ABSTRACT
A transcriptional profile of Enterococcus faecalis V583 (V583) treated with erythromycin is presented. This is the first study describing a complete transcriptional profile of Enterococcus. E. faecalis is a common and nonvirulent bacterium in many natural environments, but also an important cause of nosocomial infections. We have used a genome-wide microarray based on the genome sequence of V583 to study gene expression in cells exposed to erythromycin. V583 is resistant to relatively high concentrations of erythromycin, but growth is retarded by the treatment. The effect of erythromycin treatment on V583 was studied by a time course experiment; samples were extracted at five time points over a period of 90 min. A drastic change in gene transcription was seen with the erythromycin-treated cells compared to the untreated cells. Altogether, 260 genes were down-regulated at one or more time points, while 340 genes were up-regulated. Genes encoding hypothetical proteins and genes encoding transport and binding proteins were the two most dominating groups of differentially expressed genes. The gene encoding ermB (EFA0007) was expressed, but not differentially, which indicated that other genes are important for the survival and growth maintenance of V583 treated with erythromycin. One of these genes is a putative MsrC-like protein, which was up-regulated at all time points studied. Other specific genes that were found to be up-regulated were genes encoding ABC transporters and two-component regulatory systems, and these may be genes that are important for the specific response of V583 to erythromycin.
Enterococci are commonly found in the gastrointestinal tracts of humans and animals. Most enterococci are not virulent, and some are frequently found in artisan fermented foods, in which they contribute to longer shelf life, flavor, and texture. However, enterococci are best known as antibiotic-resistant opportunistic pathogens that are commonly recovered from patients who have received multiple courses of antibiotics and have been hospitalized for prolonged periods. Innate resistance to many antibiotics is a common trait among enterococci (see, e.g., reference 7 and the references therein). They also acquire antibiotic resistance determinants, including resistance to vancomycin, very rapidly from the environment (see, e.g., references 16-18). The two species best known to be involved in infections in human and animal bodies are Enterococcus faecalis and E. faecium (6).
The complete genome sequence of the vancomycin-resistant E. faecalis V583 (V583) is now available (20). Access to the genome sequence opens new possibilities to gain basic information on the molecular biology of the organism, and one of the tools that can be used to exploit the genome sequence experimentally is the DNA microarray technology. Microarrays give us the opportunity to study all transcriptional events going on in a cell and identify which genes are involved in certain cell processes in one experiment. The main advantage of the microarray technology is the ability to study the transcription of thousands of genes in one experiment. By nature, the microarray technology is explorative and hypothesis generating. The results of microarray experiments are, in principle, snapshots of the transcriptional activities of the cell. Moreover, since transcription and translation are coupled processes in prokaryotes, the transcriptome should reflect the proteome well. Indeed, it has been shown that regulation of the majority of genes parallels the level of proteins produced (2). On a genome-wide scale, it is difficult to speculate about which bacterial genes are regulated or not during certain conditions, and it is also hard to gain such information through the use of traditional low-throughput methods. By the use of microarrays, one can relatively quickly obtain information about gene expression levels and thereby explore the responses of the cells to changing growth conditions.
V583 survives and grows in media containing relatively high levels of the commonly used macrolide antibiotic erythromycin, but the addition of erythromycin to the culture retards cell growth. In sensitive cells, erythromycin inhibits protein synthesis by binding to the large ribosomal subunit close to the peptidyl transfer center. Protein synthesis is thereby aborted during early rounds of translation, since access to the nascent peptide channel is prevented (29). In gram-negative bacteria, intrinsic resistance to erythromycin is due to the impermeability of the cellular outer membrane to this hydrophobic macrolide. Two main mechanisms of erythromycin resistance that have been identified in gram-positive bacteria are as follows: (i) Target modification by the erm (erythromycin resistant methylase) genes that encode enzymes which methylate rRNA has been described. The rRNA methylation causes conformational changes in the P site of the rRNA and prevention of macrolide binding. (erm-mediated resistance in enterococci has been described. The pAD1-like plasmid pTEF1 of V583 encodes ErmB, an rRNA adenine dimethylase family protein.) (ii) A macrolide efflux resistance mechanism, which is an energy-dependent pump, has been described for both gram-positive and gram-negative bacteria (22).
This paper describes the use of a genome-wide amplicon-based microarray based on the genome sequence of V583 to obtain a profile of the transcriptional events in V583 cells treated with erythromycin. The aim of the work was to obtain an overview of how erythromycin affects transcriptional events and bacterial growth, aside from described resistance mechanisms, and how this bacterium tolerates stress. Taking into account the observed phenotypic effects of erythromycin on V583, it was expected that genes involved in general stress responses, genes involved in protein synthesis, and genes encoding (multi)drug resistance would be among the genes affected by the antibiotic treatment.
MATERIALS AND METHODS
Bacterial strain and growth conditions.The bacterium studied was E. faecalis V583 (V583), which was the first vancomycin-resistant clinical isolate reported in the United States (24) and whose genome sequence has been completed (20). V583 samples were grown overnight in flasks with brain heart infusion (BHI) medium (Difco) at 37°C on a rotary shaker (300 rpm). Cultures were then diluted 50× and grown in BHI, as described above, for 1 h. Cultures were split in two, and erythromycin (Sigma) was added to one of the cultures. The final concentration of erythromycin was 50 μg/ml. The two cultures (BHI and BHI plus erythromycin) were then incubated further, and 3-ml samples of each culture were collected immediately after the addition of erythromycin (t0), and after 15 (t15), 30 (t30), 60 (t60), and 90 (t90) min. Growth of the cultures was monitored spectrophotometrically (optical density at 600 nm [OD600]). Samples were centrifuged for 5 min at 4°C, and pellets were flash frozen in liquid N2 prior to RNA extraction.
Construction of V583 microarrays.PCR primers for all V583 open reading frames (The Institute for Genomic Research, May 2002 update) were designed using Primer3 (freeware; http://frodo.wi.mit.edu/cgi-bin/primer3/primer3_www.cgi ), which has commonly been used for primer design during microarray construction (3, 13). Primers were designed to amplify full-length or ∼500-bp amplicons from genomic DNA. PCRs (50-μl reaction volume) were run with 50 mM KCl; 10 mM Tris-HCl (pH 9.0); 0.1% Triton X-100; 2.5 mM MgCl2; 2 μM of each primer; 0.2 mM of each dATP, dCTP, dGTP, and dTTP; 10 to 20 ng genomic DNA; and 1.25 U Taq polymerase (Promega). Amplifications were performed with Perkin-Elmer 9600 thermocyclers with an initial denaturation at 94°C for 3 min and then 35 cycles of denaturation at 94°C for 1 min, annealing at 54 to 56°C for 1 min (the annealing temperature varied depending on the melting temperatures of the primers), and an extension at 74°C for 1 min. The PCR was finalized with incubation at 74°C for 10 min. The quality of the PCR products was assessed by agarose gel electrophoresis. The PCR products were cleaned up using a QIAquick 96 PCR purification kit (QIAGEN). Of a total of 3,337 predicted open reading frames in the V583 genome, 3,160 PCR products were obtained. The purified PCR products were eluted in 3× SSC (1× SSC is 0.15 M NaCl plus 0.015 M sodium citrate)-0.01% sodium dodecyl sulfate (SDS) and printed on Corning UltraGaps slides. All PCR products, representing the predicted V583 open reading frames, were printed in five copies on the slides. In addition, negative controls (three genes from Arabidopsis thaliana, buffer [3× SSC, 0.01% SDS], and empty spots) and digested genomic DNA from V583 (positive control) were spotted on the arrays. Altogether, 3,502 samples were spotted on the slides.
RNA isolation.Total RNA extractions from the samples collected at the time points described above were performed with RNeasy Mini columns (QIAGEN), with DNA digestions being done on the columns by the addition of 82 Kunitz units of RNase-free DNase (QIAGEN) and incubation at room temperature for 15 min. The integrity of and concentration of RNA samples were measured using a RNA 600 Nano LabChip kit and a Bioanalyzer 2100 (Agilent Technologies). For both strains, 10 μg of each RNA sample was used in separate hybridization experiments on identical arrays.
cDNA synthesis, fluorescent labeling, hybridization, and microarray data analysis.For reverse transcription and cDNA synthesis, 30 μg of random hexamers (Amersham) and 10 μg of total RNA was initially preheated at 70°C for 10 min and was incubated for 10 min at 4°C and then for 2 h at 42°C in an Eppendorf Mastercycler. The reverse transcription (RT)-PCR mix was 1× first-strand buffer, 10 mM dithiothreitol; 380 units of Superscript II RT; 500 μM concentrations of dATP, dCTP, and dGTP; 300 μM dTTP (all from Invitrogen), and 200 μM amino-allyl-labeled dUTP (Sigma). After hydrolysis with 10 μl of 1 M NaOH and 10 μl of 0.5 M EDTA for 15 min at 65°C, the samples were neutralized with the addition of 25 μl of 1 M Tris-HCl (pH 7.4) and cleanup was performed with Microcon 30 filters (Millipore). The fluorescent monofunctional N-hydroxysuccinimide-ester dyes cyanine-3 and cyanine-5 (Amersham) were coupled to the cDNAs originating from cultures grown without erythromycin and with erythromycin, respectively, for 1 h, quenched with 1.5 M hydroxylamine (Sigma), mixed, and finally cleaned with a QIAquick PCR purification kit (QIAGEN). The samples (30 μl) were then dried and used for hybridization within 12 h. Hybridizations to the microarrays were conducted as follows. Slides were prehybridized by incubation at 50°C for 30 min in a solution containing 1% bovine serum albumin (Calbiochem), 3.5× SSC, and 0.1% (wt/vol) SDS. The dried fluorescently labeled cDNA samples were resuspended in the following hybridization solution (40 μl): 5× SSC, 0.1% (wt/vol) SDS, 1.0% (wt/vol) bovine serum albumin, 50% (vol/vol) formamide, and 0.01% (wt/vol) single-stranded salmon sperm DNA. The resuspended probes were added to the arrays and incubated, in darkness, at 42°C for 6 h. After hybridization, excess hybridization solution and unspecific bound probe were washed away by four washing steps and gentle shaking, in darkness, for 2 min in 2× SSC-0.1% SDS, 1 min in 1× SSC, 1 min in 0.2× SSC, and 30 seconds in 0.05× SSC. Immediately after washing, the arrays were dried by centrifugation at 600 rpm for 5 min in an Eppendorf 5810R tabletop centrifuge. Three replicate hybridizations were performed with two separate batches of RNA. The two batches of RNA were obtained in two separate growth experiments.
Hybridized arrays were scanned at wavelengths of 532 nm (cyanine-3) and 635 nm (cyanine-5) at a 10-μm resolution to obtain two TIFF images with a ScanArrayExpress Microarray scanner (Packard Bioscience). Fluorescent intensities and spot morphologies were analyzed using the QuantArray program ver. 3.0 (Packard BioScience), and spots were excluded based on slide or morphology abnormalities.
Raw data from each array was preprocessed independently. A Lowess smoothed background was subtracted from all foreground intensities, and a cross-validated Lowess method was used in an intensity-dependent normalization of every array. The log2 ratios for each spot were further analyzed using a mixed model (30) to detect differentially expressed genes. A mixed model was fitted to the data for each of the five sample times (0, 15, 30, 60, and 90 min) separately. Data for the three arrays at every sample time were described by yijk = μi + uij + eijk where yijk is the observed log2 ratio of gene i (1, …, 3,502) on array j (1, 2, 3) and in spot k (1, …, 5) on that array, μi is the expected log2 ratio for gene i, uij is a random effect of gene i on array j, and eijk is the remaining noise. The variance components were estimated under the assumption of Gaussian errors using a restricted maximum likelihood approach coping with the unbalanced data due to missing spots. Differentially expressed genes were identified by testing the hypothesis H0, defined as μi= 0, against H1, defined as μi ≠ 0. A chi-square test for every gene resolves this for the model in the yijk equation (4), and a Bonferroni-corrected rejection level of a P value of <0.01 was used throughout. If H0 (μI = 0) was rejected, and μi is >0, genes were considered to be up-regulated in the erythromycin-treated cells. If H0 was rejected, and μi is <0, genes were considered to be down-regulated. All data analysis algorithms were programmed by using Matlab (MathWorks, Inc.), but a subset of the data was also analyzed by the SAS system (SAS Institute, Inc.) to check the validity of the code.
Confirmation of expression levels of specific genes by real-time RT-PCR.To confirm independently the differential gene expression observed by microarray experiments, four genes were selected for analysis by real-time quantitative RT-PCR (RTQ). Primers and probes for the RTQ were designed using the Assays-By-Design file builder (ver. 2.0; http://www.appliedbiosystems.com/support/software/assaysbydesign/installs.cfm?prod_id=1541 ; Applied Biosystems). Primers and probes were synthesized by and purchased from Applied Biosystems. The genes selected for RTQ analyses were EF0633 (tryS-1, encoding tyrosyl-tRNA synthetases), EF2653 (encoding a transcriptional regulator of the Cro/CI family), and EF0105 (argF-1, encoding ornithine carbamoyltransferase). EF1964 (gap-2, encoding glyceraldehyde-3-phosphate dehydrogenase), which is constitutively expressed, was used to normalize the TaqMan data. The RTQ analyses were run on an ABI PRISM 7700 sequence detector (Applied Biosystems). cDNA was synthesized as above, with 15 ng RNA as template. Real-time PCR was performed using the TaqMan Universal PCR Master Mix (Applied Biosystems), according to the manufacturer's instructions. To ensure that the cDNA was not contaminated by genomic DNA, reactions without reverse transcriptase were also included. Differential expression was determined by calculating the change in threshold cycle (ΔCt) for each gene, with RNA isolated from cells grown with and without erythromycin and harvested at the five time points mentioned above.
RESULTS
We have studied the transcriptional responses of E. faecalis V583 to treatment with the macrolide antibiotic erythromycin through a time course experiment utilizing a genome-wide microarray based on the V583 genome sequence. By applying a time course experiment in this manner, we were able to investigate both an immediate response to this stress and the cells' adaptation to a more permanent presence of erythromycin. In the following, we focus on genes that were differentially expressed in the untreated and the erythromycin-treated cells.
We applied a stringent confidence level (a P value of <0.01 plus Bonferroni correction for multiple comparisons) for the determination of significant differential transcription. Genes were “scored” as significantly up- or down-regulated as described above, with a threshold P value of <0.01 and the conservative Bonferroni correction for multiple comparisons, ensuring a very low number of false negatives.
Circa 600 (18%) of the predicted V583 genes were found to be differentially transcribed at one or more of the five time points examined during the 90-min time course experiment. Obviously, the erythromycin exposure seriously affected the transcriptional events in these enterococci: 260 genes were found to be significantly up-regulated (induced; log2 ratios of stressed cells/nonstressed cells were different from and higher than 0) and 340 genes down-regulated (repressed; log2 ratios of stressed cells/nonstressed cells were different from and lower than 0) at one or more time points. Among the up-regulated genes, four were plasmid encoded, while 27 plasmid-encoded genes were down-regulated. The numbers of differentially expressed genes, sorted by their cellular roles, are shown in Table 1. The total number of V583 genes represented on the array was 3,160 (out of total of 3,337 genes in the V583 genome), which means that ca. 8% of all V583 genes were up-regulated and ca. 10% down-regulated in the erythromycin-exposed cells. This shows that transcriptional events in these enterococcal cells are strongly altered by erythromycin exposure. During the 90-min time course, the number of down-regulated genes is higher than the number of induced genes at all time points except t90 (the last time point) (Tables 2 and 3), which indicates a general decrease in transcriptional activity by exposure to erythromycin.
Number of genes that were differentially expressed at one or more time points, sorted by cellular rolea
Number of up-regulated genes in each functional category at all time points studied
Number of down-regulated genes in each functional category at all time points studied
Growth of erythromycin-treated E. faecalis V583.Cell samples were collected at five time points after the addition of erythromycin (50 μg/ml) to one of the cultures. We chose this level of erythromycin in the cultures on the basis of preliminary experiments (results not shown), where we found that growth of V583 was retarded but still significant. (V583 does grow in medium with >400 μg erythromycin/ml.) Our time course experiment was performed over a period of 90 min, and in this first part of the growth, the erythromycin-treated cells grow more slowly than the untreated cells. Thus, this work describes the response of V583 to erythromycin during the first cycles of growth. We chose a relatively low level of erythromycin for the experiments, according to the observation by Hutter et al. (11) that lower concentrations are better for obtaining good microarray results. During the experiments, growth was measured spectrophotometrically. After 15 min, the OD of the erythromycin-treated culture was approximately 82% of the OD of the untreated (control) culture; after 30 min, the OD of the erythromycin-treated culture was approximately 69% of that of the untreated culture; after 60 min, the OD of the erythromycin-treated culture was approximately 54% of that of the untreated culture; and after 90 min, the OD of the treated culture was 40% of the OD of the untreated culture. The growth curves of erythromycin-treated and untreated V583 are shown in Fig. 1.
Growth of E. faecalis V583 treated with erythromycin (50 μg/ml; solid line) and of untreated E. faecalis V583 (broken line). Growth was measured spectrophotometrically at a λ of 600 nm. Arrows indicate time points at which samples were collected for RNA extraction and microarray experiments.
EFA0007, the ermB gene.EFA0007, which was referred to as an rRNA adenine dimethylase family protein encoded on pTEF1 (20) is, in fact, ErmB (>99% identity with ErmB from Bacillus cereus, Streptococcus pyogenes, E. faecium [BlastP]). This gene was expressed in cultures both with and without erythromycin, but the gene was not differentially expressed at any time point. This is in accordance with earlier observations that the ermB is constitutively expressed in enterococci (B. Murray, unpublished results). The retarded cell growth in the presence of erythromycin confirms that EFA0007 does not fully protect the bacterium against this antimicrobial compound and suggests that additional genes are involved in the cells' protection against this stress factor. Alternatively, the apparent lack of differential expression may be explained by posttranscriptional regulation of this gene; ermB may be posttranscriptionally regulated by a stalling on the ribosome during translation.
Up-regulated (induced) genes.The number of up-regulated genes in each functional category at various time points is shown in Table 2. Among the up-regulated genes, the genes encoding hypothetical proteins represented the highest percentage (85 genes; around 32% of all up-regulated genes), while genes related to transport and binding functions represented around 17% (44 genes) of the genes induced by the erythromycin stress. Moreover, genes encoding proteins involved in energy metabolism (10 genes), protein synthesis (13 genes), synthesis of cell envelope components (12 genes), and regulatory functions (13 genes) were up-regulated. The majority of up-regulated genes related to protein synthesis are genes encoding ribosomal proteins. In addition, genes classified in the category of genes of unknown function represent a considerable part of the induced genes (17 genes).
Among the up-regulated genes, three genes encoded drug resistance proteins. These genes were EF0420 (up-regulated at t30 and t60), EF1370 (up-regulated at t30, t60, and t90), and EF1078 (up-regulated at t15, t30, t60, and t90). Four genes only were induced at all time points. These genes were EF1400 (predicted to encode a cadmium-translocating ATPase), EF1413,EF1916 (encoding a GTP-binding protein), and EF2720 (encoding an ABC transporter). The gene encoding a putative MsrC-like protein (EF1413) is of particular interest. MsrC is widely spread among Enterococcus faecium, and MsrC is believed to be an efflux pump involved in low-level macrolide resistance in this species (23, 25). However, the EF1413 is only 40% identical (60% similar) to the E. faecium MsrC, and it is therefore not clear whether the EF1413 encodes a functional MsrC-like protein. Table 4 gives the full list of genes found to be induced (up-regulated), with log2 ratios. Genes predicted to encode hypothetical proteins and plasmid encoded genes are excluded from this list.
Log2 ratios of genes whose transcription increased at one or more of the five time points (altogether 90 min) at which the effect of erythromycin treatment was studied, sorted by functional category (cellular role)
Down-regulated (repressed) genes.The numbers of down-regulated genes in each functional category at various time points are shown in Table 3. Among the down-regulated genes, those encoding hypothetical proteins are the dominant category (115 genes; 35%), followed by genes encoding transport and binding proteins (38 genes; 11%), genes with unknown function (23 genes), genes encoding cell envelope-related proteins (24 genes), and genes involved in energy metabolism (29 genes), regulatory function (13 genes), protein synthesis (14 genes), and cellular processes (23 genes). The majority of down-regulated genes related to protein synthesis are eight tRNA synthetases (cysteinyl-tRNA synthetase, tyrosyl-tRNA synthetase, alanyl-tRNA synthetase, aspartyl-tRNA synthetase,tryptophanyl-tRNA synthetase [EF2228], threonyl-tRNA synthetase, valyl-tRNA synthetase, and seryl-tRNA synthetase). The tryptophanyl-tRNA synthetase gene encoded by EF2679 (trpS) was significantly up-regulated at all time points except t0. Transcription was found to be repressed at all five time points examined for 10 genes. Table 5 lists all of the down-regulated genes, with log2 ratios. Genes encoding hypothetical proteins and plasmid genes are not included.
Log2 ratios of genes whose transcription decreased at one or more of the five time points (altogether 90 min) at which the effect of erythromycin treatment was studied, sorted by functional categorya
Validation of microarray data by real-time quantitative RT-PCR analyses of selected genes.Verifications of the expression levels of four genes observed by microarrays were performed by real-time quantitative RT-PCR analyses (TaqMan assays). The genes selected for these RTQ analyses represented various functional categories and different patterns of expression. The transcription profiles of the genes examined by RTQ correlated well with the transcription profiles observed by the use of microarrays (r = 0.83).
DISCUSSION
The present study focuses on functional genomics of Enterococcus faecalis V583, applying a genome-wide microarray based on the genome sequence of V583. In general, the knowledge of gene expression in enterococci is scarce. This is the first paper describing a global transcriptional profile of Enterococcus, and more studies are definitely needed to obtain a deeper understanding of gene expression in this species. We used a microarray to study the transcriptional responses of V583 to the macrolide antibiotic erythromycin. V583 is resistant to relatively high levels of erythromycin, but exposure to erythromycin retards its growth. The work was performed to (i) gain insight into the mechanisms involved in resistance to erythromycin by V583, (ii) obtain clues about transcriptional responses to general stress in bacteria, and (iii) generate hypotheses for subsequently gaining deeper knowledge on antibiotic resistance and general stress in this bacterium.
In the experiments, V583 was treated with erythromycin (50 μg/ml), which decreased the growth of the cells significantly compared to the untreated control during the 90-min time course experiment. The concentration of antibiotic to apply in such experiments is, of course, a matter of discussion. Hutter et al. (11) claimed that compound concentrations should be at concentrations that are just low enough not to affect the growth of the organism. This probably varies with the strains and compounds being studied.
We chose a stringent confidence level plus Bonferroni correction to score for significant differential expression, in the data analysis. Consequently, the number of false positives is, per definition, very low. We preferred this method, although some information might have been lost. Moreover, we decided to score for significant up- and down-regulation instead of using the common fold change cutoff values. Thus, the conclusions are based on experimental data precision (noise levels and number of replicates); this is appropriate since microarrays probably underestimate actual mRNA induction ratios (2). Also, treating each array as a random sample from a population of all possible arrays makes all conclusions more general. Even our simple mixed model gives a quite accurate description of the variance structure of normalized microarray data from repeated experiments and allows for an extended information extraction from every experiment. Our analysis differs from that by Wernisch et al. (30) in that we have excluded the array main effect. Our argument for this is that all arrays have been normalized separately, and thus any array main effects will be ignorable, which is also the conclusion drawn by Wernisch et al. (30) in analyzing similar data. Another variant of our mixed model approach is taken by Wolfinger et al. (31). They use one variance component per gene, which makes the model much more complex. More research must be done in this area of bioinformatics before we can come to a conclusion as to which strategy is appropriate. For now, we have chosen to use a simple model, in line with the principle of parsimony.
The target of erythromycin and other macrolides is protein synthesis, more specifically the 50S (large) ribosomal subunit. These antibiotics block the ribosome exit tunnel, which prevents the movement and release of the nascent peptide chain. The ermB gene (EFA0007) was not found to be significantly up- or down-regulated at any time point. Hence, this gene is probably constitutively expressed and contributes to the erythromycin resistance of V583. However, based on the high level of differential transcription and slow growth observed for erythromycin-exposed cells compared to untreated cells, it appears that a battery of other genes respond expressionally to the presence of erythromycin. These responses are not necessarily all due to resistance, but rather an adaptation to changes in growth conditions. Thus, these genes contribute to maintenance of the growth during erythromycin exposure. Three genes encoding drug resistance proteins (EF0420, EF1078, and EF1370), on the other hand, appear to play a role for the ability of V583 to survive erythromycin treatment. Their influence must, however, be limited, since the growth of V583 is so retarded by erythromycin.
It was expected that the exposure of V583 to erythromycin would affect genes involved in protein synthesis, despite the expression of ermB. The induction of eight genes encoding ribosomal proteins (r-proteins), along with two r-protein-Ala-acetyltransferases, might be an indication that the overproduction of ribosomal proteins is one way to evade the effect of erythromycin. To balance the number and function of the ribosomes, the cells may have to compensate by up-regulation of these genes. Five of the induced r-protein genes and one of the r-protein-Ala-acetyltransferase genes are induced after 15 min of growth, i.e., their induction is a quick response. Closely related to the known target of erythromycin in sensitive cells is the fact that eight tRNA synthetase genes are repressed. The repression of tRNA synthetase genes may be seen as a logical consequence of the reduced ability of a cell's capacity to synthesize proteins, when an antibiotic binds a ribosomal subunit. The up-regulation of genes encoding r-proteins and two translation elongation factors, as well as down-regulation of tRNA synthetase genes, in response to translation inhibitors, has been noticed also by other authors (5, 19, 28).
Although the V583 cells are defined as resistant to erythromycin and will survive in the presence of relatively high levels of erythromycin, the cell growth is impaired, and, as shown above, the transcriptional activities are greatly altered. The differential transcription reflects the effect of erythromycin as well as the adaptation of V583 to the general stress. The retarded growth follows as a consequence of a slow-down of many cellular activities when important processes, such as protein synthesis, apparently are repressed. In our experiments, we found that a lower number of genes was induced than repressed at all time points except for t90 (see Tables 1 to 5). At t90, it is tempting to look upon the altered transcriptional activity in the stressed cells as a consequence of an adaptation to the presence of erythromycin. At this point, the cells may have adapted to the stressing conditions and may have overcome the critical effects of the drug.
Ng et al. (19) report the induction of several genes related to purine biosynthesis in Streptococcus pneumoniae R6 exposed to various translation inhibitors, including erythromycin. This effect on purine synthesis is not clear for V583; only purK is up-regulated, and purA is down-regulated. Three genes related to pyrimidine biosynthesis (pyrD-1, pyrD-2, and pyrC) are induced, however. Similarly to what was found with S. pneumoniae R6 (19), the purine salvage gene xpt had a relative increase in transcription in V583 in response to the erythromycin treatment. Therefore, it looks as if the responses of the two relatively closely related bacteria E. faecalis and S. pneumoniae to erythromycin have common themes, but most of the transcriptional responses appear distinct.
Transport and binding proteins represent the second most dominant group of differentially expressed genes (Tables 1 to 5). This group of genes appears to be heavily influenced by various stressors in many bacteria (see, e.g., references 19, 21, and 26). The altered transcription of transport and binding proteins indicates that altered transport is an important part of general stress response mechanisms. The large number of differentially expressed genes encoding ABC transporters and permeases emphasizes the importance of such genes in stress responses.
Another interesting group of induced genes are those related to fatty acid and phospholipid metabolism (six genes). As mentioned briefly above, erythromycin resistance in gram-negative bacteria is usually mediated by a low permeability of the outer membrane to the hydrophobic macrolide. One may speculate as to whether a slightly altered cell membrane may have contributed also to the erythromycin adaptation in the gram-positive V583 cells.
In time course experiments like ours, one should draw attention to genes that are differentially expressed at all time points examined. Our results showed that 14 genes were regulated, 4 up and 10 down, at all five time points. Among these, the most interesting gene is EF1413, which encodes a putative MsrC protein, although the sequence identity with the described MsrC is not more than ca. 40%. EF1413 was up-regulated at all time points and showed the strongest differential expression of all genes on the microarray (Table 4). EF1413 was not differentially expressed in experiments with chloramphenicol treatment of V583 (Å. Aakra, unpublished results). The msrC is distributed in many isolates of E. faecium and has even been suggested to be specific for this species (23, 25). Comparative genomic hybridizations with five E. faecalis strains indicate the presence of the putative msrC in these genomes, as well (Aakra, unpublished). Singh et al. (25) found that the msrC conferred an increased resistance to macrolides among E. faecium isolates. Our observation of strong induction of EF1413 (log2 ratio of 1.88 at t0; log2 ratio of 3.44 at t90) in erythromycin exposed E. faecalis V583 support the involvement of an MrsC homologue in macrolide resistance in this species as well.
EF1732 and EF1733 encode two ABC transporters belonging to the MDR family. These two genes were up-regulated at all time points except t0. (In a similar study, the same two genes were also strongly up-regulated in response to chloramphenicol treatment [Aakra, unpublished].) Taken together, these results make it tempting to speculate that the efflux of erythromycin by proteins encoded by EF1413, EF1732, and EF1733 is an important part of the survival mechanism for V583 when exposed to this antibiotic.
At t0, i.e., immediately after the addition of erythromycin, 18 genes were up-regulated and 56 genes were down-regulated, which were considerably fewer genes than those regulated at the other time points under study. It must be assumed that the genes that show an immediate response also are crucial for the growth and survival of erythromycin-exposed V583 cells.
Several studies on stress responses (and genes involved therein) of E. faecalis have been published (e.g., see references 1, 8, 9, 14, and 15). Some papers focus on two-component signal transduction pathways, which commonly are related to bacterial stress (10, 15). A few of the genes found to have a significantly differential expression in our work have been discussed in these papers, e.g., genes encoding the Gls24 proteins (EF0079 and EF0080), two histidine kinases and their cognate response regulators (EF3290 and EF3289, EF1051 and EF1050), l-lactate dehydrogenase (EF0641), l-serine dehydratase (EF0098, EF0099), carbamate kinase (EF0106), and superoxide dismutase (EF0463). In this work, we found that the gls24 genes (EF0079, EF0080) were repressed at all time points except t0, which is in contrast with the hypotheses of Giard et al. (9) that these genes are induced under stress. The repression of the gls24 genes was also found in our study of the transcriptional responses of V583 to chloramphenicol treatment (Aakra, unpublished). Similarly to Giard et al. (9), the EF0604 which encodes another Gls24 protein was not found to be differentially expressed at any time point of our study. Therefore, the role of Gls24 in stress responses in E. faecalis is not obvious. Regarding the two-component signal transduction pathways, the histidine kinase (HK) EF3290 and the cognate response regulator (RR) EF3289 were both down-regulated at t90. The function of this HK-RR pair is not known (10), and it remains to be seen whether the repression of EF3290/EF3289 at t90 in this study is stress related. The other pair of differentially expressed HK-RR (EF1051/EF1050) is similar to known HK-RR systems in Listeria monocytogenes and S. pneumoniae (10). EF1051 and EF1050 were both induced at t30. Recently, the EF1051/EF1050 pair was described by Teng et al. (27), who named this pair of genes etaRS (enterococcal two-component system a). Teng et al. (27) showed the involvement of EtaRS in both stress responses and virulence, and our results support their hypothesis on the function of etaRS. EF0641 encoding l-lactate dehydrogenase (ldh) was down-regulated at t60 and t90. Previously, it was shown that ldh was induced under stress (8), and it was speculated that also this gene could be involved in stress responses (8), but the function is unclear. The repression of ldh in this study indicates that the possible involvement of this gene in stress is not general. Likewise, sodA (encoding superoxide dismutase), arcC (encoding carbamate kinase), and sdhA and sdhB (encoding l-serine dehydratase) were down-regulated in our work, while they were induced in the study by Giard et al. (8). Based on these comparisons, it appears that a considerable number of the genes affected by environmental stress in E. faecalis are not general stress genes but, rather, are specified by certain conditions.
Interestingly, EF1078, which was induced at all time points except t0, is identical to emeA (enterococcal multidrug resistance efflux), which was characterized by Jonas et al. (12). emeA is probably a homolog of the norA found in Staphylococcus aureus (12). Jonas et al. (12), showed that emeA is involved in resistance to many toxic compounds, among them erythromycin (12). Thus, our study also supports the results of Jonas et al. on the function of emeA.
We have presented the transcriptional profile of erythromycin-exposed E. faecalis V583. The addition of erythromycin to this bacterium causes numerous events of differential transcription. Efflux of the macrolide molecules from the cells may be an important part of the survival mechanism, in addition to the resistance conferred by ErmB. This work adds information to the growing archive of condition-specific bacterial transcription signatures, which, in the future, will aid the elucidation of microbial physiology, metabolism, ecology, pathogenesis, etc. With a comprehensive archive of such transcription signatures, it will also be possible to pay more attention to all genes encoding hypothetical proteins or putative genes. This group represents a significant part of prokaryotic genomes and a dominant group of regulated genes in many microarray studies published so far. Profiling of transcriptional events in V583 under other stress conditions is in progress. Hopefully, this will lead to a deeper understanding of the biology of this bacterium.
ACKNOWLEDGMENTS
This work was financially supported by grant 142656/140 from the Research Council of Norway to I.F.N.
We thank Linda Godager and Merete Lunde for assistance with the RTQ analyses. Bjørn E. Kristiansen, of the Norwegian Microarray Consortium, Oslo, is acknowledged for printing of the microarray slides.
FOOTNOTES
- Received 1 November 2004.
- Returned for modification 16 January 2005.
- Accepted 27 February 2005.
- Copyright © 2005 American Society for Microbiology