Computational and experimental analyses of alanine racemase suggest new avenues for developing allosteric small-molecule antibiotics

Abstract

Given the ever-present threat of antibacterial resistance, there is an urgent need to identify new antibacterial drugs and targets. One such target is alanine racemase (Alr), an enzyme required for bacterial cell-wall biosynthesis. Alr is an attractive drug target because it is essential for bacterial survival but is absent in humans. Existing drugs targeting Alr lack specificity and have severe side effects. We here investigate alternative mechanisms of Alr inhibition. Alr functions exclusively as an obligate homodimer, so we probed seven conserved interactions on the dimer interface, distant from the enzymatic active site, to identify possible allosteric influences on activity. Using the Alr from M. tuberculosis (MT) as a model, we found that the Lys261/Asp135 salt bridge is critical for catalytic activity. The Lys261Ala mutation completely inactivated the enzyme, and the Asp135Ala mutation reduced catalytic activity eight-fold. Further investigation suggested a potential drug-binding site near the Lys261/Asp135 salt bridge that may be useful for allosteric drug discovery.

Introduction

Alanine racemase (Alr) is a PLP-dependent enzyme that racemizes L-alanine to D-alanine, a key building block in peptidoglycan/cell-wall biosynthesis.^1,2 It is an attractive antibacterial target because it is essential for bacterial survival but is absent in humans. Prior research identified structural analogs of D-alanine as Alr inhibitors, leading to the discovery of the FDA-approved drug cycloserine (brand name Seromycin).^3-5 Unfortunately, like other inhibitors that target the Alr active site, cycloserine lacks specificity because it acts on other PLP-containing enzymes in humans, especially in the central nervous system.^6,7 The use of cycloserine as an antibiotic is thus limited due to its severe side effects, including confusion, seizure, speech disorder, dizziness, and even coma. Designing alternative orthosteric Alr inhibitors has been challenging because the conserved active site entryway is too narrow to allow access to drug-like molecule larger than alanine, the native substrate.^8,9 The presence of PLP in the substrate binding pocket also leaves little additional space for larger drug-like molecules, thereby further limiting the drug design efforts.^10,11 These challenges suggest the need for drugs that inhibit Alr via alternative mechanisms that do not target the PLP-dependent active site.

To enable future broad-spectrum antibiotic drug discovery, we sought to identify an allosteric mechanism of Alr inhibition independent of the PLP-dependent active site. With few exceptions, Alrs are obligate homodimers formed by the head-to-tail association of two monomers, and many of the residues at the dimer interface are highly conserved over a broad spectrum of bacterial species.^8,12-14 In theory, molecules that bind at the dimer interface could inhibit Alr activity either by dimer disruption or allostery, strategies that have been successfully applied to other drug targets.^15-18

In this work, we use computational methods to identify non-active site residues associated with conserved interactions that may be critical for dimer association and thus enzyme activity. We mutate each residue to alanine and characterize the corresponding mutant enzymes for enzymatic activity. Eliminating one of the conserved salt bridges via mutagenesis resulted in the near complete abolition of catalytic activity, perhaps via impaired dimerization or an uncharacterized allostery-mediated conformational change. We further use computational methods to show that this salt bridge is adjacent to a previously unknown druggable pocket, suggesting a pharmaceutical strategy for drug design that interferes with a critical anchor point.

Materials and Methods

All reagents and chemicals used as buffers and substrates of the highest grade chemically available were purchased from Sigma-Aldrich, Fisher, Acros, or Alfa Aesar. UV-Vis spectrophotometric data were collected using Cary 60 UV-Vis Spectrophotometer (Agilent) and Multiskan GO (Fisher Scientific). Protein purifications were conducted using BioLogic DuoFlow (Bio-Rad). Alr from M. tuberculosis (MT-Alr) and L-alanine dehydrogenase from S. coelicolor (Ald) were cloned, expressed, purified, and characterized in the lab.^19,20

Predicting critical residues at the dimer interface

We followed a multi-step protocol to identify candidate residues at the dimer interface that are both distant from that active site and conserved across a broad spectrum of bacterial species. First, we used visual molecular dynamics (VMD)²¹ to identify Alr residues that came within 4 Å of the opposite monomer but were at least 8 Å from the catalytic site. The 6SCZ²² crystal structure was used for this analysis because its active site is fully occupied (i.e., PLP is conjugated to cycloserine), and we used this conjugate to judge distances to the catalytic site. Second, we used the Consurf server²³ to identify those residues with above-average conservation scores. Third, we applied Robetta,²⁴ a program for computational alanine scanning, to the 6SCZ structure and retained residues predicted to decrease the binding energy by at least 1.0 kcal/mol when mutated to alanine. Finally, we used BINANA^25,26 and visual inspection to identify residues that participated in specific (e.g., salt-bridge, hydrogen-bond, π-stacking) interactions with opposite-monomer residue(s).

Mutagenesis

The plasmid pET28b-MT_Alr²⁰ served as the DNA template to generate seven MT-Alr single mutants (Glu72Ala, Asp135Ala, Asn139Ala, Lys261Ala, Glu267Ala, Arg371Ala, and Arg373Ala) using a Q5 site-directed mutagenesis kit (NEB). The corresponding primers were designed using the NEBase Changer tool (http://nebasechanger.neb.com/; Table 1). The PCR mixture (10 μL) contained 1 μL of 10 ng/μL template DNA plasmid, 1 μL each of 5 μM forward and reverse primers, 5 μL of Q5 HS master mix, and 2 μL of ddH₂O. The PCRs were performed in a BioRad T100 Thermal Cycler with the following parameters: 98 °C for 30 s followed by 25 cycles of 98 °C for 10 s, T_a °C (suggested annealing temperature for the primers) for 20 s, and 72 °C for 3 m, and a final extension time of 2 m at 72 °C. The PCR reactions were confirmed by DNA gel electrophoresis. The KLD reactions were performed at room temperature for 10 min with 0.5 μL of amplified PCR product, 2.5 μL of KLD reaction buffer, 1.5 μL of ddH₂O, and 0.5 μL of KLD enzyme mixture. 2.5 μL of KLD mixtures were chemically transformed into 5-alpha competent E. coli cells. The mutations were confirmed by DNA sequencing (Genewiz).

Table 1.

Primers used to generate single-alanine mutants with corresponding annealing temperatures (T_a).

Primer	Sequence	Ta (°C)
E72A-FP	5'- CACCGTCGACGCGGCGCTAGCGC -3'	72
E72A-RP	5'- GCGACGCCGAGTTCGGCC -3'
D135A-FP	5'- CGTCAAGGTGGCGACCGGGCTGAACC -3'	72
D135A-RP	5'- GTCACCGTCGCCGTCCGG -3'
N139A-FP	5'- TACCGGGCTGGCGCGCAATGGCG -3'	66
N139A-RP	5'- TCCACCTTGACGGTCACC -3'
K261A-FP	5'- TGCGCTGGTGGCGTCGATTCGTG -3'	58
K261A-RP	5'- ACAGCACATTTCACGGTC -3'
E267A-FP	5'- TCGTGCGGGGGCGGGCGTGTCGT -3'	72
E267A-RP	5'- ATCGATTTCACCAGCGCAACAGCACATTTCACG -3'
R371A-FP	5'- CACCAGCCCGGCGGGACGTATCAC -3'	62
R371A-RP	5'- ACCACTTCGTAGTGGATG -3'
R373A-FP	5'- CCCGCGAGGAGCGATCACCAGGAC -3'	61
R373A-RP	5'- CTGGTGACCACTTCGTAG -3'

Protein expression and purification

The MT-Alr mutant proteins were cloned with an N-terminal His-tag and were expressed in E. Coli BL21 (DE3) cells. A typical large-scale purification involved a bacterial culture grown in 2 X 1 L of terrific broth (TB) medium with 50 μg/mL kanamycin. This broth was inoculated with 5 mL starter culture and shaken at 180 rpm at 37 °C until the OD₆₀₀ reached 0.5. The incubator temperature was then decreased to 30 °C and shaken until the OD₆₀₀ reached 0.6, at which point expression was induced to the final concentration of 1 mM isopropyl-β-D-thiogalactoside (IPTG). The culture was incubated for a further 24 hours at 30 °C with shaking at 180 rpm. Cells were harvested by centrifugation at 7,000 rpm for 5 min. The cell pellets were dissolved in 20 mL of cold buffer (20 mM Tris, 100 mM NaCl, pH 8.0) with 20 μL Halt^™ Protease Inhibitor Cocktail (100x), added only during this step. Cells were lysed by sonication (10 min, 5 s on and 5 s off), and the lysate was cleared by centrifugation. The supernatant was applied to a 1 mL IMAC column (Bio-Rad) and eluted with a linear gradient (100 mL) of 250 mM imidazole buffered with 20 mM Tris and 100 mM NaCl (pH 8.0). Fractions containing pure protein were collected and dialyzed against 20 mM Tris-HCl, 100 mM NaCl (pH 8.0), and then stored at −80 °C. The purity of the protein fractions was verified by SDS-PAGE (Fig S1).

Enzyme characterization

After determining the optimal enzymatic assay condition for the wild-type MT-Alr enzyme,²⁰ we measured the activity of the mutant enzymes in the D- to L-alanine direction. We monitored the production of NADH (ε₃₄₀ = 6.22 mM⁻¹ cm⁻¹) at 340 nm (30 °C, pH 9.0) as the L-alanine was converted to pyruvate and ammonia by L-alanine dehydrogenase (Scheme S1). The reaction mixture (250 μL) contained 0.01 – 50 mM D-alanine (dissolved in buffer), NAD (906 μM), L-alanine dehydrogenase (2.76 μM), and mutant MT-Alr (0.6 μM). The steady-state kinetic constants were determined by fitting the kinetic data (triplicate) to the Michaelis-Menten equation using GraphPad Prism7 (GraphPad Software, Inc.).

Homology modeling and structural alignment

Homology models of Alrs from Enterococcus faecium (EF-Alr; Model S1), Neisseria gonorrhoeae (NG-Alr; Model S2), and Klebsiella pneumoniae (KP-Alr; Model S3) were generated using the target-template sequence-alignment tool available through the SWISS-MODEL Workspace (https://swissmodel.expasy.org/; Fig S2). We used a crystal structure of E. faecalis Alr (PDB ID 3E6E²⁷) as the template to model EF-Alr (63.07% sequence identity), a crystal structure of P. aeruginosa Alr (PDB ID 1RCQ⁸) to model NG-Alr (51.99% sequence identity), and a crystal structure of E. coli Alr (PDB ID 2RJH²⁸) to model KP-Alr (89.97% sequence identity). We used Chimera MatchMaker²⁹ to structurally align these three models as well as crystal structures of Alrs from M. tuberculosis (MT-Alr; PDB ID 1XFC⁹) and Streptococcus pneumoniae (SP-Alr; PDB ID 3S46¹¹).

Predicting druggable pockets

To predict druggable pockets, we first processed a crystal structure of MT-Alr (PDB ID 1XFC)⁹ using Maestro 13.2.128 (Schrodinger 2022-2; Model S4). Missing internal (e.g., Arg264-Pro278 of chain B) and terminal residues in each of the monomers were constructed using Prime,^30,31 and the newly built protein structure was preprocessed using the Protein Preparation Wizard.³² We then removed all water molecules from the energy-minimized structure and analyzed it for potential druggable pockets using Schrödinger’s SiteMap program³³; the FTMap web server³⁴; and FPocketWeb,³⁵ a web-based implementation of fpocket.^36,37

Measuring the pocket volumes of Alrs from different species

To identify other Alr structures deposited in the Protein Data Bank,³⁸ we searched the PDB for proteins with sequences similar to 1XFC⁹ (E-value cutoff of 0.1). There were 66 such proteins. We discarded the seven proteins not listed as Alrs, as well as 3HUR, which did not include a homodimer of the expected configuration.

Each Alr is a dimer with two copies of a putative allosteric site identified using FTMap³² (see Results and Discussion). We aligned the Alr proteins so that all pockets were located at the same position in 3D space, using 1XFC chain A as a template. We then used the POVME2 algorithm to measure the volume of each pocket. We used a PointsInclusionSphere (radius 10 Å) and a ContiguousPocketSeedSphere (radius 3 Å), both positioned at the center of the putative pocket. The POVME2 GridSpacing and ContiguousPointsCriteria parameters were 1.0 Å and 3, respectively. All points that fell within 1.09 Å of any protein atom were removed, as were any that fell outside the convex hull defined by the positions of the protein atoms. See the original POVME2 paper for details.³⁹

Many of the species were associated with multiple PDB files. For each species, we identified the single pocket with the largest volume, excluding those crystal structures that were missing residues near the location of the putative allosteric binding pocket.

Results and Discussion

We used the alanine racemase from M. tuberculosis (MT-Alr) as a model Alr enzyme because much prior research has focused on its structure, both with and without inhibitors bound to the active site.^9,22 Like most other Alr enzymes, MT-Alr is a functionally obligate homodimer with a conserved dimerization motif.^9,20 Probing this motif should therefore provide insight into potential alternative mechanisms of inhibition.

Identifying critical residues at the dimer interface

Protein-protein interfaces are important but challenging targets for inhibitor design. A crucial first step is identifying potential sites along these interfaces where interface-perturbing drugs might bind and disrupt critical interactions. These sites are typically associated with the small subset of interfacial residues that contribute substantially to the overall protein binding energy.⁴⁰

Using computational and structural analyses, we identified eight candidate interfacial residues at the dimer interface: Arg140, Tyr271, Asp320, Asn139, Lys261, Arg371, Glu365, and Arg373. We selected these residues based on four criteria. (1) The residues line the dimeric interface but are distant from the catalytic pocket and so are unlikely to impact catalysis directly. (2) Each residue participates in specific interactions with residue(s) of the opposite monomer (e.g., salt bridges and hydrogen bonds, not just hydrophobic contacts). (3) The residues are at least modestly conserved among Alr proteins across many species, suggesting they play an essential role in protein function. (4) Computational alanine scanning suggests each residue contributes at least 1.0 kcal/mol to the binding energy.

Of these eight residues, we selected four that were distributed roughly evenly across the dimeric interface: Asn139, Lys261, Arg371, and Arg373. To this list of four, we added Asp135, which forms a salt bridge with Lys261; Glu72, which forms a salt bridge with Arg373; and Glu267, which forms a hydrogen bond with Asn139. These three additional residues also line the dimeric interface, are distant from the catalytic site, and are at least modestly conserved.

The seven selected residues can be divided into three clusters based on their locations and interaction patterns. Two clusters, 1A and 1B, are identical and located at the front and back ends of the dimer interface. Cluster 2 is in the middle of the interface (Fig 1 and Table 2).

Fig 1.

MT-Alr homology model. (A) The two monomers of the Alr dimer are shown in pink and blue ribbon, respectively. Predicted critical residues are shown in stick representation with text labels. The three hotspot clusters are labeled with braces. (B) FTMap-identified druggable hotspots near the Asp135/Lys261 salt bridge (orange), orthosteric cycloserine-binding site (green), and central Arg371/Arg371 interaction (yellow). The pocket marked with an asterisk may be artefactual given that an adjacent segment of the model (Arg264-Pro278) was predicted rather than crystallogrpahically resolved.

Table 2.

Potential dimer-promoting clusters at the MT-Alr dimer interface.

Cluster	Chain A	Chain B
1A	Lys261	Asp135
1A	Glu267	Asn139
1B	Asp135	Lys261
1B	Asn139	Glu267
2	Glu72	Arg373
2	Arg371	Arg371
2	Arg373	Glu72

Assessing the importance of the identified residues using mutagenesis and enzymatic assays

We used mutagenesis to evaluate the impact of Glu72, Asp135, Asn139, Lys261, Glu267, Arg371, and Arg373 on catalytic activity. Because Alr is an obligate homodimer, each residue is present on both sides of the interface, so each mutation had a twofold impact. We generated a single-point alanine mutant for each of the seven residues. We chose alanine because it eliminates the amino-acid side chain beyond the β carbon but does not alter the main-chain conformation (as can glycine and proline).

We determined the steady-state rate constants for all mutant proteins under the same conditions used for wild-type MT-Alr²⁰ (Table 3). To examine the impact of the Asp135/Lys261 salt bridge, we considered the Asp135Ala and Lys261Ala mutants. The Asp135Ala mutation reduced k_cat and increased K_m, resulting in an 8-fold decrease in catalytic activity. The Lys261Ala mutation resulted in a complete loss of catalytic activity. Given that these residues are distant from the catalytic pocket, we hypothesize that disruption of the Asp135/Lys261 salt bridge allosterically impedes catalysis via active site conformational changes or perhaps dimer disruption, leading to an inactive monomeric enzyme.⁴¹ This finding is particularly impactful because small molecules that similarly disrupt the Asp135/Lys261 salt bridge could serve as allosteric inhibitors of Alr activity.

Table 3.

Enzymatic rate constants for MT-Alr wild-type and single-alanine mutants.

Enzyme	kcat (s−1)	Km (mM)	kcat/Km (s−1 mM−1)
Wild type	0.94 ± 0.03	0.69 ± 0.08	1.36
Glu72Ala	0.92 ± 0.04	0.81 ± 0.02	1.14
Asp135Ala	0.46 ± 0.03	2.61 ± 0.40	0.18
Asn139Ala	0.95 ± 0.09	0.76 ± 0.30	1.25
Lys261Ala	ND	ND	ND
Glu267Ala	1.49 ± 0.10	0.42 ± 0.20	3.55
Arg371Ala	1.74 ± 0.20	0.20 ± 0.08	8.70
Arg373Ala	1.12 ± 0.10	0.57 ± 0.20	1.96

ND: non-detectable.

In contrast, we found that the remaining mutations either improved catalytic efficiency or had no impact. For example, the Glu72/Arg373 salt bridge is not critical for catalytic activity because the Glu72Ala and Arg373Ala mutants both exhibited similar k_cat and K_m values as the wild type. In contrast, the impact of the Asn139/Glu267 interaction is ambiguous; the Asn139Ala mutation had minimal impact on catalytic activity, but the Glu267Ala mutation increased k_cat and decreased K_m (i.e., improved catalytic efficiency). Finally, the Arg371/Arg371 interaction—possible because the planar-parallel orientation enables van der Waals interactions and water/Glu365 mitigate the electrostatic repulsion-appears to inhibit catalysis.^42-45 An Arg371Ala mutant resulted in a six-fold increase in catalytic efficiency. Although these mutations certainly warrant further study, we do not pursue them further in this work, given they are less likely to play a role in inhibitory (i.e., pharmacologically useful) allostery.

Identification of druggable targets: FTMAP

To assess whether carefully designed small molecule ligands could mimic the effect of these mutations on Alr activity, we used three methods to search for druggable Alr pockets. First, we used the FTMap server³⁴ to computationally dock a set of small molecular fragments with drug-like chemical properties onto the entire surface of a MT-Alr (1XFC)⁹ structure based on PDB ID 1XFC⁹ (Model S4; see Materials and Methods). Locations on the surface where fragments tend to congregate are known as druggable hotspots and often correspond to known ligand-binding pockets. FTMap identified eleven hotspots that can be generally divided into three groups. Five of the eleven hotspots fell near one of the two known cycloserine-binding catalytic sites, providing a positive control (Fig 1B, in green). FTMap automatically removes non-standard residues, so this pocket did not include the conjugated PLP molecule of the original structure. Three additional hotspots were located near the identified Asp135/Lys261 salt bridge (Fig 1B, in orange). Compounds that bind at this position could potentially disrupt the salt bridge, leading to allostery-mediated inhibition. We note that one of these FTMap-identified hotspots (Fig 1, marked with an asterisk) may be artefactual; we used Schrödinger’s Prime module^28,29 to reconstruct missing 1XFC residues (Arg264-Pro278, chain B), which likely reduced the accuracy of the model in this region (see Model S4). However, the other two Asp135/Lys261-adjacent hotspots—which represent the same pocket duplicated on both sides of the homodimer—are not impacted by the modeling artefact; indeed, one is on the opposite side of the homodimer. This predicted pocket is lined by residues Ser109, Ser110, Leu111, Asp135, Arg140, Asn141, Gly142, Val143, Gly144, Gln147, Leu259, Val260, Lys261, and Ser262 (Fig 2 in bold, under curly braces). Several of these residues are conserved: Asp135, Arg140, Gly142, Val143, Gln147, Lys261.

Fig 2.

Comparison of MT-Alr and homologs. A multiple sequence alignment of M. tuberculosis (MT-Alr), E. faecium (EF-Alr), S. pneumoniae (SP-Alr), N. gonorrhoeae (NG-Alr), K. pneumoniae (KP-Alr), P. aeruginosa (PA-Alr), and A. baumanni (AB-Alr). The active site residues are boxed in blue, and the Asp135/Lys261 hotspot residues are in red. Residues that come within 5 Å of the FTMap-predicted pocket near the Asp135/Lys261 salt bridge are shown in bold, under curly braces. A structure-based alignment of MT-Alr, EF-Alr (Model S1), SP-Alr (PDB ID 3S46), NG-Alr (Model S2), and KP-Alr (Model S3) shows the MT-Alr Asp135/Lys261 interaction is largely conserved.

Finally, three hotspots were located near the central Arg371/Arg371 interaction (Fig 1B, in yellow). Given that mutating these two arginine residues to alanine improves catalytic activity, compounds that disrupt the Arg371/Arg371 interaction are less likely to have pharmacological potential.

Confirmatory identification of druggable targets: SiteMap and FPocketWeb

As a second assessment of Alr druggability, we used Schrödinger's SiteMap program³³ to identify druggable locations on the surface of the same modeled MT-Alr protein. SiteMap identified five potential pockets. Of these, one corresponded to a cycloserine-binding catalytic site (Fig S3, in green), even though SiteMap retained the conjugated PLP bound in that site. This finding again serves as a positive control. Three of the five predicted sites were again adjacent to the Asp135/Lys261 salt bridge (Fig S3, in orange), including the top-ranked pocket, and one of the five corresponded to the Arg371/Arg371 interaction (Fig S3, in yellow).

As a third assessment of Alr druggability, we used FPocketWeb,³⁵ a web-based implementation of fpocket,^36,37 to identify MT-Alr cavities. Among the top eight FPocketWeb-identified pockets, the top two corresponded to a cycloserine-binding catalytic site (Fig S4, in green), which did not include a bound PLP due to FPocketWeb processing. Two predicted sites were again adjacent to the Asp135/Lys261 salt bridge (Fig S4, in orange), and one corresponded to the Arg371/Arg371 interaction (Fig S4, in yellow).

An extended-spectrum antibiotic drug target

To assess whether compounds that bind near the Asp135/Lys261 salt bridge might serve as extended-spectrum antibiotics, we used ClustalW to align the amino acid sequences of the seven Alrs from different pathogenic bacteria listed as the highest antibiotic resistance threats in the United States by the Centers for Disease Control and Prevention. The pathogens include drug-resistant M. tuberculosis (MT-Alr), gram-positive Vancomycin-resistant Enterococcus faecium (EF-Alr), gram-positive multidrug-resistant Streptococcus pneumoniae (SP-Alr), gram-negative multidrug-resistant Neisseria gonorrhoeae (NG-Alr), gram-negative Carbapenem-resistant Klebsiella pneumoniae (KP-Alr), gram-negative multidrug-resistant Pseudomonous aeruginosa (PA-Alr), and gram-negative multidrug-resistant Acinetobacter baumannii (AB-Alr).

As a control, we first considered the residues involved in catalytic alanine racemization (boxed in blue). These are conserved in all Alrs, as expected given their central importance to the catalytic mechanism. Next, we considered the residues analogous to Asp135/Lys261 (boxed in red). These are either completely (Asp) or semi-completely (Lys/Arg) conserved in all Alrs (Fig 2). Structural alignment of these Alrs further supports spatial conservation (Fig 2). This conservation suggests that the Asp135/Lys261 salt bridge is not unique to MT-Alr; a similar salt bridge may also be critical for catalytic efficiency in other species.

To further assess whether the putative allosteric site located near the Asp135/Lys261salt bridge is conserved across species, we searched the Protein Data Bank for other Alrs and measured the volume of any pockets at the predicted location. Although Alr from M. tuberculosis had the most voluminous pocket by our measurements, many other species also have Alr enzymes with sizable pockets at the same location (Table 4), suggesting that the identified pocket may be structurally conserved across many bacteria. Compounds that bind at this site and interfere with the Asp135/Lys261 salt bridge in MT_Alr may therefore be effective against many bacterial Alrs, including those from gram-positive, gram-negative, and mycobacteria. If this potential mechanism of action is ubiquitous, it could enable the discovery of novel broad-spectrum antibiotics.

Table 4.

Volumes of the putative allosteric pocket across Alr enzymes from multiple species.

PDB ID	Chains	Volume (Å3)	Species
1XFC	AB	755	M. tuberculosis
6G58	AB	701	S. aureus
4LUY	AB	685	C. difficile
3E5P	AC	671	E. faecalis
5YYC	AC	628	A. pseudofirmus
4Y2W	AB	609	C. subterraneus
6Q72	AB	597	B. subtilis
3CO8	AB	593	O. oeni
4QHR	AB	561	A. baumannii
5FAJ	AB	532	S. coelicolor
3B8V	AB	518	E. coli
3UW6	BC	510	G. stearothermophilus
6A2F	AB	504	P. aeruginosa
2ODO	AB	485	P. fluorescens
1VFS	AB	457	S. lavendulae
3S46	AB	452	S. pneumoniae
3HA1	AB	418	B. anthracis
4BF5	AB	127	A. hydrophila
4BEQ	AB	4	V. cholerae
3KW3	AB	0	B. henselae

“Chains” indicates which PDB chains comprise the dimer we used for our pocket-volume analysis. Volume is given in ų, as measured using POVME2.

Conclusion

We report experimental and computational analyses of the Alr enzyme that suggest a novel avenue for antibacterial drug discovery. The steady-state kinetic parameters we measured for wild-type and mutant MT-Alr show that the two Asp135/Lys261 salt bridges at the Alr dimeric interface are critical for enzymatic activity, likely because they impact activity through allosteric regulation. We identified two identical pockets near each salt bridge; compounds that bind in these pockets may similarly disrupt this critical interaction. The work is significant because existing drugs bind to the orthosteric Alr active site and have unacceptable off-target effects. Our study suggests an allosteric approach to Alr inhibition that may bypass these shortcomings.

Sequence and structural alignments of several Alrs from diverse pathogenic bacteria show that this critical salt bridge is generally conserved. If the hypothesized allosteric mechanism is similarly conserved, compounds that bind in the identified pocket could enable extended-spectrum antibacterial drug discovery.

Abbreviations:

Alr

Alanine racemase

MT

Mycobacterium tuberculosis

PLP

Pyridoxal 5-phosphate

PCR

Polymerase chain reaction

SDS-PAGE

Sodium dodecyl sulfate-polyacrylamide gel electrophoresis

IMAC

Immobilized metal affinity chromatography

KLD

Kinase ligase DpnI enzymes

OD

Optical density