Triazole-substituted phenylboronic acids as tunable lead inhibitors of KPC-2 antibiotic resistance

Inhibition of β -lactamases is a promising strategy to overcome antimicrobial resistance to commonly used β -lactam antibiotics. Boronic acid derivatives have proven to be effective inhibitors of β -lactamases due to their direct interaction with the catalytic site of these enzymes. We synthesized a series of phenylboronic acid de- rivatives and evaluated their structure-activity relationships as Klebsiella pneumoniae carbapenemase (KPC-2) inhibitors. We identified potent KPC-2 inhibitors 2e & 6c (Ki = 0.032 μ M and 0.038 μ M, respectively) that enhance the activity of cefotaxime in KPC-2 expressing Escherichia coli . The measured acid dissociation constants (pKa) of selected triazole-containing phenylboronic acids was broad (5.98 – 10.0), suggesting that this is an additional property of the compounds that could be tuned to optimize the target interaction and/or the physi-cochemical properties of the compounds. These findings will help to guide the future development of boronic acid compounds as inhibitors of KPC-2 and other target proteins.


Introduction
Extensive use of β-lactam antibiotics for the prophylaxis and treatment of bacterial infections has resulted in the rapid spread of multidrug resistance (MDR) to previously susceptible organisms [1]. The production of β-lactamase enzymes is a key factor in the emergence of resistance [2,3]. Worryingly, an expansion of resistance to carbapenems has been observed in Escherichia coli, Klebsiella and Enterobacter species due to the expression of carbapenemase β-lactamases with an extended spectrum of activity [4,5]. The most widely distributed serine carbapenemases are Klebsiella pneumoniae carbapenemases KPC-2 and KPC-3 [4,5]. These class A β-lactamases [6] hydrolyses almost all β-lactams and are especially difficult to inhibit using current β-lactamase inhibitors. This appears to be a consequence of the shallow active site of KPC proteins which allows hydrolysis of bulkier β-lactams, it also has low sequence identity compared to other class A β-lactamases (e.g. 50% CTX-M-1, 39% TEM-1 and 35% SHV-1) [7]. Recent findings also suggest a more hydrophobic active site compared to CTX-M and TEM β-lactamases might contribute to the ability of KPCs to hydrolyze a broader range of β-lactams [8].
'First generation' BLIs, which included clavulanic acid, sulbactam and tazobactam ( Fig. 1), mimic the fused ring system of β-lactams and were approved by the FDA for use in combination with selected β-lactam antibiotics. However, resistance evolved rapidly against these structurally similar BLIs. Furthermore, they are inactive against KPCs [7]. The 'second generation' BLI avibactam is a diazabicyclooctane (DBO) (Fig. 1) and was approved by the FDA in 2015 for the treatment of complicated infections in combination with ceftazidime [11]. Avibactam forms an inhibitory acyl-enzyme complex with several serine β-lactamases, however, resistant KPC mutants including KPC-2 and KPC-3 have been identified [12,13]. The structurally related DBO relebactam ( Fig. 1) was approved by the FDA in 2019 for use in combination with imipenem and cilastatin (a renal dehydropeptidase inhibitor). Although evidence suggests relebactam is a potent inhibitor against KPC-2, the effectiveness of the relebactam-imipenem combination appears to be similar to that of avibactam-ceftazidime against KPC-2, 3 and 4 producing Klebsiella pneumoniae [14,15]. Cyclic boronates are among the newest approved BLIs. They inhibit KPCs by interacting covalently with the catalytic serine residue of the β-lactamases via their vacant boron p-orbital, mimicking the tetrahedral 'transition state' intermediate of the substrate-enzyme complexes (Scheme 1) [16]. There have been a number of reports in the past decade which suggest that boronate BLIs have inhibitory activities against not only β-lactamases but also   penicillin-binding proteins (PBPs) [17,18]. Vaborbactam co-formulated with meropenem was the first cyclic boronate approved by the FDA in 2017 [19]. Taniborbactam (VNRX-5133), a bicyclic boronate analogue has recently completed phase III clinical trials in combination with cefepime; it has broad-spectrum activities against both serine and class B metallo β-lactamases unlike the compounds above [20]. X-ray crystal structures of taniborbactam in complex with serine (CTX-M-15) and metallo (NDM-1) β-lactamases revealed that, as expected, the boronic acid acts as an electrophile and binds covalently to serine β-lactamases whereas the boronic acid and carboxylate moieties coordinate via their oxygen atoms to metallo β-lactamases [5,21]. Although no clinical resistance has been reported to date, this may emerge as use becomes more widespread [9,22].
Despite significant progress, the identification of new classes of BLIs is still required to overcome resistance. The majority of the boronic acidtype transition state analogues are either glycylboronic acids (Fig. 2a) or cyclic boronates (Fig. 2b). Only a few phenylboronic acids (Fig. 3) have been reported with low to moderate efficacy against specific β-lactamases (e.g. AmpC and OXA24/40). Additionally, most previous work was directed towards elucidating structure-activity relationships (SARs) Fig. 2. a. Selected glycylboronic acid inhibitors and their reported of β-lactamase inhibitory activities. b. Selected cyclic boronates and their reported of β-lactamase inhibitory activities [17].
Our previous work describes a promising phenylboronic acid scaffold based on the known KPC-2 inhibitor 3-nitrophenylboronic acid (3-NPBA) (Fig. 3). We identified inhibitors of KPC-2 from a small compound library that were capable of reversing cefotaxime (CTX) resistance in KPC-2 overexpressing bacteria [30], a model system that mimics clinical strains of E. coli that express KPC-2 [31]. Herein, we report the design and synthesis of a new generation of phenylboronic acid inhibitors which builds on this previous work. The subsequent SAR studies identified several structures with promising KPC-2 inhibitory activity. Moreover, the pK a of the phenylboronic acids may also influence KPC-2 binding affinities and be a tunable property in the future development of this class of compounds.

Screening of phenylboronic acid ligands
Initially we screened a series of eleven commercially available phenylboronic acids (BAs) and 3-NPBA to determine their ability to potentiate the activity of cefotaxime (CTX) against E. coli BL21 (DE3) that express plasmid-mediated KPC-2 that confers β-lactam resistance.
Following the previously reported protocol, the compounds (50 μg/disk) were combined with CTX (30 μg/disk) and exposed to the E. coli in a disk diffusion assay [30]. The combinations were defined as susceptible (S), intermediate (I), or resistant (R) based on their zone of inhibition according to CLSI guidelines [32]. Compounds with phenyl substituents performed well (BA1, 2 zone of inhibition >29 mm) as did three out of  four compounds with a meta-NO 2 substituent (3-NPBA, BA10, 11).
Bulky ortho-substituents were not well tolerated (BA4, 8), whereas the smaller ortho fluorine substituents performed better (BA5, 7, 10, 11) ( Table 1). The data confirm that 3-NPBA and its analogues, including fluorine substituted analogues are promising leads for reversing β-lactam resistance due to KPC-2 overexpression. The compounds were docked covalently into a molecular model of KPC-2 derived from a co-crystal structure with 3-NPBA (PDB ID: 3RXX) [33]. First, 3-NPBA was docked into the binding site adjacent to the catalytic serine residue (S70), giving a reasonable overlay with the ligand conformation in the co-crystal structure as described in our previous work [30]. We docked the remaining compounds to the protein in an analogous manner. The docked conformations predicted that bulky substituents at the meta and para positions of the ligands occupy hydrophobic sites within the binding pocket ( Fig. 4). However, ortho-substituents are accommodated less well, and compounds BA4 and BA8 did not fit into the binding site defined in the crystal structure. The docking data are broadly consistent with the enhancement of CTX activity observed ( Table 1), suggesting that KPC-2 inhibition may play a role in the reversal of resistance. However, the efficacy of the compounds in the disk diffusion assay reflects not only ligand binding but also their solubilities, and bacterial membrane permeabilities, therefore the inhibition of KPC-2 enzymatic activity was also evaluated for subsequent compounds.

Design and synthesis of novel analogues of phenylboronic acid
Next, we explored the SAR of meta-and para-substituted phenylboronic acid compounds for inhibition of KPC-2. We demonstrated previously that analogues of phenylboronic acids containing a 1,4disubstituted 1,2,3-triazole at the meta-position of a phenylboronic acid could form a hydrogen bond with residue Thr237 when covalently docked in silico [30]. To explore the optimization of the linker between the thiophenyl and the phenyl ring, we aimed to replace the triazole in the 2 series compounds and illustrate their SAR regarding the inhibition of KPC-2. In addition, we evaluated a series of fluorine substituted analogues (5 and 8 series compounds) and explored the SAR of these compounds in relation to inhibition of KPC-2.
A synthetic route consisting of 1-3 steps was used to furnish the desired compounds. Azidomethyl derivative 1 was synthesized as described in our previous work by S N 2 substitution of the corresponding bromomethyl derivative. The 1,4-substituted triazole analogues 2a-g and the new analogues 2h-j were synthesized via copper (I) catalyzed 'click' reaction using the conditions described previously [30]. Compound 2k was synthesized by hydrolysis of the ester 2j under basic conditions (Scheme 2a). Compounds 4a-f from our previous study were synthesized in a similar manner from 4-azidomethylphenylboronic acid 3 [30]. 3-Azidophenylboronic acid 5a was produced from the aniline by diazo transfer and substitution, and then the triazole synthesis followed a similar route to the methyl-1,4-substituted 1,2,3-triazole analogues (Scheme 2b). A small set of amide derivatives 7a-f were also synthesized as triazole isosteres through coupling reactions of the aniline with carboxylic acids (Scheme 3). Since the compounds can be synthesized through a straightforward coupling reaction, inhibitors with bulkier substituents (e.g. benzothiophene) were added to this sub-library to explore potential hydrophobic contacts further. We also synthesized a selection of fluorine substituted derivatives of 2e and 6c from commercially available phenylboronic acid precursors (5b-d). Hence, we synthesized compounds 8b-d using the preferred 3-thiophenyl substituent at the 4-position of the triazole (Scheme 4a). We chose 5-bromomethyl-2-fluorobenzene boronic acid pinacol ester as a starting point to generate 5e by S N 2 substitution with azide following by hydrolysis of the boronate ester. Copper (I) catalyzed click reaction was used to afford compound 8e (Scheme 4b).

Structure activity relationships 2.3.1. Evaluation of methyl-1,4-substituted 1,2,3-triazole analogues
Initial evaluation of our previously synthesized 2a-g (meta) & 4a-f (para) analogues (Fig. 5a) indicated that they have promising antimicrobial activity against KPC-2 [30]. Therefore, we synthesized 2h-k to probe the SAR further. To investigate the KPC-2 enzyme inhibitory activity, assays were developed using nitrocefin, a chromogenic β-lactam substrate that undergoes an increase in UV absorbance at 482 nm (ϵ 482nm = 17,400 M − 1 cm − 1 ) when hydrolysed by β-lactamases [29]. The molecular mass of purified KPC-2 was determined by MALDI-MS (28,  [7,13]. Analysis of the K i data for the test compounds (  [30]. The in silico docking of 2h-k to KPC-2 suggested that they may be unable to form hydrogen bonds with Thr237 (Supporting info Fig. S1). By comparing the docked structures of, for example, 2e & 4e, the meta analogue 2e appears to fit in the binding pocket by forming a key hydrogen bond between the 2-N of the triazole and Thr237 of KPC-2. On the contrary, the para-substituent of compound 4e formed an alternative π-π stacking interaction with Trp105 (Fig. 5b).

Evaluation of triazole and amide analogues
Based on the previous docking results, we proposed that the key difference between meta-and para-analogues was the ability of the triazole 2-N to form a hydrogen bond with Thr237 [30]. Docking studies suggest that by truncating the methylene linker in 2 series (6 series) or replacing it with an amide (7 series), the H-bond acceptor 2-N of the triazole or carbonyl oxygen of the amide would be located closer to the H-bond donor Thr237 (Fig. 6). To explore this, we synthesized 6a-h and amide analogues 7a-f.
The 6 and 7 series compounds were generally less active, although compound 6c is one of the most potent triazole-containing molecules with a K i of 0.038 μM for KPC-2, confirming the important role of the 3thiophenyl substituent. Interestingly, the 2-thiophenyl analogues with either a triazole or amide linker (6d & 7c) are less active (6d, K i = 0.71 μM, 7c 2.9 μM). Kinetic studies on the 6 and 7 series revealed a difference in the IC 50 and K i values between the 3-thiophenyl and 2-thiophenyl substituted compounds: approximately 19-fold 6c c.f. 6d, and 3-fold 7b c.f. 7c (Table 3). The compounds retained activity in disk diffusion assays in combination with cefotaxime. Only compound 7a displayed intermediate sensitivity.

Evaluation of fluorine substituted phenylboronic acid derivatives
Guided by the screening results and the SAR data, fluorinecontaining derivatives of the compounds (5b-e & 8b-e) were designed and synthesized. KPC-2 enzyme inhibition assays showed that compounds 5a, c, and e had K i values ≈ 0.5 μM. Amongst the azido analogues 2,3-di-F and 2,4-di-F substitution patterns were less preferred than 2-F (48-and 14-fold decrease in K i values, 5d or 5b c.f. 5c), whereas 2-F substitution was tolerated with a minor change in activity (5c 0.57 μM c.f. 5a 0.45 μM). Compared to the azido analogues, triazole derivatives including di-F-substituents have improved activity (4-25-fold) ( Table 4). This confirmed our finding that by constructing metasubstituted analogues of phenylboronic acid with a 1,2,3-triazole linker and a 3-thiophenyl substituent at the 4-position of triazole, we achieved consistently good activity against KPC-2. However, overall the fluorine substituted compounds were less active than the non-fluorinated triazole analogues. To further explain the SAR among fluorinated derivatives with KPC-2, the analogues were docked into the KPC-2 protein (PDB id: 3RXX). The results indicate that the fluorine substituents did not significantly alter the docked conformations relative to the unfluorinated compounds (e.g. comparing 6c, Fig. 4a with 8b, Supporting info Fig. S1d). Hence, there are factors beyond the predicted bound conformations that may account for the significant reduction in activity for these analogues.
The acid dissociation constant (pK a ) value of phenylboronic acids can vary substantially with different substitution patterns (ranging from 4 to 10) [21,34]. Therefore, the relationship between the pK a value of selected compounds and their KPC-2 inhibitory activity was evaluated (Table 5). First, the change in UV absorption at 268 nm (where the difference in absorbance between acidic and basic forms is the greatest; Supporting information, Fig. S4) was used to determine the change from the trigonal form of the boronic acid to the tetrahedral form of the boronate conjugate base for the nitrophenylboronic acid compounds (3-NPBA, BA11). We then measured the pK a value of selected compounds (Table 5; Supporting information, Fig. S5) and the values were comparable to those reported in the literature reported for 3-NPBA (7.24 measured cf. 7.1 reported) and BA11 (6.03 measured cf. 6.0 reported) [34]. Subsequently, the pK a value of other analogues (NO 2 , N 3 & triazoles) were determined following a similar procedure. The in vitro KPC-2 inhibitory activity ranking of the phenylboronic acid derivatives is in the order H > 2-F > 2,4-di-F > 2,3-di-F (compounds 5a-d) and the pK a values decrease in this order from 8.2 to 6.0 (Table 5). We found pK a values increased in all cases when converting a nitro group to an azide (by 0.2-1 units), and when azides are modified to triazoles the pK a value increased by 1-2 units (e.g. 5c to 8c & 5e to 8e). These initial findings provide a possible further consideration for the SAR of phenylboronic acids as KPC-2 inhibitors. By derivatising the scaffold and increasing the pK a of the phenylboronic acid, we may expect a stronger covalent binding profile, although this requires evaluation with a broader series of analogues. Moreover, an optimised pK a may contribute to solubility and cell permeability in cellulo and in vivo.

Susceptibility testing
Our earlier work used disk diffusion assays to identify a significant difference between the regioisomers 2 and 4 against KPC-2 plasmidmediated CTX-resistant E. coli [30]. However, the scope of this synergy with other β-lactam antibiotics (e.g. other cephalosporins and penem antibiotics) and the use of lower inhibitor concentrations required further evaluation. Therefore, we investigated the synergistic effect of inhibitors in combination with CTX or meropenem (MEM), a penem-type β-lactam, to validate whether synergy could be achieved by inhibiting the β-lactamase in cellulo.
As expected from the enzyme kinetic studies, compound 2k failed to increase the growth inhibition (clearance) zone compared to 3-NPBA in combination with CTX, whereas other novel 2 analogues (2h-j) achieved a significantly larger zone of inhibition compared to 3-NPBA (Table 2). In addition, the difference of the clearance zone of 2k compared to the control CTX disc soaked in DMSO (solvent) was significantly smaller than that of 2j (Student's t-test: p = 0.0008). This suggested that a carboxylic acid is less preferred than the ethyl ester at the 4-position of the triazole for activity in cellulo, although the trend was not maintained for 6h and 6g. The truncated 1,2,3-triazole 6 and amide 7 analogues were less active as inhibitors of KPC-2, however, the 6 analogues were comparable to 3-NPBA in combination with CTX (Tables 1 and 3). Consistent with our previous observations for 2 analogues, the clearance zone of 6c (3-thiophenyl analogue) is significantly larger than that of 6d (2-thiophenyl isomer, Student's t-test: p = 0.0375), and the clearance zone of 7b (3-thiophenyl analogue) was significantly larger than its 2thiophenyl isomer 7c (Student's t-test: p = 0.0019).
The previous observations on the KPC-2 inhibitory activities of the fluorine substituted analogues were confirmed in the disk diffusion assays with CTX. The 2,3-difluoro-analogues 5d & 8d failed to sensitize the strain against CTX, supporting the hypothesis that they are less preferred than the 2,4-isomers 5b or 8b, respectively. All other fluorine substituted compounds potentiated CTX activity against the E. coli strain successfully (Fig. 7a). The correlation between these results and the in vitro data indicates the disk diffusion assay can be employed as a cellbased screening method for phenylboronic acid inhibitors with appropriate antimicrobial partners. Generally, the compounds performed poorly in combination with meropenem compared to CTX in the disk diffusion assay (Fig. 7) [35].
To further explore the synergistic effects of these compounds with the antibiotics, MIC tests were conducted as reported previously [30].
The compounds do not have an antimicrobial effect (MIC >64 μg/ml) in the absence of an antibiotic. However, as part of a fixed-dose (50 μg/ml) combination most of the compounds were able to sensitize the phenotypic strain against both CTX and MEM, except for 5b which failed to potentiate MEM ( Table 6). The SARs of the inhibitors was not clear at this test concentration, so we reduced the inhibitor concentration to 5 μg/mL and the MIC values of CTX & MEM were remeasured. Even though the MIC of most tested inhibitors falls into the 'susceptible' category of the CLSI guidelines, a differentiated profile was observed. Amongst the compounds, 2e, 2f and 8e are the most active in cellulo, decreasing the MIC values of CTX as well as MEM from 16 or >64 μg/ml to ≤0.06 μg/ml (~512 to >1000-fold increase in susceptibility), respectively. These findings correlate with the KPC-2 inhibitory activities of the compounds which indicated that 2e (K i = 32 nM) was the best inhibitor of KPC-2. Some inhibitors showed divergent synergistic effects with CTX vs. MEM. For instance, 7b and 8b are potent in combination with CTX (over 128-fold increase in susceptibility), whereas the activity when partnered with MEM was reduced (16-fold increase).
With the activities of compounds confirmed against the inducible KPC-2 expressing E. coli strain, we then moved on to test selected inhibitors in combination with β-lactams against clinical isolates of Gramnegative bacteria (E. coli, Proteus mirabilis and K. pneumoniae) expressing β-lactamases belonging to different classes to determine the resistance reversing activity (Supporting information, Table S1). Firstly, the antimicrobial activity of the KPC-2 inhibitors alone was tested against both Gram-positive (Staphylococcus aureus) and Gram-negative (E. coli) control strains. The data indicate that they do not possess antimicrobial activities on their own. The compounds were then tested against E. coli expressing class A and C β-lactamases (TEM-1 and CMY-4 respectively). These strains were susceptible to meropenem (MEM) and resistant to cefotaxime (CTX). The combination of CTX and inhibitors i.e. 5a (MIC 1 μg/ml) and 8b (MIC 0.12 μg/ml) were able to potentiate CTX against these resistant strains. This may be due to the inhibition of CMY-4 (a class C β-lactamase) by these compounds, and this finding could guide the future design of broader spectrum BLIs. The MIC values of CTX and MEM were also investigated against P. mirabilis and K. pneumoniae strains harbouring β-lactamases from all four classes. The P. mirabilis clinical isolate was susceptible to MEM (MIC 0.25 μg/ml) and K. pneumoniae clinical isolate harbouring OXA-48 was susceptible to both CTX (MIC 2 μg/ml) and MEM (MIC 0.5 μg/ml). However, K. pneumoniae strains harbouring KPC-3 and NDM-1 were resistant to both antibiotics, because there was no significant difference in antimicrobial activity in the presence or absence of inhibitors. This may indicate that the compounds are inactive against the β-lactamases expressed in these strains.

Cytotoxicity
Selected compounds from 1-8 and 3-NPBA were tested for their ability to inhibit the growth of human cells (HEK-293, a human embryotic kidney cell line) to give an indication of their potential toxicity. The results of the MTT assays at concentrations of 5 and 50 μg/ ml suggested that the compounds were well tolerated in the presence or absence of 30 μg/ml cefotaxime (>80% viability after 24 h) (Supporting info Table S2). This suggests that the compounds of this type might be suitable for further in vivo evaluation and preclinical development.

Conclusion
Here, we have presented a screening approach based on the crystal structure of KPC-2 with a phenylboronic acid inhibitor (3-NPBA) that lead to the development of a new generation of phenylboronic acid inhibitors with micromolar to nanomolar activities (e.g. 2e & 6c K i = 0.032 μM and 0.038 μM, respectively) against KPC-2. A clear SAR was identified for KPC-2 inhibition which included indications of a relationship between pK a (that ranged from 5.98 to 10.0) and the K i of phenylboronic acid that warrants further investigation. These findings were further supported by susceptibility tests on the compounds in combination with cefotaxime or meropenem in cellulo (e.g. the combination of 2e with CTX or MEM had MICs of <0.06 μg/mL). The lead phenylboronic acids were not cytotoxic in the presence of cefotaxime against a representative human cell line. These findings should help to guide the development of novel boronic acid inhibitors of β-lactamases to tackle drug resistant Gram-negative bacterial infections.

Author contributions
The manuscript was written through contributions of all authors. All authors have given approval to the final version of the manuscript.

Funding sources
The authors acknowledge UCL School of Pharmacy for financial support.

Declaration of competing interest
The authors declare that they have no known competing financial interests or personal relationships that could have appeared to influence the work reported in this paper.