Article pubs.acs.org/ac
Identification of Inhibitors of the Antibiotic-Resistance Target New Delhi Metallo-β-lactamase 1 by both Nanoelectrospray Ionization Mass Spectrometry and Ultrafiltration Liquid Chromatography/Mass Spectrometry Approaches Xin Chen,†,‡,∥ Lixin Li,‡,∥ Shuai Chen,†,‡ Yintong Xu,†,‡ Qiang Xia,†,‡ Yu Guo,‡,§ Xiang Liu,‡ Yanting Tang,‡ Tanjie Zhang,‡ Yue Chen,‡,§ Cheng Yang,*,‡,§ and Wenqing Shui*,†,‡ †
College of Life Sciences and Tianjin State Laboratory of Protein Science and §College of Pharmacy and State Key Laboratory of Medicinal Chemical Biology, Nankai University, Tianjin 300071, China ‡ High-Throughput Molecular Drug Discovery Center, Tianjin Joint Academy of Biotechnology and Medicine, Tianjin 300457, China S Supporting Information *
ABSTRACT: Mass spectrometry-based platforms have gained increasing success in discovery of ligands bound to therapeutic targets as drug candidates. We established both a nanoelectrospray ionization mass spectrometry (nanoESI-MS) assay and an ultrafiltration liquid chromatography/mass spectrometry (LC/MS) assay to identify new ligands for New Delhi metallo-β-lactamase 1 (NDM-1), responsible for worldwide antibiotic resistance. To alleviate nonspecific binding of hydrophobic compounds and eliminate false positives typically encountered in the indirect LC/MS-based assay, we introduced a blocking protein in the control, which remarkably enhances the selectivity and accuracy of the indirect approach. Side-by-side comparison of the two MS-based approaches for the first time further reveals unique advantages of the indirect approach, including better reproducibility and tolerance of interference. Moreover, the success of fishing out a potent ligand from a mixture of small-molecule fragments demonstrates great potential of the indirect LC/MS-based approach for constructing a robust screening platform against combinatorial libraries or natural product extracts. More importantly, by combining the results of MS-based analyses, enzymatic activity assay, competition experiments, and structural simulation, we discovered a new compound as a promising drug candidate targeting NDM-1.
N
discovery. A variety of analytical tools including isothermal titration calorimetry (ITC),11 surface plasmon resonance (SPR),12 and nuclear magnetic resonance (NMR) spectroscopy13 have been employed to discover and characterize ligands specifically bound to target proteins. More recently developed mass spectrometry (MS)-based assays constitute a valuable addition to the arsenal of drug discovery techniques and distinguish themselves by high sensitivity, selectivity, and rapid and simultaneous measurement of multiple ligands.14−17 Two major approaches have been developed by utilizing electrospray ionization mass spectrometry (ESI-MS) as the central technique for examination of protein−ligand interactions. One is the direct ESI-MS assay based on direct detection of the protein−ligand complexes for evaluating the binding stoichiometry and affinity. By quantifying the fractions of free and ligand-bound proteins, it allows for estimation of the
ew Delhi metallo-β-lactamase 1 (NDM-1) is a novel class B1 metallo-β-lactamase (MBL) capable of conferring upon bacteria nearly complete resistance to all β-lactam antibiotics, even including “our last line of defense” carbapenems.1−3 Ever since NDM-1 was first identified in a Swedish patient in 2008,4 it has spread almost all over the world at an alarming rate5 and poses a formidable threat to human health.6 Despite the clinical significance of NDM-1, there has been little advance in the discovery of relevant inhibitors7 except for two recently reported compounds, L and D diastereomers of the mercaptocarboxamide inhibitor captopril. L- and D-captopril were both able to suppress the NDM-1 activity of hydrolyzing β-lactam antibiotics in vitro, and Lcaptopril was further shown to bind to the active site of NDM-1 mainly through the zinc-ion cofactors by crystallography.8−10 Undoubtedly, discovery of new potential drug candidates targeting NDM-1 is of substantial significance for combating diseases caused by antibiotic-resistant bacterial infection. The molecular basis of protein−ligand interactions, particularly the relationship between the structure and binding selectivity and affinity, is of significant value for early-stage drug © 2013 American Chemical Society
Received: June 10, 2013 Accepted: July 17, 2013 Published: July 17, 2013 7957
dx.doi.org/10.1021/ac401732d | Anal. Chem. 2013, 85, 7957−7965
Analytical Chemistry
Article
HPLC-grade or better and were purchased from Merck (Darmstadt, Germany), while centrifugal ultrafiltration filters (Amicon Ultra-0.5 mL 10 kDa) were purchased from Millipore (Bedford, MA). The known inhibitor L-captopril was purchased from Sigma−Aldrich (Buchs, Switzerland). All the new compounds assayed in our study were synthesized in our laboratory; their chemical structures are provided in Table S1 in Supporting Information, and their synthetic methods will be published elsewhere. Nineteen fragments used in the mixture experiment were selected at random from a combinatorial library in our laboratory (their structures are listed in Table S2 in Supporting Information). Protein Expression and Purification. NDM-1 with an Nterminal glutathione S-transferase (GST) tag was expressed in Escherichia coli BL21 (DE3) at 37 °C, and the cells at OD600 = 0.4−0.6 were induced by 0.1 mM isopropyl thio-β-D-galactoside (IPTG) for 16−20 h at 25 °C. Cells were harvested by centrifugation and lysed with sonication in lysis buffer of 50 mM 2-(N-morpholino)ethanesulfonic acid (MES) (pH 6.5), 150 mM NaCl, and 5% glycerol. The supernatant after centrifugation was loaded twice onto a GST column (GE Healthcare). After cleavage of the GST tag from the fusion protein with PreScission protease overnight at 4 °C, NDM-1 was further purified on a Resource Q anion-exchange column (GE Healthcare) and eluted with a solution of 50 mM MES (pH 6.5), 500 mM NaCl, and 5% glycerol. The purified protein was concentrated to 6 mg/mL in a buffer containing 50 mM MES (pH 6.5), 150 mM NaCl, and 5% glycerol for both direct and indirect MS-based assays. The protein concentrations were determined by the Bradford method. Enzymatic Activity Assay. The activity of NDM-1 in the absence or presence of specific compounds was measured by an established spectrophotometric method.38 The final mixture for the assay consisted of 50 mM N-(2-hydroxyethyl)piperazineN′-ethanesulfonic acid (HEPES, pH 7.0), 15 μM ZnSO4·7H2O, 200 μM imipenem, and 20 nM NDM-1. The fluorescence readouts were recorded on Varioskan Flash (Thermo Scientific). Only compounds showing an inhibition rate over 80% were subjected to IC50 measurement. Data analysis for calculation of inhibition rates and IC50 values were performed with Prism 5.0 (Graphpad, San Diego, CA). NanoESI-MS-Based Analysis (Direct Approach). The NDM-1 protein was buffer-exchanged into 10 mM ammonium acetate (pH 6.5), and stock solutions of individual compounds (20 mM in dimethyl sulfoxide, DMSO) were diluted with 10 mM ammonium acetate (pH 6.5). NDM-1 at a final concentration of 25 μM was incubated with individual compounds at a protein/ligand concentration ratio (P/L) of 1:1 or 1:2 at room temperature for 1 h. The concentration of DMSO in all incubations was kept below 0.5%. The protein− ligand mixture was then loaded into an off-line nanospray emitter (Waters) prior to MS analysis. Mass spectra were acquired on a Waters SYNAPT G1 highdefinition mass spectrometer (Milford, MA) equipped with a nano-ESI ion source. The optimal instrument parameters during analysis of free NDM-1 and its complexes were spray voltage 1000−1400 V, cone voltage 30 V, source block temperature 50 °C, and nano gas flow rate 0.25−0.5 L/h. Full-scan mass spectra of the positively charged ions were recorded in the profile mode, covering the mass range 1000− 4500 m/z. All data were processed and transformed into the actual mass scale via the MassLynx software 4.1 from Waters.
dissociation constants (Kd) of complexes that are in good agreement with those determined by many established assays.18,19 Its sensitivity can be further raised to the level of picomolar protein consumption when utilizing a nanoESI-MS technique.14,20,21 A handful of elegant studies have applied this approach to screening libraries of small molecules, peptides, and carbohydrates for discovery of potential therapeutic agents.22,23,21 Despite its unique technical features, a few weaknesses are recognized for the direct ESI-MS assay. For example, the binding assay conditions (e.g., buffer, pH, detergents used) have to be compatible with MS analysis, which is difficult to fulfill in certain cases. In addition, gas-phase dissociation of complexes or nonspecific binding that occurs during the ESI process would generate false negatives24 or false positives,25−27,18,28 respectively. The other, indirect approach is known as affinity liquid chromatography−mass spectrometry (LC/MS) assay. It relies on LC/MS-based detection of the ligands released from the protein−ligand complex to evaluate the binding specificity. The ligand-bound complex can be separated from the unbound compounds via ultrafiltration,16b,22b,29 dialysis,30 affinity purification, size-exclusion chromatography, etc.31−33 One of the major strengths of the indirect approach is that it completely preserves the native protein−ligand interactions in solution.34,35 Additionally, combinatorial libraries or natural products can be screened in a cost-effective and high-throughput manner for bioactive (ligand) identification.22b,16b Although it may not be as quantitative as the direct approach with respect to Kd estimation, ranking a relative order of binding affinities for the target is feasible and informative.14,36 Limitations of this approach are the inability to determine binding stoichiometry and analyze other structural features of the complex, difficulty in distinguishing false negatives (transient or covalent interactions) and false positives (ligands bound to nonfunctional sites of a target), and the requirement that ligands be detectable by MS.37 Interestingly, with their distinctive pros and cons, these two types of MS-based assays have not been performed in the same protein−ligand interaction study for side-by-side comparison. Neither have they been reported for identifying ligands or inhibitors of the NDM-1 target. In this study, we assessed the interaction between NDM-1 and nine novel compounds synthesized in our lab by both nanoESI-MS and ultrafiltration LC/MS assays. The known inhibitor L-captopril was also tested as a positive control for method evaluation. The direct and indirect approaches were compared in terms of accuracy, selectivity, reproducibility, and generalizability. Notably, the background signal of the ultrafiltration LC/MS-based assay was minimized by introducing a blocking reagent to reduce nonspecific binding of compounds to the filter membrane as well as distinguish false-positive detection. Binding affinity as well as inhibition of NDM-1 activity was measured on different ligands. For the most potent inhibitor identified in our study, competition experiments and structural simulation of the complex were performed to reveal the binding site and molecular basis of the protein−ligand interaction. Finally, the inhibitor was effectively captured from a mixture of smallmolecule fragments by the indirect approach.
■
EXPERIMENTAL SECTION Materials. The plasmid of NDM-1 was constructed as previously described.8 Ammonium acetate and NaI were purchased from Sigma−Aldrich. All organic solvents were 7958
dx.doi.org/10.1021/ac401732d | Anal. Chem. 2013, 85, 7957−7965
Analytical Chemistry
Article
Figure 1. Schematic of (upper) direct and (lower) indirect approaches for detection of ligands bound to NDM-1. The nanoESI-MS-based assay analyzes free protein (P) and protein−ligand complexes (P+L) directly to assess binding selectivity and affinity. In the indirect LC/MS-based assay, protein−ligand complexes were first purified via ultrafiltration, and the ligand was dissociated from the complex and identified by LC/MS. Highresolution mass spectrometry (HRMS) facilitated confident identification of ligands.
For Kd estimation in cases where the protein (P) possesses multiple (n) binding sites for the ligand (L), we first used eq 1 to determine an for the particular P−L complexes based on the modified single-point measurement:39 an =
I(PLn) I(P) + I(PL) + I(PL 2) + ... + I(PLn)
centrifugal tube, and the ligands were dissociated from NDM-1 with 90% methanol in deionized water. The released ligands were then separated from the denatured protein by centrifugation at 13000g for 10 min at 20 °C. The supernatant containing the ligands was evaporated by use of a centrifugal evaporator, reconstituted in 80% methanol, and analyzed by LC/MS. The original control was prepared by using the binding buffer substitute for NDM-1 during incubation. In the new control experiments, a blocking protein (GST) at the same concentration of NDM-1 (25 μM) was present in the incubation. All the samples were prepared in duplicate and analyzed separately to obtain experimental replicates. In competition experiments, L-captopril was added to the mixture of NDM-1 with ligand 14, or ligand 14 was added to the mixture of NDM-1 with L-captopril, both at 150 μM during incubation. Additionally, we preincubated NDM-1 with 0.25 or 25 μM ethylenediaminetetraacetic acid (EDTA) before the addition of ligand 14 to determine the effect of metal chelation on complex formation. In the small-molecule mixture experiment, an equimolar mixture of 19 fragments plus ligand 14 (each at 25 μM) was incubated with NDM-1 (50 μM) for 1 h at 25 °C. The rest of the sample preparation was carried out as described above, and a control with GST protein was used. LC/MS Analysis. Aliquots of reconstituted ultrafiltrates were analyzed on a Waters Synapt G1 high-definition mass spectrometer (Milford, MA) equipped with a Waters Acquity UPLC system. Ultra-performance liquid chromatography (UPLC) separation was carried out on a Waters Acquity UPLC BEH C18 column (2.1 mm × 50 mm, 1.7 μm) at a flow rate of 200 μL/min, with the mobile phases water/0.1% formic acid (A) and acetonitrile/0.1% formic acid (B). The LC method was as follows: 0−1 min, B at 2%; 1−2 min, B at 2%− 50%; 2−4 min, B at 50%; 4−4.1 min, B at 50%−2%; and ending with equilibration at 2% B for 4 min. For ligand 14, an
(1)
where I is the MS peak intensity of free protein (P) or the protein complexes with different numbers of bound ligands (PL, PL2, ..., PLn). In our case, the intensities of MS peaks at charge state of 8 were used for calculation. The dissociation constant Kd,m for the protein attached with m (= 1, 2, ..., n) molecules of L can be calculated using eq 2:
where [P]0 and [L]0 are initial concentrations of protein and ligand, and am (1 ≤ m ≤ n) is calculated from eq 1 with m replacing n. Similar methods for binding-affinity estimation based on nanoESI-MS analysis are described in detail elsewhere.21 Sample Preparation for Ultrafiltration LC/MS-Based Analysis (Indirect Approach). A mixture of binding buffer consisting of 50 mM MES (pH 6.5), 150 mM NaCl, and 5% glycerol, the compound at a final concentration of 50 μM (or 25 μM in certain cases), and NDM-1 protein at 25 μM was prepared and incubated for 1 h at 25 °C. Each mixture was then filtered through a 10K molecular weight cutoff ultrafiltration membrane by centrifugation at 13000g for 20 min at 4 °C. The complexes of NDM-1 and specific ligands were washed three times with 10 mM ammonium acetate (pH 6.5) followed by centrifugation to remove the unbound compounds. The resulting solution of the complex was transferred to a new 7959
dx.doi.org/10.1021/ac401732d | Anal. Chem. 2013, 85, 7957−7965
Analytical Chemistry
Article
Figure 2. Representative nanoESI mass spectra acquired for solutions of (a) free NDM-1 at 25 μM, (b−e) NDM-1 incubated with specific compounds at two different P/L ratios, and (f) NDM-1 incubated with two compounds of severe ion suppression. Shown are the protein peaks with the charge state of 8. In panels b−e, the mixing P/L ratio (1:1 or 1:2) for preparing the complexes is indicated in each spectrum. NDM-1 is normally coordinated with two zinc ions. The number of ligands bound to dizinc NDM-1 is denoted above each peak of the complex. (*) NDM-1 coordinated with different numbers of zinc ions (1−4); (▲) NDM-1 with DMSO adduct ions; (△, panels b and c) Peaks for NDM-1 + 3Zn2+ in complex with one corresponding ligand and for NDM-1 + 4Zn2+ with two ligands.
8-min gradient was needed from 2% to 40% B. The regular ESI ion source operated in either positive- or negative-ion mode, depending on the ionization tendency of different compounds. Mass spectra were acquired within a mass range from 100 to 800 m/z, with capillary voltage 3.0 kV in positive-ion mode or
2.6 kV in negative-ion mode, sample cone voltage 54 V (positive mode) or 45 V (negative mode), extraction cone voltage 4.7 V (positive mode) or 4.0 V (negative mode), desolvation temperature 350 °C, source temperature 100 °C, and desolvation gas flow 500 L/h. External calibration with a 7960
dx.doi.org/10.1021/ac401732d | Anal. Chem. 2013, 85, 7957−7965
Analytical Chemistry
Article
Table 1. Binding Stoichiometry and Affinity of Different Compounds to NDM-1 Evaluated by Direct and Indirect MS-Based Approaches direct (1:2)a compd 14 13 L-cap 40 21 22 17 10 6 4
b
binding no. 1 1−6 1 NDh 2 1, 3 NDh 1−4 1−3 1−4
direct (1:1)a
Kd, μM c
2.0 NDg 4.6 ND 71.8 20.6 ND 6.6 4.8 1.3
binding no. 1 2 1 ND NBi NBi ND NBi 1 1
indirecta
Kd, μM 0.6 4.7 1.6 ND NB NB ND NB 6.7 16.8
d
accurate m/z 292.062 531.086 218.086 192.052 174.095 199.052 119.018 119.017 431.134 296.094
+
[M + Na] [2M + Na]+ [M + H]+ [M − H]− [M − H]− [M − H]− [M − H]− [M − H]− [M +H]+ [M + Na]+
S/Ne (1:2)
S/Ne (1:1)
IC50 f
10.5, 11.3 7.2, 8.1 5.5, 6.4 5.3, 4.6 2.4, 3.1 1.6, 1.9 1.5, 1.5 1.3, 1.6 1.7, 2.0 1.4, 1.1
8.2, 9.0
1.81 ± 0.13 μM 62.26 ± 7.81 μM 7.31 ± 0.93 μM 13.41 ± 1.56 μM 20.65 ± 1.24 μM 21.61 ± 2.31 μM 81.31 ± 10.06 μM 56.38 ± 10.93 μM 68.06% 3.51%
1.5, 1.3 1.3, 1.5 1.2, 1.9
Mixing ratios of protein/ligand (P/L) are indicated for two types of assays with a constant protein concentration (25 μM) during incubation. b Number of ligands bound to one NDM-1 molecule. cDissociation constants estimated by direct nanoESI-MS analysis. dAccurate m/z of compounds detected in a specific ionized form. eLC/MS signals of ligands detected from the active protein complex vs background signals of the control by use of a blocking protein. Experiments with P/L = 1:1 were performed only for compounds 14, 22, 10, and 4, and their results are shown. Values shown in italic type represent specific binding of the ligands with S/N above a 2-fold cutoff. fMean ± SD from three replicates of the enzymatic activity assay; the inhibition rates of compounds 6 and 4 (shown in the table) were lower than 80% and they were not subject to IC50 measurement. gCannot be determined with the complicated spectrum of multiple binding species. hUnable to assign MS peaks due to ion suppression. iNo binding detected. a
concentration of 25 μM. Representative mass spectra recorded for free protein and for protein in complex with specific ligands are shown in Figure 2. The crystal structure of wild-type NDM1 reveals its coordination with two zinc ions in the catalytic site.10,8 Our nanoESI-MS analysis detected the dizinc form of the free protein as its most abundant species, together with the mono- and trizinc forms at much lower abundances (Figure 2a), indicating its active conformation was preserved during the direct MS analysis. All the incubations were prepared in the presence of 0.5% DMSO; therefore, it is not surprising to observe protein peaks with DMSO adducts (Figure 2). Binding stoichiometry and affinity measured on different compounds by the direct approach are listed in Table 1. Notably, changing the P/L mixing ratio could dramatically affect estimation of stoichiometry and affinity for most compounds tested. Similar MS spectra and consistent Kd estimation for samples prepared at two P/L ratios were acquired only for L-captopril and a tightly bound ligand 14 (Figure 2b,c and Table 1). Compounds 13, 6, and 4 were in complex with the protein in multiple bound species at one P/L ratio but were bound to the protein with a constant stoichiometry at the other ratio (Table 1). Binding of 1−6 molecules of compound 13 to one NDM-1 molecule at the P/L ratio of 1:2 even hindered Kd calculation due to the highly complicated MS spectrum (Figure 2d, left). The protein complexes with 21, 22, and 10 were detected only at the higher P/L ratio of 1:2 (Table 1). For example, 10 was bound to the protein with stoichiometries ranging from 1 to 4 at the 1:2 P/L ratio (Figure 2e, left), and the estimated Kd for one ligand binding was 6.6 μM (Table 1). In contrast, it was found to be a nonbinder at the 1:1 P/L ratio (Figure 2e, right). We speculate that the contradictory results for the same complexes by nanoESI-MS analysis are attributed to different extent of nonspecific gas-phase interactions between the protein and ligands when they are mixed at different ratios during the ESI process. Significant nonspecific gas-phase interactions would lead to detection of false-positive binding and underestimation of Kd.18,26,27,25,28 To minimize the nonspecific interactions, a series of P/L ratios need to be tested so as to find the optimal
solution of sodium formate achieved mass accuracy within 10 ppm. The mixture of small-molecule fragments and ligand 14 was separated with a longer LC gradient: 0−2 min, B at 2%; 2− 5 min, B at 2−30%; 5−10 min, B at 30−45%; 10−15 min, B at 45−65%; and 15−17 min, B at 65−85%. ESI-MS operated in positive-ion mode, with the other parameters set as described above. LC/MS chromatograms for specific compounds were extracted by use of MassLynx software (v4.1, Waters) based on accurate mass measurement of the compound with a tolerance of 0.02 Da. For display of multiple extracted ion chromatograms, the raw LC/MS data were exported into Origin 75 (Original Lab) for chromatogram reconstruction.
■
RESULTS AND DISCUSSION Both the direct and indirect MS-based approaches were employed in our study for identification of NDM-1 ligands from a panel of rationally designed novel compounds based on the structure of L -captopril (Table S1 in Supporting Information). Our experimental workflow is outlined schematically in Figure 1. In the nanoESI-MS-based assay, incubations of NDM-1 and individual compounds were exchanged to an MS-compatible buffer (10 mM ammonium acetate, pH 6.5) prior to direct analysis of the protein−ligand complex by nanoESI-MS. MS signals of the free protein (P) and protein− ligand complexes (P+L) were used to assess binding selectivity and affinity. The same incubations of protein and ligands were also subjected to the ultrafiltration LC/MS-based assay. Protein−ligand complexes were first isolated from excess unbound compounds by ultrafiltration, and then the ligands were dissociated from the complexes before identification by LC/MS. High-resolution mass spectrometry (HRMS) with high mass accuracy (
