High resolution crystal structure of the catalytic domain of MCR-1

The newly identified mobile colistin resistant gene (mcr-1) rapidly spread among different bacterial strains and confers colistin resistance to its host, which has become a global concern. Based on sequence alignment, MCR-1 should be a phosphoethanolamine transferase, members of the YhjW/YjdB/YijP superfamily and catalyze the addition of phosphoethanolamine to lipid A, which needs to be validated experimentally. Here we report the first high-resolution crystal structure of the C-terminal catalytic domain of MCR-1 (MCR-1C) in its native state. The active pocket of native MCR-1C depicts unphosphorylated nucleophilic residue Thr285 in coordination with two Zinc ions and water molecules. A flexible adjacent active site loop (aa: Lys348-365) pose an open conformation compared to its structural homologues, suggesting of an open substrate entry channel. Taken together, this structure sets ground for further study of substrate binding and MCR-1 catalytic mechanism in development of potential therapeutic agents.

Scientific RepoRts | 6:39540 | DOI: 10.1038/srep39540 C-terminal catalytical domain of MCR-1 (aa: 219-541, MCR-1C for short) determined at 1.45 Å resolution. Our results illustrated the detailed structure of MCR-1C which will help in the understanding of the catalysis mechanism of MCR-1 and potential drug design against colistin resistance.

Results
Overall structure of MCR-1. MCR-1 contains two domains, one N-terminal transmembrane domain and one periplasmic C-terminal catalytic domain 33 . Several attempts were made to express full-length MCR-1 that resulted in insoluble protein whereas two truncations (aa: S181-R541 and T219-R541) were successfully expressed and crystallized. Crystals of the longer and shorter truncations diffracted to 2.2 Å and 1.45 Å, respectively. Thus, the shorter truncation was used for structure determination whereas the longer one was considered for ICP-MS analysis. The protein was crystallized in space group P2 1 2 1 2 1 having one MCR-1C molecule in the asymmetric unit with a solvent content of 42.5%. The structure was solved by single-wavelength anomalous scattering (SAD) method using the anomalous signal of zinc ions with well-defined electron density map for all the residues in the construct.
The MCR-1C adopted an α /β /α topology with seven β -strands, sandwiched between two layers of α -helices (Fig. 1a). Based on structural superimposition, MCR-1C is similar to its homologues LptA and EptC (Fig. 1b,c) with a r.m.s.d of 1.58 Å and 1.56 Å (all C-alpha), respectively. We observed that the central β sheets topology is well-superimposed to LptA and EptC while loops adjacent to the active site and the C-terminal fragment are variable (Fig. 1b,c). The conformational variations of the loops adjacent to the active site may infer to the substrate specificity of these enzymes. During review of this manuscript, the other crystal structure of MCR-1 (the C-terminal domain of MCR-1 from residue D218 to residue R541, cMCR-1) was released with PDB code 5K4P 42 . Superposition of current structure to cMCR-1 (5K4P) gives an r.m.s.d value of only 0.49 Å for 323 C-alpha atoms, indicating that these two structures are almost identical. The most different region between these two structures is a loop located at the tip of the structure (aa: 416-422).
MCR-1C consists of three disulfide bridges as revealed in the crystal structure (Fig. 1a, shown as ball-and-stick models). In contrast, LptA and EptC retain five and three, respectively. Moreover, only two disulfide bridges, Cys281-Cys291 and Cys414-Cys422 retain their sequence and structural conservation among all the three structures. The Cys414-Cys422 bridge stabilized a loop (aa: Lys409-Glu423) located at the top of the protein whereas Cys281-Cys291 completes the catalytic site by locking a smaller α -helix (aa: Thr285-Met292) to the central β sheet. It is worth mentioning that residue Thr285 is a catalytic residue for all phosphoenthanolamine transferases 33,34,42 . The formation of disulfide bond Cys281-Cys291 might restrain helical flexibility to accelerate the catalysis reaction. The third disulfide bridge Cys356-Cys364 is only shared between MCR-1C and LptA and it seems to arrest the conformational freedom of the loop (aa: Lys348-365) (Fig. 1b) and facilitate substrate entry. In contrast, there is no such a loop in this region of EptC (Fig. 1b), which may allow the entry of many different substrates. This is consistent with the multiple reactions catalyzed by EptC 34 . Putative active site of MCR-1C. To study the putative active site, we over-expressed MCR-1C in E. coli BL21(DE3) plysS and the metal ions bound to the protein were characterized by ICP-MS method. In regular LB medium, the protein can bind Fe, Zn, Mg and Mn ions but prefers Zn as compared to other metal ions ( Table 2). In a 50 μ M ZnCl 2 supplemented LB medium, MCR-1C can bind up to four zinc ions ( Table 2) where three of them can be located in the structure as determined by SHELXD 43 based on the anomalous signal. The fourth zinc ion might bind the protein non-specifically due to which we cannot identify it in the structure. This is also revealed in the newly published cMCR-1 (5K4P) 42 where 10 zinc ions were identified in the structure with most of them locate on the surface coordinating to waters with low occupancy.
In our study among the three identified zinc ions in MCR-1C, two of them were located at the active site ( Fig. 1a) and the third one located near residue Cys291. Similar to LptA and EptC, the first zinc ion (Zn1) is highly conserved in active site contacts and is coordinated to residues Glu246, Thr285, Asp465 and His466 33,34,44 (Fig. 2). Beside the conserved orientations of these coordination residues among MCR-1C, LptA and EptC, we found that unlike LptA, EptC and even cMCR-1 42 , Thr285 (a putative nucleophilic attacking residue in catalysis) is un-phosphorylated. The phosphorylated Thr285 is considered a catalysis intermediate during phosphate transfer 33,34,42,45 thus stating MCR-1C the first native structure without any bound substrate or reaction intermediate. In contrast, Thr285 in cMCR-1 is phosphorylated and the coordination of Zn1 is very similar to that of LptA.
Like the Zinc-soaked LptA (PDB: 4KAY), the second Zinc ion (Zn2) coordinates to two conserved histidine residues (His395 and His478 in MCR-1C) among some phospho-form transferases. However, the MCR-1C crystals uptake Zn ions from the medium for cell culture rather than soaking, suggesting a natural co-factor which is strong enough not to be washed off during protein purification. Due to un-phosphorylated nature, except the two histidine residues, Zn2 ion coordinates to water molecules and carboxylate moiety of Glu300 from a neighboring MCR-1C (Fig. 2a). The Glu300 interaction might not be observed in solution since MCR-1C is a monomer as determined by multi-angle light scattering method (Fig. S1). Moreover, this coordination could be fulfilled by the incoming substrate and possibly be a part of the catalytic mechanism. Crystals of the newly released cMCR-1 were obtained in a solution containing 0.2 M zinc acetate and Zn2 in this structure coordinates to the phosphate group and only one histidine residue His395, but not His478 33,42 , which is quite different from both LptA and our MCR-1C structure and needs further validation.

Discussion
The newly identified mcr-1 is a mobile colistin resistant gene and has been shown to rapidly spread among various bacterial strains through horizontal transfer and assisted its host in colistin resistance 14 . Since the broad-spectrum β -lactam antibiotics resistant gene, such as blaNDM-1, has spread all over the world, colistin is regarded as an important resort antibiotic against these superbugs 12 . Several studies have reported the presence of blaNDM-9 and mcr-1 in chicken meat samples 27 including carbapenem and colistin double resistant E. coli strains 19,29,30 . Furthermore, mcr-1-harbouring plasmids have been reported in multi-drug resistant E. coli 46,47 , arising more threat to our healthcare system 1 .
Based on the published reports, it is imperative to study the molecular features and catalysis mechanism of MCR-1. This is the first study that reports the high-resolution crystal structure of the C-terminal catalytical domain of MCR-1 in native state. The overall structure of MCR-1C closely resembles the other two phosphoethanolamine transferases LptA and EptC. The intriguing difference among them is found to be at the loop (aa: Lys348-365) adjacent to the active site and might regulate substrate entry and specificity (Fig. 1b). Interestingly, the absence of this loop in EptC is anticipated to allow entry of different substrates (Fig. 1b). Our notion is supported by the fact that EptC retains the ability of catalyzing multiple reactions 34 .
At the putative active site, the presence of several conserved residues in phosphorylated transferases indicates that MCR-1C might also be a phosphoethanolamine transferase which needs to be confirmed by further biochemical and functional studies. The phosphorylated nature of active site residue Thr285 is considered an intermediate form in phospho-transfer reaction 33,34,42,45 , therefore we believe the unphosphorylated active site of MCR-1C to be in a native state before substrate binding. The high resolution structure of MCR-1C sets up the primary basis for further study of substrate binding and catalysis mechanism. Future studies are required to search and validate MCR-1 inhibitors and block its enzymatic activities. The crystallographic and biophysical studies reported here lay the groundwork for developing new therapeutic strategies and chemicals against colistin resistant superbugs. The crystals were harvested with a nylon loop and immediately cooled in liquid nitrogen. Excitation scan was performed at beamline BL17U1 at Shanghai Synchrotron Radiation Facility (SSRF) at the Zinc absorption edge to confirm the identity of bound metal ions. The diffraction data were collected and processed using HKL2000 49 . Solvent content and molecule number of MCR-1 per asymmetric unit were calculated using Matthews 50 .

Materials and Methods
The structure was solved using the SAS protocol of Auto-Rickshaw: the EMBL-Hamburg automated crystal structure determination platform 51 . The input diffraction data were prepared and converted for use in Auto-Rickshaw using programs of the CCP4 suite 52 . FA values were calculated using the program SHELXC 53 . Based on an initial analysis of the data, the maximum resolution for substructure determination and initial phase calculation was set to 1.7 Å. All of the 3 heavy atoms requested were found using the program SHELXD 43 . The correct hand for the substructure was determined using the programs ABS 54 and SHELXE 55 . Initial phases were calculated after density modification using the program SHELXE. 95.36% of the model was built using the program ARP/wARP 56,57 . The model was refined with PHENIX 58 and Refmac 59 in CCP4 suite and cycled with rebuilding in COOT 60 . Solvents were added automatically in Coot and then manually inspected and modified. The final model was analyzed with MolProbity 61 , showing that 323 over 331 residues were in the favored region of Ramachandran plot and only one residue in outlier region (Ser330). The electron density map at residue S330 was very clear and there should be no doubt about its conformation. Data collection and model refinement statistics were summarized in Table 1. The coordinates and structure factors were deposited in PDB with entry code 5GRR. Structural figures were drawn with PyMOL 62 .

Metal ion identification. Metal ions bound by MCR-1 were identified through ICP-MS assays using
Orbitrap Fusion Tribrid mass spectrometer (MS, Thermo Scientific) with the LCQ Fleet MS detector (Thermo Scientific) in the core facility of our university. Two different samples of MCR-1 were assayed and the contents of bound metal ions were compared, i.e. MCR-1 expressed in regular LB, MCR-1 expressed in LB supplemented with 50 μ M ZnCl 2. Three independent assays were performed for each sample and Table 2 listed the ion contents for each sample.