Synthesis and biochemical evaluation of cephalosporin analogues equipped with chemical tethers

Molecular probes typically require structural modifications to allow for the immobilisation or bioconjugation with a desired substrate but the effects of these changes are often not evaluated. Here, we set out to determine the effects of attaching functional handles to a first-generation cephalosporin. A series of cephalexin derivatives was prepared, equipped with chemical tethers suitable for the site-selective conjugation of antibiotics to functionalised surfaces. The tethers were positioned remotely from the β-lactam ring to ensure minimal effect to the antibiotic's pharmacophore. Herein, the activity of the modified antibiotics was evaluated for binding to the therapeutic target, the penicillin binding proteins, and shown to maintain binding interactions. In addition, the deactivation of the modified drugs by four β-lactamases (TEM-1, CTX-M-15, AmpC, NDM-1) was investigated and the effect of the tethers on the catalytic efficiencies determined. CTX-M-15 was found to favour hydrolysis of the parent antibiotic without a tether, whereas AmpC and NDM-1 were found to favour the modified analogues. Furthermore, the antimicrobial activity of the derivatives was evaluated to investigate the effect of the structural modifications on the antimicrobial activity of the parent drug, cephalexin.


Introduction
The controlled functionalisation of surfaces is imperative for the preparation of functional materials. This a key step in the preparation of many (bio)sensors, for example, which are designed for selective and sensitive detection of analytes. The immobilisation or bioconjugation of a molecular probe oen requires structural modication to introduce a functional handle, able to react with a desired substrate. A common approach is to attach a linker to the probe, such as a bifunctional polyethylene glycol (PEG), with orthogonal functional groups that allows for controlled reaction with the probe and with the desired substrate. 1 However, the effect of such modi-cations on the function of a probe are oen not evaluated, even though binding interactions are likely to be affected. Consequently, opportunities to gain insights into the structureactivity relationships (SAR) are missed.
With the difficulties faced in the development of novel antibiotics and the increasing challenges of ghting against antibiotic resistance, 2,3 investigations into the SAR of antibiotic analogues could reveal valuable information. The b-lactam antibiotics are widely used and are typically considered to be one of the safest classes of antibiotics. 4 Since their discovery, there has been extensive research carried out into the derivatisation of the b-lactam scaffold, resulting in the successful development of numerous antibiotics. 5,6 Recently, we demonstrated the ability of surface bound blactam drugs to be recognised by the therapeutic target proteins as well as enzymes produced by resistant bacteria. 7 In order to attach the antibiotic molecule to the surface, an analogue of cephalexin (1) was prepared featuring a maleimide group attached via a PEG linker (2), Fig. 1. Studies of the surfacebound antibiotic demonstrated that b-lactamases and a penicillin binding protein (PBP) were able to recognise and bind the immobilised drugs. Here, we set out to investigate the effect of the addition of a chemical tether on the properties of the parent compound, thus contributing new SAR information.

Results and discussion
Chemistry Compound design. The modications to 1 described herein were performed via the amine of the molecule. This position was chosen to minimise the effect of the tether on the pharmacophore, the b-lactam ring. The amine motif provided a functional handle for the addition of the desired chemical tether through amidation reactions, allowing access to a selection of compounds with reactive pendant groups. In addition to the previously reported compounds 2 and 3, 7 a further seven analogues were synthesised, Fig. 1. Compounds 2-6 were designed to include commonly employed functional groups that are used in biocompatible reactions with surfaces and other substrates of interest.
Compounds 2 and 3 both feature a maleimide; this motif is a prevalent functional handle used for selective reaction through conjugation with thiol groups. 8 Another commonly used bio-compatible reaction is the copper-catalysed click reaction in which an alkyne, such as in compound 4, reacts with an azide to form a triazole. 9 For the direct attachment to gold surfaces disuldes, such as the lipoic acid tether of compound 5, are ubiquitous. 10 Whereas, for the attachment to a protein, or another source of amine functional groups, N-hydroxysuccinimide (NHS) esters are common. 11 Compounds with this activated ester, as in compound 6, react readily with lysine residues and other amines, providing attachment through the formation of an amide bond.
NHS-esters, such as compound 6, are known to have a short half-life in aqueous media; 11 thus, the hydrolysis product with the carboxylic acid was prepared for comparison, compound 7. Three further control compounds were synthesised: one to test the effect of a small modication with the acetyl group of compound 8, whereas the tert-butyloxycarbonyl (Boc) group of compound 9 was designed to test the effect of a large sterically demanding group in this position. Lastly, compound 10 was included featuring a small aliphatic tether with a terminal amine group. By attaching the tethers via an amide linker in compounds 2-9, the amine of the parent antibiotic is lost. Therefore, compound 10 was included to determine the effects of introducing an amide and a short exible tether, while maintaining an amine group.
Compound synthesis. The two maleimide analogues (2 and 3) were prepared as previously reported. 7 All analogues were prepared by addition of the tethers to the parent antibiotic, as described in Scheme 1. Compounds 4 and 5 were prepared from reaction of 1 with the NHS-esters of the corresponding tether, intermediates 16 and 18. In the preparation of compound 6, the previously reported acid (20) 12 was used to allow for the desired amidation via the acid chloride, while maintaining the NHSester. Compounds 7, 8, and 9 were prepared by reaction with the required anhydride reagent: glutaric anhydride, acetic anhydride and Boc anhydride, respectively. Finally, compound 10 was prepared in three steps from the N-Boc protected 5aminovaleric acid 21, Scheme 1. All compounds were characterised by 1 H, 13 C NMR, IR spectroscopy, and HRMS. Purity of compounds for in vitro and in vivo testing was determined using HPLC or QNMR. Biological evaluation PBP binding assay. A thermal shi assay was used to investigate the binding of the cephalexin analogues to the therapeutic target, PBP, in vitro. This assay provides a fast and inexpensive method of detecting a ligand binding to a protein of interest. Briey, the protein is gradually heated in the presence of a uorescent dye that binds non-specically to the hydrophobic surfaces. As the protein unfolds, hydrophobic surfaces within the protein are revealed, causing the dye to uoresce. A change in the melting temperature (T m ) when the protein is heated in the presence of a ligand, indicates a protein-ligand interaction. Fig. 2 shows the results of the thermal shi assay carried out using recombinant PBP3 and PBP4 with compounds 1-10. The PBPs can be divided into two main categories: high molecular mass (HMM) and low molecular mass (LMM). 13 PBP3 was chosen as a representative HMM PBP, and PBP4 as a representative LMM PBP. Cephalexin (1) has been previously reported to show good binding affinities for both PBP3 and PBP4. 14 Penicillin (11) and cefpodoxime (12) were included as positive controls.
The results from this assay showed that all of the analogues tested caused an increase in the T m with PBP3, thus increasing the thermostability of this particular PBP upon binding. Whereas, with PBP4 there was either no shi or a small decrease in the T m observed, indicating the ligands are facilitating the unfolding of the protein. Previously reported thermal shi assays with PBPs have noted both increases and decreases in T m of the protein aer the binding of different b-lactam analogues. [15][16][17] Further thermal shi studies were carried out to approximate the affinities of compounds 1-12 with PBP3. PBP3 was selected for this study as this particular PBP is essential for cell division in E. coli, making PBP3 an important target for b-lactam antibiotics. 13 By measuring the T m values of PBP3 with compounds 1-12 at a range of ligand concentrations, aer a constant incubation time, the equivalents of each ligand required to cause a shi greater than half the T m value (T m1/2 ) were determined, Fig. 3. Using this assay, it was determined that 75 equivalents of compound 1 was required to achieve >T m1/2 but with analogues 2-10, fewer equivalents were required. This study also showed that penicillin (11) has a great affinity for PBP3 than cephalexin (1), which is consistent with previously reported data. 14 The results, as shown in Fig. 3, suggest that addition of the tethers has improved the binding affinity of analogues 2-10 with PBP3, compared to that of the parent antibiotic 1.
b-Lactamase kinetics studies. Previously reported SAR around the b-lactam class of antibiotics has shown that increased steric bulk at the 7-position of cephalosporins can reduce the rates of drug hydrolysis by the b-lactamases. 18 Many of the b-lactam antibiotics from the later generations, for example cefpodoxime (12, Fig. 4), which are stable in the presence of many b-lactamases, feature sterically demanding groups in this position. The SAR of cephalosporins suggested that increasing the steric bulk at the 7-position of cephalexin (1) will provide resistance to b-lactamase-mediated hydrolysis. Therefore, it was hypothesised that analogues 2-10, which all feature tethers at this position, would have improved stability, compared to that of the parent drug 1.
To investigate the effect that the addition of the chemical tethers reported herein had on the rate of b-lactamase-mediated hydrolysis, a previously reported absorbance assay was employed. 19 The rates of hydrolysis were measured by the decrease in absorbance at 260 nm caused by hydrolysis of the blactam ring. The rates of hydrolysis were measured and the initial  Relative affinities for compounds 1-12 with PBP3. Ligand equivalents required to cause >T m1/2 shift. Lower equivalents conveys higher affinity. The assay protocol is provided in the Experimental section. Fig. 4 The cephalosporin structure. Cefpodoxime (12), has increased stability in the presence of b-lactamases due to the increased steric bulk at the 7-position.
velocity of each determined. The high K m values prevented determinations of the V max values, therefore the catalytic efficiencies (k cat /K m ) were determined using v ¼ (k cat /K m )[E][S]. 20,21 The relative catalytic efficiencies, with respect to the (k cat /K m ) value for the parent antibiotic cephalexin (1), were used to compare the reaction specicities of four selected b-lactamases with the tethered analogues, Table 1.
The b-lactamase enzymes are well studied, with over 2500 unique proteins identied. 22 This vast family of enzymes can be categorised into subsets (class A, B, C, or D), based on their protein sequence. Initial tests were carried out using TEM-1, one of the most common b-lactamases found in Gramnegative bacteria. 23 This class A b-lactamase is able to hydrolyse penicillins and the early generations of cephalosporins. Of the ten analogues tested, compound 9 was the preferred substrate with TEM-1, suggesting that the large Boc group attached to this analogue does not hinder the hydrolysis. The least preferred substrate with TEM-1 was compound 2, which is the largest of the compounds tested with a PEG tether. However, there was no clear trend in the relative catalytic efficiencies of compounds 1-10 with TEM-1.
As with TEM-1, compound 2 was also observed to be the least preferred substrate with CTX-M-15, an extended spectrum blactamase (ESBL) from class A. ESBLs are plasmid-encoded enzymes that confer increased antibiotic resistance to commonly used antibiotics. 24 CTX-M-14 and CTX-M-15 are the most prevalent ESBLs and known to contribute towards many cases of multidrug resistant infections. 25,26 The relative catalytic efficiencies showed that the two more preferred substrates of CTX-M-15 had no tether (compound 1) and the smallest tether tested, the acetyl group (compound 8). This trend suggests that modications to this point of the antibiotic has a direct effect on the stability against hydrolysis by CTX-M-15. Thus structural modications to this point of the antibiotic could be an effective strategy in the production of antibiotics with increased stability to this ESBL.
The relative catalytic efficiencies of compounds 1-10 with AmpC demonstrated that introduction of the tethers resulted, surprisingly, in increased substrate activity against the modi-ed b-lactam. AmpC is a class C b-lactamase known to confer resistance to the cephalosporins such as cephalexin (1), as well as cephamycins and carbapenems. 22 Compound 2 was less favoured than 1, but all other analogues had a signicantly higher catalytic efficiency than the parent antibiotic, particularly compound 7, which featured a glutaric acid tether. This observation suggests that modications to this point of the antibiotic results in reduced stability to AmpC mediated hydrolysis. This is in contrast to the trend observed with CTX-M-15, showing that the effect of these structural modications, to the 7-position of cephalexin (1), is specic to the b-lactamase under investigation.
Further tests were carried out to investigate the effects of the tethers on the hydrolysis by a metallo-b-lactamase, NDM-1. The metallo-b-lactamases are known to hydrolyse penicillins, cephalosporins and even last resort carbapenems. These class B blactamases hydrolyse b-lactams by an alternative mechanism that relies on one or two zinc ions present in the active site. 27 These enzymes are oen produced by clinical strains with multiple forms of resistance that are only susceptible to last line antibiotics. 28 The relative catalytic efficiencies of compounds 1-10 with NDM-1 demonstrated that introduction of the tethers to the antibiotic resulted in increased substrate activity, the same trend that was observed with AmpC. These results suggest that modications to this point of the antibiotic results in reduced stability to NDM-1 mediated hydrolysis and should be avoided in developing cephalexin (1) analogues stable to NDM-1.
In vivo assays. The in vitro assays demonstrated that the modied analogues were able to interact with the therapeutic target proteins (PBP3 and PBP4) and four selected b-lactamases (TEM-1, CTX-M-15, AmpC, and NDM-1), as shown in Fig. 2, 3 and Table 1. Further testing was carried out in vivo to determine whether the tethered compounds had any antibacterial activity. Table 2 summarises the MIC values determined for 1-12, MBC 50 data is included in the ESI. † The activity of 1-12 was investigated with growth assays against Staphylococcus aureus (NCTC 6571). S. aureus is a virulent Gram-positive pathogen that can cause a wide range of infections in humans. 29,30 Each compound was tested at 400-6 mM concentrations, with lower concentrations tested as required. All tethered analogues were still able to inhibit the growth of this strain of S. aureus; however, the tethers were detrimental to the antimicrobial activity against this Grampositive bacteria, with all but one MIC value (compound 9) determined to be greater than that of the parent antimicrobial 1.
Further growth assays tested 1-12 against Escherichia coli BW25113, a K-12 strain. E. coli K-12 is the workhorse of many microbiology laboratories; however, E. coli is a versatile Gramnegative bacterium and the evolution of pathogenic E. coli can cause a number of harmful infections in humans. 31 Only the parent antibiotic 1 and the two control compounds 11 and 12 were observed to have antimicrobial activity against E. coli. Thus showing that the addition of the chemical tethers has a signicant detrimental effect on the activity of the antibiotic against E. coli.
Physicochemical properties. Since the abilities of 2-10 to bind the therapeutic target had been conrmed in vitro (Fig. 2), the in vivo results suggest that the tethered analogues may be unable to access the PBPs within the periplasm of E. coli. To gain a further insight into the effects of the chemical tethers, the physicochemical properties of each compound was evaluated. The compounds were scored on the properties favourable for accumulation in Gram-negative bacteria, based on the observations of Richter et al. who reported that there are three key properties in predicting the accumulation: a primary amine, low globularity, and high rigidity. 32,33 In order to score the likelihood of the compounds accumulating, each was scored based on the "eNTRy rules" reported by Richter et al.: number of primary amines (NoA) $ 1; globularity (Glob) # 0.25; number of rotatable bonds (NoRB) # 5, Table 3. The predicted accumulation scores were determined to be below ideal for all of the tethered analogues (2-10), primarily due to the loss of the amine group used in the introduction of the tethers via an amide bond. Compound 10, which maintained an amine group, failed based on the increased NoRB. Chemical tethers are oen designed with exibility to favour the subsequent reactions with other substrates, particularly in examples that form a monolayer on a surface. However, this exibility increases the NoRB, reducing the rigidity of these analogues, which reduces the likelihood of accumulation. Based on the predicted accumulation scores of Table 3, all of the tethered analogues are expected to have lower activity than that of 1 against Gram-negative bacteria. As shown in Table 2, this was found to be true for all tethered analogues, 2-10. Compounds 8 and 10 both scored two but for different reasons: compound 8 lacked the required amine whereas compound 10 exceeded the limit for NoRB. The growth assay results showed that both compounds were inactive against E. coli demonstrating the importance of these two criteria of the "eNTRy rules". By using more rigid tethers and including an amine group, it may be possible to produce tethered analogues that are active against Gram-negative bacteria.
However, the criteria compared in Table 3 fail to predict the measured antimicrobial activity of 11 and 12 against E. coli. A reliable method of predicting antimicrobial activity is very much coveted, but it is yet to be achieved due to the complex factors involved. [32][33][34][35] Summary and conclusions A series of cephalexin (1) analogues equipped with chemical tethers was evaluated for binding to the therapeutic target, the penicillin binding proteins, and shown to maintain binding interactions in vitro. Further investigations with four b-lactamases (TEM-1, CTX-M-15, AmpC, and NDM-1) were carried out and revealed that the modications affected each enzyme's catalytic rates differently. With TEM-1, there was no clear trend in the catalytic efficiencies of 1-10. CTX-M-15 was found to favour hydrolysis of the parent antibiotic without a tether, thus demonstrating that modications to this position of 1 could produce antibiotics with increased stability to this ESBL. Conversely, both AmpC and NDM-1 were found to favour the modied analogues suggesting that these types of structural modications should be avoided in the design of analogues stable to AmpC and/or NDM-1. The tethers were found to lower the antimicrobial activities when testing against S. aureus and cause complete loss of activity against E. coli. The loss of activity against E. coli was consistent with previously reported observations linking the globularity, rigidity, and amine functionality of antibiotics with accumulation in Gram-negatives.
These results show that the addition of the tethers directly affected the properties of the antibiotic, thus highlighting the importance of evaluating the changes that occur from modifying molecular probes. Most notably, the effect of the tethers on the rate of b-lactamase-mediated hydrolysis was specic to the b-lactamase under investigation. This suggests that  (11) and cefpodoxime (12). modifying the 7-position of 1 could be key in the development of surface-bound antibiotics for the selective detection of blactamases associated with multidrug resistant infections, such as NDM-1.

Experimental
Chemistry General experimental. Analytical thin layer chromatography (TLC) was performed with EM Science silica gel 60 F254 aluminium plates. Visualisation was carried out using a UV lamp (254 nm) and by immersion in potassium permanganate (KMnO 4 ), followed by heating using a heat gun. Organic solutions were concentrated by rotary evaporation at 40-45 C. Purication of reaction products by ash column chromatography was carried out using Fluka Silica, pore size 60 A, 220-440 mesh, 35-75 mm.
Materials. Unless otherwise noted, all purchased materials were used without purication. All standard solvents were purchased from Sigma Aldrich. All standard acids, bases, and drying agents were purchased from Fisher Scientic. NHS and N,N-diisopropylethylamine (DIPEA) were purchased from Acros Organics. Cephalexin monohydrate and Boc 2 O were purchased from Fluorochem. Pentynoic acid, N,N 0 -dicyclohexylcarbodiimide (DCC), lipoic acid, glutaric anhydride, oxalyl chloride, and Et 3 N were purchased from Sigma Aldrich. DMAP and 5-(Boc-amino)pentanoic acid were purchased from TCI.
Instrumentation. 1 H and 13 C NMR spectra were recorded on a Jeol ECS 400 (400 MHz for 1 H, 101 MHz for 13 C) at ambient temperature. Chemical shis are reported relative to residual solvent peaks and coupling constants (J) are given in hertz. High-resolution ESI mass spectra were recorded on a Bruker microTOF electrospray mass spectrometer. Infrared (IR) spectra were recorded on a PerkinElmer Spectrum Two (ATIR). Analytical HPLC measurements were performed on a Shimadzu HPLC system (Prominence) equipped with a LC-20AD pump, SIL-20A autosampler, DGU-20AS degasser, CTO-20AC column oven, CBM-20A communication bus module and SPD-M20A diode array detector using an Athena C18-WP column (100 A, 4.6 Â 250 mm, 5 mm). Eluent gradient: 5-95% MeCN/H 2 O with a 0.1% formic acid modier, over 15 minutes.  (15) (100 mg, 1.020 mmol, 1 equiv.) in anhydrous DCM (1 mL, 1 M) at 0 C was added NHS (123 mg, 1.071 mmol, 1.05 equiv.). A solution of DCC (221 mg, 1.071 mmol, 1.05 equiv.) in DCM (1 mL) was added slowly. The reaction mixture was then allowed to warm to room temperature. Aer 16 h, the urea precipitate formed during the reaction was ltered off, and the lter cake washed with DCM. The ltrate was concentrated under reduced pressure to afford 2,5-dioxopyrrolidin-1-yl pent-4-ynoate (16), which was used without further purication. The residue was dissolved in anhydrous MeCN (10 mL) and cephalexin monohydrate (338 mg, 0.927 mmol, 0.9 equiv.) was added. The mixture was cooled in an ice bath followed by the slow addition of DIPEA (443 mL, 2.550 mmol, 2.5 equiv.). Once the addition was complete, the ice bath was removed, and the reaction was allowed to stir at room temperature for 2 h at which time the reaction appeared complete by TLC. The reaction mixture was concentrated under reduced pressure, dissolved in EtOAc (20 mL), and washed with 0.1 M HCl (2 Â 20 mL). The organic layer was collected and concentrated to a cream solid, which was the puried by trituration from diethyl ether to afford (6R,7R)-3-methyl-8-oxo-7-[(2R)-2-(pent-4-ynamido)-2-phenylacetamido]-5-thia-  .311 mmol, 1.3 equiv.) in DCM (10 mL) was added slowly. The reaction mixture was then allowed to warm to room temperature. Aer 16 h, the urea precipitate formed during the reaction was ltered off, and the lter cake washed with DCM. The ltrate was concentrated under reduced pressure to afford a yellow residue. Recrystallisation from EtOAc : hexane (1 : 1) afforded 2,5-dioxopyrrolidin-1-yl 5-(1,2-dithiolan-3-yl)pentanoate (18) 37 2,5-Dioxopyrrolidin-1-yl 5-(1,2-dithiolan-3-yl)pentanoate (18) (294 mg, 0.971 mmol, 1 equiv.) was dissolved in anhydrous MeCN (20 mL, 0.05 M) and cephalexin monohydrate (235 mg, 0.647 mmol, 0.66 equiv.) was added. The mixture was cooled in an ice bath followed by the slow addition of DIPEA (169 mL, 0.971 mmol, 1 equiv.). Once the addition was complete, the ice bath was removed and the reaction was allowed to stir at room temperature for 6 h. The reaction mixture was then concentrated under reduced pressure and triturated using 5% DCM/ diethyl ether to afford (6R,7R) 158 mmol, 1.5 equiv.) was then added portion-wise and the resulting reaction mixture was allowed to warm to room temperature. Aer a further 4 h at room temperature, the reaction was concentrated under reduced pressure. The residue was dissolved in EtOAc (60 mL) and washed with 0.1 M HCl (2 Â 30 mL) followed by brine (1 Â 30 mL). The organic layer was concentrated to afford 5-[(2,5-dioxopyrrolidin-1-yl)oxy]-5-oxopentanoic acid (20) (592 mg) as a colourless oil, which was used without further purication. The prepared NHS ester (20) (500 mg, 2.183 mmol, 1 equiv.) was then dissolved in anhydrous DCM (10 mL) under an atmosphere of N 2 and cooled in an ice bath. Oxalyl chloride (225 mL, 2.621 mmol, 1.2 equiv.) was then added followed by one drop of DMF. The resultant reaction mixture was allowed to return to room temperature and stirred for 16 h, aer which it was concentrated under reduced pressure to afford the acid chloride. In a separate ask, cephalexin monohydrate (598 mg, 1.637 mmol, 0.75 equiv.) was suspended in anhydrous MeCN (20 mL) then anhydrous pyridine (329 mL, 4.093 mmol, 1.9 equiv.) was added. The previously prepared acid chloride was then dissolved in anhydrous MeCN (10 mL) and added to the cephalexin/pyridine mixture. The reaction was then stirred at room temperature for 24 h aer which it was concentrated under reduced pressure and the residue was triturated in EtOAc to afford a cream solid. The solid was then recrystallised from MeCN to afford (6R,7R)    Cultures for protein expression were grown in 1 L of Luria-Bertani (LB) broth at 37 C on an orbital shaker. Expression was induced by addition of 0.01% L-arabinose during mid-log phase of growth. Cultures were further incubated for 16 h at 20 C, then the cells were harvested by centrifugation. Cell pellets were resuspended in 50 mM KPi pH 7.8, 200 mM NaCl, 10 mM imidazole, 20% glycerol with 1 mM phenylmethylsulfonyl uoride followed by sonication. The lysate was claried by centrifugation before loading onto a HisTrap HF column (GE Healthcare). To remove any pre-bound ligands, refolding purication was performed by initially washing with the protein unfolding buffer [ Thermal shi assay. The thermal shi assay was carried out using the Protein Thermal Shi™ assay kit (Applied Biosystems). 3 mM of puried E. coli PBP3 was incubated with the blactam analogues at 300 mM concentration in a mixture containing the Protein Thermal Shi™ Dye. The samples were then heated in a StepOnePlus™ Real-Time PCR System from 25 to 95 C at a rate of 1 C min À1 . Tests were carried out in triplicate and the averages plotted as the negative rst derivative vs. temperature. Reference wells, i.e. solutions consisting only of only PBP3 with dye, PBP3 only, and dye only, were used as controls. Melting temperature (T m ) values were determined with and without each compound, and the change in melting temperature (DT m ) was obtained, Fig. 2. The thermal shi assay with PBP4 was carried out using the protocol described for PBP3, with the follow deriviatisation: 500 nM of puried Bacillus subtilis PBP4 was incubated with 50 mM of the b-lactam analogues.
To determine the approximate affinity of PBP3 for compounds 1 to 12, 2 mM of puried E. coli PBP3 was incubated with a range of concentrations of each compound. The mixtures were incubated at room temperature for 25 minutes prior to the start of the protein thermal shi assay program on the qPCR machine. The signals from the hydrophobic uorescent dye were monitored as the mixtures were heated from 35 to 70 C at a ramp rate of 0.3%. The T m values of PBP3 incubated with each compound at the various concentrations were recorded. The relative changes in the T m values, relative to the highest T m change seen for each compound, are reported as a ratio in the ESI. † The concentration at which the T m rose to >T m1/2 was used to approximate the relative affinities.
Expression and purication of TEM-1. TEM-1 (Uniprot ID: A5PHA6) was cloned into pBKR, a pBADcLIC2005 derivative with a kanamycin resistance cassette in place of the ampicillin resistance cassette. The resulting plasmid, pBKR-TEM1, was transformed into the expression strain E. coli MC1061. Starting cultures were grown in 10 mL LB broth at 37 C overnight with shaking. A litre of expression culture was prepared and grown to an OD 600 of between 0.4-0.6 before induction with 0.01% (w/v) L-arabinose. The induced culture was allowed to grow for 18 h at 30 C. The resulting culture was harvested by centrifugation at 5000g. The pellet was then resuspended in 30 mL of sterile SET buffer (0.5 M sucrose, 5 mM EDTA, 50 mM Tris-HCl pH 7.8). 13 mg of lysozyme was added to the mixture before incubation for 1 h at 30 C. To isolate the periplasmic fraction, the treated sample was then claried by centrifugation at 27 000g. The supernatant was dialysed into 50 mM KPi pH 7.8, 200 mM NaCl over 18 hours at 4 C. Aer dialysis, the periplasmic fraction was loaded onto an equilibrated 5 mL HisTrap column. The bound His-tagged TEM-1 were washed with 10 column volumes of wash buffer (50 mM KPi pH 7.8, 200 mM NaCl, 20% glycerol, 40 mM imidazole). To elute the bound protein, 5 column volumes of elution buffer (50 mM KPi pH 7.8, 200 mM NaCl, 20% glycerol, 500 mM imidazole) was owed through the column while collecting the ow through. For downstream analysis, the protein was buffer exchanged into 50 mM KPi pH 7.8, 200 mM NaCl using the HisTrap desalting column.
Expression and purication of CTX-M-15. CTX-M-15 (Uniprot ID: Q9EXV5) was synthesised using the protocol described above in the expression and purication of TEM-1, with the following deviations: the gene coding for NDM-1 was synthesised as a gBlock (IDT). Following induction with L-arabinose, 1 L cultures were incubated for 20 h at 20 C.
Expression and purication of NDM-1. NDM-1 (Uniprot ID: C7C422, residues G29-R270) was synthesised using the protocol described above in the expression and purication of TEM-1, with the following deviations: the gene coding for NDM-1 was synthesised as a gBlock (IDT). Following induction with Larabinose, 1 L cultures were incubated for 20 h at 20 C.
Enzyme kinetics. The rate of b-lactamase-mediated hydrolysis was montiored using a previously reported absorbance assay. 19 In a 96-well plate, 200 mL of the test compounds at the desired concentrations with the relevant b-lactamases were incubated and the absorbance at 260 nm was monitored. Experiments using NDM-1 included 1 equivalent of ZnCl 2 . All studies were carried out using one batch of puried enzyme. Experiments were carried out at 37 C and measurements were taken every 60 s (Epoch 2 Microplate Spectrophotometer, Bio-Tek) for 1 h. Each sample was carried out in triplicate and the average was used for further calculations. k cat /K m was determined from v ¼ (k cat /K m )[E][S]. The initial velocity for each compound was determined and used as "v".
MIC growth assay. Stock solutions of each compound were prepared in 50% DMSO/water at Â100 the nal concentration. In a 96-well plate, 2 mL of the test compounds were added to each well. Each plate included the positive control (750 mg mL À1 chloramphenicol), the negative control (50% DMSO/water) and media only wells. To each test well was then added 198 mL of the required bacteria stock (OD 0.05). E. coli assays were carried out using LB broth, S. aureus were carried out using tryptic soy broth (TSB). Plates were incubated at 37 C and the OD 600 was measured (Epoch 2 Microplate Spectrophotometer, BioTek) every 30 min over a 17 h period. Tests were carried out in triplicate, the background absorbance was removed using reference wells, and the averages were determined. Dose curves were plotting using the OD 600 aer 16 h growth vs. compound concentration. MICs were dened as the minimum concentration of compound at which no signicant growth was observed aer 16 h incubation. MIC values were determined using GraphPad Prism (version 8.3.0).

Conflicts of interest
There are no conicts to declare.