An evolutionarily conserved allosteric site modulates beta-lactamase activity

Abstract

Declining efficiency of antibiotic-inhibitor combinatorial therapies in treating beta-lactamase mediated resistance necessitates novel inhibitor development. Allosteric inhibition offers an alternative to conventional drugs that target the conserved active site. Here, we show that the evolutionarily conserved PWP triad located at the N-terminus of the H10 helix directly interacts with the allosteric site in TEM-1 beta-lactamase and regulates its activity. While point mutations in the PWP triad preserve the overall secondary structures around the allosteric site, they result in a more open and dynamic global structure with decreased chemical stability and increased aggregation propensity. These mutant enzymes with a less compact hydrophobic core around the allosteric site displayed significant activity loss. Detailed sequence and structure conservation analyses revealed that the PWP triad is an evolutionarily conserved motif unique to class A beta-lactamases aligning its allosteric site and hence is an effective potential target for enzyme regulation and selective drug design.

Keywords:

• Allostery
• class A beta-lactamases
• inhibition
• PWP conservation
• structure function relationship
• TEM-1

Introduction

Beta-lactamases are ancient enzymes that have been synthesized as a response to the action of beta-lactam antibiotics. With the pressure arising from the irresponsible misuse of beta-lactam antibiotics, these enzymes continue to evolve to increase enzyme fitness. This evolutionary persistence of beta-lactamase enzymes makes antibiotic resistance a global health threat1–6. Currently, there is a vast variety of beta-lactamase inhibitors7 that are co-administered with antibiotics to enhance drug efficacy. Unfortunately, many of the over 1300 different beta-lactamases can rapidly hydrolyze these inhibitors rendering them ineffective. In this context, continued search for novel inhibitors is an area of intense research8,9.

Beta-lactamases are classified into four classes (A, B, C and D) based on sequence similarity10,11. Although they all share a common fold, their active site characteristics differ sufficiently to allow for recognition of different ligands and different catalytic mechanisms. Classes A, C and D are all serine hydrolases, while class B beta-lactamases are zinc binding enzymes. Classification based on ligand recognition is consistent with the Ambler classification but shows that the active site of class A and C beta-lactamases recognizes similar ligands while class B and D beta-lactamases each have their own specific ligands different from classes A and C7,11,12.

Class A beta-lactamases are the most prevalent class in hospital infections. Members belonging to this class have a common fold with a very well conserved active site. Current class A beta-lactamase inhibitors in use, all target this active site. TEM-1 beta-lactamase, which is the most extensively studied member of this class, has been examined using mutagenesis13, X-ray crystallography14, NMR spectroscopy15 and molecular dynamics (MD) simulations16. It has sequence identities ranging from 53 to 79% in pairwise alignments with the ancestral forms of beta-lactamase and share the conserved beta-lactamase fold17.

Allosteric inhibition is an attractive alternative for drug design studies because allosteric inhibitors are more selective than drugs that target the active site, which is conserved across the enzyme family16,18,19. The presence of an allosteric binding site for TEM-1 was first recognized by Horn and Shoichet (2004) who showed that the H10 helix moves away from the protein core to allow binding of allosteric inhibitors to a hydrophobic pocket formed by the two helices, H10 and H11 on one side and the β-sheet composed of β3, β4, β5 on the other. This binding site is located about 16 Å away from the active site serine, S70, of TEM-1 beta-lactamase18. The H10 helix C-terminus forms part of the active site, specifically K243 participating in catalysis and substrate binding, while P226–W229–P252 (PWP triad) residues at the N-terminus participate in aromatic ring stacking that stabilize the helix. Together with the T-shaped aromatic interaction between W229 and W29014, this extensive interaction network modulates the flexibility of H10 and its interactions with other secondary structural elements in the hydrophobic inhibitor binding pocket, suggesting the contribution of these residues to allostery16. Interestingly, the crystallization adjuvant CYMAL-6 binds to this hydrophobic pocket in the crystal structure of SHV-1, which is another class A beta-lactamase20. Binding of CYMAL-6 to the ridge between the H10 and H11 helices of SHV-1 resulted in approximately 20% reduction in reaction velocity21. Similarly, MD simulations and free energy calculations performed with wild type and W229A mutant forms of TEM-1 and SHV-1 beta-lactamases showed that beta-lactamase inhibitor protein (BLIP) binding to the active site was less favorable for the mutant, possibly due to loss of the PWP triad stacking interaction16. The binding of the allosteric inhibitors to the hydrophobic cavity between H10 and H11 and the β-sheet formed by β3-β4-β5, as well as the MD simulation results suggest that the relative orientation of H10 with respect to the β-sheet and its flexibility are critical for beta-lactamase activity. Even a slight disruption to the structural integrity around H10 may be amplified through the secondary structural elements, resulting in a significant impact on the allosteric site in the hydrophobic core.

To assess the significance of the PWP triad for enzyme activity and allosteric regulation, we created point mutations within the PWP triad and performed functional assays and structural characterizations to study the impact of these mutations on enzyme structure and activity. We also examined the conservation of these residues within the primary, secondary and tertiary structures of the beta-lactamases using structural bioinformatics to evaluate the significance and generalizability of our findings.

Materials and methods

Cloning of bla_TEM-1 gene, mutagenesis and transformation

The designed forward (Fwd) and reverse (Rev) bla primers given in Table 1 were used to amplify the bla_TEM-1 gene of pUC18 (our laboratory stock) using polymerase chain reaction with the following conditions: initial denaturation at 95 °C for 3 min, followed by 35 cycles of denaturation at 95 °C for 1 min, annealing at 56 °C for 1 min, extension at 72 °C for 2 min and a final extension for 10 min at 72 °C. The amplified gene was digested with XhoI and NcoI enzymes (Fermentas, Waltham, MA) and ligated into the corresponding sites in the pET28a(+) cloning vector (our laboratory stock). The resulting gene product contained a C-terminal 6xHis tag. The presence of bla_TEM-1 gene in the recombinant plasmid, pET-FGAbla, was verified by sequencing.

Table 1. Primers used in this study.

Name		Sequence
bla	Fwd	5′ CGA CCA TGG ATG AGT ATT CAA CAT TTC CG 3′
	Rev	5′ CAG TGA AGC CTC GAG CCA ATG CTT AAT 3′
W229A	Fwd	5′ CG GCC CTT CCG GCT GGC GCG TTT ATT GCT GAT AAA TCT GG 3′
	Rev	5′ CC AGA TTT ATC AGC AAT AAA CGC GCC AGC CGG AAG GGC CG 3′
W229F	Fwd	5′CG GCC CTT CCG GCT GGC TTC TTT ATT GCT GAT AAA TCT GG 3′
	Rev	5′ CC AGA TTT ATC AGC AAT AAA GAA GCC AGC CGG AAG GGC CG 3′
W229Y	Fwd	5′ CG GCC CTT CCG GCT GGC TAC TTT ATT GCT GAT AAA TCT GG 3′
	Rev	5′ CC AGA TTT ATC AGC AAT AAA GTA GCC AGC CGG AAG GGC CG 3′
P226A	Fwd	5′ CTG CGC TCG GCC CTT GCG GCT GGC TGG TTT ATT GC 3′
	Rev	5′ GC AAT AAA CCA GCC AGC CGC AAG GGC CGA GCG CAG 3′
P252A	Fwd	5′ C ATT GCA GCA CTG GGG GCA GAT GGT AAG CCC TCC 3′
	Rev	5′ GGA GGG CTT ACC ATC TGC CCC CAG TGC TGC AAT G 3′

The point mutations W229A, W229F, W229Y, P226A and P252A were introduced into the bla_TEM-1 gene cloned in pET28a(+) using QuikChange Site Directed Mutagenesis Kit (Agilent Technologies, Santa Clara, CA) following the manufacturer’s protocol. The designed primers were purchased from Integrated DNA Technologies (IDT, Coralville, IA). Primers used for mutagenesis are listed in Table 1. Sequences of constructed mutant plasmids were verified by DNA sequencing.

All the constructed plasmids were transformed using CaCl₂ method into Escherichia coli XL1 cells (our laboratory stock) for archiving and into E. coli BL21 (DE3) cells (TUBITAK-GEBI) for expression22.

TEM-1 beta-lactamase expression and purification

E. coli BL21 (DE3) cells with any of the bla_TEM-1 carrying plasmids were grown in Luria Bertani (LB) medium supplemented with 50 μg mL⁻¹ kanamycin. Precultures were prepared overnight at 37 °C and 180 rpm and 1 mL of preculture was used to inoculate 100 mL of fresh LB. Protein expression was induced with 0.1 mM isopropyl β-d-1-thiogalactopyranoside when OD₆₀₀ of cells reached 0.8 and cells were allowed to grow for 5 h at 30 °C and 180 rpm. Cells were harvested and periplasmic proteins were extracted using osmotic shock procedure23. TEM-1 beta-lactamase in the crude extract was purified by Ni²⁺-affinity chromatography using HisLink^TM Protein Purification Resin (Promega, Fitchburg, WI). Beta-lactamase containing fractions were eluted using a buffer of 50 mM NaH₂PO₄ and 300 mM NaCl with 250 mM imidazole (pH 8.0). Protein concentration was measured with the method of Bradford24. Purity of the protein samples was verified by electrophoretic analysis.

Refolding with urea

After extraction of periplasmic proteins, remaining pellets were resuspended in 5 mL 50 mM Tris buffer (pH 7.8). The samples were disrupted by sonication on ice. Cytoplasmic extracts containing intracellular proteins were removed and the remaining cell pellet was resuspended in 10 mL 8M urea solution to dissolve protein aggregates. Solubilized protein aggregates were electrophoretically analyzed on SDS polyacrylamide gels.

Measurement of beta-lactamase activity

Beta-lactamase was assayed by monitoring the hydrolysis of its substrate CENTA (Calbiochem, San Francisco, CA) at 405 nm as described previously8. The reaction was carried out at 25 °C in a total volume of 1 mL in 50 mM K⁺PO₄ buffer (pH 7.0) with 47 μM CENTA. One Unit (U) of beta-lactamase activity was defined as the amount of enzyme that hydrolyzed 1 μmol of CENTA per minute at 25 °C.

Electrophoretic analysis of proteins and zymogram

Both SDS polyacrylamide gel electrophoresis (PAGE) and native PAGE were used to analyze periplasmic proteins. Native PAGE was carried out using the standard SDS-PAGE protocol25, but without SDS in the gels and the samples. Gels were stained with Coomassie Brilliant Blue G-250 dye to visualize the proteins26. For the zymograms, native gels were equilibrated in 50 mM potassium phosphate buffer (pH 7.0) at room temperature for 45 min (3 × 15 min). Beta-lactamase activity was visualized by the appearance of yellow bands after overlaying the gel with 50 mM K⁺PO₄ buffer (pH 7.0) containing 47 μM CENTA.

Sample preparation for structural studies

Wild-type and mutant beta-lactamase enzymes were exhaustively dialyzed against 20 mM Na₂HPO₄ buffer (pH 7.0) to remove imidazole and NaCl and were spin filtered to prepare the samples used for structural studies. A single 350 μL of protein sample was prepared and used for each enzyme construct to perform circular dichroism, fluorescence spectroscopy and size exclusion chromatography (SEC). Protein concentration was determined by ultraviolet-visible absorbance at 280 nm using an extinction coefficient of 28 085 M⁻¹ cm⁻¹ for constructs retaining W229 and 22 585 M⁻¹ cm⁻¹ for constructs with W229 point mutations, as well as SDS-PAGE gel band quantification.

Circular dichroism

Far-UV CD spectra were acquired in a 2 mm cuvette from 185 to 250 nm with 1 nm steps using a 1 nm bandwidth, and 10 s averaging time, at 20 °C, on an AVIV 62 DS CD Spectrometer equipped with Peltier temperature controller (Boston University Medical Center, Department of Physiology and Biophysics). The thermal melts were performed at 222 nm from 5 °C to 80 °C in 1 °C steps, with a 30 s averaging time for each temperature step.

Tryptophan fluorescence

For each enzyme construct, the sample was excited at 295 nm using a 5 mm slit width. The corresponding fluorescence emission was recorded from 300 to 450 nm in 0.5 nm increments at 20 °C on a Varian Cary Eclipse (Boston University Medical Center, Department of Physiology and Biophysics).

Size exclusion chromatography

After completion of the CD and fluorescence experiments, the samples were applied onto a Superdex-75 analytical column (GE Healthcare, Little Chalfont, UK) in 20 mM sodium phosphate buffer, pH 7.5 and 150 mM NaCl running buffer at a flow rate of 0.5 mL min⁻¹ using the Bio-Rad Dual Chromatography system (Wellesley College, Department of Chemistry).

Sequence and structure conservation

The conservation of W229, P226, P252 and W290 was examined using the Consurf server27,28. TEM-1 beta-lactamase crystal structure (PDB ID: 1ZG4) was used to obtain homologs using CS-Blast with maximal 95% and minimal 35% ID between sequences for homology. The homologs were collected from UNIREF90. The sequences were aligned using MAFFT and the conservation score of each residue was calculated using Bayesian algorithm. Residues with high conservation received a score of 9, while those with low conservation received a score of 1. To check the structural conservation of the PWP stacking interaction, swisspdb29 was used to search for the PWP motif in the 90% nonredundant set of 14 341 X-ray structures. Structural features around the PWP triad were examined for a representative set of beta-lactamases.

Results and discussion

Point mutations of the tryptophan residue in the PWP triad cause subtle structural changes that alter enzyme activity significantly

To assess the individual contributions of the residues within the PWP triad to the structural integrity and activity of the enzyme, two sets of point mutants were studied. The first set consisted of alanine mutants of the three residues in the triad P226A, W229A and P252A and these mutants were used to determine the relative importance of each member of the triad to enzyme activity. The second set consisted of aromatic point mutants of the tryptophan residue in the triad, W229F and W229Y, and were used in conjunction with W229A to gain additional insight into the impact of amino acid size, aromaticity, and polarity for the structurally and functionally important interactions within class A beta-lactamases.

Zymograms of the native gels for W229F and W229Y mutant enzymes possessed partial activity, as indicated by the darker regions on the gel resulting from CENTA hydrolysis, while the W299A mutant lacked activity (Figure 1A and B). These bands corresponded to beta-lactamase bearing regions on the native gels. However, enzymatic activities of the purified enzymes, also measured using CENTA as the substrate, showed a significant loss in the beta-lactamase activity for all three W229 mutants (Table 2). Tryptophan is the largest amino acid occupying a volume of 240 ų. Substitution of tryptophan with tyrosine and phenylanaline (each occupying a volume of 203 ų) is expected to result in a significant decrease in van der Waals interactions or a significant molecular rearrangement to reoptimize these interactions around this side chain without impacting aromaticity. Comparing the amount of activity between wild-type and the W229F or W229Y mutants reveals that both substitutions result in a similar and significant loss of activity indicating that the size but not the polarity of the side chain has an important effect on enzyme activity. Removal of the aromatic side chain from this position, as reflected by the W229A mutant, resulted in an almost complete activity loss, underscoring the significance of size and possibly the aromaticity of the side chain at this position. Hence activity measurements were consistent with the zymogram analysis. Overall, this finding highlights the absolute conservation of the tryptophan residue at this location to ensure optimal arrangement of the structural elements extending into the active site through the hydrophobic pocket of the allosteric region.

Figure 1. Analysis of purified W229 beta-lactamase mutants (W229A, W229F and W229Y) and wild-type beta-lactamase; (A) native gel stained with Coomassie Brilliant Blue G-250, (B) zymogram of the native gel and (C) far-UV CD spectra of the beta-lactamase constructs. The ellipticity data for the mutant samples were adjusted for concentration to be equivalent to the concentration of wild-type (∼5 μM). The spectra were acquired in a 2 mm cuvette from 185 to 250 nm with 1 nm steps using a 1 nm bandwidth, and 10 s averaging time, at 20 °C, on an AVIV 62 DS CD spectrometer equipped with Peltier temperature controller, and (D) fluorescence emission spectra of the beta-lactamase constructs. The samples were excited at 295 nm and the emission was recorded from 300 to 450 nm in 0.5 nm increments at 20 °C. To allow for direct comparison of the fluorescence emission data, the fluorescence intensity for the mutant samples were normalized to wild-type to reflect equimolar (∼5 μM) concentrations.

Table 2. Activity of wild-type and mutant beta-lactamases.

	% Relative activity
Wild-type	100 ± 13.8
W229A	2 ± 0.6
W229Y	16 ± 2.3
W229F	18 ± 3.4
P226A	8 ± 2.4
P252A	15 ± 3.1

Similar activity assays performed on enzymes harboring point mutations flanking the central tryptophan residue in the PWP triad (P226A and P252A) also showed about five- to 10-fold loss in enzyme activity (Figure 2A and B, Table 2). These results clearly demonstrate that the integrity of the entire stacking interaction network established by the PWP triad is essential for optimized enzyme activity.

Figure 2. Analysis of purified P252A and P226A beta-lactamase mutants and wild-type beta-lactamase; (A) native gel stained with Coomassie Brilliant Blue G-250 and (B) zymogram of the native gel.

Analysis of purified enzymes harboring a mutation at W229 on native electrophoretic gels indicates slight differences in the migration profile of each mutant compared to the wild-type, as well as to each other. The mutant forms travelled close to their monomeric molecular weights, however exhibited a more diffuse and slower migration than the wild-type (Figure 1A). We observed similar alterations in the migration profiles of also the proline point mutants on native electrophoretic gels (Figure 2A).

Analytical SEC data for the three W229 mutants confirmed an increase in the apparent molecular weight of the mutant enzymes indicating a slightly larger hydrodynamic radius compared to the wild type. SEC data also revealed diminished chemical stability in the mutants, as evidenced by the presence of multiple proteolysis products that were absent from the wild-type sample, as well as significant increase in the aggregation propensity for the mutants (data not shown). Taken together, these results suggest a more open and dynamic global structure with decreased chemical stability and increased aggregation propensity for the mutants compared to a more tightly packed, chemically stable and soluble wild-type enzyme.

To investigate if the point mutations at W229 resulted in any major unfolding of the secondary structures in TEM-1, we performed CD experiments on the wild-type and as on the three W229 mutants, W229A, W229F and W229Y. During sample preparation, it was observed that all mutant enzymes displayed a significant decrease in solubility and were much more prone to aggregation. However, the data collected on dilute, soluble enzyme samples displayed CD spectra that were virtually superimposable on that of wild-type after appropriate concentration correction, indicating no significant change in the secondary structure of the enzyme between the wild-type and the mutants (Figure 1C).

The thermal melt experiments performed on the wild-type enzyme and the three W229 point mutants showed that for the first major unfolding of the enzyme, the T_m was around 48 °C occurring over a temperature range of ∼23 °C (35–58 °C) for both the wild-type and the three W229 mutants, indicating that there is not a significant loss in thermal stability for the point mutants compared to the wild-type enzyme.

Tryptophan is a dominant intrinsic fluorophore in proteins and its emission maximum is highly dependent upon polarity and/or local environment. In a completely apolar environment, tryptophan displays a characteristic blue-shifted structured emission, whereas a tryptophan residue exposed to water or one with an indole ring engaged in specific H-bonds, displays an emission shifted to longer wavelengths. A protein that contains multiple tryptophan residues is expected to display an emission spectrum that reflects the average environment of all the tryptophan residues, but their contributions may be unequal. TEM-1 beta-lactamase contains four tryptophan residues (W165, W210, W229 and W290). Based on the crystal structure of TEM-1, W229 and W290 are in close proximity within the alpha/beta domain housing the allosteric site, while W165 and W220 are located on the other all-alpha domain of the enzyme. Solvent accessibility analysis for these four tryptophan residues using the program DSSP indicated that two of these residues W210 and W229 are highly solvent inaccessible while W165 and W290 are relatively solvent accessible. In addition, analysis of the TEM-1 crystal structure, which contains bound water molecules, indicates that W210 and W290 are engaged in H-bonds to neighboring water molecules. Therefore, W210 and W229 are expected to display a more blue-shifted emission compared to W210 and W290.

TEM-1 also contains four tyrosine residues that can potentially contribute to the emission spectrum during fluorescence experiments. To avoid the excitation of the four tyrosine residues and to ensure that the collected emission spectrum reflects only the average environment of the tryptophan residues, each beta-lactamase construct was excited at 295 nm and the emission spectrum from 300 nm to 450 nm was recorded. The wild-type beta-lactamase displayed an emission maximum at 347 nm, which is slightly blue-shifted compared to tryptophan in water (350 nm). All the mutant enzymes displayed the expected reduction in emission intensity due to the loss of a W229 in the sequence, but they also revealed both an additional blue-shift of the emission maximum compared to the wild-type, as well as a significant structuring of the spectra (Figure 1D). This finding indicates that at least one of the remaining tryptophan residues experienced a much more hydrophobic environment in the mutant enzymes. Based on the structural information available for TEM-1 the most likely reporter of this change in local environment would be W290.

Taken together, the SEC, CD and fluorescence data suggest that even though the mutant enzymes preserve their secondary structures, there are important local rearrangements impacting both the overall packing of the enzyme resulting in a more open and flexible global 3D structure in addition to altering the hydrophobicity around the allosteric site.

Point mutations in the PWP triad cause a 10-fold decrease in enzyme expression levels

Both the wild-type and the mutant enzymes were expressed and purified identically using E. coli BL21 (DE3) expression system followed by Ni²⁺ affinity chromatography (Figure 3). While the wild-type enzyme was expressed and purified efficiently (∼0.89 mg L⁻¹ culture), expression levels of the mutant enzymes were 10-fold lower (∼0.08 mg L⁻¹ culture). To rule out the possibility that beta-lactamase was partitioning to the cytoplasmic extracts during purification, the remaining pellets for the mutants were resuspended after the extraction of periplasmic proteins, in 5 mL 50 mM Tris buffer (pH 7.8) and sonicated on ice. Electrophoretic analysis showed that neither the cytoplasmic extracts nor the urea solubilized protein aggregates post extraction from cell pellet contained any detectable beta-lactamase.

Figure 3. Electrophoretic analysis of TEM-1 beta-lactamase on denaturing gels. Protein standard (25 kDa and 35 kDa) (lane 1), wild-type (lane 2) and W229A (lane 3) beta-lactamases following purification with Ni²⁺-affinity chromatography.

One other explanation for the low yield for mutant enzymes could be that only a small fraction of the synthesized mutant beta-lactamase acquires a native-like conformation, while the remaining polypeptides fold into alternative conformations with decreased chemical stability, resulting in a faster clearance in the cell via intrinsic proteases.

The PWP triad is sequentially and structurally conserved in class a beta-lactamase family

The sequence conservation of the PWP triad was examined using the Consurf server27,28. In the Consurf scoring scheme, 0 represents no conservation and 9 represents 100% conservation. Of the 415 homologs of TEM-1 beta-lactamase (PDB ID: 1ZG4) identified using CS-Blast, all the members annotated in Uniprot (The UniProt Consortium 2015)30, belonged to the class A beta-lactamases. The calculated individual conservation scores after alignment of these sequences were 9 for both P226 and W229 indicating that these residues of the PWP triad are absolutely conserved in class A beta-lactamases, similar to other functionally conserved regions such as the active site catalytic residues (S70, K73, S130, E166, K/R244) or the omega loop ligand binding residues (R164, E166) that also have ConSurf scores of 9 (Figure 4). P252 has a conservation score of 8 and it is conserved in 97.75% of the examined sequences.

Figure 4. The PWP triad in class A beta-lactamase TEM-1 and the proline-tryptophan rich region in class C beta-lactamase AmpC; (A) TEM-1 beta-lactamase structure¹⁴ colored according to conservation score as calculated by Consurf²⁷. The backbone is shown with cartoon representation and P226, W229 and P252 side chains are shown as sticks. The most variable positions (ConSurf scores: 1–3) are colored red (light gray in print), intermediately conserved positions (ConSurf scores: 4–6) are white while the most conserved positions (ConSurf scores: 7–9) are colored blue (dark gray in print), (B) AmpC (class C, PDB ID: 1L2S) structure shown in cartoon representation with tryptophan and proline residues shown in stick representation. Tryptophan residues are in green and proline residues are in orange in the online version. The structure figures were created using VMD⁵⁵. (The color figure can be found in the online version of the manuscript.)

To determine the prevalence for the structural conservation of the PWP stacking interaction in proteins, swisspdb29 was used to search for the PWP motif in the 90% nonredundant set of 14 341 X-ray structures in the Protein Data Bank (PDB). When the search was conducted using the entire triad (P226–W229–P252), all of the nine proteins identified (Beta-lactamase from Scaphirhynchus albus 1BSG, Cephalosporinase 1HZO, BS3 1I2S, Toho-1 1IYS, TEM-1 1JTD, 1M40, SHV-2 1N9B, SHV-1 2G2U, CTX-M-9 2P74, Sed 3BFF) were class A beta-lactamases. When the search was repeated for paired selection of P226–W229 or W229–P252 only a small subset of the identified structures (13/339 and 9/31, respectively) belonged to the class A or class C beta-lactamases with the remaining structures representing a diverse set of scop folds. These results suggest that while proline–tryptophan or tryptophan–proline interaction is common in many different protein–protein or domain–domain interaction interfaces31, the PWP triad formation identified in TEM-1 is unique to class A beta-lactamases.

Table 3 shows the list of representative beta-lactamase structures for which we examined the structural features around the PWP triad. The comparison of the two crystal structures of TEM-1 beta-lactamase with and without the allosteric inhibitor (1PZO and 1ZG4, respectively) shows a significant change in the orientation of the H10 helix and an accompanying decrease in the distances between the three residues of the PWP triad in the presence of the inhibitor, further indicating that the PWP triad and the hydrophobic groove along H10 communicate with each other.

Table 3. Representative beta-lactamase structures examined.

Name	Class	PWP conservation	PDB code	Reference
TEM-1	A	Yes	1ZG4, 1BTL, 1M40, 1PZO	14,45,46,18
Toho-1	A	Yes	1IYS	47
CTX-M9a	A	Yes	2P74	48
SHV-1	A	Yes	1SHV, 2ZD8	21,49
NDM-1	B	Different fold	4EYB, 4RL0	50,51
AmpC	C	Prolines are conserved. Tryptophan is not conserved	1L2S, 1KE4, 2FFY	33,52,53
OXA-10	D	Helix10 missing	3LCE, 1K55	54,35
PNCA (3GA)	Precambrian	Yes	4C6Y	37
GCNA (2GA)	Precambrian	Yes	4B88	36
ENCA (1GA)	Precambrian	Yes	3ZDJ	36
ALL-CON	Precambrian	Yes	4C75	37

We have previously shown by ligand based network analysis that class A and class C beta-lactamases bind to similar inhibitors7, suggesting that the active site is structurally similar for ligand recognition. On the other hand, class A active site inhibitors such as clavulanic acid have lower effect on class C beta-lactamases32. Since the two classes have been reported to display similar ligand binding specificities to their active sites with varying affinities, we performed an additional sequence and structural analysis, this time focusing on the allosteric binding site, the PWP triad and its surrounding local environment. Our findings indicate that in class C beta-lactamase structures, represented by ampC beta-lactamase (Figure 4B)33, both H10 and one of the prolines (corresponding to P252) are strictly conserved elements, but the PWP triad residues are sequentially rearranged to preserve the overall hydrophobic environment at the H10 N-terminus. Based on the sequence alignment of 15 selected class C beta-lactamases including AmpC, CMY-2, CMY-6, CMY-10 and FOX-7 as well as DHA-1, the W229 position of the class A PWP repeat is either conserved or replaced by Y or L, while replacements and rearrangements around the P226 position allow for direct hydrophobic contributions from neighboring W312, which can also be present as Y or L. Our findings show that while the PWP triad is unique to class A beta-lactamases, the properties and structural arrangement of corresponding residues in class C beta-lactamases create a similar hydrophobic environment to that of class A beta-lactamases. These results have important implication for drug design, as they clearly highlight the importance of the hydrophobic environment in allosteric ligand binding for these two classes of beta-lactamases.

On the other hand, class B beta-lactamases, also called metallo beta-lactamases represented by NDM-1 (PDB ID: 3ZR9)34 have a completely different fold. Class D beta-lactamases, represented by the OXA family (PDB ID: 1K55)35, lack the H10 helix and therefore these two classes are not expected to respond to allosteric inhibitors developed against the H10 helix site.

The PWP triad residues and their conformation are conserved even in the four resurrected precambrian beta-lactamases, GNCA36, ENCA36, PNCA37 and ALL-CON37. Furthermore, MD simulations on both TEM-1 and precambrian beta-lactamases show that H10 and H11 helices and W229 display similar flexibility while P226 and P252 are rigid17 suggesting the importance of the PWP triad dynamics in modulation of allostericity.

The contribution of CH-π hydrogen bonds and specifically proline–tryptophan interactions to protein–protein interactions and to protein stability has been recognized for over forty years38–41. In most of the domain–proline rich motif interactions, the proline residues and the hydrophobic residues are coplanar31 and the tryptophan residues interact with proline via the tryptophan “face” rather than the “edge”42. Moreover, in most CH-π H-bonds that involve tryptophan–proline interaction, the rings are aligned in an off-centered or displaced parallel stacking conformation43. These tryptophan–proline interactions established in the stacked and L-shaped configurations are very stable and can contribute to the stability of the folded state via intramolecular contacts as well as to intermolecular recognition44. The PWP triad geometry is also in an off-centered stacking interaction in class A and the precambrian beta-lactamases, suggesting that this stable and highly conserved interaction is critical for the function and dynamics of beta-lactamases in the evolutionary process.

Conclusion

Beta-lactamases are ancient enzymes and as such their structure, evolution and catalytic mechanism have been examined in extensive detail but the discovery of an allosteric site formed by the outward rotation of H10 from the enzyme core in TEM-1 and SHV-1 has opened a new gate toward the design of novel inhibitors that target this hydrophobic binding site. In this work, we used computational and experimental analysis to focus on the PWP triad located at the N-terminus of the H10 helix to determine its significance for enzyme structure and activity. Our data shows that the PWP triad communicates with this allosteric site and regulates its structure, dynamics and stability. Point mutations within this triad result in significant or complete loss of enzyme activity as well as a more open and dynamic global structure with decreased chemical stability and increased aggregation propensity. The decrease in the overall solubility of the enzyme limited experimental studies to low working enzyme concentrations but the importance of size, but not polarity, was clearly established by detailed biophysical analysis of the W229Y and W229F mutants. The relative importance of the loss of aromaticity versus further reduction in size in almost a complete loss of activity for the W229A mutant remains as an answered question. Although the mutant enzymes preserve their secondary structures, local arrangements in the structure alter the overall packing of the enzyme resulting in a more open and flexible global 3D structure as well as decreased hydrophobicity around the allosteric site, which explains the significant loss in enzyme activity.

Both the sequence analysis of the triad using ConSurf server and the analysis of the existing crystal structures with swisspdb showed that while the pairwise proline-tryptophan or tryptophan–proline interactions were common in many different protein–protein or domain–domain interaction interfaces, the PWP triad is highly conserved and unique to class A beta-lactamases. The binding of the allosteric inhibitors to the groove formed between H10 and H11 suggests that H10 is a mobile element. Indeed, we showed that the dynamics of this region is critical for the communication with the active site16. Orthosteric inhibitors that recognize class A beta-lactamases7 can also bind class C beta-lactamases7, although with weaker affinity. On the other hand, conservation of the hydrophobic environment at the H10 N-terminus in both classes suggests that specific allosteric inhibitors directed for the H10 groove might be designed to recognize and inhibit not only class A beta-lactamases but also class C beta-lactamases and their specificity and selectivity might be optimized for each class with minor adjustments to the same core design. These computational analyses do not provide a direct mechanistic explanation of how the activity is modulated by the PWP triad, but its conservation at the sequence and structure level provide valuable clues about its evolutionary significance. Taken together, the findings of this work underscore the importance of the PWP triad in the regulation of enzyme activity in class A beta-lactamases through a unique interaction network at the recently identified allosteric site and highlight the potential for this signature region to become an effective inhibitor design target.

Declaration of interest

The authors report no declarations of interest. This work was supported by TUBITAK Research Grants 113M533 (BSA) and 109M229 (EO), 114M179 (EO), Marmara University Research Foundation FEN-C-YLP-181208-0290 (BSA) and Boğaziçi University Research Foundation 14A05D7 (EO).

Acknowledgements

We thank Dr. Olga Gursky for providing Circular Dichroism and Fluorescence Spectroscopy facilities and for many helpful discussions.