It has been known for more than three decades that boronic acids are good inhibitors of serine proteases and the considerable efforts to find inhibitors of β-lactamases have now been extended in direction of PBPs by several groups. In 2003, Pechenov et al. [67] developed a series of transition state analogs, amongst which they identified the peptide boronic acid Boc-L-Lys(Cbz)-D-boroAla 1 (Figure 5) as an excellent inhibitor of three LMW-PBPs (N. gonorrhoeae PBP 3: K_i = 0.37 µM).

The crystal structure of compound 1 in complex with E. coli PBP5 revealed, as expected, the boron covalently attached to the active serine. The complex mimics the transition-state intermediate during the deacylation step of the enzyme-catalyzed reaction [68]. More recently, the peptidyl boronic acid 2 with a diaminopimelic acid like side chain was developed to target the R39 active site. A K_i value of 32 nM was obtained, which correlated well with the tight fit of the diaminopimelic side chain into the enzyme active site and the strong interactions made by this side chain ammonium and carboxylate groups with residues bordering the active site cleft, as revealed by the X-ray structure [69].

These peptidyl boronic acids certainly helped in the understanding of the underlying mechanisms of the enzyme DD-carboxypeptidase/DD-endopeptidase activity. But as they proved to be excellent inhibitors of their target PBP, further efforts were made to synthesize non-peptidyl boronic acids and assays were extended to target more medically relevant PBPs (S. pneumoniae PBP2x, S. aureus PBP2a, and E. faecium PBP5 [53,70,71].

A set of 21 commercially available boronic acids, including phenylboronic acids, thiophenyl boronic acids, an alkylboronic acid, a bicyclic benzoxaborole and a boronic acid pinacolester were tested for inhibition of R39 [70]. The most potent inhibitor was 3, with a residual activity of 20% at 1 mM after a pre-incubation of 60 min. A library of boronic acids analogs was synthesized. Two ortho-substituted derivatives were poor inhibitors of R39 but nearly all of the meta-substituted aryl boronic acids displayed improved inhibition against R39. The most active compounds 4 have IC₅₀ values in the 20–30 µM range. Some of these compounds displayed also activities against R6 PBP2x, 5204 PBP2x (penicillin resistant) and PBP1b all of S. pneumoniae.

Amidoethylboronic acids were in general better inhibitors of R39 and structure guided development of these compounds led to inhibitors with IC₅₀ values around 1 µM, the most powerful inhibitor being 5 (K_i = 63 nM, IC₅₀ < 0.08 µM, pre-incubation 60 min) [71]. The computer aided approach for elaborating those inhibitors from a structure of R39 in complex with 6 (IC₅₀ = 33 µM, pre-incubation 60 min) was validated by the crystal structure of R39 with compound 7 (IC₅₀ = 1.8 µM, pre-incubation 60 min). The structure shows both 7 aromatic rings occupying different active site pockets, which had been identified computationally and used for structural modifications of 6 (Figure 6).

Some amidomethyl- and amidoethylboronic acids were inhibitors of S. pneumoniae PBP1b with IC₅₀ values lower than 20 µM, the most powerful inhibitors being 5 (IC₅₀ = 4.9 µM, pre-incubation 60 min) and 8 (IC₅₀ = 6.9 µM, pre-incubation 60 min). A series of crystal structures of amidoethyl- and amidomethylboronic acids could also be obtained with S. pneumoniae PBP1b, some of these inhibitors showing antibacterial activity against methicillin resistant S. aureus (Table 1) [52].

The structures revealed two distinct side chain binding modes: the methyl group of the S-amidoethyllboronic acids (5–7, 9 and 10) bound in the active site region similarly to the amide side chain of penicillins and cephalosporins, and thus should be more representative of substrate analogs than the amidomethylboronic acids or the R-amidoethylboronic acids, the side chain of which could bind in an alternative region of the enzyme catalytic cleft. This was even more dramatically observed in structures of R39 in complex with amidomethylboronic acids [58]. The boron atom covalently binds to the active serine Ser49 forming a monocovalent adduct but the ligand further inserts into the active site so that the boron is linked to one lysine Lys410 and two serine residues Ser49 and Ser298 and eventually makes a tricovalent adduct with the enzyme (Figure 7). On the basis of the crystallographic and kinetic results, a reaction scheme for this inhibition by boronic acids was proposed (Figure 7).

S-Amidoethylboronic acids display a greater analogy to the natural substrate than R-ethyl- or methyl-boronic acids. In R39 as well as in PBP1b, the Cα methyl group inserts into a hydrophobic pocket that plays an important role in the substrate specificity toward the D-alanine as the penultimate residue of the peptidoglycan stem pentapeptide. Thus boronic acids with larger substituent on the Cα methyl group should be very poor inhibitors of PBPs, as was shown in the case of R39 and PBP1b. In both enzymes, the binding modes that are not genuine analogs of substrate binding could represent an interesting alternative to find specific inhibitors; 2-nitrobenzamidomethylboronic acid 12 represents such an inhibitor of R39 [58].

Boronic acids are rarely used in medicinal chemistry [72]. The experimental evidence of their great PBPs inhibitory potential should encourage future developments and prospects as the fight against bacterial resistance to β-lactams continues.

Carbonyl compounds have been systematically studied by Pechenov et al., who synthesized peptide chloromethyl ketones, trifluoromethyl ketones and aldehydes as potential transition state analogs. Boc-L-Lys(Cbz)-D-Ala was chosen as the parent structure, since it is a simple dipeptide mimic of the natural PBP substrate (Figure 8) [67].

The peptide aldehydes Boc-L-Lys(Cbz)-D-Ala-H (Ki = 60 µM) and Boc-L-Lys(Cbz)-L-Ala-H (Ki = 79 µM) were identified as inhibitors of N. gonorrhoeae PBP3. A Ki-value of 60 µM was described for the inhibition of the same PBP3 by the diastomeric mixture of Boc-L-Lys(Cbz)-D,L-Ala-CF₃ while the chloromethyl ketone showed no inhibitory activity on PBPs. Trifluoromethylketone analogs of good boronic acid inhibitors of Actinomadura R39 were studied but showed no inhibitory activity [71].

Neutral phosphyl reagents are strong inhibitors of serine proteases [64]. Since the 1980s various phosph(on)ates were synthesized as inhibitors of β-lactamases. Often the DD-peptidase of Streptomyces R61 was used as a model enzyme to explore the inhibitory potential of phosph(on)ates on PBPs. Phosphonate monoesters 13 (Figure 9) were found to be the first β-lactamase inhibitors [73,74,75]. After formation of a trigonal bipyramidal transition state and departure of the leaving group L, β-lactamases form stable tetrahedral phosphonyl-enzyme adducts with phosphonate monoesters (Figure 2) [76,77,78]. The inactivation can be described by the inhibition model (ii) shown above. The formation of the covalent adduct, the acyl phosphonate, is characterized by a second-order rate constants k₂/K. The stability of the E-I complex is characterized by the rate constant k₃. Inhibitory power of phosphonate monoesters can be improved by selection of a good leaving group L and by modification of the amido side chain R. The phosphonate monoesters 14 and 15 (Figure 9) were poor inhibitors of R61 with k₂/K-values of 0.07 M⁻¹s⁻¹ [75] and 0.06 M⁻¹s⁻¹ [73], respectively. The observation that leaving group lability is an important element led to the development of acyl phosph(on)ates 16 (Figure 9). These molecules can form both acyl- and phosphor(on)yl-enzyme species. The acyl phosphate 17 was found to be a substrate of typical class A and class C β-lactamases and a good substrate of R61 (K_m = 0.2 mM, k_cat = 4.1 s⁻¹). The acyl phosphonate 18 was an irreversible inhibitor of β-lactamases, probably by phosphonylation of the serine (Figure 9), but was found to be a poor inhibitor of R61 with k₂/K-value of 5 × 10⁻³ M⁻¹s⁻¹. A very slow turnover of this molecule by R61 was observed (K_m = 0.21 mM, k_cat = 3.5 × 10⁻⁴ s⁻¹) meaning that an acyl-enzyme is formed or that the phosphoronyl enzyme is unstable [79].

Cyclic variants of phosph(on)ates like 19 were proposed. In this case the leaving group actually does not leave and avoids hydrolysis and rapid regeneration of the free enzyme [80]. The crystal structure of R61 with compounds 19 and 20 revealed that the R61 enzyme is preferentially phosphorylated rather than acylated. Both molecules inactivated R61 rather slowly (k₂/K-values of 19 and 20: 0.46 M⁻¹s⁻¹ and 24 M⁻¹s⁻¹, respectively) and form long-lived intermediates (k₃-values of 19 and 20: 2.7 × 10⁻³ s⁻¹ and 8.9 × 10⁻³ s⁻¹, respectively) [81]. Amidoketophosph(on)ates 21 inhibit typical class C and class D β-lactamases but not R61 [82].

Diisopropyl fluorophosphate (DIFP), which is efficient against serine proteases, was a poor inhibitor of E. coli PBP 5 (residual activity of 72% at a concentration of 1 mM after a pre-incubation of 1 h) [83]. A rather poor inhibition of Actinomadura R39 by phosphonic bioisoster of aminocitrate (22, residual activity of 47% at a concentration of 500 µM) and pyrrolidinone (23, residual activity of 67% at a concentration of 500 µM) (Figure 9) was described [84]. The enzymatic assay with R39 was done using a thioester assay [53] and a pre- incubation time of one hour (it should be noted that the experimental details described in the original publication—fluorescent ampicillin and a 16 h pre-incubation—are not correct). The phosphonates 24–27 show a modest inhibition of R39 (residual activity of 86%, 24 and 25, residual activity of 65%, 26 and 27 residual activity of 76%, all at a concentration of 1 mM and a pre-incubation of 1 h, respectively), [71] whereas the corresponding boronic acid of 24 [58] and 25 are good inhibitors of R39. The boronic acid analogue of 26 was a poor inhibitor (residual activity of 69% at a concentration of 1 mM). [53] These results probably reflect the structural differences between the covalent complexes formed between the enzyme and the boronic acids and the phosphonates, respectively. In phosphonate adducts the 2,4 dimethoxybenzoylamino side chain fits probably better than the 2,6 dimethoxybenzoylamino chain in the active site. Compounds 22–27 contain two hydroxyl groups which are poor leaving groups. The inhibitory power of the published compounds could probably be optimized by the introduction of a good leaving group and by the modification of the amido side chain. A crystal structure of a PBP-phosphonate complex is awaited to develop a structure based design of potent phosphonate inhibitors.

Biological active prototypes of bicyclic pyrazolidinones 47 (Table 2) were designed at the Lilly Research laboratories [93]. Modification of the substituents at C-3 (W) and C-7 (R) of 31 (Figure 10) allowed the variation of biological activities. Side chains from active β-lactam compounds like the ATMO—aminothiazoylmethoxime-side chain were introduced in C-7 position in 48 and 49 (Table 2). Compounds like 48 and 49 with strong electron-withdrawing groups in C-3 position were shown to be more sensitive to hydroxide ions and had better in vitro activities than the original prototype 31 [85].

In 1986, lactivicin (LTV) 50 (Table 3) was isolated from bacterial strains (Empedobacter lactamgenus and Lysobacter albus) by the Takeda Research group [94,95,96]. LTV, which has a unique ring structure comprising a functionalized L-cycloserinyl ring linked to a γ-lactone ring, was the first natural PBP inhibitor without a β-lactam ring. LTV is active against a wide range of Gram-negative bacteria and highly active against Gram-positive bacteria. LTV derivatives, where the 4-aminolactivicinic acid LTV nucleus was acylated by various residues, were synthesized to increase its antibacterial activity against Gram-negative bacteria and to overcome its relatively strong toxicity (Table 3) [87,97,98,99].

Crystallographic analyses of S. pneumoniae PBP1b with LTV and a more potent analog, phenoxyacetyl-lactivicin (PLTV, 52) reveal that inhibition of PBPs involves the opening of both monocyclic cycloserine and γ-lactone rings and the formation of a stable covalent adduct with the active site serine (Figure 11). PLTV 52 was a more efficient inactivator of 5204 PBP2x of penicillin resistant S. pneumoniae with an IC₅₀-value of 7.9 µM (pre-incubation: 60 min) compared to 150 µM (pre-incubation: 120 min) for LTV 50. PLTV showed antimicrobial activity against drug-sensitive (MIC: 2 µg/mL) and various drug-resistant S. pneumoniae strains (MIC: 10 or 20 µg/mL) [100].

A screening of chemical libraries led to a series of rhodanines, which were found to be inhibitors of class C β-lactamases in the micromolar range [101]. Similar compounds were found to be inhibitors of various PBPs including PBP2x of a penicillin resistant strain of S. pneumoniae and S. aureus PBP2a. Furthermore in vitro activities against various bacterial strains including S. aureus (MRSA), S. pneumoniae (PRSP), and E. faecium (VRE) strains were described. A detailed kinetic study with the most active compounds 55 and 56 (Figure 12) shows that the inhibitor activity of these slow-binding inhibitors is non-competitive, which means that they probably do not bind to the active site of β-lactam binding enzymes [54]. The inhibition by compound 55 is detergent-sensitive and inhibition activity cannot be detected in presence of 0.01% Triton-X-100 (personal communication) underlining that the detection of non-competitive inhibitors -promiscuous inhibitors- can be avoided by addition of detergents in the enzymatic assay as previously mentioned.

Aminothiadiazole derivate 57 and two ortho-phenoxyl diphenylurea compounds 58 and 59 are inhibitors of R6-PBP2x and 5204-PBP2x both of S. pneumoniae strains [R6 (penicillin sensitive) and 5204 (penicillin resistant)], Table 4, Figure 12 [102].

The discovery of these compounds was the result of a structure-based virtual screening. A pharmacophore model developed from an in-house database of active compounds against PBP2x was used to screen the NCI database containing 260,071 compounds.

The development of high throughput screening assays for PBPs has allowed the screening of compound libraries. After a screening of an in-house bank of 250 compounds from different non-reactive chemical classes, 60 (Figure 12) was identified as an initial hit, which inhibited PBP2a with a -positive strains including E. faecium ATCC 19434 (MIC: 64 µg/mL), various S. pneumoniae promising IC₅₀ of 97 µM, and 5204 PBP2x with a not less interesting IC₅₀ of 391 µM (Table 4). It also showed in vitro antibacterial activities against some Gram strains (MIC: 1 µg/mL), as well as both penicillin sensitive and resistant S. aureus strains (MIC: 32 µg/mL). A small library of structurally related compounds was obtained by performing computational similarity searches based on the structure of 60 as a starting point and using the ChemBridge bank of compounds containing more than 800,000 compounds [103]. From a series of naphthalene sulfonamides (four compounds), 61 (Figure 12) was a moderate inhibitor of PBP5fm but had no significant antibacterial activity. Various naphthalene sulfonamides were synthesized by the same authors to clarify the structure-activity relationship for PBP inhibition. Some inhibitors of PBP2a were found with IC₅₀-values in the micromolar range. The best inhibitor was 62 (Figure 12, Table 4). Unfortunately all naphthalene sulfonamides described in this study were only poor inhibitors of bacterial growth [104].

5-Bromo-2-(3-propoxybenzamido) benzoic acid (63) is a promising anthranilic acid inhibitor of PBP2a and 5204 PBP2x (Figure 12, Table 4). This compound showed a good antibacterial activity against Gram-positive bacterial strains, including E. faecium ATCC 19434 (MIC: 16 µg/mL), various S. pneumoniae strains (MIC: 1 µg/mL), as well as both sensitive and resistant S. aureus (MIC: 32 µg/mL). 63 was detected by performing computational similarity searches based on the structure of an analogue of 60, where the sulfonamide bond was replaced by an amide, as already described in paragraph 3.3.3 [103]. Chemical synthesis of a library of anthranilic acid analogs of 63 allowed the discovery of inhibitors of PBP2a in the micromolar range. The best of these inhibitors were 64 and 65 (Figure 12, Table 4). Both compounds showed good antibacterial activities against Listeria innocua, L. monocytogenes, S. epidermidis and B. subtilis strains, with MICs of 4 µg/mL or 8 µg/mL. Furthermore with 65, the growth of two MRSA strains, with MICs of 4 µg/mL and 2 µg/mL, was 16- to 32-fold more efficiently prevented than that of the penicillin-sensitive S. aureus strain, which had a MIC of 64 µg/mL. The author suggested that compounds 64 and 65 exert their in vitro antibacterial activities by targeting other proteins besides their inhibition of PBPs, because the MIC values are significantly lower than the IC₅₀ ones.

Cibacron Blue 66 and Erie Yellow 67 (Figure 12, Table 4) were identified as inhibitors of PBP2a, with IC₅₀-values of 24 µM and 13 µM, respectively [48]. The screening was done without addition of detergents like Triton-X-100 or the presence of bovine serum albumin to avoid the detection of promiscuous inhibitors. The kinetics of the inhibition mechanism were not studied and no crystallographic data are available. It is unclear whether these molecules competitively bind to the active site.

A disulfide bond containing cyclic heptapeptide (sequence NH₂-CYHFLWGPC-COOH), first selected as a β-lactamase inhibitor, was described as an inhibitor of some PBPs [105]. Competition experiments with fluorescent ampicillin were used to measure IC₅₀-values in the micromolar range. The kinetic mechanism of inhibition and the influence of detergents were not studied. In the absence of crystallographic data, it is not clear if this peptide binds competitively to the active site.

4-Quinolones were found to be noncovalent inhibitors of PBPs of E. coli and B. subtilis. The initial lead, an inactive 4-quinolone, was composed from fragments and docked into the active site of E. coli PBP5. Series of 4-quinolones were designed and synthesized. Dissociation constants (K_i) with membrane-bound PBPs of E. coli in the micromolar range were detected and inhibiton of PBPs of B. subtilis was observed. The most efficient binding was observed with compound 68 (Figure 12), with Ki values around 30 µM for each high molecular mass PBP of E. coli (PBP1a/1b, PBP2 and PBP3). All active 4-quinolones had no in vitro antibacterial activities against E. coli or B. subtilis [106].