Structural and biochemical characterization of the environmental MBLs MYO-1, ECV-1 and SHD-1

Abstract Background MBLs form a large and heterogeneous group of bacterial enzymes conferring resistance to β-lactam antibiotics, including carbapenems. A large environmental reservoir of MBLs has been identified, which can act as a source for transfer into human pathogens. Therefore, structural investigation of environmental and clinically rare MBLs can give new insights into structure–activity relationships to explore the role of catalytic and second shell residues, which are under selective pressure. Objectives To investigate the structure and activity of the environmental subclass B1 MBLs MYO-1, SHD-1 and ECV-1. Methods The respective genes of these MBLs were cloned into vectors and expressed in Escherichia coli. Purified enzymes were characterized with respect to their catalytic efficiency (kcat/Km). The enzymatic activities and MICs were determined for a panel of different β-lactams, including penicillins, cephalosporins and carbapenems. Thermostability was measured and structures were solved using X-ray crystallography (MYO-1 and ECV-1) or generated by homology modelling (SHD-1). Results Expression of the environmental MBLs in E. coli resulted in the characteristic MBL profile, not affecting aztreonam susceptibility and decreasing susceptibility to carbapenems, cephalosporins and penicillins. The purified enzymes showed variable catalytic activity in the order of <5% to ∼70% compared with the clinically widespread NDM-1. The thermostability of ECV-1 and SHD-1 was up to 8°C higher than that of MYO-1 and NDM-1. Using solved structures and molecular modelling, we identified differences in their second shell composition, possibly responsible for their relatively low hydrolytic activity. Conclusions These results show the importance of environmental species acting as reservoirs for MBL-encoding genes.


Introduction
The class B MBLs are enzymes with the ability to hydrolyse virtually all b-lactam antibiotics, including carbapenems. 1 Various MBLs, including NDM, VIM and IMP, are associated with mobile genetic elements and widespread among clinically important Gramnegative pathogens. Phylogenetically, MBLs can be grouped into three subclasses, B1 to B3. 2 While enzymes belonging to subclasses B1 and B3 carry two Zn(II) binding sites (Zn1 and Zn2), B2 MBLs are mono-Zn(II) enzymes. 2,3 In subclass B1, Zn1 is coordinated by three histidine residues (His/Gly116, His118 and His196), while the Zn2 binding site is coordinated by Asp120, Cys221 and His263. 2,[4][5][6][7] In B2 MBLs, the Zn1 binding site displays one altered residue (Asn116, His118 and His196), whereas the Zn2 site is identical to that of the subclass B1 MBLs. 8,9 The subclass B3 MBLs exhibit a variety of different Zn1 binding sites (His/Gln116, His118 and His196) and a distinct Zn2 binding site, which does not contain a cysteine residue (Asp120, His121 and His263). The Zn(II) ions are bridged by a hydroxide ion most likely attacking the b-lactam ring. 4 Recently, 76 novel B1 MBL genes were predicted through largescale screening of genomic and metagenomics data. 6 Some of these enzymes exhibited sequence identities as low as 28% compared with widespread MBLs like NDM-1. 6 Carbapenemase activity was experimentally confirmed for 18 of 21 tested MBLs when expressed in Escherichia coli. 6,10 This shows that there is a vast environmental reservoir of MBL genes that could potentially be horizontally transferred into pathogenic bacteria and further compromise the effect of b-lactam antibiotics. Here, we investigated three of these B1 MBLs, 6 SHD-1, MYO-1 and ECV-1, in comparison with the clinically widespread enzyme NDM-1. ECV-1 originated from Echinicola vietnamensis, which has previously been isolated from sea water. 11 SHD-1 was identified in Shewanella denitrificans, a genus that is known as the possible origin of resistance genes, including genes encoding b-lactamases. 12 MYO-1 was encoded on a tet(X)-harbouring plasmid in Myroides odoratimimus, a widely distributed bacterium in natural environments. [13][14][15][16] The plasmid also encoded a type IV secretion system, which could make it conjugatable. 17 M. odoratimimus is not considered pathogenic under normal circumstances; 18 however, it has been reported to cause opportunistic infections [19][20][21][22] and treatment options are limited since most strains display MDR. 21,[23][24][25][26][27]

Strains and MIC determination
All strains used for MIC determination have been published previously. 6 In short, the candidate B1 MBL genes bla MYO-1 , bla ECV-1 and bla SHD-1 were synthesized and sub-cloned into the pZE21-MSC1 vector (Expressys, Ruelzheim, Germany). Recombinant plasmids were transformed into E. coli C600Z1 (Expressys). 6,28 For MIC determination, single colonies were incubated overnight on Mueller-Hinton II agar (Becton Dickinson, Franklin Lakes, USA) containing 25 mg/L kanamycin and subsequently suspended in 0.85% saline to a cell density with a turbidity equivalent to that of a 0.5 McFarland standard (1.5%10 7 cells/mL). The McFarland solution was uniformly dispersed with a swab onto the agar plates containing 100 ng/ mL anhydrotetracycline (Sigma-Aldrich, St Louis, MO, USA). Gradient diffusion strips (Liofilchem, Roseto degli Abruzzi, Italy) were applied and the MICs were determined after 19 h of incubation at 37 C.

Enzyme expression, purification and molecular mass verification
For enzyme expression, we used synthetic and codon-optimized genes of bla MYO-1 , bla ECV-1 and bla SHD-1 in a pDest17 vector (Thermo Fisher Scientific, Waltham, USA) with a TEV cleavage site placed prior to the bla genes. The genes were based on the bla genes found in M. odoratimimus, 23 S. denitrificans and E. vietnamensis (GenBank accession numbers CP013691.1, NC_007954.1 and NC_019904.1, respectively). The expression vectors were electroporated into E. coli BL21-AI (Invitrogen, Carlsbad, USA). For protein expression, cultures were induced with L-arabinose (0.1%; Sigma-Aldrich) at an OD 600 of 0.5. Expression was performed in Terrific Broth including 100 mg/L ampicillin (Sigma-Aldrich) at 15 C and 225 rpm. TEV cleavage and purification were done as previously described. 29 Due to the TEV cleavage site and expression without the signal peptide, the protein sequences start at position Gln30, Gly18, Val25 and Gly25 for MYO-1, ECV-1, SHD-1 and NDM-1, respectively (additional glycine at the start). NDM-1 was expressed and purified as described previously. 30 For ESI-MS, the buffer was changed to 0.1% formic acid (Merck Millipore, Burlington, USA) in centrifugal molecular cut-off filters (Merck MilliPore, 10000 Da) and concentrated to 0.25 g/L. The protein masses were verified using an Orbitrap Fusion Lumos (Thermo Fisher Scientific). Proteins were injected using an EASY-nano LC (Thermo Fisher Scientific) with a 15 cm C18 EASY-Spray column. Masses were calculated using the BioPharma Finder 3.0 protein deconvolution software (Thermo Fisher Scientific).

Thermostability
Fluorescence-based thermal stability of the enzymes was determined. 31 In short, purified enzymes were diluted to 0.2 mg/mL using 50 mM HEPES buffer pH 7.5 supplemented with 100 lM ZnSO 4 (Sigma-Aldrich) and 250 mM NaCl (VWR, Radnor, USA). For the fluorescence signal, 12.5% SYPRO orange (Sigma-Aldrich) was used. Melting curves were recorded across a temperature gradient (10-75 C). Tests were performed in an MJ Minicycler (Bio-Rad, Hercules, USA) and melting temperatures were calculated by using the Bio-Rad CFX Manager (v. 3.1). All experiments were carried out in a final volume of 25 lL and at least in triplicate. Purified NDM-1 was included as a control.

Crystallization and structure determination
For ECV-1 (5 mg/mL), crystals were grown from reservoirs with 25%-26% PEG3350 (Sigma-Aldrich), 0.1 M BIS-TRIS buffer pH 6 (Sigma-Aldrich) and 0.2 M sodium acetate (Sigma-Aldrich) at 4 C. Crystal-containing drops were diluted with 10 lL of reservoir solution and microcrystals were created. Microcrystals were seeded into drops of 2 lL containing the same composition and 5 mg/mL purified protein. For MYO-1 (5 mg/mL), crystals were grown in 32%-36% PEG4000 (Sigma-Aldrich) and 0.2 M ammonium sulphate at 4 C (drop size 2 lL). Crystals were flash-frozen in liquid nitrogen using 10% ethylene glycol (Sigma-Aldrich) in addition to the reservoir solution. Since crystallization of SHD-1 was not successful, we used SWISS-MODEL and the solved structure of TMB-1 (PDB ID: 5MMD) with sequence identity of 58%, to obtain a homology-modelled structure. 29,32 Diffraction data were collected at ID30A-3, at the European Synchrotron Radiation Facility (ESRF), France, at 100 K, wavelength of 0.961 Å , and the diffraction images were indexed and integrated using XDS. 33 AIMLESS was used for scaling. 34 For scaling, we aimed for high completeness, a CC 1/2 >0.5 in the outer resolution shell and a mean <I> above 1.0 (Table 1). Both structures were solved by molecular replacement using PDB ID: 1ZNB (ECV-1) and 1HLK (MYO-1) as search models and refined using Phenix 1.12. 35 Modelling was done using Coot. 36 Figures were prepared using PyMOL version 1.8 (Schrödinger).

Results
Environmental MBLs decrease susceptibility to b-lactams in E. coli The sequence identity of MYO-1, ECV-1 and SHD-1 was as low as 28% compared with the widespread MBL NDM-1 ( Figure 1). We identified differences in their loop regions L3 (residues 56-66), L8 (residues 151-160) and L10 (residues 220-237), which are involved in Zn(II) binding and defining substrate specificity. 4 In addition, MYO-1 and ECV-1 harboured in total three cysteine residues (positions 69, 121 and 221) within their active site. To explore if the differences in the amino acid sequence could potentially influence the substrate specificity, we performed susceptibility testing of E. coli expressing MYO-1, ECV-1 and SHD-1. The respective genes (not codon-optimized) were sub-cloned into pZE21-MSC1 and expression was induced with anhydrotetracycline in E. coli C600Z1 ( Table 2). NDM-1 was included for comparison. All three enzymes showed the characteristic MBL profile, increasing the MIC of all b-lactams except for aztreonam. MBL activity was also confirmed by inhibition with EDTA. SHD-1 conferred the highest increase in carbapenem MICs, with a 64-, 4-and 1024-fold increase for ertapenem, imipenem and meropenem, respectively (compared with E. coli C600Z1). The observed effect on carbapenem   Fröhlich et al. Figure 1. Multiple sequence alignment based on the MBL numbering system. 72 For calculating the secondary structure elements, we used the published structure of NDM-1 (PDB ID: 3ZR9). 40 Sequence identity compared with NDM-1 was determined for MYO-1 (28%), ECV-1 (33%) and SHD-1 (33%). The alignment shows conserved (filled boxes) and semi-conserved (grey font) residues within the selection. 72 TT and TTT indicate b-turns and a-turns, respectively. All enzymes showed catalytic activity against the tested b-lactams (Table 3). In general, SHD-1 showed the lowest enzymatic activity. Against penicillins and carbapenems, the catalytic activity of SHD-1 was usually 2-to 4-fold lower compared with MYO-1 and ECV-1. The activities of MYO-1 and ECV-1 were generally comparable to each other. The cefepimase and ceftazidimase activity of MYO-1 was 10-fold higher than that of SHD-1. The lower activity of SHD-1 towards cephalosporins was due to both lower affinity (K m >300 lM) and lower turnover (k cat 10 s #1 ). In line with the MIC results, the b-lactamase activities of the environmental MBLs were lower, ranging from <5% to 70%, compared with NDM-1 ( Figure 2). For MYO-1, the catalytic activity tended to be higher and its carbapenemase activity reached up to 70% to that of NDM-1. On the contrary, SHD-1 displayed the weakest comparative carbapenemase and cephalosporinase activity, with values generally below 10%. In addition, ECV-1 demonstrated high catalytic activity towards meropenem (65% compared with NDM-1), whereas imipenem, penicillins and cephalosporins were hydrolysed to a lower degree (10%-40%).

First shell, second shell and substrate binding residues
The structures of MYO-1 and ECV-1 were successfully solved by Xray crystallography to 2.17 and 1.33 Å , respectively (Figure 3a and b and Table 1). For MYO-1 we found two molecules (chains A and B) in the asymmetrical unit with R work and R free of 0.22 and 0.25 (space group P6 5 ). Due to lack of electron density in chain B, the regions of N60 to K66, L93 to I96 and K104 to S105 could not be built. The structure of ECV-1 was refined to an R work and R free of 0.16 and 0.19, respectively, with one molecule in the asymmetrical Table 2. MICs (mg/L) for E. coli C600Z1 expressing bla MYO-1 , bla ECV-1 and bla SHD-1 sub-cloned into the pZE21-MSC1 expression vector; bla NDM-1 was included as a comparator and empty vector was included as a control  Fröhlich et al.

Discussion
Here, we present two new crystal structures and one homology model of MBLs identified in environmental bacteria. 6 Expressed in E. coli, all three enzymes conferred decreased susceptibility to carbapenems, cephalosporins and penicillins. Compared with NDM-1, the expression of MYO-1, ECV-1 and SHD-1 led to lower MICs, especially those of carbapenems (Table 2). We determined the catalytic efficiency using purified enzymes. Generally, the enzymatic activity ranked MYO-1 > ECV-1 > SHD-1. We found the largest differences in catalytic efficiency towards cephalosporins, where MYO-1 exhibited up to 44-fold higher activity against cefepime compared with SHD-1. Interestingly, SHD-1 conferred the highest MIC values when expressed in E. coli, but the lowest catalytic efficiencies (purified enzyme). SHD-1 was identified in a Gammaproteobacterium, while the natural hosts of MYO-1 and ECV-1 belong to the distant phylum of Bacteroidetes 6 and hence may not be expressed efficiently in the periplasm of E. coli. Work on the subclass B1 SPM-1 has shown different drug selectivity when tested in the periplasm, in enzyme kinetic assays and in an MIC setup. 39 In addition, the expression of the same class B and D b-lactamases in different hosts exhibited a lack of correlation between MICs and the catalytic efficiency of these enzymes. 43,44 Hence, phenotypic variation can be due to differences in catalytic efficiency in the periplasmic conditions, but expression level, protein folding and translocation to the periplasm can also play a role. 43 ECV-1 and SHD-1 exhibited thermostabilities 3 and 8 C higher than MYO-1 and NDM-1. Studies have shown that lower thermostability was accompanied by higher flexibility, facilitating cephalosporin hydrolysis in b-lactamases. 45,46 Interestingly, the more thermostable SHD-1 and ECV-1 showed lower catalytic efficiency, especially against oxyimino cephalosporins. However, due to the low sequence identity (28%) further studies have to be conducted exploring the structure-activity relationships and a possible correlation with thermostability.
Since second shell residues have been reported to be under evolutionary pressure and their substitutions have created variants with changed enzymatic activity, 38,47,48 we investigated the structures of MYO-1, ECV-1 and SHD-1. We found the positions 69, 121 and 262 differed from the second shell residues of NDM-1. In NDM-1, mutational studies of Ser69 and Lys121 revealed that a cysteine replacement, as present in MYO-1 and ECV-1, reduced bacterial fitness towards cefotaxime and imipenem, while Ala69 and Arg121 (SHD-1) improved bacterial survival after selection. 49 The amino acid position 121 is semi-conserved as it is directly situated 'below' the Zn2 binding site. While crystallographic occupancy correlated with reduced Zn(II) affinity for MBLs carrying Arg121 (e.g. BcII, VIM-2 and BlaB), 50,51 high occupancy was seen for MBLs carrying serine or cysteine at this position, e.g. IMP-1 and CcrA. [52][53][54] Mutational studies of BcII:R121C showed a marginal increase in occupancy compared with WT BcII. 55 In contrast, C121R in CcrA resulted in a variant with lower Zn(II) affinity. 56 Arg121 interacts with Asp120 in BcII and data suggest that R121C may affect the pK a of Asp120, thus changing the pH-dependent activity of the enzyme. 57 Interestingly, in BcII:R121C a network of water molecules populates the active site as a replacement for a guanidinium group of arginine that usually preserves its shape. 55 G262S differentiates IMP-1 from IMP-6 and has been shown to also enhance catalytic efficiency in both IMP and BcII. 58 Precursor enzymes of IMP-1 have therefore been reported to be less active against, for example, ampicillin, ceftazidime and imipenem. 59,60 In addition, an amino acid substitution of G262S in IMP-1 suggested reduced mobility of His263 by the formation of an H-bond network allowing the accommodation of cephalosporins. 61
carbapenems, penicillins and cephalosporins. 70 The amino acids located at 224 and 233 have been reported to be important in substrate recognition and hydrolysis. 41,42 In conclusion, this work presents the structure and activity of three MBLs from environmental sources. We showed that these enzymes act as carbapenemases exhibiting increased catalytic activity and conferring elevated MICs when expressed in E. coli. The lower activity towards cephalosporins and carbapenems could be, at least partially, explained by their second shell residues. These residues have been previously shown to be under selective pressure in other enzymes, and amino acid substituents may alter Zn(II) binding and extend their substrate specificity. 38,47,48,[56][57][58][59][60] Mobilization and horizontal transfer of genes expressing these or similar enzymes into clinical strains may render those strains less susceptible towards carbapenems and carbapenemase inhibitors acting as Zn(II) chelators. 71 Funding This work was funded by the Swedish Research Council (2013-08633 and 2018-02835).

Transparency declarations
None to declare.