Subinhibitory Concentrations of Ciprofloxacin Enhance Antimicrobial Resistance and Pathogenicity of Enterococcus faecium

ABSTRACT

Enterococcus faecium has emerged as a major opportunistic pathogen for 2 decades with the spread of hospital-adapted multidrug-resistant clones. As members of the intestinal microbiota, they are subjected to numerous bacterial stresses, including antibiotics at subinhibitory concentrations (SICs). Since fluoroquinolones are extensively prescribed, SICs are very likely to occur in vivo, with potential effects on bacterial metabolism with subsequent modulation of opportunistic traits. The aim of this study was to evaluate globally the impact of SICs of ciprofloxacin on antimicrobial resistance and pathogenicity of E. faecium. Transcriptomic analysis was performed by RNA sequencing (RNA-seq) (HiSeq 2500; Illumina) using the vanB-positive reference strain E. faecium Aus0004 in the absence or presence of ciprofloxacin SIC (0.38 mg/liter, i.e., 1/8 of the MIC). Several genetic and phenotypic tests were used for validation. In the presence of ciprofloxacin SIC, 196 genes were significantly induced, whereas 286 genes were significantly repressed, meaning that 16.8% of the E. faecium genome was altered. Among upregulated genes, EFAU004_02294 (fold change, 14.3) encoded a protein (Qnr of E. faecium [EfmQnr]) homologue of Qnr proteins involved in quinolone resistance in Gram-negative bacilli. Its implication in intrinsic and adaptive fluoroquinolone (FQ) resistance in E. faecium was experimentally ascertained. Moreover, EFAU004_02292, coding for the collagen adhesin Acm, was also induced by the SIC of ciprofloxacin (fold change, 8.2), and higher adhesion capabilities were demonstrated phenotypically. Both EfmQnr and Acm determinants may play an important role in the transition from a commensal to a pathogenic state of E. faecium that resides in the gut of patients receiving fluoroquinolone therapy.

INTRODUCTION

Although considered to be slightly virulent since they are natural inhabitants of the gastrointestinal tract of humans, enterococci have emerged as major opportunistic pathogens, now being responsible for 5 to 10% of hospital-acquired infections and numerous outbreaks (1). In addition, there is a spread of vancomycin-resistant enterococcus (VRE) clinical isolates, especially within the species Enterococcus faecium, which is increasingly reported (2). Notably, E. faecium is part of the ESKAPE (E. faecium, Staphylococcus aureus, Klebsiella pneumoniae, Acinetobacter baumannii, Pseudomonas aeruginosa, and Enterobacter spp.) multidrug-resistant (MDR) opportunistic pathogens (3). The epidemiological success of E. faecium mainly results from the worldwide dissemination of a subpopulation of hospital-adapted clones that belong to the so-called clonal complex 17 (CC17) (4). Recently, a comparative analysis of the genomic sequence of 51 E. faecium strains has revealed that the CC17 strains belong to an epidemic hospital-adapted lineage (clade A1) that is genetically distant from the community-associated lineage (clade B) and that emerged approximately 75 years ago from the animal-associated lineage (clade A2) after the introduction of antibiotics (5). Note that these CC17 isolates have common characteristics, such as high-level resistance to ampicillin and fluoroquinolones (6).

Besides its selective advantage provided by intrinsic and acquired antibiotic resistance traits, E. faecium is able to cope with numerous environmental stresses thanks to its great genomic plasticity and its versatile metabolism (1). As an opportunistic pathogen, the transition from a commensal to a pathogenic state is a key element in the pathogenesis of CC17 (clade A1) E. faecium isolates, but the nature of the intrinsic and extrinsic changes involved in this process has been poorly investigated (7). Among bacterial stress responses, the presence of antibiotics, particularly at subinhibitory concentrations (SICs), i.e., concentrations below the MIC that do not affect growth of the organism tested (8), is an environmental condition that has an important role in the evolution and dissemination of antibiotic resistance (9, 10). In addition, the exposure to antibiotic SICs also has significant effects on bacterial physiology and morphology, with subsequent modulation of fitness and pathogenicity traits (10). Indeed, there is an increasing number of reports of numerous phenotypic changes induced by SICs of different antibiotics in several bacterial species (8, 10, 11). However, only a few studies have been conducted in enterococci, all in Enterococcus faecalis, with SICs of ampicillin (12), vancomycin (12), chloramphenicol (13), or tigecycline (14), while no study has been performed in E. faecium. Also, only a few global transcriptomic analyses using RNA sequencing (RNA-seq) have been reported to study the bacterial response to antibiotic SICs (15–17). However, SICs of antibiotics are often present in body compartments and tissues in patients under antimicrobial therapy, which can be due to many reasons, such as suboptimal dosing regimens, poor drug pharmacokinetics, use of low-activity drugs, and/or poor patient compliance (10).

Fluoroquinolones are synthetic antibacterial agents that exhibit potent activity against a broad range of Gram-negative and Gram-positive bacteria (18). They are used to treat a great variety of infections, including urinary tract, bone, joint, respiratory tract, and enteric infections (18). Even if fluoroquinolones are poorly or moderately active against enterococci and generally not used for the treatment of enterococcal infections, they can have profound collateral effects on bacterial populations of the microbiota, including enterococci (19). This is particularly true for the gut microbiota of hospitalized patients for which exposure to antibiotics results in major modifications, with facilitation of colonization by drug-resistant E. faecium (20, 21). Indeed, the impact of fluoroquinolones on this major opportunistic pathogen, even at SICs, is very likely.

In the study described here, we analyzed for the first time, to our knowledge, the transcriptome of a hospital-adapted E. faecium clinical isolate in response to antibiotic stress. This provides an opportunity to evaluate the consequences of environmental adaptation with the identified fluoroquinolone-responsive genes. We used the method of deep sequencing-based transcriptome analysis (RNA-seq) to monitor the levels of all transcripts in bacterial cells grown in the presence and the absence of ciprofloxacin, the most commonly used fluoroquinolone.

RESULTS AND DISCUSSION

Transcriptome of E. faecium Aus0004 determined by RNA-seq.

For transcriptomic analysis, we used the reference strain E. faecium Aus0004, a vanB-positive sequence type 17 (ST17) clinical isolate recovered from a bloodstream infection in Australia in 1998 (22). This strain corresponds to the first complete E. faecium genome, which contains a 2.9-Mb circular chromosome and three circular plasmids (22). The chromosome harbors 2,753 open reading frames (ORFs), including several putative virulence factors, such as enterococcal surface protein (Esp), hemolysin, and collagen-binding adhesion (Acm), and a single copy of Tn1549 that contains the vanB operon (22).

RNA-seq was used to compare transcriptomes of strain Aus0004 grown in the absence or presence of a subinhibitory ciprofloxacin concentration (−Cip or +Cip, respectively), which we determined at 0.38 μg/ml (i.e., 1/8 the MIC) (see Fig. S1 in the supplemental material). Around 20 million reads were obtained for each cDNA library, of which 69% to 92% mapped to the genome of E. faecium Aus0004, corresponding to average genome coverage from 244× to 274× (Table S2). The reproducibility of experimental duplicates was satisfactory (r², >0.91) under both conditions (Fig. S2), with an average number of reads per coding sequence (CDS) between 986 and 5,056 (Table S2). To date, the RNA-seq approach has been used only twice in E. faecium: a multiomic analysis to reveal phenotypic and genetic changes after spaceflight (23), and a study investigating the response to subinhibitory chlorhexidine exposure (24).

Differentially expressed genes in the presence of ciprofloxacin.

The expression of each annotated gene of strain Aus0004 in bacterial cells grown under −Cip and +Cip conditions is shown in Fig. 1A (see also Table S3A in the supplemental material). To assess the reliability of RNA-seq in determining the relative abundance of individual transcripts under both conditions, we used the identical total RNA samples and determined by reverse transcription-quantitative PCR (qRT-PCR) the mRNA levels of six upregulated genes and five downregulated genes (Table 1). Those genes were chosen according to their levels of expression fold changes and their putative functions. The ratios of the transcripts from −Cip and +Cip samples determined by RNA-seq and compared to those obtained by qRT-PCR resulted in an excellent concordance, with a Pearson correlation value of 0.9772 (Fig. 1B). Therefore, RNA-seq appeared to be a reliable method for global transcriptomic analysis in E. faecium under the conditions tested in this study.

FIG 1.

Transcriptional response of E. faecium to SIC of ciprofloxacin. (A) Global analysis of transcript levels in E. faecium Aus0004 by RNA-seq. Conditions −Cip and +Cip refer to bacterial growth in the absence and presence of ciprofloxacin (concentration at 0.38 μg/ml), respectively. Green and blue circles correspond to the mean expression of each gene (as calculated by differential expression sequencing [DESeq]) in bacteria grown under −Cip and +Cip conditions, respectively. The red line in the gray circle represents the baseline, and thin gray circular lines represent 4-fold changes (or log₂ = 2) in expression of each gene (see Table S3A). Outermost circle represents the full E. faecium Aus0004 genome with a ×50 magnification of the genes for which the expression was confirmed by qRT-PCR (see Table 1). (B) Validation of RNA-seq results by qRT-PCR on 11 selected genes. Mean log₂ ratios of values determined in the qRT-PCR experiments are plotted against the mean log₂ ratios of the RNA-seq experiments.

TABLE 1.

Selected genes used for qRT-PCR validation

Gene no.	Gene name	Product name	Gene start	Gene end	RNA-seq fold change	Adjusted P value
EFAU004_00325		β-Glucosidase	322212	323627	−28.86	3.1E−16
EFAU004_00481		Galactokinase	496188	497375	−765.26	2.6E−48
EFAU004_00494		LysM domain-containing protein	509550	510158	18.94	5.9E−11
EFAU004_00691		MerR family transcriptional regulator	723563	723955	23.91	2.1E−8
EFAU004_01169		Formate acetyl-transferase	1194938	1197169	−18.92	1.3E−7
EFAU004_01501		Iron-containing dehydrogenase	1518035	1518832	−127.49	1.3E−24
EFAU004_02077		SorC family transcriptional regulator	2102863	2103900	8.90	1.5E−8
EFAU004_02266	msmK	ABC transporter ATP-binding protein	2298900	2300009	−23.75	2.2E−5
EFAU004_02292	acm	Collagen-binding protein	2329088	2331253	8.20	1.3E−5
EFAU004_02294	Efmqnr	Qnr-like protein	2332398	2333030	14.33	2.2E−7
EFAU004_02458		Preprotein translocase subunit SecE	2476151	5476321	4.38	8.2E−5

The analysis of RNA-seq data identified 482 genes (16.8% of the genome) with statistically significantly (i.e., adjusted P value, < 0.05) altered transcription levels, with 196 genes and 286 genes showing an increase and decrease of at least 4-fold (i.e., log₂ >2) in mRNA amounts in cells grown in the presence of subinhibitory ciprofloxacin concentration compared to those grown in the absence of antibiotic, respectively (Tables S3B and C). The 482 differentially expressed genes were classified into functional categories, with approximately 32% coding for hypothetical proteins of unknown function (Fig. 2 and Tables S3B and S3C). In the presence of subinhibitory ciprofloxacin concentration, genes coding for proteins involved in nucleotide biosynthesis and metabolism, translation, posttranscriptional modification, degradation, and genes for noncoding RNAs and cell wall proteins were significantly induced, whereas genes coding for proteins involved in carbon compound metabolism, energy metabolism, transcription, RNA processing and degradation, transport of small molecules, and genes coding for putative enzymes, were significantly repressed (Fig. 2). It seems that exposure to ciprofloxacin SIC perturbs bacterial metabolism globally, especially with a decrease in metabolic activity.

FIG 2.

Classification in functional categories of the 482 genes significantly induced (n = 196) or repressed (n = 286) by SIC of ciprofloxacin. Percentages of genes with a change in expression level lower or greater than 2 log₂-fold are represented in white or black bars, respectively. Statistically significant differences (using Fisher's exact test) are indicated as follows: *, P < 0.05; ***, P < 0.001.

High-level alteration of carbon metabolism caused by SIC of ciprofloxacin.

As assessed by RNA-seq, the most significant alteration in gene expression caused by subinhibitory ciprofloxacin concentration concerned genes involved in glycerol catabolism, with very strong repression of the glpK and glpO genes (fold changes, −187 and −404, respectively) as well as strong repression of the dhaK and dhaL genes (fold changes, −11.5 and −40.7, respectively) (Table S3A). These data revealed the repression of the two pathways involved in glycerol catabolism. Glycerol has been shown to be an important substrate for E. faecalis in vivo, and its metabolism appeared to be induced during mice peritonitis (25). There also was a strong repression of genes involved in galactose metabolism (EFAU004_00481 to EFAU004_00483), with fold changes between −765 and −523, as well as several phosphotransferase system (PTS) and ABC transporters (Tables 1 and S3A). This pointed out that the stress caused by the presence of even a low antibiotic concentration deeply modifies the carbon flow, with a concomitant alteration in interconnected metabolic pathways and specific use of energy sources available for growth.

Induction by ciprofloxacin of a qnr-like determinant involved in fluoroquinolone resistance.

Among genes significantly upregulated by the SIC of ciprofloxacin, EFAU004_02294 was identified (fold change, 14.3) (Table 1). This gene (633 bp) coded for the Qnr protein in E. faecium (EfmQnr) previously identified in E. faecium ATCC 35667 (26), which exhibited 14% to 33% amino acid identity with plasmid-mediated Qnr determinants described in Gram-negatives and chromosomal Qnr-like proteins of Gram positives (Fig. S3) (27, 28). Since qnr in E. faecalis (Efsqnr) was demonstrated to be involved in the intrinsic low-level fluoroquinolone resistance in E. faecalis (28), we hypothesized that Efmqnr was also associated with fluoroquinolone resistance in E. faecium.

First, we determined the genetic organization of the Efmqnr transcriptional unit. Rapid amplification of cDNA ends-PCR (RACE-PCR) experiments identified a unique promoter upstream of the Efmqnr gene, and we demonstrated that Efmqnr was part of a two-gene operon since it was cotranscribed with a downstream gene, EFAU004_02295, coding for a hypothetical protein likely belonging to the CAAX amino terminal protease family (Fig. 3). The role of these proteins is mostly unknown, even if some data suggest that these proteins are putative metal-dependent proteases potentially involved in protein and/or peptide modification and secretion (29). By genomic comparison with all publicly available E. faecium genomes, it was demonstrated that Efmqnr was conserved in all chromosomes and then part of the E. faecium core genome, and that Efmqnr and EFAU004_02295 were always part of the same operonic structure (data not shown). Interestingly, no homolog of EFAU004_02295 was found in E. faecalis and Efsqnr is not cotranscribed with other genes within this species. Of note, a putative LexA binding site was identified in the promoter region (Fig. 3) (30), strongly suggesting an SOS-dependent regulation of the Efmqnr expression. This regulation has only been described for the qnrB expression in Gram-negative bacteria (31, 32), whereas no LexA binding site has been found so far for other qnr and qnr-like genes, including Efsqnr in E. faecalis. Because fluoroquinolones introduce DNA strand breaks, these compounds are inducers of the SOS response (33). In this study, we showed that the increasing concentration of ciprofloxacin was correlated with the induction of Efmqnr transcription (Fig. 4A), which is also in agreement with an SOS-mediated regulation of this gene. Note that the transcription of the recA gene was induced more than 3-fold when cells were in the presence of the SIC of ciprofloxacin (Table S3A), which is expected since the upregulation of recA is a hallmark of the SOS response (33).

FIG 3.

Genetic environment of the Efmqnr gene in E. faecium Aus0004. Open reading frames (ORFs) are indicated by horizontal arrows according to the strain Aus0004 annotation (GenBank accession no. NC_017022). The transcription start site (+1) determined experimentally is underlined, as well as the deduced −35 and −10 promoter boxes. The putative LexA binding site is represented in bold. The cotranscription of Efmqnr and EFAU004_02295 was evaluated by reverse transcription-PCR (RT-PCR) using primers designed to amplify specific regions of Efmqnr (A) or EFAU_02295 (B), intergenic region (C), the long cotranscript (D), and a negative control (see Table S1). Lanes 1 and 2 on agarose gel correspond to PCR amplifications using chromosomal DNA (used as control) or cDNA as matrix, respectively.

FIG 4.

(A) Expression levels of Efmqnr (black bars) and acm (gray bars) genes evaluated by qRT-PCR with increasing concentrations of ciprofloxacin. Mean expression ratios (± standard deviations) for the Aus0004 strain in the presence of ciprofloxacin compared to the same strain without antibiotic stress (+Cip/−Cip ratios) are indicated. (B) MICs of ciprofloxacin determined by Etest in the absence (−Cip, left side) or in the presence (+Cip, right side) of ciprofloxacin. The fluoroquinolone was incorporated into the solid agar at the concentration of 0.38 μg/ml.

In order to demonstrate the role of Efmqnr in fluoroquinolone resistance in E. faecium, we cloned this gene along with its own promoter into the pAT29 shuttle vector and expressed the recombinant plasmid in both Escherichia coli and E. faecium cell recipients. The presence of Efmqnr was responsible for a 2- to 8-fold and 1.5- to 3-fold increase in the MICs of fluoroquinolones in E. coli and E. faecium, respectively (Table 2). This increase in the MICs of fluoroquinolones is lower than that previously described in E. coli (8- to 32-fold increase), where Efmqnr was cloned under the control of the lac promoter, allowing high-level constitutive expression (26). We also showed that the clean deletion of Efmqnr in E. faecium was responsible for a 1.5- to 4-fold decrease in the MICs of fluoroquinolones, confirming the role of Efmqnr in intrinsic low-level resistance of E. faecium to fluoroquinolones, as previously described for Efsqnr in E. faecalis (28). Note that no role of EFAU004_02295 was evidenced in fluoroquinolone resistance in E. faecium, since its chromosomal deletion did not impact fluoroquinolone susceptibility profiles (Table 2). Finally, we showed that the presence of the SIC of ciprofloxacin was responsible for a 3-fold increase in the MICs of ciprofloxacin (Fig. 4B). It may be suggested that the low concentrations of ciprofloxacin could be favorable to E. faecium, allowing it to cope with higher fluoroquinolone concentrations, through the transcriptional induction of Efmqnr. In this context, the frequent use of fluoroquinolones to treat infections caused by pathogens other than E. faecium may, however, constitute a risk factor for colonization of individuals by VRE belonging to the CC17. Note that no reduced susceptibility was observed for other antibiotics tested (i.e., ampicillin, gentamicin, vancomycin, teicoplanin, linezolid, daptomycin, and tigecycline) (data not shown). Very recently, a global transcriptomic study in E. faecium showed that a subinhibitory concentration of the antiseptic chlorhexidine induced genes encoding VanA-type vancomycin resistance (vanHAX) and those associated with reduced daptomycin susceptibility (liaXYZ) (24). Altogether, these results highlight the significant effect of antimicrobial agents (both biocides and antibiotics) on E. faecium metabolism, with adaptive responses potentially enhancing opportunistic traits of this major human pathogen in vivo.

TABLE 2.

MICs of fluoroquinolones for E. coli and E. faecium strains

Strain	MIC (μg/ml)a
	NOR	OFX	CIP	LVX	MXF
E. coli
TOP10_pAT29	0.016	0.012	0.002	0.004	0.002
TOP10_pAT29ΩEfmqnr	0.032	0.064	0.016	0.023	0.016
TOP10_pAT29ΩEfmqnr-EFAU004_02295	0.032	0.032	0.006	0.016	0.012
E. faecium
Aus0004_pAT29	4	8	1.5	2	1
Aus0004_pAT29ΩEfmqnr	8	12	4	3	2
Aus0004_pAT29ΩEfmqnr-EFAU004_02295	6	8	3	2	1
Aus0004ΔEfmqnr	2	6	0.75	1	0.25
Aus0004ΔEFAU004_02295	4	8	1.5	2	1
Aus0004ΔEfmqnr-EFAU004_02295	2	6	0.75	1	0.25

^aNOR, norfloxacin; OFX, ofloxacin; CIP, ciprofloxacin; LVX, levofloxacin; MXF, moxifloxacin.

Ciprofloxacin upregulates the expression of the collagen-binding adhesion Acm.

Few virulence factors have been characterized in E. faecium, and only three genes/loci have been shown to be associated with virulence in animal models: acm (34, 35), the ebp_fm operon (36), and esp (37, 38). Apart from a gene coding for a putative hemolysin (EFAU004_01337) that is expressed in many E. faecium bacteremic isolates, only the acm and esp genes (EFAU004_02292 and EFAU004_02750, respectively) have been identified in E. faecium Aus0004.

In the presence of a subinhibitory ciprofloxacin concentration, the acm gene was significantly upregulated by a magnitude of 8.20-fold. On the other hand, the fold change in expression of the genes coding for hemolysin and esp was lower (2.41 and −2.64, respectively) and not considered significant (Tables 1 and S3A). Ciprofloxacin-induced expression of acm was phenotypically confirmed by collagen-binding assays (Fig. 5A). Since Acm was shown to bind collagen types I and IV and to contribute to E. faecium pathogenesis in an experimental model of endocarditis (34), this observation of acm induction by ciprofloxacin in vitro may have clinical consequences. Note that a functional copy of the acm gene is almost exclusively carried by multidrug-resistant E. faecium isolates (39). In contrast, the level of biofilm production was not significantly different in the presence or the absence of ciprofloxacin (Fig. 5B), which is consistent with the fact that the expression of the esp gene, encoding a factor enhancing biofilm formation, was not significantly altered. Finally, the promoter region of acm was determined by RACE-PCR, and no LexA binding site was identified (Fig. S4), suggesting that this gene is not member of the SOS regulon. This is in accordance with the absence of ciprofloxacin concentration-dependent induction of level of acm transcription, as opposed to what was observed with Efmqnr (Fig. 4A).

FIG 5.

(A) Adhesion of the E. faecium Aus0004 strain to collagen in the absence (−Cip) or presence (+Cip) of ciprofloxacin (concentration of 0.38 μg/ml) at different time points (2, 6, and 18 h). Adherence level is expressed as mean log₁₀ values (± standard deviations) of the number of adherent E. faecium cells. Statistical comparison between −Cip and +Cip conditions at each time point was performed using the unpaired t test, and corresponding P values are indicated. (B) Ability of the strains to form biofilm on a polystyrene surface is shown after 24 and 48 h of incubation. Results are expressed as OD₆₀₀ measurements, and median and interquartile range values are shown. Statistical comparison between −Cip and +Cip conditions at each time point was performed using the unpaired t test, and corresponding P values are indicated.

Impact of ciprofloxacin on other important genes in E. faecium Aus0004.

Interestingly, the expression of seven transcriptional regulators was significantly induced in the presence of ciprofloxacin, which should play an important role in the genetic response. It is interesting to note that among the most important differences in expression between the two conditions was the strong induction of the EFAU004_00691 gene (fold change, 23.9) that putatively codes for a MerR-like (metal-responsive) transcriptional regulator (40). The role of this regulatory protein in the response to antimicrobial exposure will be further investigated.

Last, no alteration was found in the expression of other important genes functionally characterized in E. faecium, such as asrR (fold change, 1.2), coding for a major global regulator modulating antimicrobial resistance and pathogenicity of E. faecium (7), or vanB operon genes (fold change of vanB, −1.3) (Table S3A).

In conclusion, the use of the approach RNA-seq allowed us to have a global picture of genes whose transcript levels were significantly influenced by the presence of ciprofloxacin, an antibiotic with well-known effects on bacterial populations of the gut microbiota. Therefore, we showed that the antibiotic SIC induced the expression of both antimicrobial resistance and pathogenicity determinants, namely, Efmqnr and acm.

MATERIALS AND METHODS

Bacterial strains, plasmids, and growth conditions.

The vanB-positive E. faecium Aus0004 reference strain (22) was grown under ambient air at 37°C in brain heart infusion (BHI) broth or on BHI agar (Difco Laboratories, West Molesey, UK). Escherichia coli TOP10 (Life Technologies, Saint-Aubin, France) and E. coli EC1000 (41) were grown under ambient air at 37°C in Luria-Bertani (LB) broth or on LB agar (Difco Laboratories). The plasmids pAT29 (42) and pWS3 (43) were used for cloning and knockout deletion, respectively.

Determination of MIC and SIC values.

MICs of fluoroquinolones (norfloxacin, ofloxacin, ciprofloxacin, levofloxacin, and moxifloxacin) and other antibiotics (ampicillin, gentamicin, vancomycin, teicoplanin, linezolid, daptomycin, and tigecycline) were determined on Mueller-Hinton (MH) agar using Etest strips (bioMérieux, Marcy l'Etoile, France), according to manufacturer's recommendations. Experiments were repeated at least three times.

The growth kinetics of E. faecium Aus0004 was assessed in BHI broth with different concentrations of ciprofloxacin (from 0.06 to 2 mg/liter) under ambient air at 37°C for a 24-h period. Experiments were repeated at least three times. The SIC was defined as the highest concentration of antibiotic for which no significant effect on bacterial growth was observed.

Oligonucleotides, PCR, and DNA sequencing.

All synthesized oligonucleotides were obtained from Sigma-Aldrich, France (Table S1), and confirmatory DNA sequencing was performed by GATC Biotech (Constance, Germany). PCR techniques were performed in accordance with standard protocols, and DNA fragments were purified with a QIAquick PCR purification or QIAquick gel extraction kit (Qiagen, Courtabœuf, France).

RNA isolation.

Bacterial cells of E. faecium Aus0004 were cultured at 37°C to the late-exponential-growth phase in BHI broth, and total RNA was extracted from using the ZR fungal/bacterial RNA miniprep kit (Zymo Research, Irvine, CA). Residual chromosomal DNA was removed by treating samples with the Turbo DNA-free kit (Life Technologies).

For RNA-seq, the Ribo-Zero magnetic kit for Gram-positive bacteria (Epicentre, Madison, WI) was used, according to the manufacturer's recommendations, to remove the 23S, 16S, and 5S rRNAs from the total RNA samples. To evaluate the degree of rRNA depletion, the samples were analyzed using the Agilent 2100 Bioanalyzer.

RT-PCR, quantitative RT-PCR, and 5′-RACE PCR.

For both RT-PCR and quantitative RT-PCR experiments, cDNA was synthesized from total RNA (25 ng) using the QuantiTect reverse transcription kit (Qiagen), according to the manufacturer's instructions. For operon mapping, PCRs were then carried out under standard conditions using specific primers (Table S1). For gene expression, transcript levels were determined by the ΔΔC_T method using the adk gene as a housekeeping control gene (Table S1), and each experiment was done in triplicate.

The transcription start site (TSS) and promoter sequences were determined using the 5′-RACE system kit (Life Technologies SAS) using specific primers (Table S1), according to the manufacturer's instructions.

Transcriptomic analysis by RNA-seq.

Both cDNA synthesis and preparation of the library for high-throughput sequencing were done by the ProfileXpert company (Lyon, France). The strand-specific library preparation was performed using the NEXTflex directional RNA-Seq kit (dUTP-based) (Bioo Scientific, Austin, TX), and the samples were sequenced using the Illumina HiSeq 2500 platform (multiplex protocol, single-end 50-bp reads). Two independent growth experiments were conducted for each condition (i.e., with or without ciprofloxacin).

The reads were mapped against the genomic sequence of E. faecium Aus0004 (GenBank accession no. CP003351) using the CLC Genomics Workbench software version 8.1 (Qiagen). The number of reads uniquely aligned to each genomic position was determined using the BEDTools program version 2.16.2 (44). To verify the reproducibility in duplicate, the expression of genes in different RNA-seq samples was compared using normalized RPKM values calculated as follows: (no. of reads for the gene × 10⁹)/(no. of total of reads × size of the gene). Bioinformatics analysis was performed using the R package DESeq R package (45). Differentially expressed genes were identified using a log₂ absolute fold change greater than 2, and fold changes (FC) of expression were considered statistically significant if the adjusted P value was <0.05 (the adjusted P value was calculated by DESeq using the Benjamin-Hochberg correction). Mean expression and FC values were plotted and visualized as a circle using the Circos program (46). Products of differentially expressed genes were classified by functional category according to the KEGG pathway database.

Construction of deletion mutants and cloning experiments.

Three different deletion mutants (Efmqnr, EFAU004_02295, and operon Efmqnr-EFAU004_02295) were derived from E. faecium Aus0004 by allelic exchange with a truncated copy of the gene using the pWS3 suicide vector and specific primers (Table S1), as previously described (43).

Both Efmqnr and the Efmqnr-EFAU004_02295 operon (with the promoter region) were cloned in the spectinomycin-resistant shuttle vector pAT29 (42) using specific primers (Table S1). The recombinant plasmids were introduced into both electrocompetent E. coli TOP10 and E. faecium Aus0004 strains. The transformants were selected on medium containing 60 mg/liter (E. coli) or 300 mg/liter (E. faecium) spectinomycin.

Collagen-binding assay.

High-binding microtiter plate (Immulon 2 HB; Corning) wells were coated overnight at 4°C with 10 μg/ml collagen (collagen from human fibroblast; Sigma-Aldrich, France) and bovine serum albumin (BSA; Sigma-Aldrich), used as a negative control. Two cultures of E. faecium Aus0004 were grown in BHI broth for overnight with or without ciprofloxacin (0.38 mg/liter) and then were subinoculated under the same conditions to an optical density at 600 nm (OD₆₀₀) of 0.8. Cells were centrifuged and resuspended in 1 ml of phosphate-buffered saline (PBS), and 100 μl of cells (10⁸ CFU/ml) was added to the wells and incubated at 37°C for 2, 6, and 18 h. After washing and scratching, serial dilutions were made to count the cells on BHI agar plates. Each assay was repeated three times in at least three independent experiments. Statistical comparison at each time point (2, 6, and 18 h) was performed using the unpaired t test.

Biofilm formation.

Biofilm formation was measured as previously described (7). Each assay was repeated three times in at least three independent experiments. Statistical comparison conditions at each time point was performed using the unpaired t test.

Accession number(s).

Raw and processed data generated in this study have been submitted to the Gene Expression Omnibus (GEO) repository at the National Center for Biotechnology Information (NCBI) and are available under accession no. GSE94507.