Non‐Hydrolytic β‐Lactam Antibiotic Fragmentation by l,d‐Transpeptidases and Serine β‐Lactamase Cysteine Variants

Abstract Enzymes often use nucleophilic serine, threonine, and cysteine residues to achieve the same type of reaction; the underlying reasons for this are not understood. While bacterial d,d‐transpeptidases (penicillin‐binding proteins) employ a nucleophilic serine, l,d‐transpeptidases use a nucleophilic cysteine. The covalent complexes formed by l,d‐transpeptidases with some β‐lactam antibiotics undergo non‐hydrolytic fragmentation. This is not usually observed for penicillin‐binding proteins, or for the related serine β‐lactamases. Replacement of the nucleophilic serine of serine β‐lactamases with cysteine yields enzymes which fragment β‐lactams via a similar mechanism as the l,d‐transpeptidases, implying the different reaction outcomes are principally due to the formation of thioester versus ester intermediates. The results highlight fundamental differences in the reactivity of nucleophilic serine and cysteine enzymes, and imply new possibilities for the inhibition of nucleophilic enzymes.

Bacterial peptidoglycan biosynthesis is an antibiotic target of immense clinical importance. [1][2][3] Peptidoglycan, an extracellular network of polysaccharides and cross-linked peptides, plays ac ritical structural role in the bacterial cell wall. [4] In many Gram-negative bacteria, cross-links occur primarily between meso-diaminopimelate (meso-Dap) and d-Ala residues.F ormation of these meso-Dap-d-Ala cross-links is catalyzed by the d,d-transpeptidase activities of the PBPs. However,t he peptidoglycan of some bacteria (for example, Mycobacterium tuberculosis in stationary phase) consists substantially of cross-links formed between meso-Dap residues,catalyzed by the Ldts. [5,6] Thetranspeptidase activities of PBPs and Ldts both proceed via acyl-enzymes ubstrate complexes formed by reaction with adonor substrate peptide; the acyl-enzymec omplex subsequently reacts with an amino group of an acceptor peptide to give the cross-linked product. While the acyl-enzyme complex is formed by reaction with an ucleophilic serine residue in the PBPs,t he Ldts instead employ anucleophilic cysteine.
TheP BPs are established targets of b-lactam drugs,t he most widely used class of antibiotic. [7] b-Lactams ( Figure 1A) inhibit PBPs by covalently modifying the nucleophilic serine ( Figure 1B). Theu se of b-lactams is compromised by bacterial resistance mechanisms,m ost importantly the production of serine b-lactamases (SBLs). TheS BL mechanism involves the reaction of the b-lactam ring with anucleophilic serine to form an acyl-enzymecomplex ( Figure 1B), which is then efficiently hydrolyzed. Three of the five clinically used SBL inhibitors (for example,c lavulanic acid) react irreversibly to form hydrolytically stable acyl-enzyme complexes, whilst the other two (avibactam and vaborbactam) react reversibly. [8] Thenucleophilic cysteine of the Ldts can also be acylated by b-lactams ( Figure 1B). [9,10] In particular, carbapenems and penems ( Figure 1A)a re potent Ldt inhibitors, [9,11] and manifest antibiotic activity against M. tuberculosis. [12,13] However,t he Ldts are often considered insensitive to the other classes of b-lactam antibiotics (for example,penicillins). [6,14,15] Recent work has suggested that the acyl-enzyme complexes formed by some b-lactams with Ldts behave differently than the analogous complexes formed with PBPs or SBLs ( Figure 1B). [9,12,16] However,i ti su nclear whether the difference in reactivity arises from the cysteine-for-serine substitution, or if it is due to other active site variations.B ased on studies of Ldt Mt2 (from M. tuberculosis)and cysteine variants of the OXA-48 and KPC-2 SBLs,w er eport that the difference in reactivity is principally due to the identity of the nucleophilic residue,a nd arises from the different chemical properties of the ester and thioester intermediates.T his is arare example where the identity of the nucleophilic residue changes the reaction outcome of an enzyme.T he results suggest that the biological use of specific nucleophiles may be related to function, and raise the potential of developing new nucleophilic-enzyme inhibitors (including for serine and cysteine proteases) and antibiotics,i .e., which are poised to undergo fragmentation reactions,yielding stable acyl-enzyme complexes.
To compare the reactivity of the Ldts with the nucleophilic serine enzymes (i.e., PBPs and SBLs), mass spectrometry (MS) was used to monitor the adducts formed by Ldt Mt2 with ap anel of b-lactam antibiotics.W hile screening representative penicillins,m ass shifts were observed that are consistent with the formation of acyl-enzyme complexes with oxacillin, piperacillin, and ticarcillin, but not with ampicillin ( Figure 1C and Supporting Information, Figure S1). Over time,an ew species 159 Da smaller than the intact acyl-enzyme complexes accumulated for Ldt Mt2 in the presence of ampicillin, oxacillin, and piperacillin ( Figure 1C, and Supporting Information, Figures S1 and S2). This mass difference is consistent with the cleavage of the C5-C6 bond of the penicillin-derived Ldt Mt2 acyl-enzymec omplex, yielding Ldt Mt2 modified by an acetyl group bearing the penicillin C6 side chain. This is similar to what has been observed previously for ampicillin with other Ldts. [16,17] In the case of faropenem, apeak corresponding to Ldt Mt2 acylatedwith a3hydroxybutyrate fragment was observed ( Figure 1D). [9,12] A similar adduct was observed following treatment of Ldt Mt2 with the penem sulopenem (Supporting Information, Figure S1). Incubation of Ldt Mt2 with the cephalosporins cephalothin, cefotaxime,and cefapirin resulted in adducts in which the C3' leaving groups were lost (Supporting Information, Figure S1). Fragmentation was not observed for the carbapenems meropenem, imipenem, and ertapenem (Supporting Information, Figure S1), and acylation was not observed for the monobactam aztreonam and the oxacephem moxalactam. NMR experiments indicated that cephalothin, meropenem, aztreonam, and moxalactam were not efficiently degraded by Ldt Mt2 under the MS assay conditions.N oa cylation of the Ldt Mt2 C354S variant (where the nucleophilic cysteine is replaced with aserine) by b-lactams was observed under the conditions used for MS (Supporting Information, Figure S3).
Thefragmentation of penicillins and penems with Ldt Mt2 , which is not typically observed for PBPs and SBLs,may result from the different chemical properties of the thioester (Ldts) and ester (PBPs and SBLs) acyl-enzymec omplexes.A lternatively,t hese differences could arise from other features of the active site;for example,there is less than 25 %sequence identity between PBP5 from Pseudomonas aeruginosa and Ldt Mt2 from M. tuberculosis (active sites are compared in Figure S4 in the Supporting Information). To investigate, variants of the clinically important OXA-48 and KPC-2 SBLs were prepared in which the nucleophilic serine was replaced with ac ysteine.C onsistent with previous reports on the cysteine variants of SBLs, [18][19][20] OXA-48 S70C (k cat /K M 0.098 AE 0.017 mm À1 s À1 with the fluorogenic cephem substrate FC5) [21] was not as catalytically active as the wild-type enzyme (k cat / K M 0.36 AE 0.06 mm À1 s À1 with FC5;S upporting Information, Table S1). However, the impact of this substitution on the enzymatic mechanism of the SBLs has not been previously described.
Following incubation of the penicillin ampicillin with wild-type OXA-48 or KPC-2, only hydrolysis was observed by NMR ( Figure 2A). While OXA-48 S70C and KPC-2 S70C also catalyzed ampicillin hydrolysis,additional products were formed ( Figure 2A). Based on further NMR and MS analyses,t hese products were assigned as the d-phenylglycylglycine dipeptide (derived in part from C6 and C7 of ampicillin) and adihydrothiazole (derived from the ampicillin thiazolidine ring) ( Figure 2B,Supporting Information, Figures S5 and S6, and Tables S2-S5). Thea ssignment of dphenylglycylglycine was validated by comparison with asynthetic standard (Supporting Information, Figure S7). Treatment of other penicillins (for example,piperacillin, oxacillin, and ticarcillin) with OXA-48 S70C and KPC-2 S70C resulted in the formation of both hydrolysis and C5-C6 fragmentation products ( Figure 2C and Supporting Information, Figure S8). While Ldt Mt2 -catalyzed penicillin fragmentation could not be observed by NMR (Supporting Information, Figure S8), MS analyses indicated that OXA-48 S70C and Ldt Mt2 form the same C5-C6 fragmentation products ( Figure 1C and Supporting Information, Figure S9). Although the relative levels of hydrolysis and C5-C6 fragmentation were similar for both OXA-48 S70C and KPC-2 S70C,Ldt Mt2 was observed to favor hydrolysis of the penicillin b-lactam (contrasting with Ldt fm from Enterococcus faecium,f or which fragmentation is reported to predominate). [16] Tr eatment of faropenem with OXA-48 S70C or KPC-2 S70C yielded different products than the hydrolysis products formed by wild-type OXA-48 and KPC-2 ( Figure 3A and Supporting Information, S10). Based on NMR and MS,these products were assigned as 3-hydroxybutyrate (derived in part from C6 and C7 of faropenem) and at hiazole (derived from the dihydrothiazole ring of faropenem) ( Figure 3B,Supporting Information, Figure S10 and Tables S6-S9). Further NMR analyses demonstrated that the same faropenem-derived thiazole is formed by Ldt Mt2 ;e nzyme-catalyzed faropenem hydrolysis was not observed for Ldt Mt2 or OXA-48 S70C (Supporting Information, Figure S11).
Theproducts formed by Ldt Mt2 and the SBL thiol variants with penicillins and penems are consistent with C5-C6 bond cleavage,a nd inform on the mechanism of fragmentation ( Figure 4A,B). Thep enicillin-derived dihydrothiazole and faropenem-derived thiazole products imply that the nitrogen/ sulfur lone pairs in the thiazolidine (for penicillins) or dihydrothiazole (for faropenem) rings promote C5-C6 bond cleavage.B oth reaction pathways likely proceed via at hioester-enolate intermediate involving the nucleophilic cysteine.F ormation of this thioester-enolate is expected to be more energetically favorable than the formation of the analogous ester-enolate (involving an ucleophilic serine). As the stability of these enolates appeared to relate to the acidity at the a-position of the corresponding ester/thioester,c alculations were performed to compare the pK a values for the ester and thioester species derived from b-lactam fragmentation (Supporting Information, Figure S12 and Tables S10-S13). Thec alculated pK a values for the thioesters derived from ampicillin and faropenem fragmentation (as seen with the nucleophilic cysteine enzymes) are 7.2 and 6.4 units lower, respectively,than those for the analogous ester species (which were not seen for the nucleophilic serine enzymes) (Supporting Information, Table S11). These calculated values are Figure 2. Reaction of penicillins with Ldt Mt2 and SBL thiol variants. A) 1 HNMR spectra (700 MHz) of methyl group resonances for products formed upon addition of ampicillin to OXA-48 S70C, compared to wild-type OXA-48. These signals correspond to methyl groups indicated by the asterisks in panel B. B) Scheme showing ampicillinderived fragmentation products formed by Ldt Mt2 and SBL thiol variants, and hydrolysis products formed by wild-type SBLs, SBL thiol variants, and Ldt Mt2 .C ) 1 HNMR spectra (600 MHz) showing the formation of the same dihydrothiazole fragmentation product upon incubation of OXA-48 S70C with ampicillin,p iperacillin, oxacillin, and ticarcillin (see also Figure S8 in the SupportingInformation). Note the other signals present arise from intact penicillinsand hydrolyzed products.  Table S7 in the Supporting Information), although double bond isomers are also possible. The highlighted signals correspond to methyl groups indicated by asterisks in panel B. B) Scheme showing the C5-C6 faropenem fragmentation products observed for Ldt Mt2 and SBL thiol variants, as compared to the hydrolysis products observed for wild-type SBLs. consistent with experimentally determined pK a values for ethyl acetate (pK a = 25.6) [22] and ethyl thioacetate (pK a = 20.4-21.5). [23] Thus,t he favorable formation of thioesterenolates may rationalize why enzymes with nucleophilic cysteine residues (for example,L dts and SBL thiol variants) catalyze b-lactam C5-C6 fragmentation, while those with nucleophilic serine residues (for example,P BPs and SBLs) typically do not.
Fragmentation was not observed for the acyl-enzyme complexes derived from carbapenems with Ldt Mt2 (Figure 4C), contrasting with the rapid fragmentation of the penems faropenem and sulopenem, suggesting the heteroatom at position 1i si mportant. While the C5-C6 bond cleavage of the faropenem-derived acyl-enzymec omplex forms an aromatic thiazole ( Figure 4B), the product immediately formed following C5-C6 cleavage of ac arbapenemderived complex would not be aromatic (although it would likely rapidly tautomerize to give an aromatic pyrrole). Furthermore,t he pyrroline ring of the carbapenem-derived acyl-enzyme complex can tautomerize,i mpacting on the availability of the b-lactam-derived nitrogen lone pair. Analogous C6-C7 bond cleavage of the cephalosporinderived acyl-enzymec omplexes was also not observed (Fig-ure 4D). Instead, elimination of the cephalosporin C3' substituent occurs,l imiting the ability of the nitrogen lone pair in the dihydrothiazine ring to promote C6-C7 cleavage.
These results imply that the nature of the nucleophilic residue can determine the fate of the acyl-enzymec omplex; however, other active site features can be relevant. As we and others have observed, the Ldt Mt2 C354S variant does not form covalent complexes with the carbapenems and penems tested. [12] Furthermore,t he C5-C6 fragmentation of benzylpenicillin has been described for aP BP from a Streptomyces spp., [24] while the formation of at hiazole product from faropenem was observed for some class CSBLs. [25] It could be that as yet unidentified features in the active sites of some nucleophilic serine enzymes promote C5-C6 fragmentation. Thea cyl-enzyme complexes of SBLs with some other blactams favor alternative (i.e., not C5-C6) fragmentation reactions,a sr eported for the SBL inhibitors clavulanic acid, sulbactam, and tazobactam. [7,8] Alternative enzyme-catalyzed mechanisms not involving CÀCfragmentation have also been observed, including the degradation of carbapenems by class DS BLs via lactone formation. [26] There are few examples where substitution of the nucleophilic residue determines the type of reaction catalyzed by an enzyme.W hile the wild-type protease subtilisin catalyzes hydrolysis,t he subtilisin S221C variant also manifests transpeptidase activity. [27] More typically,substitution of an ucleophilic serine with cysteine can stabilize the acylenzyme complex. [28,29] However,a cylation by the natural substrate may not occur following the substitution of serine with cysteine (and vice versa), as observed by us (for Ldt Mt2 C354S) and others. [30,31] TheL dts are antibiotic targets in pathogens,n otably M. tuberculosis. [6,12] These results indicate that the different reactivities of the Ldts with b-lactam antibiotics,ascompared to the PBPs and SBLs,a re substantially governed by the nucleophilic cysteine,w hich facilitates fragmentation reactions not normally observed for enzymes employing nucleophilic serines.T hese fragmentation reactions can be rationalized mechanistically by the relatively favorable involvement of at hioester-enolate intermediate in fragmentation. These results highlight ap otential new approach for the development of antibiotics targeting the Ldts,f or example,t he identification of b-lactams that undergo C5-C6 (or C6-C7 for cephalosporins) fragmentation to give stable acyl-enzyme complexes.They may also inspire the development of related inhibitors targeting other classes of enzymes with nucleophilic cysteine residues (for example,t he cysteinep roteases). The potential for toxic and/or immunomodulatory effects of the blactam fragments/complexes formed by the Ldts should also be considered. [32] More generally,these results suggest that it is of interest to consider links between the biological roles of nucleophilic enzymes and the nature of the nucleophilic residue employed, including in the many biologically important serine/cysteine proteases and related enzymes (for example,t hose involved in ubiquitin biology). [33] Thee nhanced nucleophilicity of thiols compared to alcohols,a nd the increased reactivity of thioesters versus esters are often cited as factors for the use of thiols in biological systems.O ur results,combined with prior  (Figure 2and Supporting Information, Figure S8), while only fragmentation of faropenem was detected (Figure 3and Supporting Information, Figure S11). C) C5-C6 fragmentation was not observed for the carbapenems (Supporting Information, Figure S1), potentially in part due to delocalizationo fthe pyrroline nitrogen lone pair or due to tautomerization of the pyrroline ring (as well as the lack of aheteroatom at position 1o fthe ring). Note the hypotheticalp roduct would be expected to tautomerizet oapyrrole. D) C6-C7 fragmentation was not observed for cephalosporins;h owever,elimination of the C3' leaving group occurs (Supporting Information, Figure S1). Note that carbapenem and cephalosporinhydrolysis by Ldt Mt2 was not observed under our NMR conditions. studies on the nucleophilic residues of PBPs and SBLs,imply that both the reaction rate (including if the reaction occurs or not), and the fate of the acyl-enzymecomplex depend on the identity of the nucleophile and the context in which the ester or thioester exists.T hus,t he use of nucleophilic serine or cysteine residues may reflect chemical evolutionary pressures subject to their biological roles and the types of reactions catalyzed. Thel atter may include factors relating to competing pathways/by-products,which the present work shows can differ with the different types of nucleophile.