Crystal Structure of the Catalytic Domain of MCR-1 (cMCR-1) in Complex with d-Xylose

Abstract

The polymyxin colistin is known as a “last resort” antibacterial drug toward pandrug-resistant enterobacteria. The recently discovered plasmid-encoded mcr-1 gene spreads rapidly across pathogenic strains and confers resistance to colistin, which has emerged as a global threat. The mcr-1 gene encodes a phosphoethanolamine transferase (MCR-1) that catalyzes the transference of phosphoethanolamine to lipid A moiety of lipopolysaccharide, resulting in resistance to colistin. Development of effective MCR-1 inhibitors is crucial for combating MCR-1-mediated colistin resistance. In this study, MCR-1 catalytic domain (namely cMCR-1) was expressed and co-crystallized together with d-xylose. X-ray crystallographic study at a resolution of 1.8 Å found that cMCR-1-d-xylose co-crystals fell under space group P2₁2₁2₁, with unit-cell parameters a = 51.6 Å, b = 73.1 Å, c = 82.2 Å, α = 90°, β = 90°, γ = 90°. The asymmetric unit contained a single cMCR-1 molecule complexed with d-xylose and had a solvent content of 29.13%. The structural model of cMCR-1-d-xylose complex showed that a d-xylose molecule bound in the putative lipid A-binding pocket of cMCR-1, which might provide a clue for MCR-1 inhibitor development.

1. Introduction

Antimicrobial resistance among Gram-negative bacteria, especially the multidrug-resistant enterobacteria which are the leading cause of human clinical infections, is a global healthcare concern [1]. The carbapenemase-producing carbapenem-resistant Enterobacteriaceae (CRE), such as Klebsiella pneumoniae strains expressing the KPC-2 enzyme and Enterobacteriaceae strains expressing the NDM-1 enzyme, are of special clinical importance [1].

Polymyxin is often employed as the final therapeutic option to treat CRE-caused clinical infections because of its low resistance and high efficiency among CRE [2]. Polymyxins (colistin, polymyxin B) are cationic polypeptides which could bind the lipid A moiety of bacterial lipopolysaccharide and disrupt the bacterial cytomembrane subsequently [2]. Bacterial polymyxin resistance was considered to be very low and primarily caused by genomic mutations associated with specific two-component regulatory systems, which either modify lipid A or lead to complete loss of the lipopolysaccharide [2].

Recently, a novel mobile colistin resistance mechanism, led by a protein named MCR-1 (a phosphoethanolamine (PEA) transferase that confers colistin resistance by catalyzing the transference of phosphoethanolamine to lipid A moiety of lipopolysaccharide), has been discovered [3]. The gene encoding MCR-1 (mcr-1) has been shown plasmid-located and self-transmittable between various bacterial strains [2]. Until now, mcr-1 has already been detected within a broad range of pathogenic isolates from humans and animals worldwide, which poses a huge threat to the sustaining effectiveness of colistin against CRE-caused clinical infections [2]. Development of effective MCR-1 inhibitors might be the only way to extend the usage of colistin as a reserved antibacterial drug to treat CRE infections [4].

Although several structures of MCR-1 catalytic domain (namely cMCR-1) have been determined [5,6,7,8], few effective inhibitors for MCR-1 are known. A recent co-crystallization study [9] showed that two substrate analogues of MCR-1, ethanolamine and d-glucose, could specifically bind to cMCR-1. Here, the crystallization and primary structure analysis of cMCR-1 complexed with d-xylose is reported. The structure determined showed that a d-xylose molecule bound in the putative lipid A-binding pocket of cMCR-1, which might provide a clue for MCR-1 inhibitor development.

2. Materials and Methods

2.1. Recombinant cMCR-1 Production

The sequence of mcr-1 gene is available in GenBank (GenBank accession no. KY685070). Based on the secondary structure predictions, the MCR-1 catalytic domain (namely cMCR-1) includes 326 amino acids, from Pro216 to Arg541. The partial mcr-1 gene sequence encoding cMCR-1 with NcoI/XhoI restriction sites incorporated at the 5′/3′ ends was commercially synthesized and cloned into NcoI/XhoI restriction sites of the expression vector pET-28a(+) (Novagen), creating pET-28a(+)-mcr-1. In construct pET-28a(+)-mcr-1, a histidine tag (HHHHHH) was fused to the C-terminus of cMCR-1 (

Table 1. Production specifics for cMCR-1.

Source	Escherichia coli
DNA	Synthesized DNA
Forward primer 1	5′-CATGCCATGGCCAAAAGATACCATTTATCAC-3′
Reverse primer 2	5′-CCCTCGAGGCGGATGAATGCGGTGCGGTC-3′
Expression vector	pET-28a(+)
Host	E. coli BL21(DE3)pLysS
Recombinant protein sequence 3	MGPKDTIYHAKDAVQATKPDMRKPRLVVFVVGETARADHVSFNGYERDTFPQLAKIDGVTNFSNVTSCGTSTAYSVPCMFSYLGADEYDVDTAKYQENVLDTLDRLGVSILWRDNNSDSKGVMDKLPKAQFADYKSATNNAICNTNPYNECRDVGMLVGLDDFVAANNGKDMLIMLHQMGNHGPAYFKRYDEKFAKFTPVCEGNELAKCEHQSLINAYDNALLATDDFIAQSIQWLQTHSNAYDVSMLYVSDHGESLGENGVYLHGMPNAFAPKEQRSVPAFFWTDKQTGITPMATDTVLTHDAITPTLLKLFDVTADKVKDRTAFIRLEHHHHHH

¹ The NcoI site noted. ² The XhoI site noted. ³ The cloning artifacts are underlined.

).

Escherichia coli BL21(DE3)pLysS was transformed with pET-28a(+)-mcr-1 and grown at 310 K, 200 rpm rotation in LB liquid medium containing 50 μg mL⁻¹ kanamycin for cMCR-1 expression. Confluent cultures (OD₆₀₀~0.6) were then treated with 0.3 mM (final concentration) IPTG at 298 K with shaking (180 rpm) for 20 h. Cells were collected by 20 min of centrifugation (4500 g, 277 K) and pellets were kept at 193 K for subsequent use.

The cell pellets were lysed with 10 mM Tris-HCl pH 8.0, 200 mM NaCl, 5% (v/v) glycerol, 0.3% (v/v) Triton X-100, 1 mM DTT, and 0.1 mM PMSF in a French press. Cell wastes were excluded by centrifuging the lysates at 12,000 g for 30 min at 277 K, and the supernatant was clarified using a 0.45 μm filter and then passed through a pre-equilibrated Ni-NTA affinity column (GE Healthcare). The affinity column was washed thoroughly using 10 mM Tris-HCl pH 8.0 containing 50 mM imidazole to remove the miscellaneous proteins. The target proteins were eluted using 10 mM Tris-HCl pH 8.0 containing 200 mM imidazole. Concentrated protein was then loaded onto a MonoQ 5/50 GL anion exchange column (GE Healthcare) and chromatographed at 1 mL min⁻¹ using a linear NaCl gradient generated with 10 mM Tris-HCl pH 8.0 (buffer A) and 10 mM Tris-HCl pH 8.0, 1 M NaCl (buffer B). Peak fractions were pooled and run through a Superdex 200 10/300 GL column (GE Healthcare) equilibrated with buffer A at a flow rate of 1 mL min⁻¹. Peak fractions were recovered and concentrated to 10 mg mL⁻¹ for crystallization. The purity of the final protein (cMCR-1) was checked by SDS-PAGE. All specifics for recombinant cMCR-1 production are present in Table 1.

2.2. Crystallization

The cMCR-1 was crystallized at 277 K using the sitting-drop vapor-diffusion method as described by Wei et al. [9]. The 0.5 μL sitting drops consisting of 0.25 μL cMCR-1 solution and 0.25 μL reservoir solution were equilibrated against 30 μL reservoir solution in 96-well MRC plates (Molecular Dimensions). The best crystals were achieved in 10% (w/v) PEG 1000, 5% (w/v) PEG 8000.

2.3. Data Collection, Structure Solution, and Refinement

The cMCR-1 crystals were incubated in mother liquor containing 100 mM d-xylose for 10 s to form the cMCR-1-d-xylose complex. The cMCR-1-d-xylose co-crystals used for diffraction data collection were incubated in cryoprotectant (mother liquor containing 20% (v/v) glycerol) for 10 s before flash cooling in streams of liquid nitrogen. Data for cMCR-1-d-xylose complex were acquired at 100 K using an ADSC Q315r detector at beamline BL17U1 of Shanghai Synchrotron Radiation Facility (SSRF), China; 360 frames were taken with 1.0° oscillations. The data were indexed, integrated, and scaled using HKL-2000 (HKL Research, Inc., Charlottesville, VA, USA) [10] and iMosflm programs [11]. The structure was solved by molecular replacement with Phaser [12] using a single monomer of cMCR-1 (PDB entry 5GRR [7]) as the search model. The structure model was constructed using alternating manual building in Coot [13] and restrained refinement in PHENIX [14]. The final model was optimized on PDB_REDO web server [15] and validated by MolProbity [16]. All figures were prepared by PyMOL (Schrödinger).

Table 2. Data-collection and crystallographic statistics of cMCR-1-d-xylose complex.

Diffraction Source	BL17U1, SSRF
Wavelength (Å)	0.9792
Temperature (K)	100
Detector	ADSC Q315r
Crystal-to-detector distance (mm)	350
Total rotation range (°)	360
Rotation range per image (°)	1.0
Exposure time per image (s)	0.5
Space group	P212121
a, b, c (Å)	51.6, 73.1, 82.2
α, β, γ (°)	90, 90, 90
Resolution range (Å)	43.71–1.82 (1.88–1.82) 1
Total number of reflections	56014 (5514)
Number of unique reflections	28258 (2789)
Mosaicity (°)	0.5
Multiplicity	2.0 (2.0)
Completeness (%)	99.0 (99.0)
Mean I/σ(I)	10.61 (4.94)
Rmerge (%)	3.5 (11.1)
CC1/2	0.997 (0.949)
Wilson plot overall B factor (Å2)	13.74
Reflection number, working set	28223 (2785)
Reflection number, test set	1997 (196)
Rwork	0.139
Rfree	0.178
Ramachandran favored region (%)	98
Ramachandran allowed region (%)	1.75
Ramachandran outliers (%)	0.25
Rotamer outliers (%)	0.65
R.m.s.d. bond lengths (Å)	0.006
R.m.s.d. bond angles (°)	0.86
Average B factor (Å2)	17.55

¹ Outer shell values.

summarizes data-collection and crystallographic statistics of cMCR-1-d-xylose complex. Coordinates and structure factors of the cMCR-1-d-xylose complex have been deposited in the Protein Data Bank (PDB) under accession code 5ZJV.

3. Results and Discussion

As stated in the Introduction, development of effective MCR-1 inhibitors is crucial for combating the threat of colistin resistance mediated by MCR-1. A recent co-crystallization study [9] showed that two substrate analogues of MCR-1 (ethanolamine and d-glucose) could specifically bind to MCR-1 catalytic domain (cMCR-1). Both d-glucose and lipid A are hexacyclic compounds. Thus, we tried many other hexacyclic compounds for co-crystallization with cMCR-1 (unpublished). The only other co-crystal structure was that obtained for the complex formed between cMCR-1 and d-xylose at a resolution of 1.8 Å.

MCR-1 belongs to the phosphoethanolamine (PEA) transferase family. It contains 541 amino acids with an N-terminal five-helix transmembrane domain (amino acid residues 1–215) and a C-terminal periplasmic catalytic domain (amino acid residues 216–541) [9]. In order to investigate the potential interactions between d-xylose (and other hexacyclic compounds) and MCR-1, cMCR-1 (MCR-1 catalytic domain) was expressed and purified using a combine of affinity, anion exchange and gel filtration chromatography (Figure 1a), as stated in Section 2.1. The purity of the purified cMCR-1 was confirmed with SDS-PAGE (Figure 1b) and subsequent Western blot analysis (Figure 1c). The cMCR-1 was effectively crystallized using the sitting-drop vapor-diffusion method as described by Wei et al. [9], which generated diffraction-quality crystals with a longest dimension of 0.2 mm (Figure 2). SDS-PAGE showed that the obtained crystals were protein crystals and had the same molecular weight as the purified cMCR-1 protein (Figure 1b). The cMCR-1 crystals were incubated in the mother liquor supplemented with 100 mM d-xylose for 10 s to form the cMCR-1-d-xylose complex.

Diffraction data for the cMCR-1-d-xylose complex was collected to 1.8 Å resolution (Figure 3) and on its basis, the cMCR-1-d-xylose co-crystals fell under space group P2₁2₁2₁, possessing unit-cell parameters a = 51.6 Å, b = 73.1 Å, c = 82.2 Å, α = 90°, β = 90°, γ = 90°. The asymmetric unit contained a single cMCR-1 molecule complexed with d-xylose. The data set of X-ray diffraction had a resolution range from 43.71 Å to 1.82 Å with 3.5% R_merge and 99.0% completeness. To elucidate the structure of cMCR-1-d-xylose complex, we employed molecular replacement method using cMCR-1 monomer (PDB entry 5GRR [7]) as a search model and obtained a clear solution. We confirmed the occurrence of a single protein molecule in the asymmetric unit by cross-rotation and translation-function calculations; the corresponding solvent content was 29.13%. Initial structure refinement using PHENIX [14] yielded a model (Figure 4a) with an R_work of 13.9% and an R_free of 17.8%.

As shown in the Fo-Fc map (contoured at 3.0 σ level) (Figure 4b) and 2Fo-Fc map (contoured at 1.0 σ level) (Figure 4c), a d-xylose molecule bound in the putative lipid A-binding pocket of cMCR-1. According to the PDB structure validation report, real space correlation coefficient (RSCC) and real space r-value (RSR) are 0.94 and 0.13, respectively, for the ligand d-xylose. The d-xylose, Pro481, and Tyr287 formed a sandwich structure with the d-xylose molecule in the middle (Figure 4d). Obviously, the hydrophobic stacking interaction played a crucial role in d-xylose recognition.

The structure of cMCR-1-d-xylose complex is similar with that of cMCR-1-d-glucose complex (PDB entry 5YLF [9]). Meanwhile, there are still many differences between the two structures. First, alignment of the two structures (Figure 4f) showed that both d-xylose and d-glucose bound to the same pocket of cMCR-1 and formed a π-π-conjugated interaction with Pro481 and Tyr287 of cMCR-1, but the skeletons of d-xylose and d-glucose were in the opposite positions. Next, we analyzed the cMCR-1-d-xylose and cMCR-1-d-glucose interactions by using PISA server (http://www.ebi.ac.uk/pdbe/pisa). The accessible surface area (ASA) = 269.67 Ų, buried surface area (BSA) = 194.27 Ų, solvation energy effect (ΔⁱG) = −2.37 kcal/mol between d-xylose and cMCR-1, and the ASA = 303.17 Ų, BSA = 204.82 Ų, ΔⁱG = −2.97 kcal/mol between d-glucose and cMCR-1. It suggested that the interaction between d-glucose and cMCR-1 is greater than that between d-xylose and cMCR-1. The d-xylose and d-glucose molecules also bound to cMCR-1 through a large number of hydrogen bonds. The O1, O2, O3, and O4 atoms of d-xylose hydrogen-bonded to Ser284 OG, Thr283 OG1, N/Ser284 N, OG, Thr283 OG1, N, and Asn482 N, respectively (Figure 4d). The O1, O2, O3, and O4 atoms of d-glucose hydrogen-bonded to Asn482 N, Thr283 N, Ser284 N, OG/Thr283 OG1, and Ser284 OG, respectively (Figure 4e).

We also conducted a comparison of cMCR-1-d-xylose complex with phosphoethanolamine transferase A (EptA) from Neisseria meningitides (PDB entry 5FGN [17]), the only structure of a full-length phosphoethanolamine transferase so far. The structure of cMCR-1-d-xylose complex can be well superimposed with the structure of EptA catalytic domain, with a Cα root-mean-square deviation of 2.0 Å as revealed by Dali server (http://ekhidna.biocenter.helsinki.fi/dali_server/start) [18] (Figure 4g). Anandan et al. [17] have shown that detergent dodecyl-β-d-maltoside (DDM) could bind in a substrate pocket of EptA, and the pocket bound by DDM was probably the phosphoethanolamine (PEA) binding pocket near the putative lipid A-binding pocket.

In conclusion, our finding that d-xylose could bind in the putative lipid A-binding pocket of cMCR-1 is interesting, which might provide a clue for MCR-1 inhibitor development. In vitro inhibitory assay is currently in progress to confirm if d-xylose could inhibit colistin resistance mediated by MCR-1.