been shown to be short, in the range of 2.5-2.6 Å[6-10]. Additionally, Asp233 resides on the β3 strand that forms part of the active site, including the oxyanion hole crucial to the stabilization of the reaction transition states. These structural features, together with the sequence conservation[7, 10], suggest that the Asp-Asp HB may play a special role in the structure and function of Class A β-lactamases, even though it is not directly involved in substrate binding or the chemical reaction.

Figure 1. Chemical structures of β-lactam substrates and a tetrazole-based β-lactamase inhibitor.

The enzymatic mechanism of Class A β-lactamases includes an acylation and deacylation step, both of which involve a proton transfer event that is facilitated by general-acid/base catalysis[11, 12]. Following substrate binding, the acylation reaction begins with a nucleophilic attack of the substrate β-lactam ring by Ser70, producing a covalent acyl-enzyme complex. During the deacylation step, a water molecule reacts with the acyl-enzyme linkage, leading to the release of the hydrolyzed product and regeneration of the free enzyme. Due to its importance in bacterial antibiotic resistance, the hydrolysis reaction catalyzed by Class A β-lactamases has been intensely studied, yet many mechanistic and structural details remain elusive[13-16].

CTX-M Class A β-lactamases are the most common clinically observed extended spectrum β-lactamases (ESBLs) and are capable of hydrolyzing third-generation cephalosporins (e.g., cefotaxime, Figure 1) and other common β-lactam antibiotics such as penicillins[17, 18]. They provide a well-characterized model system for investigating Class A β-lactamase structure and catalysis. In our recent study of CTX-M β-lactamase using sub-Angstrom resolution X-ray crystallography, we have observed a short HB (2.47 Å) between Asp233 and Asp246, including a hydrogen atom that is apparently shared by the two carboxylate groups[19]. In order to study the importance of the Asp-Asp pair in Class A β-lactamase structure and function, we constructed several CTX-M mutants, and used X-ray crystallography and biochemical studies to gain a better understanding of the structural and functional contribution of this conserved interaction outside the active site. Interestingly, our results indicate that the Asp-Asp HB is important for stabilizing both the 213-219 loop and β3 strand in Class A β-lactamases, which in turn are critical features for maintaining the active site integrity required for substrate recognition and catalysis. These results provide valuable insights into the contribution of Asp-Asp pairs to protein structure and function, as well as the role of protein flexibility in the evolution of drug resistance.

Results

Functional analysis of the D233N mutation

To investigate the Asp233-Asp246 interaction, particularly its contribution to enzyme stability and activity, several mutants (D233N, D246N, D246I) were constructed using site-directed mutagenesis, in order to disrupt the short HB between the Asp-Asp pair. Both Asp233 and Asp246 were mutated to asparagine but D246N failed to fold properly during protein expression and we were only able to purify the D233N mutant. We were primarily interested in Asp→Asn mutations in order to introduce minimal changes, while still maintaining a HB between these two residues, so that we may be able to determine whether the short HB between Asp233 and Asp246 plays any special role in protein structure and function compared with a standard HB. Because Asp246 is replaced by isoleucine in TEM-1 and SHV-2 Class A β-lactamase[20, 21], we next investigated the D246I mutant, which behaved similarly to the D246N mutant with the inability to properly fold. Since Asp246 is buried close to the tightly packed protein core, it is possible that both the D246N and D246I mutations may have caused steric clashes with surrounding residues. Although the result of D246N initially appeared surprising, close examination of the CTX-M wild type (WT) structure suggests that the additional hydrogen atom on Asn246-Nδ2 may clash with the Asn246-Cα group if the asparagine side chain adopts the same conformation as Asp246. The WT structure also indicates that smaller side chains, such as Ala, can be tolerated at this position, although such mutants were not pursued in the current study.

Our studies subsequently focused on the CTX-M-14 D233N mutant. In order to gain a clear understanding of the importance of the Asp-Asp interaction with respect to enzyme function, we performed biochemical kinetic studies of the mutant and compared them with the WT CTX-M-14 enzyme (Table 1). The activity of the enzyme was tested using nitrocefin, ampicillin, or cefotaxime as substrate (Figure 1). Whereas nitrocefin is a standard substrate for assaying β-lactamase activity and may represent the early-generation cephalosporins, cefotaxime is a third-generation cephalosporin and a so-called ESBL antibiotic. Its bulky side chain prevents it from being hydrolyzed by narrow-spectrum β-lactamases such as TEM-1 (Figure 1). However, CTX-M and other ESBLs have evolved to accommodate these antibiotics as substrates and efficiently catalyze their hydrolysis[17]. Interestingly, the D233N mutant displayed K_m and k_cat values similar to WT when tested using the nitrocefin substrate. On the other hand, when tested with ampicillin and cefotaxime as substrate, the D233N mutant exhibited lower activity compared to the WT, with respectively a 3-fold and 6-fold reduction in k_cat/K_m. We considered whether these substrate-dependent differences in k_cat/K_m could be attributed to the effects of the D233N mutation in accommodating different substrates. Interestingly, for ampicillin, the change in activity derived primarily from a ~2-fold decrease in k_cat, whereas for cefotaxime, a ~7-fold increase in K_m was observed. For the hydrolysis reaction catalyzed by Class A β-lactamases, the deacylation process is usually the rate-limiting step. In these cases, k_cat and k_cat/K_m reflect the efficiency of deacylation and acyl-enzyme formation respectively, as described in previous studies[22, 23]. It appears that the D233N mutation had some small effect on the deacylation step for ampicillin, but a rather significant effect on the acylation reaction for cefotaxime (i.e., both substrate binding and the formation of the acyl-enzyme linkage).

Table 1.

Biochemical characterization of CTX-M-14 β-lactamase and mutants.

	Nitrocefin			Ampicillin			Cefotaxime			1
	Km (μM)	kcat (s−1)	kcat/Km (μM−1s−1)	Km (μM)	kcat (s−1)	kcat/Km (μM−1s−1)	Km (μM)	kcat (s−1)	kcat/Km (μM−1s−1)	Ki (μM)
WT	41.2±10.3	1262.7±53.7	30.8±3.9	44.2±14.5	47.2±4.3	1.22±0.60	83.8±2.3	50.2±1.1	0.60±0.01	21.4±7.6
D233N	46.7±2.3	1138.7±146.8	24.6±4.2	54.6±23.1	19.0±7.7	0.36±0.06	631.8±282.8	57.0±7.3	0.10±0.03	226.6±36.6
T216A	187.8±16.1	4242.9±234.3	22.6±0.7	376.7±0.7	58.2±1.2	0.15±0.01	>1 mM	ND	ND	768.2±242.7
D233N / T216A	215.5±17.3	2312.8±16.6	10.8±0.9	247.0±91.2	21.7±5.6	0.09±0.02	>1 mM	ND	ND	> 2 mM

^*ND, not determined. Due to limited ligand quantities, Ki values were averaged data sets of two trials.

To further probe the D233N mutation’s influence on ligand binding, we studied inhibition of the WT and mutant by a known non-covalent, tetrazole-based inhibitor, (compound 1, Figure 1). The compound showed approximately 10-fold weaker inhibition of the mutant than the WT protein, with the K_i value increasing to 226.6 μM from 21.4 μM. These results indicate that the D233N mutation directly impacts non-covalent interactions between the protein and ligand in the active site.

We then studied whether the activity decrease may have resulted from change in protein stability through melting temperature experiments. Surprisingly, the D233N mutant displayed slightly increased protein stability compared with the WT (55.1±0.6°C vs 54.3±0.5°C, Figure 2), suggesting that the Asp-Asp interaction may be needed to maintain a local structural feature rather than the global stability.

Figure 2. Melting temperature analysis of CTX-M-14 WT and D233N mutant.

X-ray crystallographic structure of the D233N mutant

To understand the structural basis for the decrease in activity, and slight increase in stability, of the D233N mutant, we attempted to crystallize the protein. Using potassium phosphate buffer and microseeding methods, previous CTX-M proteins were crystallized in both the P2₁ and P3₂21 space groups[24]. However, we were only able to crystallize D233N in the P3₂21space group. These observations hinted at structural differences between D233N and WT. The WT CTX-M-14 apo crystal structure and the D233N mutant, both in the P3₂21 space group, were solved to 1.8 Å and 2.0 Å resolutions, respectively (Figure 3A and 3B). The two structures are nearly identical, with an r.m.s.d value of 0.216 Å when superimposing all residue atoms. However, there are significant differences near the active site, particularly the movement of residues 214-218 (in the 213-219 loop) and surrounding residues (Figure 4). Residues 214-218 shift ~0.9 Å away from the β3 strand, which consists of residues 231-240[25]. The biggest movement (1-1.6 Å) is observed in the side chain of Thr216. These conformational changes also lead to smaller shifts (~0.4 Å) in nearby residues, including Tyr129 and Ser130, and the loss of the HB between Thr216 and Thr235 on the β3 strand, accompanied by a small movement of Thr235 as well. The residue movements may explain why the mutant could not be crystallized in the P2₁ space group, because some of these residues are involved in crystal packing in the P2₁ crystal form, but not the P3₂21 space group. In addition, the movement of Thr235, and to a lesser extent Ser130, may be responsible for the loss of the phosphate molecule (from the crystallization buffer) in the protein active site, as both Thr235 and Ser130 are involved in coordinating the phosphate. In previously determined structures, Thr235 and Ser130 have also been found to interact with the C3(4)-carboxylate group of the β-lactam substrate[18, 26].

Figure 3. Active site of CTX-M-14 Class A β-lactamase D233N mutant vs. WT apo active site.

2F_o-F_c electron density map is shown in blue and contoured at 1.5 σ. Water molecules are shown as red spheres. Wat1 is the catalytic water. Potential HBs are represented by black dashed lines. A) CTX-M-14 WT (green) determined at 1.8 Å resolution in the P3₂21 space group. B) CTX-M-14 D233N mutant (cyan) determined to 2.0 Å resolution in the P3₂21space group.

Figure 4. Conformational changes caused by the D233N mutation.

Potential HBs in the mutant structure are represented by black dashed lines. A) Superimposed structures of CTX-M-14 WT (green) and D233N mutant (violet), highlighting the new interaction between the mutant Asn233 and backbone O atom of Gly217, and the shift in the 213-219 loop position. B) Superimposed active sites of WT and D233N mutant.

An examination of the structure at the Asn233-Asp246 position provides clues to the conformational shift of residues 214-218. In the mutant structure, the short hydrogen bond between Asp233 and Asp246 has been abolished, with the Asn233Nδ₂ now within a standard hydrogen bonding distance of 3.0 Å with Asp246Oδ₁(Figure 4A). A series of other structural changes have also taken place due to the alteration of Asp233Oδ₂ to Asn233Nδ₂. In the WT structure, Asp233 is located close to Gly217, with Asp233Oδ2 placed 3.6 and 4.6 Å away from the Gly217Cα and Gly217O respectively. In the mutant structure, the distances between the corresponding Asn233Nδ2 and Gly217Cα/Gly217O have changed to 3.9/3.4 Å respectively, largely due to a shift in the Gly217 position as described above. We hypothesize that this change of contacts between Asp/Asn233 and Gly217 may be the trigger for the movement of residues 214-218 and originates from potential favorable and unfavorable interactions between Asp233Oδ2/Asn233Nδ2 and Gly217Cα/Gly217O. Asp233Oδ2 and Gly217Cα may be involved in a special CH•••O weak HB that has been well documented, particularly involving Cα atoms (even though its biological significance has been debated)[27, 28]. The distance between Asp233Oδ2 and Gly217Cα, and the locations of the oxygen lone pair and Hα atom, seem to support such a HB. Although the strength of this special HB may not be particularly strong, it is at least favorable. The interaction also places Gly217O away from Asp233Oδ2 and reduces electrostatic repulsion between the two partially negative charged atoms. In contrast, replacing the oxygen lone pair of Asp233Oδ2 with a hydrogen atom on Asn233Nδ2 can cause unfavorable interactions with the Hα atom of Gly217. The movement of Gly217 not only positions Gly217Cα away from Asn233Nδ2, but it also brings Gly217O closer to Asn233Nδ2, with a distance of 3.4 Å, allowing for favorable interactions and possibly a weak HB. Taken together, these potential favorable and unfavorable interactions, albeit individually weak, may have led to the shift in the Gly217 position and the rest of the 213-219 loop. In addition, the D233N substitution has also weakened the HB with Asn214, with the length increasing from 3.3 Å between Asp233Oδ1 and Asn214Nδ2, to 4.0 Å between Asn233Oδ1 and Asn214Nδ2 (Figure 4A). This is accompanied by a slight rotation in the Asn214 end group, enabling it to form a HB with the backbone carbonyl group of Asn233, with a distance of 3.1 Å.

Analysis of the temperature factors (B factors) of the structure, an indicator of structural heterogeneity and thermal motion, also revealed increased mobility of the 213-219 loop and the catalytic water. The average B factors of the WT and D233N mutant structure are 29.2 and 34.0 Ų respectively, with the mutant value 4.8 Ų (16%) higher and consistent with its slightly lower resolution (2.0 Švs 1.8 Šfor the WT). In comparison, the B factor values of residues 213-219 are 19.2 and 36.9 Ų for the WT and mutant respectively, with the mutant value 17.7 Ų (92%) higher than the WT. A similar trend is observed for active site residues and the catalytic water, albeit to a lesser extent. Particularly, residues 235-237 have an average B factor of 11.1 and 15.9 Ų in the WT and D233N mutant respectively, with an increase of 4.8 Ų (43%). The catalytic water molecule exhibits a B factor of 8.5 and 13.2 Ų for the WT and D233N mutant respectively, an increase of 4.7 Ų (55%). The average B factors for water molecules are 37.4 and 31.6 Ų for the WT and mutant respectively, although these values do not provide a good reference point due to the smaller number of water molecules in the mutant structure resulting in a lower average B factor value.

X-ray structure of the D233N mutant complexed with tetrazole ligand compound 1

The complex crystal structure of compound 1 with CTX-M-14 WT was determined previously[29]. To understand how the D233N mutation may affect ligand binding, we solved the complex structure of the D233N mutant with compound 1 to 1.95 Å resolution. Compared with the D233N mutant apo structure, inhibitor binding has led to considerable conformational changes in the active site, restoring the Thr216-Thr235 HB observed in the WT–compound 1 complex. As shown in Figure 5, the binding of the tetrazole inhibitor moves Thr235 towards the active site to more closely resemble its position in the WT structure, establishing a HB between Thr235 and the tetrazole ring. This in turn causes changes in Thr216 and Gly217, moving Thr216 also to the location in the WT structure and enabling a HB with Thr216. In comparison to Thr216 and Thr235, Gly217’s position is different from those in the WT and D233N apo structures, and the density also appears to be slightly less ordered compared with the apo structures, suggesting possible alternative conformations.

Figure 5. CTX-M-14 Class A β-lactamase D233N mutant complexed with tetrazole inhibitor 1.

2F_o-F_c electron density map is shown in blue and contoured at 1.5 σ. (A) CTX-M-14 D233N mutant complexed with 1 determined to 1.95 Å in the P3₂21 space group. (B) Superposition of D233N complex (cyan) with the mutant apo structure (magenta). (C) Superposition of WT (green) and mutant D233N complex structures.

Functional analysis of the T216A mutation

As the D233N mutation caused conformational changes in the 213-219 loop, particularly the loss of a HB between Thr216 and Thr235, we next investigated the contribution of Thr216 to enzyme activity. Specifically, the T216A single mutant and the D233N/T216A double mutant were constructed using site-directed mutagenesis. The biochemical assay results demonstrated a significant decrease in activity for both mutants when tested against nitrocefin, ampicillin, and cefotaxime. Particularly, when compared to the WT, the activity of the T216A mutant exhibited approximately 4- and 7-fold higher K_m values against nitrocefin and ampicillin respectively. In the case of cefotaxime, we were unable to obtain conclusive results due to the dramatic increase in K_m values resulting in undersaturation of the enzyme, even with high substrate concentrations up to 1 mM. Meanwhile, the T216A mutation led to a ~3-fold increase in the k_cat value for nitrocefin, but no significant change for ampicillin. Consequently, there was a small (30%) decrease in k_cat/K_m for nitrocefin, and a more significant 5-fold reduction for ampicillin.

For nitrocefin, the increase in k_cat is interesting and indicates a more efficient deacylation process. Nitrocefin is a unique cephalosporin substrate in that its C3 side chain does not dissociate after formation of the acyl-enzyme complex. Based on previous structures, this cephalosporin side chain interacts with Thr216 during the formation of the Michaelis complex. The T216A mutation may therefore affect substrate binding. It is also possible that the C3 side chain of nitrocefin, missing for other substrates in the acyl-enzyme complex, may restrict the conformational changes of the substrate required for the progression of the reaction. The T216A mutation can relieve these constraints in the acyl-enzyme intermediate, increasing the deacylation rate. For ampicillin, The T216A mutation appears to have minimal effect on the deacylation process but have a significant impact on the acyl-enzyme formation, probably due to its influence on substrate binding.

Finally, we evaluated the D233N/T216A mutant. Interestingly, the trend in the activity change between D233N/T216A and T216A is similar to that between D233N and WT, suggesting that the effects of the D233N and T216A are somewhat independent of one another and can be synergistic. As the T216A mutation also affects the HB between Thr216 and Thr235, these results indicate that the influence of D233N on enzyme activity is more than impacting the contributions of Thr216 and Thr235 to substrate binding.

Discussion

The mutagenesis, biochemical, structural, and microbiological experiments described herein suggest a potentially important role of the Asp-Asp pair with respect to the functionality and stability of CTX-M specifically, and possibly in other contexts as well. These results have implications for understanding this conserved short HB interaction across Class A β-lactamases, which also highlights the critical features of the 213-219 loop in maintaining active site integrity and function.

Thr216 and loop 213-219 in Class A β-lactamase function

Our studies have shed light on the essential role of Thr216 and the loop containing residues 213-219 in the structure and function of Class A β-lactamases. Previous complex structures have shown that Thr216 forms non-polar interactions with the substrate[18, 26]. However, this contribution to ligand binding alone may not explain the dramatic loss of activity in the T216A mutant. Thr216 hydrogen bonds with Asn214 and Thr235, and, like Asn214, forms a water-mediated interaction with Lys234 (Figure 3A). Lys234 and Thr235 are two highly conserved residues on the β3 strand, which lines one side of the active site and plays a crucial role in substrate binding and catalysis. Thr235 interacts directly with the C3(4)-carboxylate group of the substrate. Lys234 serves as an electrostatic anchor for the negatively charged substrate and may also influence the pK_a of the nearby Lys73, a crucial catalytic residue hypothesized to function as the general base in the acylation reaction[11]. The interactions between Thr216 and these residues may stabilize the active site configuration necessary for substrate binding and catalysis. In TEM-1 and SHV-2, Thr216 is replaced by a valine and Asn214 is substituted for an aspartate residue. Compared with Asn214 in CTX-M, Asp214 in TEM-1 and SHV-2 forms a new water-mediated interaction with Thr235, potentially strengthening the contact with Lys234, which is also mediated by a water molecule^13, [30, 31]. The new interactions again maintain a strong linkage between the 213-219 loop and the two important active site residues of Thr235 and Lys234.

In addition to the contacts with the β3 strand, Thr216 and residues on the 213-219 loop also interact with other structural elements contributing to the active site. The Cγ2 atom of Thr216 is in van der Waals contact with Tyr129Cα and Tyr129C, whereas the side chains of Thr215 and Tyr129 form non-polar interactions with one another (Figure 4B). Both the backbone and side chain of Thr215 form HBs with Gln128O. Furthermore, Gly213-Asn214 have extensive van der Waals contacts with Gln128. These interactions may help position Ser130 for substrate binding and general acid catalysis during the acylation step, two roles suggested by previous studies[32-34].

The 213-219 loop connects two α-helices (residues 200-212 and 220-225) (Figure 4B). The two helices interact extensively with the β3 strand and the α-helix that contributes the catalytic Ser70 and Lys73. This expansive network of molecular interactions suggests that the 213-219 loop may play a crucial role in ensuring a productive configuration of the active site.

The Asp233-Asp246 pair in Class A β-lactamases

Asp233 is highly conserved in Class A β-lactamases, although the overall sequence homology among these proteins is only 20-40% (Figure 6). In most Class A enzymes such as CTX-M, Asp233 forms a short HB with Asp246. From our previous sub-Angstrom resolution X-ray crystallographic studies, this HB has a length of 2.47 Å and the hydrogen appears to be located equidistant to the two Asp Oδ atoms[19]. These observations suggest that this HB could be a low-barrier hydrogen bond (LBHB), a special type of short HB (~2.5 Å in length) where the hydrogen is equally shared between the two hydrogen bonding groups (for comparison, a typical HB is 2.8-3.2 Å)[35]. Due to their special strength in the gas phase, LBHBs have been hypothesized to play an important role in proteins[1, 36-38], although their existence and functional relevance have been intensely debated[39-43], and some putative LBHBs have been found to be crystal artifacts[44, 45]. For the Asp233-Asp246 HB, it is currently unknown whether it is indeed a LBHB, especially in solution. However, in other Class A β-lactamase structures, the Asp233-Asp246 HB length remains short at approximately 2.5 Å. These enzymes include the clinically relevant Class A enzyme KPC-2 (HB length 2.49 Å)[46] and another Class A β-lactamase from a deep-sea bacteria (HB length 2.53 Å)[6]. The short HB is even conserved in GES-2, where residue 233 is a glutamate, with a HB distance of 2.58 Å[8]. Interestingly, in a few enzymes where Asp246 is replaced by isoleucine, another aspartate residue, Asp214 (Asn214 in CTX-M), forms an alternative short HB (~2.5 Å) with Asp233 (Figure 6)[20]. In the 0.90 Å resolution SHV-2 structure, a hydrogen atom is observed in the difference electron density map close to the center of the HB (2.59 Å in length) between these two residues[21].

Figure 6. Conservation of Asp-Asp pairs in Class A β-lactamases.

A) Apo structure of SHV-2 class A β-lactamase (PDB ID: 1N9B, magenta) superimposed with CTX-M-9 cefotaxime complex (PDB ID: 3HLW, green). Asp233 forms a short hydrogen bond with Asp246 CTX-M-9, and with Asp214 in SHV-2. B) Sequence alignment of common Class A β-lactamases. Three positions for aspartate residues are highlighted. Asp233 is conserved in most Class A β-lactamases, with Asp246 also maintained in many of these enzymes. In a few proteins where Asp246 is replaced by isoleucine, Asp214 substitutes for Asn214. CTX-M-9 differs from CTX-M-14 only by a single V231A mutation.

Our mutagenesis results suggest that the Asp233-Asp246 pair is important to the enzymatic activity of Class A β-lactamases, achieved through maintaining the structural integrity of the active site involving loop 213-219 and neighboring residues. This function is particularly important for substrates with bulky side chains, such as cefotaxime, which can potentially destabilize or deform the active site. The special structural contribution of the Asp pair can be two-fold – fulfilling stringent electrostatic and steric requirements in a tightly packed region of the protein, and strengthening intramolecular interactions through a short, strong HB. Our studies offer valuable insights into the first possible role of this Asp pair, highlighting the sensitivity of protein structures to the smallest change in residues. Because the Asp pair already shares a hydrogen atom between them, a single Asp→Asn mutation adds only one additional hydrogen to these residues in terms of the overall size, without changing the net charge of the pair. Yet, D246N failed to fold and express properly, and D233N exhibited significant structural distortions around the mutation. Although this is hardly surprising due to the well-known, densely packed, nature of protein interiors, it underscores the need for specific functional groups in particular regions of protein structure for optimal folding and function, a result of protein evolution. For that purpose, the Asp pair offers a unique piece of this 3-dimensional puzzle, as demonstrated by the WT and D233N structures. It is worth pointing out that, while the evidence for a O•••H-C interaction between Asp233Oδ2 and Gly217Hα is tenuous, Gly217 is also highly conserved in Class A β-lactamases and the close contact between Asp233 and Gly217 is observed in other proteins as well. In addition, not all hydrogen-mediated carboxylate pairs are sensitive to mutations. For example, for the Glu15-Asp24 pair in the human protein DJ-1, substitutions of Gln15 and Asn24 are both well tolerated by the protein, albeit while decreasing the protein stability slightly[3].

In comparison, our experiments do not offer direct answers to a possible special role of the short HB’s strength, particularly that of LBHB, which can be several times stronger than a standard HB[1, 35, 37]. Previous studies have demonstrated that generally the shorter the HB length, the stronger its strength in non-aqueous solvent[47]. We hypothesize that the strength of a short HB can be beneficial for maintaining the active site integrity of Class A β-lactamases. This is based on the high level of conservation of the Asp-Asp interaction in many Class A β-lactamase structures that use Asp246 or alternatively Asp214 to form such a pair with Asp233. As previous experiments have shown, protein active sites, including those of β-lactamases, contain many high-energy features that only become complemented during substrate binding[48]. Two such features in CTX-M include the sub-pocket recognizing the substrate C3(4)-carboxylate group, and the oxyanion hole. The carboxylate binding site is formed by the side chains of Thr235, Ser237, and Ser130 through direct HBs, as well as by Lys234 and Lys73 through long-range electrostatic interactions[18, 26]. In apo Class A β-lactamase structures, this site is usually occupied by a phosphate ion or other negatively charged molecules from the crystallization buffer. The oxyanion hole is formed by the backbone amides of Ser70 and Ser237, the latter of which resides on the β3 strand. This subsite is crucial in stabilizing the transition state oxyanion, but the close juxtaposition of two partially positively charged amide groups is energetically unfavorable. In most apo structures, this sub-pocket is occupied by a water molecule. Interestingly, an Asp-Asp LBHB has been found close to the oxyanion hole of another unrelated enzyme, Rhamnogalacturonan acetylesterase, based on NMR analysis and the 2.47 Å short HB length in the crystal structure[49]. In addition, a disulfide bond is also very often observed near the active site of many Class A β-lactamase active sites, involving C69 and C238, located adjacent to the two residues making up the oxyanion hole, Ser70 and residue 237. These studies suggest that the added strength of the Asp-Asp short HB could be important for countering the destabilizing internal interactions of the protein active site, particularly involving the oxyanion hole of β-lactamases.

Protein flexibility in enzyme evolution and resistance development

Structural flexibility plays a myriad of sometimes contrasting roles in protein function. While it confers adaptability and tunability of substrate recognition and enzymatic activity, the entropic cost can have an adverse effect on the ligand binding affinity, as well as the availability of productive active site configurations. Previous studies on β-lactamase evolution have demonstrated that increased flexibility in the active site can enable the enzyme to better accommodate larger substrates and broaden the spectrum of activity, and that these changes are often accompanied by additional mutations improving the overall stability of the protein or rigidifying local structures[50-54]. Our current results have shown that by increasing the conformational heterogeneity of certain structural features, β-lactamases can also drastically reduce the binding affinities of non-covalent inhibitors, while incurring only a modest cost on the activity for certain substrates (e.g., nitrocefin and ampicillin). The D233N mutation does not change the lowest-energy conformations of the residues in direct contact with the tetrazole-based inhibitor and many substrates, but rather results in a rise in the mobility of these active site residues, as partly indicated by the temperature factors of the crystal structure, as well as alternative conformations. Such effects can be more consequential for non-covalent inhibitors, specifically in changes to k_off, than for covalently bound substrates. However, for some substrates, such as cefotaxime, whose large side chains may cause deformation of the active site even in the WT enzyme and/or contact directly the regions distorted by the mutation, the negative impact on enzyme activity can nevertheless be significant.

Conclusion

Asp233 and Asp246 are highly conserved in Class A β-lactamases. Our mutagenesis and structural analysis have shed light on the contribution of this pair of aspartate residues to the integrity of the active site, particularly concerning the conformation of the 213-219 loop. The differential effects on the non-covalent inhibitor and various substrates suggest that β-lactamases can potentially develop resistance against certain inhibitors by increasing its active site flexibility. Furthermore, the Asp→Asn mutations illustrate the sensitivity of protein structure to the smallest change, i.e., the addition of one hydrogen atom. This has also made it difficult to isolate and quantify the effects of the D233N mutation on specific interactions. Nevertheless, we hypothesize that the short HB between Asp233 and Asp246 is likely stronger than a standard HB, and this strength may be important for maintaining the high-energy features of a protein active site. However, this particular role of the short HB will await future studies, including computational simulations.

Materials and Methods

Mutagenesis

CTX-M-14 (UniProt ID: H6UQI0) was cloned into a modified plasmid vector pET-9a as previously described[55]. To make the CTX-M-14 mutants - D233N, T216A, and double mutant D233N/T216A - the QuikChange Lightning Site-Directed Mutagenesis Kit was used.

Protein purification, crystallization, and structure determination

The protein was purified as previously described[55] and crystallized in 1.0-1.2 M potassium phosphate buffer (pH 8.3) from hanging drops at 20°C. The final concentration of the protein in the drop ranged from 6.5 mg/mL to 10 mg/mL. Full size crystals were grown in approximately two weeks with microseeding. The complex crystal of D233N and compound 1 was obtained by soaking the crystal in a solution of 10 mM compound for 12 hours prior to flash freezing with liquid nitrogen. For all apo and complex structures, a 30% w/v solution of sucrose was added to the crystallization mother liquor as a cryoprotectant. Diffraction data were collected on the SER-CAT 22-ID-D beamline at the Advanced Photon Source (APS), Argonne, Illinois. Data were indexed, scaled, and merged with HKL2000[56]. The models for refinement were first obtained through using a rigid-body refinement by Phaser[57] in PHENIX[58] with an apo CTX-M-14 structure (PDB 4UA6). PHENIX and Coot[59] were used to complete the model building and refinement. All figures were generated using PyMol (Schrodinger Inc).

Biochemical assays

The hydrolysis reaction kinetics of CTX-M and mutants were measured using the β-lactam substrates ampicillin and cefotaxime, as well as the chromogenic compound nitrocefin, in 100 mM Tris-HCl (pH 7.0) and monitored using a Biotek Synergy Mx monochromator-based multimode microplate reader at 235 nm, 260 nm, and 486 nm wavelengths respectively, at 37°C. The enzyme concentration was kept constant at 0.5 nM between assays for WT and mutants. For inhibition analysis, nitrocefin was used as the substrate, and a concentration of 2 mM compound 1 was used as the highest concentration of inhibitor. K_m, k_cat, and K_i were calculated using the software SigmaPlot.

Thermal stability assays

Using circular dichroism, the secondary structure was monitored with a Jasco J-815 CD spectropolarimeter in conjunction with a Petiltier cell holder. CTX-M-14 WT and D233N mutant were both diluted to 0.05, 0.1, and 0.2 mg/mL in 100 mM potassium phosphate buffer pH 7.0. All spectra were collected in triplicate for each of the three protein concentrations and measured using the wavelength 222 nm with a temperature range of 25°C to 85°C to determine the melting temperature. The collected data was analyzed using the software SigmaPlot and a two-state fitting program.

Table 2.

X-ray data collection and refinement statistics

Data Collection
Structure	CTX-M-14	CTX-M-14 D233N	CTX-M-14 D233N
			Compound 1
Space Group	P3221	P3221	P3221
Cell Dimensions
a, b, c (Å)	41.71	41.82	41.62
	41.71	41.82	41.62
	232.52	231.95	231.16
α, β, γ (°)	90	90	90
	90	90	90
	120	120	120
Resolution (Å)	28.53 - 1.80	32.79-2.00	38.53-1.95
	(1.83-1.80)	(2.07-2.00)	(2.02-1.95)
No. of nique reflections	22657 (2186)	16896 (1634)	17735 (1743)
Rmerge (%)	6.0 (13.0)	10.6 (28.1)	5.4 (7.0)
<I> / <óI>	31.9 (17.3)	24.5 (10.9)	23.7 (17.2)
Completeness (%)	98.91 (99.86)	99.90 (100.00)	98.41 (99.15)
Refinement
Resolution (Å)	28.53-1.80	32.79-2.00	38.53-1.95
Rwork/Rfree (%)	18.3/23.6	19.8/23.3	18.90/25.9
No. heavy atoms
Protein	1977	1984	1956
Ligand/ion	9	0	29
Water	272	153	221
B-factors (Å2)
Protein	28.60	34.06	21.00
Ligand/ion	24.64	-	12.42
Water	37.59	31.81	22.29
R.M.S.D.
Bond lengths (Å)	0.011	0.011	0.011
Ramachandran favored (%)	97.69	96.54	96.54
Ramachandran allowed (%)	1.92	3.08	2.31
Ramachandran outliers (%)	0.38	0.38	1.15
Rotamer outliers (%)	0.00	0.00	0.00
Cβ outliers	0	0	0
Clash score	3.57	11.19	5.30
PDB ID	6D7H	6D7I	7S0V

^*Values in parantheses represent highest resolution shells.

Abbreviations:

HB

hydrogen bond

ESBL

extended spectrum β-lactamase

WT

wild-type

Data availability

The atomic coordinates and structure factors for the CTX-M-14 D233N, WT, and complex structures have been deposited in the Protein Data Bank with IDs 6D7I, 6D7H, and 7S0V, respectively. CTX-M-14 has Uniprot ID H6UQI0.