Characterization of MATE-Type Multidrug Efflux Pumps from Klebsiella pneumoniae MGH78578

We previously described the cloning of genes related to drug resistance from Klebsiella pneumoniae MGH78578. Of these, we identified a putative gene encoding a MATE-type multidrug efflux pump, and named it ketM. Escherichia coli KAM32 possessing ketM on a plasmid showed increased minimum inhibitory concentrations for norfloxacin, ciprofloxacin, cefotaxime, acriflavine, Hoechst 33342, and 4',6-diamidino-2-phenyl indole (DAPI). The active efflux of DAPI was observed in E. coli KAM32 possessing ketM on a plasmid. The expression of mRNA for ketM was observed in K. pneumoniae cells, and we subsequently disrupted ketM in K. pneumoniae ATCC10031. However, no significant changes were observed in drug resistance levels between the parental strain ATCC10031 and ketM disruptant, SKYM. Therefore, we concluded that KetM was a multidrug efflux pump, that did not significantly contribute to intrinsic resistance to antimicrobial chemicals in K. pneumoniae. MATE-type transporters are considered to be secondary transporters; therefore, we investigated the coupling cations of KetM. DAPI efflux by KetM was observed when lactate was added to produce a proton motive force, indicating that KetM effluxed substrates using a proton motive force. However, the weak efflux of DAPI by KetM was also noted when NaCl was added to the assay mixture without lactate. This result suggests that KetM may utilize proton and sodium motive forces.


Introduction
Bacterial drug efflux pumps have generally been classified into five families: the resistance nodulation cell division (RND) family, major facilitator superfamily (MFS), small multidrug resistance (SMR) family, ATP-binding cassette (ABC) family, and multidrug and toxic compound

MIC determination
The MICs of various antimicrobial agents were determined as described previously [29]. The same experiment was repeated at least five times and the most reproducible values were shown in Tables 3-6.  [30]. Cells were resuspended in this buffer containing 4',6-diamidino-2-phenyl indole (DAPI, final concentration: 5 μM). Cells were incubated at 37°C for 2.5 hr in the presence of 5 mM 2,4-dinitrophenol (DNP) in order to load DAPI.

DAPI efflux assay
The efflux of DAPI was performed as previously described [10]. Energy-starved and DAPIloaded cells were washed three times with 0.1 M 3-morpholinopropanesulfonic acid (MOPS)tetramethylammonium hydroxide (TMAH) (pH 7.0) containing 2 mM MgSO 4 and 5 μM DAPI, and were then resuspended in the same buffer to make the cell suspension OD 650 0.4. Fluorometric measurements were performed at 37°C with a Hitachi F-2000 fluorescence spectrometer. Excitation and emission wavelengths were 332 nm and 462 nm, respectively. After being preincubated at 37°C for 4 min, lactate-TMAH (pH 7.0) (final concentration 20 mM) and salt were added to the assay mixture.
RNA preparation and reverse transcriptase-polymerase chain reaction (RT-PCR) analysis K. pneumoniae cells were harvested during the exponential phase of growth. Total cellular RNA was isolated from these cells using the Qiagen RNeasy Mini Kit (Qiagen Inc., USA). The cell lysate was treated with a QIAshredder column before loading onto the RNeasy column. An RNase-Free DNase (Qiagen Inc., USA) treatment was performed according to the manufacturer's protocol. Extracted total RNA was applied to RT-PCR with the QIAGEN One- Step RT-PCR Kit (Qiagen Inc., USA). The primers used for RT-PCR are listed in Table 2. PCR without the reverse-transcriptase reaction was performed to confirm the absence of detectable DNA contamination in the extracted RNA solution. RT-PCR was repeated five times with independently prepared template RNA and a set of reproducible result was shown in the result.

Construction of K. pneumoniae SKYM
We previously described the method to disrupt a gene on the genome in K. pneumoniae [29,31]. DNA fragments were amplified with the following primers: ketM 1, ketM 2, ketM 3, ketM  (Table 2). pKD4 was used as a template for amplification with the primers ketM FRT Fw and ketM FRT Re. Chromosomal DNA from K. pneumoniae ATCC10031 was used as a template for amplification with the other sets of primers. A PCR reaction was then performed with the three kinds of PCR fragments to produce a fusion fragment.

Computer-based analysis
Alignment and phylogenetic analyses of MATE-type proteins were performed with Clustal W (version 2.1 in DDBJ). Information on the proteins used for phylogenetic analyses were listed in S1 Table. Results

Gene cloning
We previously reported the cloning of genes that elevated the resistance levels of host cells against antimicrobial compounds, and these genes were classified into seventeen groups based on the vector plasmids used for gene cloning, digestion pattern of restriction enzymes, and conferred drug resistance [28]. The partially determined DNA sequence of the insert of the hybrid plasmids categorized as groups 3, 4, and 14 were compared with the genome database for K. pneumoniae MGH78578 (http://www.ncbi.nlm.nih.gov/genome/815?project_id=57619), and we found that all cloned DNA fragments in groups 3, 4, and 14 possessed a common gene (ACCESSION YP_001335662.1, KPN_02001) encoding a predicted efflux pump classified into MATE family. We chose pDSH8 and pNTV3 for subsequent experiments because these plasmids possessed the shortest insert of the hybrid plasmids in groups 3, 4, and 14. The vector of pDSH8 was pBluescript II SK(-) while that of pNTV3 was pSTV28. The insert lengths of pDSH8 and pNTV3 were 2.5 kbp and 2.8 kbp, respectively, and these inserts did not include intact ORFs, except for the predicted MATE-type efflux pump gene. We designated the possible MATE-type efflux pump gene as ketM. ketM encoded a protein similar to NorM from V. parahaemolyticus (ACCESSION AB010463.1). A comparison with the primary structure of KetM showed 85% identity and 96% similarity to YdhE (ACCESSION AAB47941.1) from E. coli, 56% identity and 86% similarity to NorM from V. parahaemolyticus, 54% identity and 86% similarity to VcmA (ACCESSION Q9KRU4.2) from V. cholerae, and 44% identity and 81% similarity to HmrM (ACCESSION P45272.1) from Haemophilus influenzae using the software, Genetyx Version 7.0.8. These proteins, which were homologous to KetM, were typical multidrug efflux pumps in the MATE family, and KetM was also thought to be a multidrug efflux pump in this family of cluster 1. Therefore, KetM was presumed to be categorized into cluster 1 in the MATE family.

Drug susceptibility testing
The MICs of various antibiotics were measured in E. coli KAM32 transformed with the plasmid carrying ketM ( Table 3). The MIC of norfloxacin in E. coli KAM32/pDSH8 was eight-fold higher than that of KAM32/pBluescript II SK(-). The MIC of norfloxacin in E. coli KAM32/ pNTV3 was four-fold higher than that of KAM32/pSTV28. In both strains carrying ketM, the MIC of DAPI was 64-fold higher than that of the control cells and the MICs of Hoechst 33342 and acriflavine increased two-fold. Therefore, ketM increased the resistance of the host cells to several antimicrobial chemicals with different structures. We speculated that these differences in MICs to several chemicals between KAM32/pDSH8 and KAM32/pNTV3 may be attributed to differences in the copy numbers of the plasmids in a host cell.

DAPI Efflux Assay
KetM was thought to be a drug efflux pump in the MATE family. Several pumps belonging to cluster 1 in the MATE family were previously reported to be driven by the electrochemical potential of sodium [3,[7][8][9]. Meanwhile, AbeM from A. baumannii, which is also a drug efflux pump belonging to cluster 1 in the MATE family, appeared to be driven by the electrochemical potential of protons [11]. Therefore, we measured the efflux activity of DAPI in the presence or absence of sodium. The ionic radius of Li + is similar to that of Na + and can be regarded as an analogue of Na + . K + is a congener of Na + . K + also plays an important role in the formation of ΔC in the proton motive force; therefore, we investigated the effects of LiCl and KCl on DAPI efflux in addition to NaCl. Rb + has been a utilized in a crystal structure analysis for MATE-type transporters [32]. However, we assumed that Na + was unsuitable for a crystal structure analysis and used Rb + instead. Therefore, we did not investigate the effects of RbCl on KetM.
Energy-starved KAM32/pDSH8 showed stronger efflux activity than that of the control cells when the energy source was added (Fig. 1). This result indicated that DAPI was effluxed in an energy-dependent manner and the electrochemical potential of protons played a role in this efflux activity. The addition of NaCl stimulated the efflux of DAPI by KetM (Fig. 1). This was also observed with the addition of KCl or LiCl (Figs 1 and 2). The addition of KCl in particular strongly activated the efflux of DAPI.
We then investigated the dependency of the efflux of DAPI on salt concentrations (Fig. 2). DAPI efflux in E. coli KAM32/pDSH8 was accelerated by increases in NaCl concentrations, and reached a plateau at 50 mM NaCl. The similar result was obtained with KCl or LiCl.
These results indicated that KetM was an energy-dependent DAPI efflux pump. A proton motive force has been shown to drive the efflux activity of DAPI by KetM and KetM may efflux not only DAPI, but also other chemicals whose MICs were increased in ketM-carrying cells.
It was difficult to detect H + /norfloxacin antiport activity with our assay system, which used reverted membrane vesicles. An eight-fold increase in the MIC of norfloxacin was too weak to detect antiport activity. The results of MICs identified DAPI the best substrate to characterize the coupling cation of KetM. However, the fluorescence of DAPI itself disturbed the detection of fluorescence by the indicator for H + movement.  Chloride did not appear to be important because a substitution with the sulfate ion instead of chloride did not change a DAPI efflux activity by KetM (Fig. 3).

Measurement of MIC in the presence of NaCl or KCl
The efflux activity of DAPI by ketM was accelerated in the presence of Na + or K + (Fig. 1). We then investigated whether the MICs of norfloxacin, ciprofloxacin, DAPI, and kanamycin were increased in E. coli KAM32 carrying ketM in the presence or absence of 50 mM NaCl or 50 mM KCl. No significant changes were observed in the MICs of norfloxacin and ciprofloxacin in E. coli strains with the addition of the salts (S2 Table). The MICs of DAPI and kanamycin slightly increased (two-fold) with the addition of these salts in E. coli KAM32 carrying ketM (S2 Table). However, this slight increase in the MIC was also observed in E. coli KAM32 as the control (S2 Table), which suggested that the increase in the MIC in the presence of NaCl or KCl was not caused by KetM. Therefore, the acceleration of DAPI efflux activity by Na + or K + may not have been sufficiently effective to elevate the MICs of antibiotics.

Contribution of KetM to intrinsic drug resistance
The expression of mRNA was investigated in the three K. pneumoniae strains. K. pneumoniae MGH78578 is a highly drug-resistant strain that exhibits resistance to multiple antibiotics and antimicrobial chemicals. K. pneumoniae ATCC10031 possesses a nonsense mutation in acrB and its drug resistant levels were found to be markedly weaker than those of K. pneumoniae MGH78578 [33]. The measurement of MICs revealed that K. pneumoniae NCTC9632 had the intermediate drug-resistant levels of K. pneumoniae MGH78578 and ATCC10031 ( Table 4). The expression of ketM mRNA was detected in all three strains (Fig. 4A); however, the expression levels of KetM did not parallel drug-resistant levels ( Table 4).
We disrupted ketM on the genome of K. pneumoniae ATCC10031 because the expression of its mRNA was detected. We designated this ketM-disrupted strain K. pneumoniae SKYM and tested its drug-resistant levels (Table 4). However, no significant changes were observed in any of the MICs tested between K. pneumoniae SKYM and the parental strain, ATCC10031. Therefore, the contribution to intrinsic drug resistance by KetM may be negligible in K. pneumoniae, at least under the conditions we used to measure MICs.

Putative MATE efflux pumps in K. pneumonaie MGH78578
We identified three genes-dinF (KPN_04432), yeeO (KPN_02447), and ketM-that met the criteria of the MATE family by searching the K. pneumoniae genome database with comparisons to KetM and DinF from E. coli. We cloned these genes including ketM by PCR downstream of the lac promoter, and measured the MICs of several antibiotics in E. coli KAM32 and K. pneumoniae SKYM transformed with the cloned gene (Tables 5 and 6). The start codon of ketM was originally GTG while those of the other two genes were ATG, and we constructed two kinds of plasmids carrying ketM with GTG or ATG as the start codon.
E. coli carrying ketM cloned by PCR also elevated resistance against DAPI even if the start codon was GTG or ATG (Table 5). pDEM2a carrying ketM with ATG as the start codon exhibited slightly stronger resistance to several chemicals than pDEM2. Functionally cloned ketM increased MICs more than PCR-cloned ketM. This result did not change when MICs were measured in the presence of IPTG.
The MICs of ciprofloxacin and norfloxacin were slightly increased by pNTV3 and pDEM2a in K. pneumoniae SKYM (Table 6). DAPI was considered to be a good substrate for KetM in E. coli KAM32, but the increase in the MIC for DAPI was not observed in K. pneumoniae SKYM. The MIC of DAPI may have originally been too high in K. pneumoniae SKYM to detect any MIC changes. Any notable change in this MIC was not observed when E. coli KAM32 and K. pneumoniae SKYM were transformed with pDEM3 or pDEM4. However, mRNA expression for yeeO and dinF were detected in K. pneumoniae under laboratory growth conditions (Fig. 4). This was similar to that observed for spr1756 and spr1877, which we previously described as MATE-type proteins in S. pneumoniae R6 [10]. The mRNA expression of these genes was detected in Streptococcus pneumoniae R6, whereas their artificial expression in Bacillus subtilis as a host cell did not cause MIC changes. Tocci et al reported that spr1756 was related to moxifloxacin resistance in S. pneumoniae R6, which was very weak [34]. We speculated that the main roles of yeeO and dinF also differed from the expulsion of antimicrobial chemicals.

Discussion
In the present study, we characterized three MATE-type transporters from K. pneumoniae. One of the transporters, KetM elevated resistance to several antimicrobials and showed DAPI efflux activity. Therefore, it was considered to be a MATE-type multidrug efflux pump. However, cells transformed with dinF and yeeO deduced to encode the MATE-type transporter from K. pneumoniae MGH78578 did not exhibit increased drug resistance. Therefore, DinF and YeeO may have transported chemical compounds that were not investigated in the present study.
Resistance to both norfloxacin and DAPI was increased in E. coli KAM32 transformed with ketM. NorM from V. parahaemolyticus was very similar to KetM. However, more substrate specificity and higher increases in drug-resistant levels were observed when norM from V. parahaemolyticus was introduced into E. coli KAM32 [2]. YdhE from E. coli was also markedly similar to KetM but did not cause marked increases in drug resistance that were observed with norM from V. parahaemolyticus [2,35]. This slight increase in drug resistance by YdhE was not attributed to the problems associated with promoter recognition or translation efficiency because the host cell was the same E. coli as the cloned ydhE. K. pneumoniae and E. coli, belonging to Enterobacteriaceae, may not strongly depend on MATE-type transporters for self-defense against antimicrobials. This prospect appeared to be supported by K. pneumoniae SKYM and E. coli KAM32 not showing any changes to drug sensitivity [7].
A comparison of the primary structure of KetM identified that this protein belonged to cluster 1 of MATE-type multidrug efflux pumps. NorM from V. parahaemolyticus [2,3], which was the first MATE-type transporter discovered, has been categorized into cluster 1 [1]. MATE-type multidrug efflux pumps are thought to utilize Na + or H + as a coupling cation and more MATE-type transporters coupling with Na + have been reported than transporters coupling with H + in cluster 1. The transport of DAPI by KetM was facilitated by the addition of NaCl. Previous studies showed that the activity of transporters coupling with Na + were accelerated in the presence of NaCl [36,37]; therefore, we expected to detect the movement of Na + by KetM with the addition of norfloxacin. We, however, could not detect it (S1 Fig). KetM was able to efflux DAPI without Na + (Figs 1, 2, and 3), and Na + did not appear to be indispensable for this efflux activity. This characteristic differed from that of NorM from V. parahaemolyticus [3].
To complicate matters further, DAPI transport activity was enhanced by the addition of Na + , K + , or Li + in the presence of an energy source (lactate) (Fig. 2). We assumed that the acceleration with Na + and that with the other two kinds of cations should be considered separately because the addition of only NaCl without lactate facilitated the efflux of DAPI by KetM. On one hand, the addition of only KCl or LiCl without lactate could not transport DAPI (Fig. 5). These results suggested that a proton motive force may be indispensable for facilitating substrate transport for K + and Li + , whereas Na + itself could transport the substrate of KetM. NorM from V. cholerae is considered to be a MATE-type transporter that utilizes the electrochemical potential of Na + . However, van Veen et al recently reported that NorM from V. cholerae also utilized a proton motive force to transport ethidium [38]. The efflux of ethidium by NorM from V. cholerae was facilitated in the presence of Na + and this transporter could only efflux ethidium with the addition of glucose. The phenomenon observed with KetM appeared to be similar to that with NorM from V. cholerae. Therefore, we suggested that the utilization of cations (H + and Na + ) by KetM may resemble that by NorM from V. cholerae.
We compared the primary structure of KetM from K. pneumoniae to those of NorM from V. parahaemolyticus and V. cholerae, and searched for common amino acid residues between NorM from V. cholerae and KetM from K. pneumoniae only. Nineteen amino acid residues were detected in this search, of which five amino acid residues were predicted to be structurally different from those of NorM from V. parahaemolyticus. Their positions were Glu225, Leu235, Gln339, Leu342, and Ala456 in KetM (Fig. 6). According to the 3D structure of NorM from V. cholerae (PDB: 3MKU), it was presumed that Glu225 was located in a loop between transmembrane domain (TM) 6 and TM7 in KetM. Leu235 was presumed to be in the end of TM7. Gln339 and Leu342 were presumed to be located in the end, close to the periplasm, in TM9. Ala456 was located in the C-terminal on the cytosol side. The loop between TM6 and TM7 was located on cytosol side and appeared to wrap around the outside of TM3 and TM4 (PDB: 3MKU), predicted to be the lateral TMs in NorM from V. cholerae.
KetM was unsuitable for investigating whether these amino acid residues were related to the utilization of a proton motive force for substrate transport for several reasons: DAPI, the best  Amino acid residues that were common between NorM from V. cholerae (NorMVc) and KetM from K. pneumoniae (KetMKp) only, and deduced to be structurally different from NorM from V. parahaemolyticus (NorMVp) were marked in blue (NorMVc and KetMKp). As a reference, the corresponding amino acid residues in NorM from V. parahaemolyticus were marked in red. Asp367 in NorM from V. parahaemolyticus, Asp371 in NorM from V. parahaemolyticus, and Asp368 in KetM from K. pneumoniae were marked in green. Asterisks indicate amino acids that were common to all sequences at a particular position. Colons and dots indicate structurally similar amino acids as calculated by the Clustal W program.
doi:10.1371/journal.pone.0121619.g006 substrate for KetM, was unsuitable for detecting the movement of H + accompanied by substrate transport because of the fluorescence of DAPI. The fluorescence of DAPI disturbed the detection of the fluorescence of the pH indicator. The low solubility of DAPI in H 2 O was also problematic for measuring Na + movement by KetM in our assay system, and MIC increases by norfloxacin were too low to detect H + or Na + movement. Furthermore, there were no options for substrates. Therefore, NorM from V. cholerae or V. parahaemolyticus was considered appropriate for identifying the amino acid residues that were important for cation recognition.
Information on crystal structures appears conclusive, and previous findings on crystals containing the Rb + of NorM from V. cholerae strongly supported NorM from V. cholerae being a MATE-type transporter that utilizes the electrochemical potential of Na + [32,39]. However, a recent study reported that NorM from V. cholerae also utilized a proton motive force to transport ethidium [38]. They showed that the movement of H + consequent to the addition of ethidium was lost by the substitution of Asp371 to Asn in NorM from V. cholerae, and suggested that this amino acid residue may be important for H + transport [38]. We confirmed that these amino acid residues were also preserved in AbeM from A. baumannii (ACCESSION BAD89844.2) and PmpM from P. aeruginosa (ACCESSION AAG04750.1). However, we also showed that the amino acid residue corresponding to Asp371 in NorM from V. cholerae was widely preserved in MATE-type transporters driven by Na + (e. g. Asp367 in NorM from V. Phylogeny of KetM and MATE-type proteins in cluster 1. MATE-type proteins were analyzed with CLUSTAL W (the program used was ClustalW2.1). Na + -coupling MATE proteins were shown in red and H + -coupling MATE proteins were shown in blue. NorM Vc , which is considered to utilize both Na + and H + , was shown in pink. KetM and YdhE, the coupling cations of which have not yet been identified, were shown in black. The criteria to determine cation utilization by each MATE-type transporter was shown in S1 Table. The  parahaemolyticus, Asp367 in VcmA from V. cholerae, Asp373 in HmrM from H. influenzae, Asp364 in VcrM from V. cholerae, Asp362 in VmrA from V. parahaemolyticus, and Asp370 in PdrM from Streptococcus pneumoniae). The amino acid residue corresponding to Asp371 in NorM from V. cholerae was also preserved in KetM from K. pneumoniae (Asp368).
Otsuka et al previously reported that Asp367 in NorM from V. parahaemolyticus was important for substrate transport based on the findings obtained from the introduction of an artificial mutation at this site [40]. Based on these findings, we proposed that the importance of this amino acid residue cannot be limited to only one aspect (e.g. only coupling with Na + or H + or only the interaction or recognition of substrates) and may have multiple functions in the transport of substrates by MATE-type transporters. Otherwise, amino acid residues may play different roles in various bacteria.
KetM from K. pneumoniae transported DAPI with the addition of only lactate. NorM from V. cholerae also expelled ethidium with the addition of only glucose. We also observed a similar ethidium efflux pattern in a deduced MATE-type transporter from a Serratia species (manuscript in preparation).
Collectively, these findings suggest the existence of three kinds of MATE-type transporters in bacteria; transporters that only utilize a sodium motive force, those that only utilize a proton motive force, and those that utilize sodium and proton motive forces. We conducted a phylogenetic analysis of MATE-type transporters in cluster 1, and found no correlation between cation utilization and phylogenetically-close proteins (Fig. 7). Therefore, an evolutionally apparent branching point to separate MATE-type transporters based on energy utilization did not appear to exist, and the assembly of partial structures (e.g. a cluster of amino-acid residues) may determine the cation selectivity of MATE-type transporters.
We could not identify the coupling cations of KetM directly in the present study. However, we showed that a proton motive force played an important role in the transport of substrates by KetM from K. pneumoniae.
Supporting Information S1 Fig. Sodium movement with the addition of norfloxacin in E. coli cells transformed with ketM. We previously reported the details for this assay (PLoS One. 2013;8(3):e59525.). Briefly, cells were aerobically cultured in Tanaka medium containing 1% tryptone, 10mM melibiose, and 100 μg/ml ampicillin until the late exponential phase of growth at 30°C. These cells were then washed twice and suspended in 0.1 M 3-morpholinopropanesulfonic acid (MOPS)tetramethylammonium hydroxide (TMAH) buffer. In the assay mixture, 0.1 M N-[Tris(hydroxymethyl)methyl]glycine (Tricine)-TMAH (pH8.0) containing) containing 33 μM NaCl was used. The final concentration of Methyl-β-D-thiogalactoside (TMG) was 5 mM while that of norfloxacin (NFLX) was 20 μM. These reagents were added at each arrow point. The detector marker drifted upwards when sodium influxed into the cell and moved downwards when sodium was antiported by a secondary added chemical (J Bacteriol. 2000;182(23):6694-6697). Sodium/NFLX antiport activity was not detected in E. coli cells transformed with pDSH8 even though the experiment to detect the sodium movement in this cell was performed six times. The sample cells were E. coli KAM32/pBluescript SK(-) (A) and E. coli KAM32/pDSH8 (B). The calibration control was drawn as the change from 60 nmol NaCl (C). (TIF) S1 Table. MATE-type transporters whose cation utilization was investigated.