Interactions of "bora-penicilloates" with serine β-lactamases and DD-peptidases.

Specific boronic acids are generally powerful tetrahedral intermediate/transition state analogue inhibitors of serine amidohydrolases. This group of enzymes includes bacterial β-lactamases and DD-peptidases where there has been considerable development of boronic acid inhibitors. This paper describes the synthesis, determination of the inhibitory activity, and analysis of the results from two α-(2-thiazolidinyl) boronic acids that are closer analogues of particular tetrahedral intermediates involved in β-lactamase and DD-peptidase catalysis than those previously described. One of them, 2-[1-(dihydroxyboranyl)(2-phenylacetamido)methyl]-5,5-dimethyl-1,3-thiazolidine-4-carboxylic acid, is a direct analogue of the deacylation tetrahedral intermediates of these enzymes. These compounds are micromolar inhibitors of class C β-lactamases but, very unexpectedly, not inhibitors of class A β-lactamases. We rationalize the latter result on the basis of a new mechanism of boronic acid inhibition of the class A enzymes. A stable inhibitory complex is not accessible because of the instability of an intermediate on its pathway of formation. The new boronic acids also do not inhibit bacterial DD-peptidases (penicillin-binding proteins). This result strongly supports a central feature of a previously proposed mechanism of action of β-lactam antibiotics, where deacylation of β-lactam-derived acyl-enzymes is not possible because of unfavorable steric interactions.

Enzyme inhibitors remain important as drug leads.[1] Boronic acids, 1, have for quite some time now been designed and used as sources of active site-specific, anionic, tetrahedral transition state analogue complexes, 2, of n class="Chemical">serine amidohydrolases (Scheme 1). They are thus very effective inhibitors of these enzymes and potential drug candidates.[2−4] Among the enzymes that are inhibited by these compounds are the β-lactam-recognizing enzymes, the serine β-lactamases and DD-peptidases. Boronic acid inhibition of serine β-lactamases has been recognized for many years,[5,6] but only more recently have such inhibitors of DD-peptidases been identified.[7,8] The time gap between these developments may reflect the increasing awareness of the evolutionary relationship between DD-peptidases and β-lactamases and thus their close structural and functional similarity.[9−11]

Scheme 1

β-Lactamases catalyze the hydrolysis of β-lactam antibiotics and are thus an important source of bacterial resistance to these molecules.[12] The reaction (Scheme 2; shown with a n class="Chemical">penicillin) proceeds by way of a covalent acyl enzyme intermediate 4 and, therefore, through tetrahedral intermediates 3 and 5. Acyl-enzymes, analogous to 4, are formed on reaction of DD-peptidases with β-lactams but in this case hydrolyze very slowly leading to effective inhibition of these enzymes and thus interruption of bacterial cell wall synthesis. One would expect that the closest boronate analogue to a β-lactamase deacylation tetrahedral intermediate/transition state 5 would be 6, arising from reaction between the enzyme and boronic acid 7. A number of approximations to the structure 7 have been described, for example, initially, amidoalkyl boronic acids such as 8.[13,14] Subsequently, closer analogues, such as 9 and 10, were found to be very powerful β-lactamase inhibitors.[15,16] Crystal structures showed them to form the anticipated tetrahedral adducts 2 at the β-lactamase active site. To complement these developments, we describe here the syntheses of the boronic acids 11 and 12. We follow this with a description and analysis of their inhibitory activity against representative serine β-lactamases and DD-peptidases.

Scheme 2

Materials and Methods

The boronic acids 11 and 12 were synthesized as described in detail in Supporting Information. The n class="Species">Actinomadura R39 and Streptomyces R61 DD-peptidases, and Bacillus subtilis PBP4a, were generous gifts from Dr. J.-M. Frère and Dr. P. Charlier of the University of Liège, Liège, Belgium. The Escherichia coli PBP5 DD-peptidase was a generous gift from Dr. R. A. Nicholas of the University of North Carolina, Chapel Hill, NC. The AmpC β-lactamase was provided by Dr. B. K. Shoichet of the University of California at San Francisco, San Francisco, CA. The class C P99 β-lactamase from Enterobacter cloacae, the class A TEM-2 β-lactamase from E. coli W3310, and the class A Staphylococcus aureus PC1 β-lactamase were purchased from the Centre for Applied Microbiology and Research (Porton Down, Wiltshire, UK). The class A SHV-1 enzyme was a gift from Dr. Michiyoshi Nukaga of Jyosai International University, Japan.

Enzyme Kinetics Studies

DD-Peptidase Inhibition

In Solution

Experiments designed to obtain equilibrium constants of inhibition of the Actinomadura R39 DD-peptidase, n class="Species">B. subtilis PBP4a, and E. coli PBP5 in solution by compounds 11 and 12 were performed as described previously[17] from steady-state competition experiments where N-(phenylacetyl)glycyl-d-thiolactic acid was employed as a spectrophotometric (245 nm, Δε = 2500 cm–1 M–1) substrate (0.5 mM). Enzyme concentrations were between 0.1 and 0.2 μM. Initial rates of substrate hydrolysis in the presence of a range of concentrations of 11 and 12 (0–1.0 mM) were obtained.

Membrane-Bound Enzymes

Equilibrium constants for inhibition of E. coli DD-peptidases (PBPs) in membranes were obtained as described previously, employing Bocillin Fl as a fluorescent competitive β-lactam.[18] Compounds 11 (0–1.0 mM) and 12 (0–100 μM) were incubated with n class="Species">E. coli membrane preparations for 1 h prior to addition of Bocillin Fl (20 μM).

β-Lactamase Inhibition

Equilibrium constants of inhibition of the P99 and AmpC β-lactamases by compounds 11 and 12 (0–100 μM) were obtained from steady-state competition experiments where n class="Chemical">cephalothin was employed as a spectrophotometric (262 nm, Δε = 7660 cm–1 M–1) substrate (0.2 mM). The reaction conditions were 20 mM MOPS buffer, pH 7.50, 25 °C, and enzyme concentrations of 2 nM, stabilized by 0.1% bovine serum albumin in solution. Under these conditions, the Km value of the substrate was 9.0 μM for the P99 enzyme[19] and 13.8 μM for AmpC. No time-dependence of inhibition was observed in the manual mixing time frame in these experiments. Measurements of the initial rates, v, of cephalothin hydrolysis catalyzed by the β-lactamase in the absence and presence of 11 or 12 were fitted to Scheme 3 by means of eq 1 (least-squares), where v0 is the initial rate in the absence of inhibitor, to obtain the Ki values. Attempts to obtain Ki values of 11 and 12 for the class A TEM-2, SHV-1, and PC1 β-lactamases were conducted similarly.

Scheme 3

Molecular Modeling

Simulations were performed on a SGI workstation running the program Insight II, essentially as previously described.[21,22] The crystal structure of the ampC β-lactamase, inhibited by 15 [PDB entry 1MXO(23)], including the crystallographic n class="Chemical">water molecules, was modified to construct the adduct of 12 with the active site Ser 64. The pH was set to 7.0, and the total charge on the complex was zero. At this pH, the side chain of Tyr 150 was neutral, and those of Lys 67 and 315 were cationic. The partial charges of the enzyme were assigned by Insight II. Partial charges (MNDO) of the inhibitor in the complex were calculated from a model adduct with serine. Boron is not parametrized in Insight II so fixed standard crystallographic B–C and B–O bond distances and a rigid tetrahedral geometry at boron were assumed. The active site was hydrated with a 15 Å sphere of water centered on Oγ of the nucleophilic serine 64. The models were then subjected to 100 steps of steepest descent energy minimization followed by short molecular dynamics runs (40 ps) to relax the active site side chains and inhibitor. A typical snapshot from the molecular dynamics runs was selected and subjected to 1000 steps of steepest descent energy minimization, followed by 2000 steps of conjugate gradients.

A similar procedure was followed to obtain a model of the deacylation analogue adduct of 12 with the TEM-1 β-lactamase. This was derived from the crystal structure of the complex of 15 with this enzyme [PDP entry 1NXY(16)]. Note, however, that the n class="Chemical">thiophene ring of 15 has been retained in this model rather than changed to phenyl. In the TEM active site, Glu166 was neutral and Lys73 and Lys234 cationic. A generic model of a tetrahedral acylation adduct of 12 with the TEM-1 enzyme was derived directly from the crystal structure of its complex with 27 [PDB entry 1NY0(16)]. Models of benzylpenicillin at the CTX-M-9 and E. coli PBP4 active sites were built directly from the published crystal structures [PDB entries 3HUO(24) and 2EX8,[25] respectively]. In each case, the acyl forms were converted to tetrahedral intermediates by Insight modeling.

Results and Discussion

The syntheses of the boronic acids 11 and 12 are outlined in Schemes 4 and 5, respectively. In these syntheses, we made use of the recent discovery that stable α-n class="Chemical">boryl aldehydes can be prepared when N-methyl or pinene iminodiacetic acid (MIDA or PIDA) boronic acid protecting groups are employed.[26,27] Our use of MIDA thus allowed synthesis of aldehydes 13 and 14 from which the thiazolidine rings of 11 and 12 could readily be constructed. We also found that the MIDA protecting group was conveniently removed when treated with penicillamine in methanol during the last step of each synthesis, yielding 11 and 12 in unprotected form. High resolution mass spectral characterization of 11 and 12 presented difficulties since free boronic acids are generally not volatile or stable enough under the conditions required.[28] The mass spectrum of 11 as the pinacol ester, however, displayed a small peak at the appropriate mass (m/z = 406.3), but even this represented a metastable ion and was thus not suitable for accurate mass measurement. For both 11 and 12 we were able to obtain accurate mass measurements for M-BO2 cations from these pinacol esters (see Supporting Information). The presence of the boronic acid moiety in the parent boronic acids was indicated in their 1H NMR spectra by the absence of a methylene hydrogen resonance in the spectrum and the characteristic upfield shift of the hydrogen α to the boryl group.[29]

Scheme 4

Synthesis of 11

Stereochemistry is discussed in the text. Reagents and conditions: (a) Grubb’s II catalyst, CH2Cl2, reflux; (b) mCPBA, CH2Cl2, 0 °C; (c) BF3.Et2O, CH2Cl2, −30 to 0 °C; (d) d-penicillamine, MeOH-H2O, rt.

Scheme 5

Synthesis of 12

Stereochemistry is discussed in the text. Reagents and conditions: (a) nBuLi, THF, −78 °C; (b) BnOH, nBuLi, THF, −78 °C; (c) CH2Cl2, nBuLi, THF, −100 °C; (d) LHMDS, THF, −100 °C; (e) PhCH2COCl, CH2Cl2, −78 °C; (f) 1-methylpropylboronic acid, MeOH – H2O/hexane, 6 h; (g) MIDA, DMF, 80 °C, 12 h; (h) H2, Pd on carbon, MeOH, 40 psi, 12 h; (i) DMP, AcOH – CH3CN; (j) d-penicillamine, MeOH.

Stereochemistry is discussed in the text. Reagents and conditions: (a) nBuLi, THF, −78 °C; (b) BnOH, nBuLi, n class="Chemical">THF, −78 °C; (c) CH2Cl2, nBuLi, THF, −100 °C; (d) LHMDS, THF, −100 °C; (e) PhCH2COCl, CH2Cl2, −78 °C; (f) 1-methylpropylboronic acid, MeOH – H2O/hexane, 6 h; (g) MIDA, DMF, 80 °C, 12 h; (h) H2, Pd on carbon, MeOH, 40 psi, 12 h; (i) DMP, AcOH – CH3CN; (j) d-penicillamine, MeOH.

These syntheses yielded, as expected, mixtures of diasteroisomers that we did not attempt to separate since we could get a good estimate of the activity of the likely most active diastereoisomer from the activity of the mixture. The preparation of 12 yielded a mixture of two diasteroisomers in an essentially 1:1 ratio as indicated by the 1H NMR spectrum. Presumably these would have structures 12a and 12b (Scheme 6), formed in the last step (j) of the synthesis. All precedent with β-lactam-recognizing enzymes would suggest thn class="Species">at 12a, which has the stereochemistry of bicyclic β-lactams, would be the active inhibitor. The situation with 11 was more complicated. Epoxidation and rearrangement of the intermediate alkenes (Scheme 4) led to a mixture of two enantiomers with a chiral carbon α to the boronic acid group (Scheme 6, a/c and b/d) in a 4:1 ratio, where the stereochemistry of (a), α to the boryl group, is likely to be present in the major component.[27] Subsequent formation of the thiazolidine ring (step d) would presumably, as with 12, subdivide these equally into a final 4:4:1:1 mixture, as observed by 1H NMR (Supporting Information), where the likely most active 11a would represent 40% of the total. A minor component 11c might also have some activity since this stereochemistry is found in carbapenem antibiotics.[30]

Scheme 6

Boronic Acid Stereochemistry

The first important result was that neither 11 nor 12 inhibited representative DD-peptidases. In solution, no inhibition of the low molecular mass DD-peptidases of Streptomyces R61 and n class="Species">Actinomadura R39, or of B. subtilis PBP4a was observed. Gel experiments also demonstrated that these compounds did not inhibit E. coli PBP1a/1b, 2, 3, 4, and 5. These results concretely support a current general proposal for the mechanism of inhibition of DD-peptidases by β-lactams. Figure 1 shows the structure of a model of the putative tetrahedral intermediate for deacylation of the covalent acyl-enzyme formed on reaction of benzylpenicillin with E. coli PBP4. This structure was generated directly from the acyl-enzyme structure.[25] This structure is not stable (and, consequently, deacylation is very slow) because of unfavorable steric interactions between the hydroxyl group of the water nucleophile and the adjacent carbon and nitrogen atoms of the thiazolidine ring. These interactions are nicely seen in a space-filling model (Supporting Information, Figure S1). If this rationale for the slow deacylation of β-lactams from DD-peptidases were true, the tetrahedral boronate adduct 6 would also be unstable, and thus 11 and 12 would not be DD-peptidase inhibitors. Therefore, our results with these inhibitors support the mechanism of action of β-lactams described above. It is also reported that 15 does not inhibit Streptococcus pneumoniae PBP1b.[31] As would be anticipated from the above discussion, analogues of 16, unsubstituted α to the boronic acid, are more likely to be DD-peptidase inhibitors.[32−36]

Figure 1

Active site of the E. coli PBP4 DD-peptidase with benzylpenicillin bound as a deacylation tetrahedral intermediate. Modeled from the crystal structure of the acyl-enzyme.[25]

The results of Table 1 show that the new boronic acids 11 and 12 at micromolar concentrations do inhibit class C β-lactamases, the P99 and n class="Gene">AmpC enzymes. Notably, 11 is comparably effective to 12, despite the absence in the former of the amido side chain which is present in good substrates and generally thought to be important for active site recognition through hydrogen bonding.[37] Apparently the hydrophobic side chain of 11 is just as effective for inhibition as the amido side chain of 12, which we can assume binds in the usual site between the backbone carbonyl oxygen of residue 318 and the amide side chain of the conserved Asn152.

Table 1

Enzyme Inhibition by Boronic Acids 11 and 12

	Ki (μM)a
enzyme	11	12
P99 β-lactamase	0.62 ± 0.15	0.95 ± 0.11
AmpC β-lactamase	0.38 ± 0.12	1.3 ± 0.4
TEM-2 β-lactamase	NIb	NIb
PCI β-lactamase	NIb	NIb
R39 DD-peptidase	NIc	NIc

The Ki values above are not corrected for the presence of stereoisomers of 11 and 12 (see text). If it were assumed that the most likely stereoisomer in each case was the only one with activity, the values for 11 reported above would be multiplied by 0.4 and those for 12 by 0.5 (see text) to obtain the Ki values of the active isomers.

NI, no inhibition observed at the concentration 1.0 mM.

NI, no inhibition observed at the concentration 0.10 mM.

Boronic acid 12 is similar in structure to 15, an inhibitor described by Morandi et al.[23] The latter compound is a 1 nM inhibitor of the n class="Gene">AmpC β-lactamase. A crystal structure of the inhibitory complex has been published,[20] from which the active site diagram of Figure 2A has been taken. This shows the amide side chain of 15 firmly hydrogen-bonded to the protein as described above, and the boronate present as a tetrahedral anion covalently bound to the nucleophilic serine of the active site. One boronate oxygen is in the oxyanion hole (hydrogen-bonded to backbone NH groups of Ser64 and Ala318), and the other takes up the position of a leaving group or of the deacylating water molecule, depending on whether the complex is seen as an analogue of an acylation or deacylation tetrahedral intermediate (see below).

Figure 2

(A) Active site of the AmpC β-lactamase with the boronic acid 15 bound, from the crystal structure.[23] (B) An energy-minimized model of 12 bound to the AmpC active site, derived directly from the structure in A.

Finally, the carboxyphenyl group is oriented with its plane perpendicular to the chain formed by the amido side group, the alkyl n class="Chemical">boronate, and the side chain of Ser64. In this position, the carboxylate is directed above the β2-strand where it forms a hydrogen bond with the amido side chain of Asn 289. This is apparently an effective arrangement since the boronic acid 16, lacking the carboxyphenyl group, has a Ki value of 0.57 μM.[14]

A model of the complex between 12 and the AmpC β-lactamase was constructed based on the crystal structure with 15 as described in Materials and Methods. The initial structure was subjected to a short molecular dynamics simulation to relax active site interactions and a typical snapshot energy-minimized (see Materials and Methods). This procedure led to the structure of Figure 2B. This resembles the structure of the complex of 15 described above but differs in the positioning of the n class="Chemical">thiazolidine carboxylate vs the phenyl carboxylate of 15, due to the nonplanar nature of the five- (vs six-) membered thiazolidine ring and thus to the positioning of its substituents. The short MD run suggested that the thiazolidine carboxylate may prefer interaction with Arg349 over that with Asn 289. A similar dynamics run on the complex with 15 suggested that it too may be mobile in solution, with carboxylate access even to the Arg204 side chain. At any event, it appeared that the carboxylates of 12 and 15 may prefer somewhat different environments with some complementary adjustment of the active site structure. These effects presumably lead to the weaker binding of 12 than 15 to the enzyme. It is possible that a cephalosporin analog of 12, e.g., 17, which more closely resembles 15, may be a better inhibitor of class C β-lactamases: kcat values (deacylation rate constants) of cephalothin and cephalexin are larger than that of benzylpenicillin for the P99 enzyme,[28,29] suggesting that deacylation transition states of the former may be better stabilized by the enzyme.

It might be noticed in passing that the interactions of the inhibitor carboxylate group with the enzyme described above hold the n class="Chemical">thiazolidine ring of 12 from sterically impeding nucleophilic attack by water on the acyl-enzyme intermediate; the thiazolidine nitrogen is held well away from both boronate hydroxyl groups. This situation is in clear contrast to that seen in DD-peptidases[11] (see above) and presumably allows facile hydrolysis of the acyl-enzyme by β-lactamases.

It might also be noted that Gln289 is not strictly conserved in class C β-lactamases but is replaced by serine in Enterobacter homologues such as the P99 β-lactamase. The hydroxyl of serine in this position may also be able to form a hydrogen bond with the carboxylate of a penicillin substrate and of an inhibitor such as 15. Certainly 12 inhibits the P99 and AmpC β-lactamases to very similar extents (Table 1).

Perhaps more striking than the results with class C enzymes are those with class A. Neither the TEM-1, SHV-1, nor PCI β-lactamases were inhibited by 11 or 12, even n class="Species">at 1 mM concentrations. Given the fact that 12 should be the source of a precise transition state analogue structure by reaction with these enzymes (Schemes 1 and 2), much as with the class C β-lactamases, the result is quite astonishing. In stark contrast, 15 is a powerful inhibitor of the TEM-2 β-lactamase; Ki = 64 nM.[16] This profound and unexpected difference between 12 and 15 is discussed below.

First, it is necessary to recall that the geometry of the acylation and deacylation transition states (and tetrahedral intermediates, Scheme 2) is different in class A β-lactamases, unlike the situation with the class C enzymes where the acylation and deacylation tetrahedral intermediates have similar structure (see above). This situation arises because, although the hydrolytic water molecule is believed to attack the acyl-enzyme intermediate from the solution side (Re face) in class C β-lactamases, this attack comes from the protein side (Si face), catalyzed by n class="Chemical">Glu 166, in class A β-lactamases.[40,41] During acylation, the leaving group is thought to depart from the Re face of the acyl-enzyme in each case. These details are shown diagrammatically in Scheme 7.

Scheme 7

Mechanism of Formation of Tetrahedral Intermediates and Their Boronate Analogues

In this diagram, 18 and 20 represent the acylation and deacylation tetrahedral intermediates in class C β-lactamase catalysis, and 22 and 23, respectively, represent these species for a class A enzyme. The boronate analogues 21, 24, and 26 are shown below their respective intermediates, and 19 is the central acyl-enzyme. The arc adjacent to one oxygen ligand in each case represents the oxyanion hole, the presence of which serves to distinguish the two boronate hydroxyl groups.

In this diagram, 18 and 20 represent the acylation and deacylation tetrahedral intermediates in class C β-lactamase catalysis, and 22 and 23, respectively, represent these species for a class A enzyme. The n class="Chemical">boronate analogues 21, 24, and 26 are shown below their respective intermediates, and 19 is the central acyl-enzyme. The arc adjacent to one oxygen ligand in each case represents the oxyanion hole, the presence of which serves to distinguish the two boronate hydroxyl groups.

Thus, although one would expect all boronic acid complexes of class C enzymes to resemble 18 and 20 (i.e., have structure 21, the two forms of which are interconvertible by a small single bond rotation), two possibilities, 24 and 26, exist for class A β-lactamases. Indeed, examples of each kind have been obn class="Chemical">served with class A β-lactamases in crystal structures of boronate complexes.[16] Note that 22 and 23 (or 24 and 26) are not interconvertible by single bond rotations.

The crystal structure of the strongly inhibitory (64 nM) complex between 15 and the TEM-1 β-lactamase represents a mimic of a deacylation tetrahedral intermediate (Figure 3A), where one n class="Chemical">boronate hydroxyl group is in hydrogen bond contact with the deacylation catalyst Glu166.[16] In this structure, the carboxylate group of the inhibitor forms a strong hydrogen bond with the Thr235 hydroxyl group and, probably, with the Arg244 side chain via a water molecule. An energy-minimized model of a complex of 12 with this enzyme is shown in Figure 3B. In this structure, the interactions with the active site are the same as that of the complex with 15. Both structures seemed stable during short MD runs. The experimental result, that 12 (or 11) is not an inhibitor, therefore remains a puzzle.

Figure 3

(A) Active site of the TEM-1 β-lactamase with the boronic acid 15 bound, from the crystal structure.[16] (B) An energy-minimized model of 12 bound to the TEM-1 active site, derived directly from the structure in A. Note, however, that the thiophene ring of 15 has been retained in this model rather than changed to phenyl.

A possible way out of this dilemma may be found when the mechanism of formation of deacylation analogue boronate complexes is considered. Direct formation of these complexes from the free n class="Chemical">boronic acids and enzyme might be difficult since it would require prior or concerted displacement of the deacylating water molecule from Glu166. An alternative, perhaps more facile, mechanism would proceed by way of initial direct formation of an acylation tetrahedral intermediate analogue, 24. In an analogous fashion to the acylation/deacylation sequence of a substrate (Scheme 7), consecutively a neutral trigonal boronic acid intermediate 25 may be formed from 24, followed by its attack by the hydrolytic water molecule, presumably catalyzed by Glu166, to form the deacylation tetrahedral intermediate analogue 26.

The crystal structure of the complex between the class A TEM-1 β-lactamase with the n class="Chemical">boronic acid 27 resembles the acylation tetrahedral intermediate 22 (i.e., it has the structure 24) while, as noted above, that with 15 resembles the deacylation intermediate 23 (i.e., it has the structure 26). Wang et al. have suggested that the preference for the deacylation analogue structure 26 by 15 derives from the presence of the pendant carboxylate group which interacts with Arg244 in the complex, as seen in the crystal structure[16] (Figure 3A). One might expect that the boronic acids, 11 and 12, bearing the thiazolidine carboxylate, would also form stable complexes analogous to 26. Ke et al. have also discussed the relative merits of acylation vs deacylation complexes in various specific instances.[42]

We approached a structural model of acylation tetrahedral intermediate complexes by way of the crystal structure of the complex of 27 with the TEM-1 enzyme.[16] An immediate problem was that a bulky substituent α to the n class="Chemical">boronic acid, as present in 11, 12, and 15, sterically interacts unfavorably with Tyr105 (Figure 4), raising the likelihood of significant conformational adjustment (Figure 4). Models of both 12 and 15 were unstable, both to energy minimization and to MD simulation. In both cases, particularly that of 12, expansion of the lower part of the active site, comprising Tyr105, Asn132, and Glu166, occurred. In particular, Tyr 105 was “pushed away” with motion of the Asp101 – Leu108 loop. This distortion was more extreme with the bulkier (nonplanar) thiazolidine of 12. Acylation tetrahedral intermediates thus appeared likely to be less stable than the deacylation species described above. This is certainly a reasonable explanation for the two types of boronate structures obtained, the nature of the complex depending on whether a bulky substituent α to the boronic acid is present.

Figure 4

Active site of the TEM-1 β-lactamase with the boronic acid 27 bound, from the crystal structure.[16] Also shown, in the form of an added methyl group, is the general orientation of an l-α substituent, as present in 11 and 12.

This problem of the instability of the intermediate 22 (and thus the analogue 24) might also be sufficient to explain the lack of inhibition by 11 and 12. The ability of 15 to act as inhibitor but not 11 and 12 can be supposed to arise from the greater bulk of the thiazolidine substituent of the latter than the phenyl of the former (Figure S2, Supporting Information) and thus the greater difficulty of the latter to achieve the acylation analogue structure 24 required (Scheme 7) as a precursor of the likely stable deacylation analogue structure 26. This is an explanation in terms of unfavorable kinetics, but it should be noted that no inhibition of the TEM-2 enzyme was obn class="Chemical">served in 24 h.

In support of the rationalization above, it should be pointed out here that acylation tetrahedral intermediate analogues generated from acyclic boronic acids such as 11, 12, and 15 are not likely to closely resemble the structure of a “real” acylation tetrahedral intermediate 3 (Figure 5). This structure was generated from the crystal structure of a noncovalent complex between n class="Chemical">benzylpenicillin and a Ser70Ala mutant of the class A CTX-M-9 β-lactamase.[21] In this structure, the azetidine ring is still intact, with the scissile C–N bond eclipsed by the C–C bond on the opposite side of the ring. Such eclipsing will be lost for steric reasons when the four-membered ring is opened, as seen in the structures of Figure 3. In Figure 5, the carboxylate substituent of the thiazolidine ring forms hydrogen bonds with the side chains of Lys234 and Thr235. After rotation away of the thiazolidine ring with opening of the four-membered ring, this hydrogen-bond pattern is lost and replaced by the interactions seen in Figure 3. Thus, acyclic boronates cannot generate close acylation transition state analogue structures, at least of bicyclic β-lactam substrates, hence the instability of the boronate acylation complexes 24 of class A β-lactamases. On the other hand, acyclic boronic acids can form direct analogues of deacylation tetrahedral species, 23 (Scheme 7). It is possible that recently discovered cyclic borinates[43,44] may better approximate acylation complexes.

Figure 5

Active site of the class A CTX-M-9 β-lactamase with benzylpenicillin bound as an acylation tetrahedral intermediate. Modeled from the crystal structure of the noncovalent complex of penicillin with the Ser70Gly mutant.[24]

These considerations lead to another point and one that provides direct evidence for the instability of acyclic boronate adducts 24. n class="Chemical">Methyl penicilloate, 28, is known to be a substrate of a Pseudomonas class C β-lactamase,[45] and we have extended this result to the P99 β-lactamase (see Supporting Information). This compound is, however, neither a substrate nor a covalent inhibitor of the class A BCI β-lactamase,[46] and we have extended this point by observations with the TEM-2 enzyme (Supporting Information). These observations prove that while the class C β-lactamase active site can significantly stabilize the acylation tetrahedral intermediate 29 (a direct analogue of 18), the class A active site cannot (stabilize the analogue of 22), presumably for the reasons discussed above.

Summary and Conclusions

Neither 11 nor 12 [or, most likely, 15(31)] inhibit DD-peptidases, even at 0.1 mM concentrations, probably because of unfavorable steric interactions at the active site of these enzymes (Figure 1). Deacylation tetrahedral intermediates of DD-peptidase catalysis are thought to be destabilized in the same way.[9,11] The results with 11 and 12 from experiments in solution therefore strongly support the steric mechanism of inhibition of DD-peptidases by β-lactams and thus the mechanism of antibiotic action by these compounds. Previous evidence for this mechanism largely rested on inspection of crystal structures of inert complexes.

The boronic acids 11 and 12 inhibit class C β-lactamases, at micromolar concentrations, presumably by formation of covalent tetrahedral adducts (Figure 2B) that resemble the high energy tetrahedral intermediates of n class="Chemical">penicillin turnover. Compounds 11 and 12 are not, however, as effective as inhibitors as the phenyl analogue 15.[23] It is possible that 17, a cephalosporin analogue, closer in structure to 15, may be more effective than 12. The new compounds do not inhibit class A β-lactamases even at millimolar concentrations, in strong contrast to 15, which, at nanomolar concentrations, forms 26, a structural analogue of the deacylation tetrahedral intermediate 23(16) (Scheme 7). A rationale for this surprising result is offered in terms of a mechanism of formation of 26 (Scheme 7), which requires the initial formation of an acylation tetrahedral analogue 24, followed by that of a neutral trigonal boronic acid intermediate 25, analogous to an acyl-enzyme, and finally formation of 26 by intramolecular water attack on 25, presumably catalyzed by Glu166. A model of the acylation tetrahedral intermediate analogue from 12 suggests that this species may be unstable on steric grounds, precluding progress of 11 and 12 toward the deacylation analogue 26. An effective transition state analogue inhibitor requires an energetically accessible path to the inhibitory complex as well as transition state mimicry in that complex.[47]