Embelin (purity > 98%), meropenem, biapenem, cefepime ampicillin, cefradine, clavulanic acid, sulbactam, and tazobactam were purchased from Sigma Co. (United States). ATCC BAA-2146 was purchased from the American Type Culture Collection.

NDM-1 was expressed and purified as previously described (Li et al., 2013). The purity of NDM-1 assessed by SDS–PAGE and was > 95%. The purified NDM-1 was used to determine the kinetic parameter, Km as described previously and was in agreement with previously published values (Li et al., 2013). Inhibition of enzyme activity was determined by comparing the initial hydrolysis rates of meropenem between NDM-1 and a mixture of NDM-1 with different compounds. Hundreds of natural product extracts and chemicals, including embelin, curcumine and resveratrol were screened for inhibitory activity. Reactions were performed at 30°C in a total volume of 200 μL 4-(2-hydroxyethyl)-1-piperazineethanesulfonic acid (HEPES) (50 mM, pH 7.5) buffer containing 50 μM ZnSO₄. The initial hydrolysis rates of meropenem were measured using a SynergyHT spectrophotometer (Biotek Corp., United States) at a wavelength of 297 nm (Li et al., 2013). The same assay was performed with VIM-1 and IMP-1.

Prior to testing with 0.5 mM meropenem, the NDM-1 (0.5 μg/ml) enzyme was incubated with various concentrations of embelin in 10 mM HEPES (pH 7.5) buffers at 30°C for 10 min. The IC₅₀ was defined as the concentration of inhibitor needed to reduce the initial rate of hydrolysis of meropenem by 50% (Li et al., 2013), and the K_i (inhibitory constant) was determined using GraphPad Prism software. Experiments were performed independently at least three times. The same methods were used to measure the IC₅₀ and K_i of VIM-1 and IMP-1.

In vitro susceptibility tests were performed using the Mueller–Hinton broth microdilution technique as recommended by the (Clinical and Laboratory Standards Institute [CLSI], 2012) Briefly, the selected antibiotics were diluted from 4096 to 0.125 μg/ml by the broth two-fold dilution method. Different concentrations of embelin (from 1 to 64 μg/ml) were added in combination with β-lactam antibiotics. The inoculum (final size, 5 × 10⁵ CFU/ml) was prepared by the direct colony suspension method.

Fractional inhibitory concentration (FIC) values were determined using standard methods (Pillai et al., 2005). Checkerboards were set up with 8 concentrations of each meropenem and embelin in serial one-half dilutions. The minimal inhibitory concentration (MIC) for each compound was determined by the broth microdilution technique as recommended by (Clinical and Laboratory Standards Institute [CLSI], 2012). The FIC for each compound was calculated as the compound in the presence of co-compound concentration for a well showing no growth, divided by the MIC for that compound (King et al., 2014). The FIC index was taken as the sum of the two FICs. Experiments were performed three times and the means were used for calculation.

Various concentrations of embelin (from 2 to 16 μg/ml) were tested in combination with 2 μg/ml of meropenem, which is the EUCAST breakpoint for resistance (King et al., 2014). The synergistic properties of the two compounds were tested in microtiter plates. Escherichia coli ATCC 25922 was used as the control in all plates. Overall, 46 NDM-1-haboring clinical isolates (Enterobacteriaceae and Acinetobacter spp.), including 17 E. coli, 15 K. pneumoniae and 14 A. baumannii, were tested. After an 18-h incubation at 37°C, the plates were read. All strains were used to examine reproducibility and no deviation was shown.

The flexible molecular docking method AutoDock (Goodsell and Olson, 1990) was used to analyze the intermolecular interaction between the NDM-1 protein and the small-molecule ligand, embelin. In the protocol, the high-resolution crystal structure of NDM-1 complexed with its potent competitive inhibitor L-captopril was used as a template to perform the docking (Gonzalez et al., 2012; King et al., 2012); the structure was retrieved from the PDB database (Berman et al., 2000) under the accession code 4EXS. The binding site for the docking calculations was defined as those residues that could directly contact the co-crystallized L-captopril. The binding site contained 11 amino acid residues (Met67, Val73, Trp93, His120, His122, Asp124, His189, Cys208, Gly219, Asn220 and His250) and two zinc ions. To prepare the site for docking, a rectangular grid was defined that covered the binding-site residues and zinc ions. All water molecules, as well as the cocrystallized L-captopril, were removed from the crystal structure system, while the coordinated Zn²⁺ was kept in NDM-1’s active site due to its important role in stabilizing the NDM-1–substrate/ligand interaction. The docking exploration of embelin’s binding conformations was performed in the rectangular gird. Subsequently, Gasteiger and Kollman united atomic charges were assigned for the embelin ligand and NDM-1 protein, respectively. The protein was also added with polar hydrogen atom. A grid was set to accommodate the active-site region with a 0.375 Å span. The torsion and rotatable bonds in the ligands were defined, and the non-polar hydrogens and partial atomic charges were added to the bonded carbon atoms (Wang et al., 2017). The docking was carried out using the AutoDock Vina program (Trott and Olson, 2010) to evaluate ligand binding energies over the conformational search space using the Lamarckian genetic algorithm.

No datasets were generated or analyzed in the current study.

An enzyme activity-based screening, using our in-house collection of natural product extracts and chemicals was initiated in order to find a new NDM-1 inhibitor. Three compounds (20 μM), embelin (Figure 1A), curcumine (Supplementary Figure S1A) and resveratrol (Supplementary Figure S1B), showed more than 20% inhibition. Among these three compounds, embelin inhibited NDM-1 by more than 50%. Dose-dependent analyses further revealed that embelin inhibited NDM-1 at an IC₅₀ value of 2.1 ± 0.2 μM and K_i value of 0.19 ± 0.02 μM when meropenem was the substrate (Figures 1B,C). The IC₅₀ value of embelin with imipenem as a substrate was 2.3 ± 0.4 μM, similar with that with meropenem (Supplementary Figure S2). The inhibitor activity of embelin against other metallo-beta-lactamases, such as VIM and IMP was also determined. The IC₅₀ was approximately 200 μM against VIM-1 and 100 μM against IMP-1 with 0.5 mM meropenem as a substrate. These results suggest that embelin is a potent inhibitor of NDM-1, but a relatively weak inhibitor of VIM-1 and IMP-1.

Molecular docking was done to fully search the huge conformation space of the embelin molecule within NDM-1’s active site. An empirical scoring function was then employed to evaluate the relative binding potency of generated ligand conformations to protein receptors. Hundreds of potential binding conformations of the embelin ligand to the NDM-1 active site were predicted using AutoDock, and the top-10 embelin binding conformations (with the highest theoretical scores) were superposed in the NDM-1 active site (Figure 2A).

The 10 best conformations exhibited very similar binding modes toward the NDM-1 protein, with only slight fluctuations at their side groups and moieties. Interestingly, the fatty hydrocarbon chain of the embelin molecule also highly consistent in different docking modes, although this chain is highly flexible and thus may have significant swing when binding to a protein receptor. The average structures of the 10 best conformations were then clustered into one representative, which is shown in Figure 2B. As can be seen, the embelin molecule adopts its quinonyl moiety to interact with the catalytic pocket of the enzyme, while its fatty hydrocarbon chain points out of the pocket. In addition, the embelin is located near the di-Zn²⁺ cation, thus forming a strong coordination bond.

In order to elucidate the molecular mechanism involved in embelin binding to NDM-1, the non-bonded interactions at the complex interface of the NDM-1–embelin complex were determined using the LIGPLOT program (Wallace et al., 1995) and are shown in Figure 2C. Most regions of the embelin molecule are buried within NDM-1’s active site, suggesting that a wide van der Waals contact exists between NDM-1 and embelin, which should confer substantial stabilization for the complex architecture. In addition, although only limited hydrophobic forces are observed at the interface, there are two geometrically satisfactory hydrogen bonds that can be formed with residues Cys208 and Asp124; these are thought to constitute recognition specificity for the NDM-1–embelin binding. As might be expected, embelin can coordinate potently with the one of the di-Zn²⁺ through a hydroxyl group; this coordination bond was also found in the complex systems of NDM-1 with natural substrates and small-molecule inhibitors, suggesting the importance of the coordination bond in NDM-1–ligand recognition and interaction.

ATCC BAA-2146, a K. pneumoniae strain harboring the bla_NDM-1 gene was tested for susceptibility to different β-lactam antibiotics. The NDM-1 gene was verified using PCR. The selected antibiotics were diluted via serial twofold dilutions ranging from 4096 to 0.125 μg/ml. The strain was found to be highly resistant to the penicillins, cephalosporins and carbapenems tested. To detect inhibition activity, embelin, clavulanic acid, sulbactam or tazobactam was added to β-lactam antibiotics at different concentrations ranging from 1 to 64 μg/ml (Figures 3A–C and Table 1). The MIC values for the β-lactam antibiotics were not lowered by clavulanic acid, sulbactam or tazobactam (at 64 μg/ml) in the K. pneumoniae strain producing the NDM-1 enzyme. In contrast, embelin efficiently rescued the antibiotic activity of various β-lactam antibiotics. In vitro MICs against the NDM-1-producing strain (ATCC BAA-2146) were reduced 2- to 512-fold with embelin (Figures 3A–C and Table 1).

The antibacterial activity of embelin was analyzed to distinguish between NDM-1 inhibition and antibacterial activity. Embelin (128 μg/ml) alone did not inhibit the growth of ATCC BAA-2146. The MICs of various β-lactam antibiotics were very different from 4096 to 0.0625 μg/ml, when combined with an embelin concentration of 64 μg/ml. Ampicillin is a penam compound and has a high MIC (>4096 μg/ml) against ATCC BAA-2146. Only when the concentration of embelin reached 64 μg/ml, was the MIC of ampicillin reduced to 4096 μg/ml. When combined with embelin (64 μg/ml), the activity of cephem antibotics (cefradine, ceftazidime and cefepime) improved, but was not restored to susceptible levels. This may reflect a high efficiency of NDM-1 against amoxicillin and cephem antibotics, as the NDM-1 enzyme has been shown to exhibit high catalytic efficiencies for amoxicillin and cephem compound substrates (Thomas et al., 2011; Li et al., 2013). Compared with the other classes of antibotics, the MICs of carbapenems (meropenem, biapenem) were significantly decreased in the presence of embelin. When treated with 32 μg/ml embelin, the MICs of ATCC BAA-2146 against meropenem fell by up to 512-fold, 128-fold for biapenem and about 256-fold for imipenem. In the current study, the addition of embelin could reduce MIC values to below susceptibility breakpoints. Thus, the mixture of meropenem (or biapenem, or imipenem) with embelin could be as a candidate combination used to treat NDM-1-producing strains.

Systematic titration of concentrations of embelin and meropenem against NDM-1 positive clinical bacterial strains showed that embelin restored meropenem activity consistent with NDM inhibition. FIC index values were determined to be 0.05 – 0.15 for a panel of 6 clinical NDM-1 positive isolates (2 E. coli, 2 K. pneumoniae and 2 A. baumannii) tested against a combination of meropenem and embelin (FIC values of ≤ 0.5 are defined as synergistic (Pillai et al., 2005) (Figure 3D). Different concentrations of embelin (from 2 to 16 μg/ml) combined with meropenem were further investigated using 46 NDM-1 positive clinical isolates (17 E. coli, 15 K. pneumoniae and 14 A. baumannii, Figure 3E). Embelin at 2, 4, 8, and 16 μg/ml restored meropenem sensitivity (2 mg/ml) in 28.3, 47.8, 78.3, and 100% of NDM positive isolates, respectively. An additional experiment was performed to evaluate the inhibitory activity of embelin alone, using the 46 isolates. Embelin (64 μg/ml) alone did not significantly inhibit activity against any of the strains of E. coli, K. pneumoniae, or A. baumannii tested. These results demonstrate that embelin could restore carbapenem efficacy against NDM-1-positive pathogens.