In vitro testing of amoxicillin-imipenem-relebactam on M. abscessus
In this paper we have identified that M. abscessus can be made susceptible to amoxicillin by the addition of the competitive β-lactamase inhibitor, relebactam. We have further demonstrated that when used in conjunction with amoxicillin, the MIC of the imipenem-relebactam combination can be significantly reduced. This effect is not as pronounced in ceftazidime-avibactam in combination with amoxicillin. We also show that meropenem susceptibility can also be enhanced by the addition of relebactam. Firstly, we conducted disk diffusion assays with amoxicillin and meropenem, with and without relebactam (Fig. 1a). Subsequently the zones of inhibition (ZOI) were measured as a marker of susceptibility. Amoxicillin alone failed to demonstrate any kind of susceptibility, however the addition of relebactam provided a clear ZOI, demonstrating the induced susceptibility to amoxicillin by relebactam. A small ZOI was visible for meropenem, however with the incorporation of relebactam, this ZOI was enhanced, suggesting increased susceptibility upon combination with relebactam (Fig. 1a). In order to assess the clinical relevance of these observations, both amoxicillin and meropenem were tested alone and in combination with relebactam against a panel of clinical isolates of M. abscessus obtained from patients at Brighton and Sussex Medical School, in addition to M. abscessus NCTC strain. In all cases, the addition of relebactam resulted in the isolate becoming susceptible amoxicillin (n = 3) (Fig. 1b left). The same experiment was conducted with meropenem, and in most cases, an increase in susceptibility was observed (n = 3) (Fig. 1b right). All necessary controls were conducted and no ZOI was observed for relebactam alone.
In order to further validate our observation, liquid cultures were exposed to increasing concentrations of amoxicillin and meropenem with and without a dose response of relebactam (Fig. 1c). As a control, relebactam was also tested for inhibitory activity on its own. Relebactam at 128 mg/L demonstrated no inhibitory activity, giving a growth profile much the same as the compound-free control (no drug), confirming that relebactam lacks antibacterial activity against M. abscessus. 32 mg/L of amoxicillin appeared to show a moderate enhancement of bacterial growth, however when combined with 2 mg/L of relebactam, this amoxicillin concentration was found to display potent antibacterial activity against M. abscessus (Fig. 1c). The same experiment was conducted with meropenem, wherein partial growth impairment was observed at 8 mg/L, however when combined with 2 mg/L relebactam, significant antibacterial activity was observed at this concentration. End point statistical analysis was conducted using a t-test to assess the significance of differences between cultures with and without relebactam. In both cases, a statistically significant increase in susceptibility was observed upon addition of 2 mg/L of relebactam in combination with amoxicillin or meropenem.
Since relebactam is only available in combination with imipenem, experiments with amoxicillin in conjunction with imipenem-relebactam were conducted, alongside amoxicillin and ceftazidime-avibactam (Fig. 1d top and bottom respectively). Both of these β-lactam/ β lactamase inhibitor combinations are commercially approved by the FDA, however their efficacies against M. abscessus are yet to be compared. An experiment was conducted using a fixed ratio of imipenem-relebactam with and without amoxicillin, representative of the ratio of these drugs in the approved combination. At concentrations of 0.5:0.25 mg/L imipenem:relebactam without amoxicillin, growth was observed to be similar to that of the M. abscessus cells only control, whereas with the addition of 32 mg/L amoxicillin, total inhibition of growth is seen. A similar experiment was conducted using ceftazidime:avibactam with and without amoxicillin. Inhibition was observed at ceftazidime:avibactam 32:32 mg/L with the addition amoxicillin 8 mg/L, and those same concentrations with no amoxicillin were again similar to the no drug control.
In vitro antibiotic susceptibility testing was performed for imipenem, amoxicillin and relebactam, either on their own, in conjunction with each other, or as a triplicate combination, against 16 M. abscessus clinical isolates, M. abscessus containing the BlaMab constitutive overexpression plasmid pVV16-blaMab and M. abscessus pVV16 (Table 1). All but one isolate (93.75%) had MICs of >128 mg/L for amoxicillin and >32 mg/L for relebactam, indicating these isolates are not susceptible to these drugs within their reasonable clinical range. When combining amoxicillin with relebactam, 81.25% of the strains became susceptible to amoxicillin, with MICs ranging from 4–128 mg/L, requiring 0.5–16 mg/L of relebactam to produce this effect. The imipenem MICs ranged from 2–8 mg/L, therefore according to CLSI breakpoint guidelines, all of the clinical isolates were susceptible or intermediate to imipenem.
M. abscessus pVV16-blaMab had an MIC of imipenem of 32 mg/L. When combining imipenem with relebactam (MIC range 1–2 mg/L), the MICs of imipenem were reduced 2-fold in 43.75% of the clinical isolates, produced no effect in 31.25% of the clinical isolates and surprisingly, increased the MIC of imipenem 2-fold in 6.25% of the clinical isolates. The MIC of imipenem against M. abscessus pVV16-blaMab was reduced 4-fold from 32 mg/L to 8 mg/L with the addition of 4 mg/L relebactam. The MIC of imipenem against M. abscessus pVV16 was unchanged at 2 mg/L. When imipenem and relebactam (MIC range 0.25–1 mg/L) were combined with amoxicillin (MIC range 8–128 mg/L), the MICs of imipenem reduced 2-fold in 18.75% of the clinical isolates, 4-fold in 62.5% of the clinical isolates, 8-fold in 12.5% of the clinical isolates, remained the same in 6.25% of the clinical isolates and reduced 16-fold in M. abscessus pVV16-blaMab compared with imipenem alone. Briefly, the MIC ranges for imipenem alone in our isolates was 2–8 mg/L, which was reduced to 2–4 mg/L with the addition of relebactam (aside from M. abscessus pVV16-blaMab which has an MIC of 8 mg/L), and further reduced to 0.5–4 mg/L with the addition of both relebactam and amoxicillin. The MIC ranges for amoxicillin were generally higher and more variable than those of imipenem, with an MIC range of 4–128 mg/L with the addition of relebactam, with some isolates lacking susceptibility at the highest concentration tested. The addition of both relebactam and imipenem resulted in an MIC range of 4–128 mg/L. This result is replicated using the ZOI method, which clearly demonstrated drug-target interaction between relebactam and BlaMab, as no activity of amoxicillin is seen. Plate images and plate map for the subsequent resistance induced by BlaMab overexpression are shown in Supplementary Figure 1.
Although 93.75% of M. abscessus clinical isolates are resistant to amoxicillin on its own, with MICs of >128 mg/L, the results in Table 1 show that amoxicillin susceptibility in M. abscessus isolates can be seen with the addition of the β-lactam inhibitor, relebactam. It is also well established that imipenem is effective against most strains of M. abscessus, but the results from Table 1 clearly show that the efficacy of imipenem can be enhanced with the addition of relebactam, and further enhanced with the addition of relebactam and amoxicillin.
Biochemical analysis of relebactam inhibition of M. abscessus β-lactamase
In order to validate the inhibitory activity observed phenotypically in Fig. 1, we conducted biochemical analysis of the activity of relebactam on the M. abscessus endogenous β-lactamase, BlaMab. The gene was amplified by PCR, digested, ligated into pET28a and sequenced, before transformation into chemically competent E. coli BL21. BlaMab was expressed, cells harvested and the enzyme purified by IMAC. We subsequently devised a novel TLC-based assay for assessing β-lactamase activity by separating the penicillin V substrate from the penicilloic acid product. This assay enabled us to demonstrate the β-lactamase activity of our purified BlaMab (Fig. 2a), as well as assay the efficacy of inhibitors against the enzyme (Supplementary Fig. 2a,b). We used avibactam as a positive control, as its inhibitory activity against BlaMab has previously been described by Dubée et al.19. Lane 1 contained protein purification buffer only, and lane 2 had the addition of enzyme (0.01 mg/mL). Following chromatography, two lower spots are observed in lanes 1 and 2, indicative of buffer. Penicillin V was added to lane 3 and gave a characteristic spot of high Rf value, demonstrating unhydrolysed penicillin resulted in a spot just below the solvent front. Lane 4 was identical to lane 3 with the addition of BlaMab protein. The hydrolysis of penicillin V to penicilloic acid by BlaMab resulted in a spot with reduced Rf value. The pre-incubation of enzyme with avibactam in lane 5 resulted in a loss of the lower penicilloic acid spot, demonstrating the inhibition of BlaMab activity. In lane 6, relebactam alone did not resolve on the TLC, but its pre-incubation with BlaMab before addition to the penicillin substrate in lane 7 resulted in total inhibition of hydrolysis as observed with avibactam (lane 5). Finally, the inhibition of BlaMab is further validated by the repeat of lane 4 and lane 7 conditions with heat-denatured BlaMab (lanes 8 and 9 respectively). This result confirms the direct inhibition of BlaMab by relebactam (n = 5).
We investigated the kinetic parameters of the Michaelis-Menten kinetics, the apparent Ki (Ki app), the second-order carbamylation rate constant (k2/Ki) and the decarbamylation rate constant (koff) of relebactam inhibition of soluble recombinant BlaMab using a commercially available β-lactamase substrate, nitrocefin, as a reporter substrate. Nitrocefin is selectively hydrolysed by β-lactamases, resulting in an increase in absorbance which can be monitored at 486 nm. By pre-incubating BlaMab enzyme with a dose response of relebactam (from 0–100 μM), before initiation of the absorbance assay with the addition of nitrocefin substrate, we observed partial inhibition at 1 μM (0.348 μg/mL) and a complete loss of activity at 10 μM (3.48 μg/mL), confirming the direct inhibitory activity of relebactam on BlaMab (Fig. 2b).
The initial velocities (vi) of nitrocefin turnover were recorded for a range of nitrocefin concentrations (1–500 µM) over a range of relebactam concentrations (0–2.5 µM). These results were plotted as classical Michaelis-Menten curves (vi vs [S]) (Fig. 2c), resulting in Michaelis constants (Km) between 69.43–16.96 µM, before initial velocities were abolished at 2.5 µM relebactam. However, we were unable to derive a Ki value using this data. In response, we plotted the reciprocal initial velocities against relebactam concentrations as a linear equation and derived Ki app observed from the Y intercept/slope, which was then normalised for the use of nitrocefin (Fig. 2f)26. The Ki app value obtained of 183.6 nM is indicative of the high inhibitory potency of relebactam against BlaMab. The carbamylation rate constant (k2/Ki) of 2.32 ×106 M−1s−1 is similar to the observed rate for avibactam inhibition of BlaMab (4.9 ×105 M−1s−1)20. The rate of decarbamylation (koff) of BlaMab for relebactam was 0.0146 min−1, which is again similar to the rate previously obtained for that of avibactam (0.047 min−1)19. Our kinetics analysis show that, like avibactam, relebactam is a potent, competitive and reversible inhibitor of BlaMab, displaying a reasonably rapid “on” rate and slow “off” rate, with only half the enzyme recovering activity after 40 minutes in the absence of relebactam (Fig. 2e)19.
Our TLC-based β-lactamase assay enabled us to further explore the parameters of the inhibitory activity of relebactam by varying the time of pre-incubation of relebactam with BlaMab (Supplementary Fig. 2a) and the minimum inhibitory concentration required to abrogate catalytic turnover of the penicillin V substrate to the penicilloic acid product (Supplementary Fig. 2b). We found that penicillin V turnover was rapid and that only addition of relebactam at the same time as the substrate demonstrated turnover of penicillin V, suggesting competitive, reversible inhibition of BlaMab by relebactam. The dose response of relebactam demonstrated total inhibition down to 20 µg/mL, and activity of BlaMab was maintained below a relebactam concentration of 2 µg/mL. This suggested a minimal concentration of relebactam required to inhibit BlaMab in the assay is 20 µg/mL, which corresponds to a less than or equal to 100 fold stoichiometric excess of relebactam required to completely inhibit BlaMab (0.5 µM BlaMab to 57.5 µM relebactam corresponding to 20 µg/mL).
In silico analysis of relebactam binding to M. abscessus β-lactamase
In order to further investigate the mechanism of relebactam inhibition of BlaMab, we conducted molecular docking simulations in silico. 6 potential binding sites were identified (Fig. 3c). For pockets 2–6 the ligand was weakly-held and generally exited the pocket after a few tens of nanoseconds. For pocket 1 (corresponding to the main active site) the ligand reoriented itself relative to the docked conformation and thereafter remained relatively stable within the pocket. The binding interactions for the stable pose are shown in Fig. 3(a,b) and include several polar interactions with the sulphonamide moiety and hydrophobic interactions between the relebactam central piperidine ring and tryptophan 106. In addition, after the initial reorientation, the relebactam carbonyl carbon remained in the vicinity of the hydroxyl oxygen of the catalytically-active serine 71 as can be seen after 120 ns in the distance plot given in Fig. 3(d). The average distance in this period was approximately 5.5 Å and there were many close approaches. Furthermore, the corresponding O-C=O angle (Fig. 3d) in the same period was roughly 130° and as such it is reasonable to assume that it would be possible for the serine hydroxyl to attack the relebactam carbonyl and effect a ring-opening. A further molecular dynamics simulation of the enzyme in a periodic box of explicit water was undertaken but this time in the presence of ten copies of unbound relebactam. Over the course of a 200 ns simulation, one ligand found its way into the main active site (pocket 1) after 130 ns and remained stable therein. The other nine copies of the ligand found no place to reside in the enzyme.