Modulation of the Specificity of Carbapenems and Diazabicyclooctanes for Selective Activity against Mycobacterium tuberculosis

ABSTRACT

Treatment of multidrug-resistant tuberculosis with combinations of carbapenems and β-lactamase inhibitors carries risks for dysbiosis and for the development of resistances in the intestinal microbiota. Using Escherichia coli producing carbapenemase KPC-2 as a model, we show that carbapenems can be modified to obtain drugs that are inactive against E. coli but retain antitubercular activity. Furthermore, functionalization of the diazabicyclooctanes scaffold provided drugs that did not effectively inactivate KPC-2 but retained activity against Mycobacterium tuberculosis targets.

INTRODUCTION

Tuberculosis (TB) remains a major cause of mortality by infectious diseases, ranking above AIDS (1). The emergence and spread of multidrug- and extensively drug-resistant (MDR/XDR) Mycobacterium tuberculosis strains constitute a serious threat for the control of the TB pandemic. Second-line drugs used in the treatment of MDR/XDR-TB are often associated with adverse effects that frequently interrupt therapy and negatively impact the outcome (2). Drugs of the β-lactam family, in particular carbapenems associated with β-lactamase inhibitors (BLIs), have regained interest for the treatment of MDR/XDR-TB since they are well-tolerated antibiotics. Mechanistically, the broad-spectrum β-lactamase produced by M. tuberculosis (BlaC) is irreversibly inactivated by clavulanate, a first-generation BLI, and carbapenems are active on cell wall targets (3, 4). Meropenem, a β-lactam of the carbapenem class, has shown efficacy in combination with clavulanate in a phase II clinical trial (5). Carbapenems are the only β-lactam class effectively inhibiting both penicillin-binding proteins (PBPs) and l,d-transpeptidases (Ldts). Second-generation BLIs belonging to the diazabicyclooctane (DBO) family, such as avibactam, have been developed in combination with ceftazidime (a third-generation cephalosporin) to treat bacterial infections due to Gram-negative bacteria producing Ambler class A, class C, and certain class D β-lactamases (6). Although DBOs are not effective for the inhibition of BlaC since the carbamoylation is irreversible but slow (7, 8), they act as slow-binding inhibitors of Ldts (9). These enzymes are nonclassical peptidoglycan polymerases (10, 11) that are responsible for formation of the majority (70% to 80%) of the peptidoglycan cross-links in mycobacteria (12, 13) and in Clostridium difficile (14, 15).

Although carbapenems are potentially useful for the treatment of MDR/XDR-TB and are well tolerated (16, 17), these drugs have several disadvantages related to their broad antibacterial spectrum, namely, deleterious effects on the microbiota, which can potentially lead to dysbiosis and selection of carbapenemase-producing enterobacteria. Use of BLIs of the DBO family is also at similar risk, since emergence of resistance to the ceftazidime-avibactam combination resulting from impaired β-lactamase inhibition has already been reported under treatment (18, 19) and in vitro for various β-lactamases (20–22). Thus, treatment of MDR/XDR-TB by carbapenem-BLI combinations for extended time periods incurs a high risk of emergence of resistance to drugs that are the last line of defense against important multidrug-resistant pathogens, such as carbapenemase-producing enterobacteria (23). Since this risk could be mitigated by the use of drugs specific to M. tuberculosis, we investigate here a series of synthetic carbapenems (24, 25) and DBOs (9, 26) for their antibacterial activity against M. tuberculosis and Escherichia coli and their interaction with the KPC-2 carbapenemase. We show that modifications of the carbapenem and DBO scaffolds are a promising approach to develop drug combinations selectively active against M. tuberculosis.

Hydrolysis of carbapenems by KPC-2.

The routes of synthesis of the carbapenem series, containing various R₁ and R₂ side chains (Table 1), have previously been reported (24, 25), except for compound 29a (see the supplemental material). Exploiting the nucleophilicity of the C₈ hydroxyl group has been used to introduce chemical diversity in the R₂ position of carbapenems and increase similarity between the drugs and the substrates of the transpeptidases (25). Compound 29a was added to this series of carbapenems in order to determine whether introduction of a charged amine at the R₂ position instead of a phenethyl group would affect the kinetics of hydrolysis by β-lactamases and antibacterial activity. Purification of KPC-2 and determination of kinetic parameters were performed as previously described (27). Esterification of the hydroxyethyl side chain of the carbapenem scaffold (R₂) led to drastic decreases (up to 10,000-fold) of the catalytic efficacy (k_cat/K_m) of KPC-2 due to decreases in the catalytic parameter k_cat and, to a lesser extent, to increases in K_m. These results indicate that modification of the R₂ side chain can be used to minimize the selective advantage conferred by the production of KPC-2 in enterobacteria exposed to carbapenems.

TABLE 1.

Efficacy of hydrolysis of carbapenems by KPC-2^a

^aKinetic parameters k_cat, K_m, and k_cat/K_m for hydrolysis of carbapenems were determined at 20°C in 2-(N-morpholino)ethanesulfonic acid (MES; 100 mM; pH 6.4) by spectrophotometry, as previously described (27). The kinetic constants k_cat and K_m were determined by fitting the equation v_i (initial velocity) = k_cat[E][S]/K_m + [S] to data generated in a minimum of six progress curves and two independent experiments (27). Synthesis, purification, and characterization of compounds were described in reference 25, except for 2′ (24) and 29a (this work, supplemental material).

Antibacterial activity of carbapenems against E. coli.

We compared the MICs of synthetic carbapenems on the E. coli TOP10 strain and the NR698 strain, which is hyperpermeable to drugs due to defects in the outer membrane (28). MICs were determined by the broth microdilution method (29). Carbapenems obtained by modification of the R₁ and R₂ side chains were not active against E. coli TOP10 (Table 2). The absence of activity was at least partly due to reduced penetration of the drugs in the periplasm, since synthetic carbapenems were more active against the NR698 hyperpermeable strain than against TOP10. However, most synthetic carbapenems were less active than meropenem against NR698, indicating that the inhibition of PBP targets was also affected. The low antibacterial activity of the synthetic carbapenems against NR698 was achieved by introduction of a phenethyl side chain at the R₁ position. The antibacterial activity was further decreased when this modification was associated with introduction of R₂ substituents. These results indicate that modifications of the R₁ and R₂ side chains can be used to obtain drugs that are inactive against E. coli due to limited penetration in the periplasm and limited interaction with the PBP targets.

TABLE 2.

Antibacterial activity of carbapenems against Escherichia coli and Mycobacterium tuberculosis^a

^aThe antibacterial activity of carbapenems against E. coli TOP10 and NR698 strains and M. tuberculosis H37Rv was determined by the broth microdilution method (9, 29). MICs are the medians from a minimum of 6 biological replicates.

Antibacterial activity of synthetic carbapenems against M. tuberculosis.

We have previously reported that the esterification of the hydroxyethyl side chain (R₂) of the carbapenem scaffold does not compromise the efficacy of inactivation of a model l,d-transpeptidase, namely, Ldt_fm from Enterococcus faecium (25). Indeed, synthetic carbapenems with R₂ substituents were at least as active as meropenem for acylation of the purified Ldt_fm enzyme (25). This activity in an acellular assay was translated into antibacterial activity similar to that of meropenem against the E. faecium M512 strain (25), a mutant that exclusively relies on Ldt_fm for peptidoglycan cross-linking under conditions of inhibition of PBPs by ampicillin (30). This prompted us to determine the MICs of the synthetic carbapenems against the M. tuberculosis H37Rv strain that also predominantly relies on Ldts for peptidoglycan cross-linking (Table 2). MICs for M. tuberculosis were determined by the broth microdilution method (29). This analysis indicated that certain carbapenems obtained by modification of the R₁ and R₂ side chains retained antitubercular activity. In particular, the MICs of five compounds, 28a, 28c, 28e, 28l, and 29a, were similar to that of meropenem (4 or 8 μg/mL versus 8 μg/mL, respectively). Altogether, these results show that the carbapenem scaffold can be functionalized to obtain drugs that are active against M. tuberculosis but not against the model enterobacterium E. coli.

Inhibition specificity of synthetic DBOs.

Avibactam was previously reported to potentiate the activity of amoxicillin against M. tuberculosis, presumably by inhibiting a combination of cell wall targets such as classical PBPs (amoxicillin) and Ldts (9). Thus, DBOs and β-lactams are candidate scaffolds for the development of drug combinations active against M. tuberculosis. Since the use of such combinations would be at risk for the selection of resistant enterobacteria, we investigated whether the DBO scaffold could be modified to prevent inhibition of the KPC-2 β-lactamase. This would avoid the selection of mutations that would result in decreased inhibition of β-lactamases by DBOs (18–22). To that end, we tested a series of functionalized triazole derivatives of the DBO scaffold. The route of synthesis of these compounds, which contained a substituted triazole ring instead of the carboxamide present in avibactam, has previously been reported (9, 26). This modification impaired the efficacy of inhibition of purified KPC-2, as shown for the representative DBOs depicted in Table 3 (from 28- to 137-fold). This was associated with impaired potentiation of aztreonam antibacterial activity by the DBOs against E. coli producing KPC-2 (MICs, ≥32 μg/mL). The disk diffusion assay indicated that the DBOs either were noneffective or displayed reduced efficacy against β-lactamases CTX-M-15, AmpC, and OXA-48 (see Table S1 in the supplemental material). The strains remained resistant in the presence of DBOs according to the EUCAST susceptibility breakpoints except for inhibition of CTX-M-15 by DBOs 7b and 7f (Table S1). In contrast, the triazole-containing DBOs retained biological activity against M. tuberculosis since they potentiated amoxicillin activity (MICs = 16 to 32 μg/mL versus 8 μg/mL for amoxicillin-triazole-containing DBOs and amoxicillin-avibactam, respectively).

TABLE 3.

Specificity of triazole-containing DBOs

^aSynthesis of the DBOs was previously reported (9, 26). DBOs were used at a fixed concentration of 15 μM (corresponding to 4 μg/mL for avibactam). NA, not applicable.

^bThe efficacy of inhibition of purified KPC-2 (k₂/K_i) was previously determined by using nitrocefin as the substrate (26).

^cMICs were determined by the broth microdilution method as previously described for E. coli (26) and M. tuberculosis (9). Data are the medians from a minimum of six biological replicates.

Conclusion.

Carbapenems and DBOs have been developed to obtain broad-spectrum drugs and drug combinations active on β-lactamase-producing pathogens, including enterobacteria. More recently, it was recognized that carbapenems are also active against M. tuberculosis in combination with clavulanate, which irreversibly inactivates BlaC, a class A β-lactamase intrinsically produced by all members of this species. Although avibactam is a poor inhibitor of BlaC, this DBO displays antibacterial activity in combination with β-lactams, presumably by inactivating cell wall biosynthesis enzymes. These observations suggest that carbapenems and DBOs could be optimized for the treatment of tuberculosis. Accordingly, we previously showed that the carbapenem scaffold can be functionalized to obtain drugs that are superior to meropenem with respect to in vitro inactivation of the l,d-transpeptidases of M. tuberculosis (e.g., Ldt_Mt1) and limited hydrolysis by BlaC (24). In the current study, we explored another facet of drug optimization, namely, the design of drugs specifically active against M. tuberculosis to limit side effects on the intestinal microbiota in terms of dysbiosis and selection of resistances. Using KPC-2-producing E. coli as a model, we show that the carbapenem scaffold can be functionalized to obtain compounds that are essentially devoid of biological activity against E. coli since they combine ineffective inactivation of PBPs and poor penetration. In addition, the synthetic carbapenems and DBOs displayed slow hydrolysis and ineffective inhibition of KPC-2, respectively. Functionalization of the carbapenem and DBO scaffolds resulted in compounds that retained activity against M. tuberculosis and is thus a promising avenue for the development of selective antitubercular drugs. We have not directly tested carbapenems and DBOs in combination, and it is currently unclear whether the antibacterial activities of these compounds, which are potentially active against similar targets, are additive or synergistic, but that is a goal of future research.