Development of Quinoxaline 1, 4-Dioxides Resistance in Escherichia coli and Molecular Change under Resistance Selection

Quinoxaline 1, 4-dioxides (QdNOs) has been used in animals as antimicrobial agents and growth promoters for decades. However, the resistance to QdNOs in pathogenic bacteria raises worldwide concern but it is barely known. To explore the molecular mechanism involved in development of QdNOs resistance in Escherichia coli, 6 strains selected by QdNOs in vitro and 21 strains isolated from QdNOs-used swine farm were subjected to MIC determination and PCR amplification of oqxA gene. A conjugative transfer was carried out to evaluate the transfer risk of QdNOs resistant determinant. Furthermore, the transcriptional profile of a QdNOs-resistant E. coli (79O4-2) selected in vitro with its parent strain 79–161 was assayed with a prokaryotic suppression subtractive hybridization (SSH) PCR cDNA subtraction. The result showed that more than 95% (20/21) clinical isolates were oqxA positive, while all the 6 induced QdNOs-resistant strains carried no oqxA gene and exhibited low frequency of conjugation. 44 fragments were identified by SSH PCR subtraction in the QdNOs-resistant strain 79O4-2. 18 cDNAs were involved in biosynthesis of Fe-S cluster (narH), protein (rpoA, trmD, truA, glyS, ileS, rplFCX, rpsH, fusA), lipoate (lipA), lipid A (lpxC), trehalose (otsA), CTP(pyrG) and others molecular. The 11 cDNAs were related to metabolism or degradation of glycolysis (gpmA and pgi) and proteins (clpX, clpA, pepN and fkpB). The atpADG and ubiB genes were associated with ATP biosynthesis and electron transport chain. The pathway of the functional genes revealed that E. coli may adapt the stress generated by QdNOs or develop specific QdNOs-resistance by activation of antioxidative agents biosynthesis (lipoate and trehalose), protein biosynthesis, glycolysis and oxidative phosphorylation. This study initially reveals the possible molecular mechanism involved in the development of QdNOs-resistance in E. coli, providing with novel insights in prediction and assessment of the emergency and horizontal transfer of QdNOs-resistance in E. coli.

It is demonstrated that usage of olaquindox on pig farms may increase the emergence of olaquindox resistance in some pathogens [11,12]. The bacteria with resistance to olaquindox are generally co-resistant to ampicillin, tetracycline or chloramphenicol [1,13]. For the mechanism of QdNOs resistance, it is only known that plasmid-encoded multidrug efflux pump OqxAB confers resistance to ampicillin, chloramphenicol, ciprofloxacin and olaquindox in E. coli [14,15]. This OqxAB located plasmid has a high frequency of transfer among enterobacterial pathogens (Salmonella Typhimurium, Klebsiella pneumoniae, Kluyvera sp. and Enterobacter aerogenes, etc) [14,16]. It is rarely known how E. coli develop or acquire the QdNO's resistance. For the concern of animal and human safety, it is required to be elucidated whether the selection pressure of QdNOs contributes the emergence of OqxAB and whether the resistant determinants would spread to other enterobacterial pathogens. During the development of QdNOs resistance, it is an urgent need to explore the novel molecular mechanisms involved in the emergency and transfer of QdNO's resistance in pathogenic bacteria especially E. coli.
Since serial passage is a traditional way to predict the resistance development to some antibiotics [17], some QdNOs resistant E. coli strains selected by two-fold increasing concentration of olaquindox and cyadox in vitro and some clinical QdNOs-resistant E. coli strains isolated from swine farms in China where QdNOs have been used for a long time, were subjected to analyze the development of QdNOs resistance. Since conjugation is the most important way of horizontal transfer of some antibiotic resistance determinants in E. coli [18], the horizontal transfer of QdNOsresistant determinant was then assessed by conjugative experiment. Since a prokaryotic SSH PCR cDNA subtraction approach developed by De Long is a technique that can be used to identify differentially expressed genes under given conditions [19], this approach was consequently carried out to investigate the novel mechanism beyond the OqxAB-mediated QdNOs resistance. Stock solutions of OLA, MEQ and FFC were prepared in hot water; Stock solutions of CTF were prepared in phosphate buffer (pH 6.0, 0.1 mol/L). After vortexing, the solutions were sterilized by filtration through 0.22 mm pore size filters (Millipore) before storage in aliquots at 220uC. Solutions of ampicillin, tetracycline and oxytetracycline were prepared fresh on each occasion.

Strains
The porcine E. coli 79-161 strain was purchased from China Institute of Veterinary Drug Control (Beijing, China). E. coli NK5449 (CGMCC NO. 1.1437) being resistant to rifampicin at 100 mg/ml and as conjugation recipient was purchased from China General Microbiological Culture Collection Center (CGMCC, Beijing). E. coli ATCC25922 used as quality control in MIC test was kindly donated by South China Agricultural University (Guangzhou, China). All the strains were cultured at 37uC in Luria-Bertani (LB) broth medium in aerobic incubator.

In vitro Selection of High-level QdNOs Resistant Strains
The selection of high-level QdNOs-resistant strains was performed with ancestor strain E. coli 79-161 by serial transferring in LB broth containing two-fold increased concentrations of CYA or OLA. The selection concentrations of CYA and OLA ranged from 1/2MIC to 4MIC. Each step of transferring contained 5 passages of parent strain in broth with certain concentration of drug. Four increased concentrations of drug (1/2MIC, MIC, 2MIC and 4MIC) led to a total of 20 passages of parent strain. During each passage, an initial colony concentration of 10 4 ,10 5 cfu/ml of parent strain was grown at 37uC for 24 h in aerobic incubator. Cultures obtained at 5 th , 10 th , 15 th and 20 th passages were stored at 220uC. To estimate the stability of resistance, the 20 th subcultures were serially transferred in drugfree LB broth for 10 times. For each passage, a drug-free culture (growth control) and an uninoculated medium sterility control were included.

In vivo Isolation of Resistant E.coli from Swine
On swine farm in Wuhan Huaguoshan Livestock Company, six groups of swine were breeding with drugs (Group 1 with no drug, group 2 with 50 mg/ml olaquindox, group 3 with 50 mg/ml chlorotetracycline, group 4 with 50 mg/ml cyadox, group 5 with 150 mg/ml cyadox and group 6 with 250 mg/ml cyadox). The samples of anal swab were collected from each group of swine after 0, 45 th and 100 th day adding drugs to their basal diet. The E. coli strains were isolated and identified by the selective agar (MacConkey Agar) and classic biochemical test following the microbiology laboratory guidebooks published by ministry of health in China (2003). Species identification was confirmed using an ABI 3130 system (Applied Biosystem, USA). 1,2 respective E. coli strains selected from each of the six groups were subjected to MIC determination as below. The experimental procedures involving animals in the study were approved by the Animal Care Center, Hubei Academy of Medical Sciences. All efforts were made to minimize suffering of animals.

MIC Determination
MIC was determined by agar dilution antimicrobial susceptibility test according to the M7-A7 guideline of the Clinical and Laboratory Standards Institute (CLSI, 2006).Briefly, the density of E.coli suspensions was adjusted to equal the density of a 0.5 McFarland standard (1610 8 CFU/ml). After 10 fold dilution by MH broth, 1,2mL of the diluted suspension was inoculated on prepared MH agar plates containing two-fold diluted concentration of each drug. The final concentration of strains was 10 4 CFU/ point on drug-containing plates. The standard strain of E. coli ATCC25922 was used as quality control (QC) strain. Most of the QC range and susceptible breakpoint were cited from CLSI document M100-S19. The QC range of CIF was quoted from performance standard of cefotaxime and ceftriaxone. The QC range of FFC was quoted from performance standard of chloramphenicol. The resistance breakpoint of FZD (128 mg/ml) was quoted from performance standard of furantoin. The resistance breakpoints of CTF, FFC and CAR were cited from published paper [12]. Olaquindox resistance has been defined with a breakpoint of 64 mg/ml [1]. Acquisition of cyadox resistance was defined as an $4-fold increase of the MIC value comparing with its control parent strain [20].

PCR Amplification of oqxA Genes
Colony PCR was proceeded to screen for oqxA genes in the selected E.coli strains. Based on the sequences of oqxA (GeneID 5962055) in E. coli, a pair of primers (OqxA-F 59-GGGA-TAGTTTTAACGGTCGCATTG-39 and OqxA-R 59-TT CACGGGAGACGAGGTTGGT-39) was designed to amplify 266-bp oqxA gene fragment. The amplification conditions comprised a denaturation step of 95uC for 5 min, followed by 32 cycles of 94uC for 30 s, 64uC for 30 s, 72uC for 1 min and a final extension for 10 min at 72uC. Amplicons were sequenced by Nanjing Genscript Biotechnology Limited Company (Nanjing, China).

Conjugative Transfer of QdNOs-resistant Determinant
To investigate whether QdNOs-resistant determinant possess horizontal transfer risk, filter conjugation was carried out according to the method used by Yuan et al [21], in which induced high-level QdNOs-resistant strains were used as donor and E. coli NK5449 was used as recipient. The transconjugants was selected on LB agar plates containing 100 mg/ml olaquindox and rifampicin. The frequency of conjugation was calculated as following formula: CFU of transconjugant/CFU of donor. The donor, recipient and transconjugant were identified by comparison of their EcoRI RiboPrintH pattern with that in identification database [22]. Comparing to E. coli NK5449, the similarity of the RiboPrintH pattern was shown in percent.

Preparation of mRNA
Before isolation of total RNA, driver and tester stains were grown in 50-ml drug-free Luria-Bertani (LB) broth till log phase. Briefly, overnight cultures were 100 fold diluted in 50-ml of LB broth, the culture were grown at 37uC with 250 rpm shacking. When the density of culture reached to 0.6,0.8 OD 600 , the cells were harvested by centrifugation of 1-ml aliquots in a microcentriguge for 2 min. An RNeasy Protect Bacteria Mini Kit (Qiagen, Valencia, CA, USA) was used to lyse the cells and extract total RNA. The RNA was quantified by a light absorption at 260/ 280 nm. The purity and integrity of the RNA samples were verified by an electrophoresis in agarose gels using the 16S and 23S ribosomal bands as indicators. Samples of valid RNA were DNase treated with RQ1 RNase-free DNase (Promega, Madison, WI, USA). The DNA contamination was verified by PCR amplification of rpoE gene using the primers in Table 1. Reactions containing pure RNA preparations would fail to yield a product. The mRNA was isolated using a MICROBExpress bacterial mRNA enrichment kit (Ambion, Austin, TX, USA). Every 5 preparations for each substrate were pooled before ethanol precipitation to obtain at least 2 mg mRNA per preparation.
The mRNA quantity was measured by absorbance at 260 nm, and quality was assessed on agarose gel.
Prokaryotic SSH PCR cDNA Subtraction cDNA synthesis, cDNA digestion, adaptor ligation, first and second hybridization, suppression PCR and clone screening were carried out following the protocols of De Long with slight modification on the PCR-select cDNA subtraction kit(Clontech) [19].
A RNA polymerase sigma factor rpoE (the primer in Table 1) was used to detect the quality of double-strand cDNA. The amplification conditions comprised a denatured step of 95uC for 5 min, followed by 32 cycles of 94uC for 30 s, 51uC for 30 s, and 72uC for 1 min and a final extension for 10 min at 72uC. Products were resolved by gel electrophoresis on a 2% (wt/vol) agarose gel. After second-strand cDNA synthesis, two cDNA synthesis reaction mixtures were pooled and purified with DNA fragment purification kit (Takara, Kyoto, Japan). Tester and driver cDNA were digested with Rsa I and purified also with DNA fragment purification kit (Takara).
To determine the efficiency of adapter ligation and subtraction, PCR amplification were performed according to the protocol in the PCR-Select cDNA subtraction kit by use of a housekeeping gene ribosomal protein S1 (rpsA) (the primer in Table 1). The amplification conditions comprised a denatured step of 95uC for 5 min, followed by cycles of 94uC for 30 s, 60uC for 30 s, and 72uC for 1 min and a final extension for 10 min at 72uC. For test of ligation efficiency, the cycles were 32. And for test of subtraction efficiency, the cycles were 18, 23, 28, 33 and 38.
After the PCR adaptor screen, 1, 2-clones were selected and submitted to Nanjing Genscript Biotechnology Limited Company (China) for sequencing directly. The sequences were aligned to available sequences in NCBI library by blastn algorithm (www. ncbi.nlm.nih.gov. BLAST). The homological genes were subjected into molecular annotation system (http://www.capitalbio.com/ support/mas) or KEGG library for gene function annotation and pathway analysis.
Reverse Transcription, Quantitative Real-time PCR (RT-qPCR) 4,5 mg of DNase-treated total RNA (isolated as described above) was used to synthesize cDNA according the method of De Long [19]. Real-time PCR was carried out on Bio-Rad iQ TM 5 Multicolor Real-Time PCR Detection System, using SYBRH Premix Ex TaqTM (Takara) in a total volume of 25 ml: 12.5 ml SYBRH Premix Ex TaqTM, 1.0 ml template cDNA, 0.5 ml (final concentration 0.2 mM) or 0.25 ml (final concentration 0.1 mM) of each primer, and PCR-grade water to a final volume of 25 ml. Samples without template were set as negative control. At least 3 replicates were analyzed for each gene. Primer sequences were presented in Table 1.The fold change of selected genes was normalized to fold change of reference gene (16S rRNA gene). The relative expression was calculated according to the formula 2 2ggCt [23], The data were considered for significant differences by one-way ANOVA using SPSS programme (P#0.05).
Before and after adding olaquindox, chlorotetracycline and cyadox into swine diet, 21 representative E.coli strains were isolated from 6 groups of swine by anal swab. From the data shown in Table 3, there is no significant change of the MIC before and after usage of QdNOs. Before usage of drug, the 7 E.coli strains selected at 0 day exhibited resistance to olaquindox with MIC $64 mg/ml. Notably, the 3 E.coli strains from the control group is naturally resistant to olaquindox with MIC = 64 mg/ml or 256 mg/ml.

oqxA Gene Mediated Resistance
All of the in vitro selected stains presented in Table 2 were negative for oqxA gene products, indicating that other mechanisms except for OqxAB mediated the resistance in the selected strains. However, 95% (20/21) E.coli strains isolated from swine farms were positive for oqxA gene (see Table 3).The oqxA positive strains are isolated before and after the usage of QdNOs drugs on pig farm. No relationship can be observed between the occurrence of oqxA gene and usage of QdNOs drugs. The oqxA gene positive strains always exhibited resistance to olaquindox except for E100-4 strain which carried oqxA gene but is susceptible to olaquindox (MIC = 32 mg/ml).

Frequency of Conjugal Transfer
Frequencies of conjugal transfer for each strain were exemplified in Table 4. The conjugation frequencies of the three E. coli donors were at 10 210 CFU/donor. Before conjugation, the recipient is susceptible to the four QdNOs but resistant to rifampicin, the donors are resistant to the four QdNOs but susceptible to rifampicin. After conjugation, the transconjugants were resistant to both four QdNOs and rifampicin (see Table 4). The recipient and transconjugants share similar RiboPrintH pattern with the similarity .90% (see Table 4), confirming that QdNOs-resistant determinant had been horizontally transferred by conjugation in E. coli strains.

Differentially Expressed Genes Determined by Prokaryotic SSH PCR cDNA Subtraction
By prokaryotic SSH PCR cDNA subtraction, 44 unique fragments were found in the QdNOs resistant strain E. coli 79O4-2 comparing to its control strain E. coli 79,20. Among these 44 sequences, 3 sequences were with no significant similarity found in the Genbank; 40 sequences were found in genome of E. coli; 1 sequence was found in plasmid of E. coli (mobA) (see Table 5). The homology search results showed that QdNOs-resistant E. coli alters the expression of 18 genes involved in biosynthesis, 11 genes associated with metabolism, 3 genes encoding transporters, 3 genes involved in flagellar assembly, 1 gene encoding mobilization protein and 5 genes associated with hypothetical proteins (see Table 5).
In the 11 genes involved in metabolism, the number 19 and 20 genes (gpmA and pgi) participate in glycogen degradation, the number 21 to 24 genes (clpX, clpA, pepN and fkpB) take part in degradation of damaged protein or protein modulation, the number 25 to 29 genes (atpA, atpD, atpG, ubiB and yecD) contribute to electron transport and ATP biosynthesis (see Table 5 and Figure 1).
As shown in table 5, the 3 transport proteins include spermidine/putrescine ABC transporter membrane protein (potC), outer membrane-specific lipoprotein transporter subunit (lolC) and preprotein translocase (secA). The flgD, flhA and fliC are involved in assembly of flagellar (see Table 5). The mobA gene is associated with the mobilization of bacterial (see Table 5).  Up-regulation of 11 respresentative genes (mobA, rpoA, rplF, fusA, fabI, clpX, atpD, ubiB, lolC, secA and fliC) was confirmed by RT-qPCR. Fold change of the 11 genes is in the range of 1.7 to 11 fold of over-expression with the P value much less than 0.05, indicating that the methodology of prokaryotic SSH PCR cDNA subtraction is reliable to identify differentially expressed genes (see Figure 2).

Resistance Development Under QdNOs
Although the resistant E. coli strains were obtained from stepwise selection by QdNOs drugs in vitro, there is no significant change of susceptibility of the E.coli strains isolated from pigs farm before and after the usage of QdNOs drugs. It was revealed that the incidence and level of olaquindox resistance in the coliform floras isolated from pigs and pig farms was complex and largely fluctuated, the possible origin of olaquindox-resistant strains may be related to the biotypes of the drug-resistant sub-populations [24]. During the usage of olaquindox as feed additive on commercial pig farms from 1982 to 1984, the monitoring result about the development of olaquindox resistance showed that there was a low but not significant increasing incidence and level of resistance to olaquindox on drug-using farms [25]. Conclusively, it is unreliable to connect the emergency of QdNOs resistance to the usage of QdNOs drugs only by series passage in vitro. The resistance of E.coli strains isolated from the control group of pigs may due to the complexity of clinical environment.
Compared the MICs of QdNOs in the three strains (79O4-1, 79O4-2 and 79O4-3) selected by olaquindox to the three strains (79C4-1, 79C4-2 and 79C4-3) selected by cyadox, olaquindox appears to possess more potency in selection of cross-resistance to QdNOs, suggesting that long-term use of cyadox could be safer than olaquindox. The difference in the development of crossresistance induced by olaquindox and cyadox may attribute to the discrepancy in the chemical characteristics or structure of these two drugs [2].
Similar to previous studies of resistant phenotype in E.coli strains isolated from pigs and pig farms, the QdNOs resistant strains selected in our study also exhibited multi-drug resistance against tetracycline and chloramphenicol [24,25,26]. Both the result of previous studies and our study revealed that multidrug resistance in most of clinical isolates may due to the multidrug efflux pump OqxAB [14,15]. The OqxAB located plasmid was reported to has a high frequency (10 24 CFU/recipient) of transfer among enterobacterial pathogens including S.Typhimurium and K.pneumoniae, etc [14,16]. The E100-4 strain (see table 3) carried the oqxA gene but is susceptible to olaquindox, indicating that some strains may acquire the oqxA gene from the horizontal transfer in clinical environment. However, different with previous studies, the six QdNOs resistant strains selected by in vitro passage ((79O4-1, 79O4-2, 79O4-3, 79C4-1, 79C4-2 and 79C4-3) were negative in OqxAB pump (data not shown ) and showed much lower frequency of horizontal transfer (see table 4) with the ratio of transconjugant to donor only in a range of 2.5,4.0610 210 CFU/ donor, suggesting that other mechanism beyond OqxAB might confer the multidrug resistance and lower transfer risk in the selected strains in our study. A prokaryotic SSH PCR cDNA subtraction was thus carried out to find the novel molecular mechanism involved in QdNOs resistance.

Differently Expressed Genes Involved in the QdNOs Resistance
Based on networks of the drug-target induced cell death pathway found recently [27], it is supposed that the QdNOs Table 4. Transfer frequency and specific assessment of conjugation.    binding to unknown target may initiate a metabolic feedback dependent on the tricarboxylic acid (TCA) cycle, stimulate the oxidation of NADH through the electron transport chain; promote superoxide (O 22 ) formation; damage the structure of Fe-S clusters, lead to the formation of hydroxyl radicals (OH 2 ); and finally damage DNA, lipids or protein and contributes to the death of E.coli (see Figure 1). In the differently expressed genes, the iron-sulfur cluster binding protein encoding gene (gene number 1) participates in Fe-S cluster assembly. The nitrate reductase 1 encoding gene (narH) contribute to the electron transfer for Fe-S cluster [28,29]. Thus, overexpression of number 1 and 2 genes (narH) may compensate Fe-S cluster damage by assembly of Fe-S cluster (see figure 1).
The rpoA is a DNA-direct RNA polymerase which can interact with other proteins (rplF, rplC, rplX, rpsH and others) and plays importance role in transcriptional regulation [30,31]. The trmD, truA, glyS and ileS are essential enzymes for tRNA modification and aminoacyl-tRNA synthesis which also play essential role on the protein synthesis [32]. The fusA gene encoding for a translational regulator which is essential for translational elongation [33,34]. The overexpression of the translational regulator fusA and some ribosomal proteins involved in translational process may compensate protein synthesis for oxidative damage by QdNOs and contribute to the multidrug resistance against some protein target drug of tetracycline and chloramphenicol.
The number 13 gene (lipA) encodes lipoate synthase (LS) which is involved in the biosynthesis of Lipoic acid [35,36,37]. Lipoic acid is a protein-bound dithiol-containing cofactor which is commonly required for the energy metabolism in E. coli [37]. Recent studies also show that lipoic acid is a free radical scavenger and a potent antioxidant [38,39,40]. The number 17 gene (otsAB) encodes trehalose-6-phosphate synthase which is responsible for trehalose biosynthesis [41,42,43]. It has been shown that trehalose can protect proteins and cellular membranes from inactivation or denaturation caused by a variety of stress conditions, including desiccation, dehydration, heat, cold or oxidation [42,43,44]. The accumulation of trehalose can protect Saccharomyces cerevisiae from oxidative damage and enhance resistance of bacterial to H 2 O 2 , indicating that trehalose may be a good free radical scavenger [45]. Therefore, the overexpression of lipA and otsA gene may promote production of two free radical scavenger (lipoic acid and trehalose) which may protect QdNOs-resistant E. coli from oxidative damage and result in QdNOs-resistance. There is no significant change of the expression of oxyR or soxRS regulon, so this mechanism of antioxidative protection mediated by lipoic acid and trehalose may be independent on the regulation of oxyR or soxRS [46].
The number 19 gene (phosphoglyceromutase 1, gpmA) and number 20 gene (glucosephosphate isomerase, pgi) are involved in glycolysis which generates a small quantity of ATP and degrades glucose into pyroracemic acid and acetyl-CoA for tricarboxylic acid cycle (TCA) under aerobic growth condition. These two enzymes encoded by gpmA and pgi gene are also important for central carbon metabolism (CCM) pathway which may influence the fidelity of DNA replication in E.coli [47,48]. The 2octaprenylphenol hydroxylase (ubiB), F1 sector of membranebound ATP synthase a (atpA), b (atpD) and c (atpG) subunit are involved in ATP biosynthesis, electron transport chain and oxidative phosphorylation [49].
Based on the function annotation and pathway analysis, the Clp, one of the key ATP-dependent proteases in E. coli, is required during some stress conditions [50,51,52,53]. The PepN, a major aminopeptidase in E. coli, can degrade amino acids into small molecular for TCA cycle [54]. The Clp protease cleave damaged proteins into large peptides and amino acids which are further degraded by PepN [55]. The fkpB, peptidyl-prolyl cis-trans isomerase, participates in several biochemical processes including protein folding, receptors signaling, protein trafficking and transcription [56]. Therefore, up-regulation of Clp proteases (21-ClpX and 22-ClpA), aminopeptidase pepN (number 23 gene)and protein folding chaperones fkpB (number 24 gene)may play important roles in regulating levels of specific proteins and in eliminating or modifying damaged or abnormal proteins [57].
Considering the drug-target induced cell death pathway and function analysis of the differentially expressed genes, the upregulation of some genes may be response for some process of cell death pathway, but others may compensate or inhibit that pathway. See figure 1, the two genes involved in glycolysis for TCA cycle (19-gpmA and 20-pgi), four genes related to ATP biosynthesis and electron transport chain (25-atpA, 26-atpD, 27-atpG and 28-ubiB) and four genes associated with protein degradation and modulation (21-clpX, 22-clpA, 23-pepN and 24-fkpB) might be response for the metabolic feedback, hyperactivation of electron transport chain and protein damage, respectively. However, the two genes involved in Fe-S cluster assembly (1-gene and 2-narH) and several genes associated with protein biosynthesis may compensate the Fe-S cluster damage and protein damage respectively (see Figure 1). The overexpression of genes involved in biosynthesis of lipoic acid and trehalose (13-lipA and ostA) may eliminate superoxide formation and hydroxyl radical production (see Figure 1). The role of lipA and ostA on clearage of free radical and development of QdNOs resistance would be investigated in our next plan, because unpublished data obtained in our lab recently revealed that QdNOs are endowed with producing free radical in E.coli cell.
Additionally, the potC gene is involved in spermidine uptake which plays an important role in bacterial growth including cell proliferation and differentiation [58]. The lolC gene encodes an outer membrane-specific lipoprotein transporter which mediates the detachment of lipid-modified proteins and catalyzes release of outer membrane-directed lipoproteins from the inner membrane of E. coli [59,60,61]. Disruption of lolCDE can prevent release of lipoprotein from inner membrane and shows to be lethal for E.coli [60]. The secA gene encodes a peripheral membrane ATPase which participates in preproteins translocation across the bacterial cytoplamic membrane and plays an essential role in the survival of the bacterial species [62]. Recent study demonstrated that export of the superoxide dismutase SodA to periplasm was dependent on activity of SecA ATPase [63]. Therefore, overexpression of genes involved in transporters plays an essential role on the survival and growth of QdNOs-resistant E. coli. The lpxC encodes zincdependent UDP-3-O-(acyl)-N-acetylglucosamine deacetylase which catalyzes the first irreversible step of lipid A biosynthesis [64]. Lipid A biosynthesis (lpxC), cytidine triphosphate biosynthesis (pyrG), fatty acid biosynthesis (fabI) and enterobacterial common antigen biosynthesis (rffM) may be essential for viability of E.coli. The flgD, flhA and fliC genes are involved in flagellar assembly which exhibit functions in cell motility and sensibility [65]. Recent findings revealed that the flagellum participated in a variety of processes including biofilm formation, pathogenesis and even regulation of gene expression [58]. The mobA gene encodes a mobilization protein which can be induced by some drugs and mediates the transfer of bacteroides conjugation [66,67]. These genes may play important roles in physiological functions of QdNOs-resistant E. coli.

Conclusions
Results of the present study suggest that olaquindox possess more potency in selection of multi-drug resistant and crossresistant E. coli strains. The induced QdNOs-resistant determinant shows lower frequencies of conjugation than previous reports in clinical isolates. More than 90% clinical isolates carry oqxA gene, but it is negative in our in vitro selected QdNOs-resistant strains, indicating that there might be other mechanism conferring resistance. By SSH PCR subtraction, 44 genes were differentially expressed in the QdNOs-resistant strain 79O4-2 selected by olaquindox in vitro. The overexpression of genes involved in glycolysis for TCA cycle (gpmA and pgi) and genes involved in ATP biosynthesis and electron transport chain (atpADG and ubiB) may due to the metabolic feedback and hyperactivation of electron transport chain, respectively. The genes involved in Fe-S cluster assembly (1-gene and narH) and genes involved in biosynthesis of lipoic and trehalose (lipA and ostA) may play an important role in clearage of free radical and prevention of Fe-S damage. The genes involved in protein biosynthesis (rpoA, trmD, truA, glyS, ileS, rplF, rplC, rplX, rpsH, rplF and fusA) and genes involved in protein degration and modulation may participate in compensation of the protein damage (see Figure 1). The other genes including genes encoding transporter, flagellar, lipid A and mobilization protein may be essential for the growth and physiological functions of QdNOs-resistant E. coli. Conclusively, most of these genes have previously been implicated in resistance or tolerance to superoxide stress, suggesting that their altered expression may be a part of a protective response in QdNOs-resistant E. coli [68]. Likewise, other identified changes in gene expression may affect the growth and transfer capacity of QdNOs resistance. Further studies are required to investigate the roles of these genes in development of QdNOs resistance by technique of gene knockout and transcriptomics or proteomics. In addition, the molecular mechanism involved in the discrepancy of resistance development selected by olaqindox and cyadox is necessary to be investigated in future, because these information will be useful for the designation of new member of versatile QdNOs.