|Home | About | Journals | Submit | Contact Us | Français|
We investigated the influence of a low concentration of the gyrase inhibitor norfloxacin on the transcriptome of enterohemorrhagic Escherichia coli O157:H7 strain EDL933. For this purpose, we used a commercial DNA microarray containing oligonucleotides specific for E. coli O157:H7 strains EDL933 and RIMD0509952 and E. coli K-12 strain MG1655. Under the conditions applied, 5,963 spots (94% of all spots) could be analyzed. Among these, 118 spots (P < 0.05) indicated transcriptional upregulation and 122 spots (P < 0.05) indicated transcriptional downregulation of the E. coli genes present on the array. Eighty-five upregulated EDL933 genes were phage borne. Fifty-two of them could be ascribed to the Shiga toxin-encoding phages (Stx phages) BP-933W and CP-933V; the other 33 genes belonged to non-Stx prophage elements in the EDL933 genome. Genes present in the BP-933W prophage genome were induced most strongly up to 158-fold in the case of stxA2 upon induction with norfloxacin. Twenty-two additional upregulated genes appeared to be E. coli O157:H7 strain RIMD0509952-specific phage elements, and the remaining 11 genes were related mainly to recombination and stress functions. Downregulation was indicated predominantly for genes responsible for bacterial primary metabolism, such as energy production, cell division, and amino acid biosynthesis. Interestingly, some genes present in the locus of enterocyte effacement appeared to be downregulated. The results of the study have shown that a low concentration of norfloxacin has profound effects on the transcriptome of E. coli O157:H7.
Analysis of whole genomes rates high among the techniques used to characterize bacterial pathogens. At present, 185 bacterial genomes have been completely sequenced and are available at the National Center for Biotechnology Information (NCBI), Bethesda, Md., under http://www.ncbi.nlm.nih.gov/genomes/MICROBES/Complete.html. Research with whole genomes is an effective tool for obtaining a global overview about complex processes such as microbial pathogenesis (5, 9, 22, 45). Moreover, genome-wide expression profiling with DNA microarrays is a powerful method for the study of the transcriptome of a given organism after treatment with environmental stimuli (7).
In 1997, the complete genome sequence of Escherichia coli K-12 strain MG1655, which consists of 4,639 Mb, was published (4). Four years later, the complete genomes of two serious bacterial pathogens, enterohemorrhagic E. coli (EHEC) O157:H7 strains EDL933 and RIMD0509952, were sequenced; and genome analysis revealed surprising results (15, 35). In comparison to E. coli K-12, their genomes were approximately 880 and 860 kb larger, respectively. For example, E. coli O157:H7 strain EDL933 contains 1,387 genes which are not present in E. coli K-12 (35). E. coli K-12 and EDL933 share a common backbone of 4.1 Mb which is predominantly colinear. The similarity between both chromosomes is discontinued by a number of so-called K islands (present in K-12 strain M1655) and O islands (present in EDL933). A total of 177 O islands ranging in size from 50 to 88 kb and comprising a total of 1.34 Mb and 234 K islands of >50 bp comprising 0.53 Mb have been detected (35). Among these O islands, 18 prophages and prophage-like elements have been identified in the EDL933 genome, including the Shiga toxin 1 (Stx1)-encoding prophage CP-933V and Stx2-encoding prophage BP-933W. Only phage BP-933W has yet been shown to produce infectious phage particles. The other phage sequences are mainly incomplete and cryptic (35). Hayashi et al. (15) also identified 18 prophages and 6 prophage-like elements in the chromosome of E. coli O157:H7 strain RIMD0509952. Phages represent a very mobile group of genetic elements which integrate into the bacterial genome at specific sites, the most prominent of which were found to be close to tRNA genes (6). Since prophage sequences comprise approximately 5% of the E. coli O157 genome, there is evidence that phages may play a role on genomic variability in E. coli (35).
From its first description as a food-borne pathogen in 1983 (40), E. coli O157:H7 has emerged as an important cause of gastrointestinal disease that has been detected throughout the world. E. coli O157:H7 is the prototype of EHEC, a subgroup of Stx-producing Escherichia coli (STEC) causing hemorrhagic colitis and the hemolytic-uremic syndrome (HUS). EHEC and STEC strains can be found among a large number of E. coli O serotypes and represent a genetically heterogeneous group of pathogens. STEC strains are now among the strains at the top of the list of bacterial causes of food-borne infections in the United States and Germany (26, 41).
Stx is the major pathogenicity factor of EHEC and is responsible for multiple damaging effects on eukaryotic cells, and the most convincing evidence that has been provided indicates that it is responsible for the clinical symptoms of bloody diarrhea and HUS (27, 38, 49).
The production of Stx by Stx-producing E. coli strains can be modulated by antibiotics (11, 18, 20, 54, 57). Karch et al. (18) have shown that subinhibitory concentrations of trimethoprim-sulfamethoxazole increased the yield of Stx produced by Shigella dysenteriae and STEC strains. In another study, E. coli O157:H7 strains expressing either one or two Stxs were treated with 13 antibiotics. It could be shown that Stx production was modulated differently in a strain-specific manner (11). Zhang et al. (57) demonstrated that ciprofloxacin induced Stx and Stx-encoding phages (Stx phages) in a mouse model, and Matsushiro et al. (25) induced Stx phages with norfloxacin. In a study by Köhler et al. (21), the simultaneous induction of Stx and Stx phages was triggered by particular growth promoters used in animal husbandry. In contrast, azithromycin showed strong in vitro bactericidal activity without inducing Stx (32).
The general location of the stx gene in the late-phase prophage region and the recA dependency of Stx induction led several investigators to suggest that Stx production is linked to the phage growth cycle. This was proven on a molecular level by Neely and Friedman (28) and Wagner and colleagues (51, 52). These findings shed new light on the controversial discussion about antibiotic treatment for HUS. Whereas some studies have reported on the beneficial effects of early antibiotic therapy, another study (21) showed that antibiotics trigger the progression to HUS.
The effects of antibiotics have hitherto been investigated only on individual Stx phages. In order to study the impact of prophages and prophage genes on the physiology of EDL933 at the transcriptional level, we investigated the influence of a low concentration of an antibiotic substance on the whole transcriptome of E. coli O157 by DNA microarray analysis, with particular emphasis on the expression of prophage genes. We have chosen norfloxacin as a model substance since it is a representative of the group of gyrase inhibitors, it is effective against Stx production, and it is used for the treatment of bacterial infections.
Testing of the susceptibility of E. coli O157:H7 strain EDL933 (31) to norfloxacin was performed by a broth macrodilution method in test tubes according to standard DIN 58940. Mueller-Hinton broth (MHB) served as the culture medium for determination of the MIC. A stock solution of 32 mg of norfloxacin per ml in glacial acetic acid was diluted to a working solution of 32 μg/ml in MHB and then transferred to test tubes, each of which contained 1 ml of MHB, to obtain serial dilutions ranging from 16 to 0.015 μg/ml. One milliliter of MHB containing 106 CFU of EDL933 was finally added, and the test tubes were incubated at 37°C for 16 to 18 h. MICs were interpreted visually. The MIC test was performed in parallel with the same amount of glacial acetic acid without norfloxacin to investigate its influence on the MIC.
E. coli O157:H7 strain EDL933 was grown in five Erlenmeyer flasks to an optical density at 600 nm (OD600) of 0.3, as described above. Norfloxacin was added to four of the cultures to obtain final concentrations of 100, 200, 500, and 1,000 ng/ml. The fifth culture was used as a control and did not receive norfloxacin. The ODs were measured at 30-min intervals. Samples of 5 ml were taken 1 and 2 h after addition of norfloxacin. Following centrifugation at 3,740 × g for 5 min at 4°C, the supernatants were diluted in bacterial dilution buffer (Alexon Trend) and Stx production was measured by a ProspectT Shiga Toxin E. coli microplate assay (Alexon Trend), according to the recommendations of the manufacturer.
To analyze the time course of induction in more detail, tests for ODs, viable cell counts, Stx production, and plaque formation were performed in parallel with 200 ng of norfloxacin per ml. Viable cell counts were calculated after serial dilution in phosphate-buffered saline and plating on Luria-Bertani (LB) agar. Stx production was determined as described above. Concentrations of mature phage particles were determined by a double-layer plaque test, as described previously (43).
EHEC O157:H7 strain EDL933 was grown overnight at 180 rpm in a rotary shaker at 37°C in 10 ml of LB broth (pH 7.5). Six 500-ml Erlenmeyer flasks, each of which contained 100 ml of LB broth, were inoculated with an overnight culture to an initial OD600 of 0.05. The bacteria were grown at 37°C at 180 rpm to an OD600 of 0.3. Subsequently, 0.8 μl of a 25-mg/ml norfloxacin (ICN) stock solution, prepared in glacial acetic acid, was added to three of the cultures (experimental cultures), giving a final concentration of 200 ng/ml. This amount of glacial acetic acid did not alter the pH of the medium. The other three cultures served as controls and were not supplemented with antibiotics. All bacterial suspensions (both experimental and control suspensions) were further incubated for 120 min at 37°C. RNA isolation was then performed at this time.
Total bacterial RNA was isolated from 50 ml of the EDL933 cultures on three different days by using RNeasy kits (Qiagen), according to the recommendations of the manufacturer, with minor modifications. Briefly, the bacteria were centrifuged at 3,740 × g for 5 min at 4°C. The pellets were resuspended in 1 ml of fresh lysis buffer (RLT) included in the RNeasy kit. The suspensions were transferred to Biopulverizer FastRNA-Blue tubes (Qbiogen) and mechanically treated in a Fast Prep instrument (Qbiogen) for 45 s at a speed setting of 6.5 to obtain crude extracts. After an additional centrifugation at 13,000 rpm for 3 min in a benchtop centrifuge (Eppendorf), the supernatants were processed as described in the RNeasy kit manual. To remove the DNA, we digested the sample with DNase I (Roche) and purified the RNA using the same kit. Elution of RNA from the columns was performed with diethyl pyrocarbonate (DEPC)-treated distilled water. DEPC-treated distilled water were prepared as described in the Bacteria MWG-Application Guide (http://ecom2.mwgdna.com/download/arrays/manuals/manual_bacteria.pdf). The RNA preparations from the three experiments and from the three controls were pooled. Total RNA was quantified by measuring the absorbance at 260 nm, and the quality and purity were confirmed by electrophoresis on 1.2% agarose gels containing 1.9% (vol/vol) formaldehyde. Morpholinopropanesulfonic acid (Sigma-Aldrich) was used as the electrophoresis buffer.
Hybridization probes were prepared by incorporation of cyanine 3 (Cy3)-dCTP and Cy5-dCTP (NEN) into the cDNA during reverse transcription of the experimental and the control RNA, as described in the Bacteria MWG-Application Guide. For trials 1 and 2, the experimental and the control RNAs were also reverse labeled in order to control for differences in the labeling efficiencies between the fluorescence dyes.
Briefly, 6 μl of RNA solution containing 50 μg of total RNA in a volume of 6 μl DEPC-treated distilled water and 3 μl of random primer solution (3 μg/μl; Invitrogen) were mixed, incubated at 65°C for 10 min, subsequently incubated for 10 min at room temperature, and finally cooled on ice for 2 min. Two microliters of a reaction mixture containing 5 mM each dATP, dGTP, and dTTP and 2 mM dCTP (all from Invitrogen), 4 μl of fivefold-concentrated reaction buffer (Invitrogen), 2 μl of 100 mM dithiothreitol (Invitrogen), 2 μl of 1 mM Cy3-dCTP or 2 μl of 1 mM Cy5-dCTP, and 1 μl of Superscript II H− (10 U; Invitrogen) was added to the mixture; and the mixture was incubated for 2 h at 42°C. To terminate the reaction and degrade the RNA, 5 μl of 1 N NaOH was added to the sample and the mixture was incubated at 65°C for 10 min. For neutralization, we added 5 μl of 1 N HCl and 200 μl of TE (Tris-EDTA) buffer (pH 7.5) to the mixture.
Labeled cDNA was then purified with a PCR purification kit (Qiagen) in accordance with the guidelines in the kit manual to remove the deoxynucleoside triphosphates that had not been incorporated, random primers, and fluorescent dyes. cDNA was eluted twice with 40 μl of nuclease-free water (Braun).
A total of 3,000 ng of Cy3-labeled cDNA from the experimental sample (Cy3 cDNAE) and 3,000 ng of Cy5-labeled cDNA from the control sample (Cy5 cDNAC), and vice versa for reverse labeling, were combined in one tube and dried in a Speed Vac apparatus (Heto Lab). The labeled cDNA was resolved in 250 μl of salt-based hybridization buffer (MWG Biotech AG).
The labeled cDNA was hybridized overnight with the oligonucleotides present on an E. coli O157 array (MWG Biotech AG) in a hybridization chamber (Corning). The slides were then washed three times and dried under standard conditions in accordance with the protocol of the manufacturer. For optimal utilization of spot intensities, the arrays were scanned at three different amplification settings (photomultiplier tube values of 35, 45, and 55) by using an Affymetrix 428 scanner.
After the slides were scanned, crude data from each slide were analyzed with BioDiscovery ImaGene (version 5.0) image-processing software (MWG Biotech AG). For optimal spot analysis, a grid pattern was adjusted for each slide, and awry spots (messed or fizzled spots) were tagged and excluded from the analysis. The spot intensities and local backgrounds around each spot were measured for both the Cy3 and the Cy5 channels.
Further analysis was performed with MAVI software, a proprietary product of MWG Biotech AG. This software calculates the slope for each spot and extrapolates the fluorescence intensity for faint and saturated spots scanned at three different amplification settings. Following this, the spot intensity of a specific amplification was determined by linear regression of the log signals. These values were normalized because of the different incorporation efficiencies of the Cy3 and Cy5 dyes (data not shown).
For each slide, the mean of the median background intensity values of each spot (MB) and the standard deviation (SD) of MB were calculated for both channels. Intensity cuts (Icut) were determined from these data by use of the formula MB + (2 × SD) for each channel (the Cy3- and Cy5-channels) (Table (Table11).
Only spots with fluorescence values greater than Icut in either one or both channels were used. This calculation was performed for all spots on each slide (Table (Table1).1). Further analysis was performed with those spots whose intensity values in all five slides were greater than Icut (Table (Table1).1). For all five slides (two control slides and three experimental slides) that were included further in the analyses, 5,963 spots fulfilled these criteria.
The quotients of the fluorescence intensities of hybridized cDNA for each spot obtained by dividing the values of experiments and controls were calculated and represent a ratio. Spots with ratios of >2.0 were classified as upregulated genes, and spots with ratios of <0.5 were classified as downregulated genes. These numerical values for regulated genes have been adapted from other studies (12, 55), in which they were determined with confidence intervals >99.9%. P values were determined by the t test with Expressionist software (Genedata, Basel, Switzerland). Only data with P values <0.05 were included in the study.
For our studies, we used the commercially available O157 array (MWG Biotech AG). The E. coli O157 array is composed of 6,176 50-mer oligonucleotides specific for 4,288 genes of E. coli K-12 strain MG1655, 5,358 genes of E. coli O157:H7 strain RIMD0509952, and 5,336 genes of E. coli O157:H7 reference strain EDL933. The array contains oligonucleotides specific for open reading frames (ORFs) of all three strains, two of the strains, or only one of them. Gene sequences with identities more than 92% (less than four mismatches) are represented by only one oligonucleotide. For example, oligonucleotide ecov2#4432 reports 18 homologous endolysin genes of different prophages (see https://ecom.mwgdna.com/services/cgx/cgx-search). On the other hand, certain genes, e.g., mosaic genes or truncated ORFs, may be specified by two oligonucleotides.
The prefixes Z (E. coli O157:H7 strain EDL933), B (E. coli K-12), and Ecs (E. coli O157:H7 strain RIMD0509952) describe oligonucleotides specific for each of the strains. In addition, 68 Arabidopsis control genes and 92 replicated spots increased the total number of spots to 6,336.
RNA isolation and cDNA synthesis for real-time PCR were performed as described above, but without cyanine dye labeling. Real-time PCR was accomplished in a final volume of 20 μl in glass capillaries with a LightCycler instrument (Roche Diagnostics). PCR was done with a FastStart DNA master kit SYBR Green I (Roche Diagnostics), according to the protocol of the manufacturer, with a final MgCl2 concentration of 4 mM. The cycling conditions were as follows: preincubation at 95°C for 10 min and 40 cycles of 95°C for 10 s, 55°C for 5 s, and 72°C for 10 s. Subsequently, a melting curve was calculated at temperatures ranging from 65 to 95°C. The specificity of the PCR for all amplicons was confirmed by agarose gel electrophoresis.
To obtain a standard curve, real-time reverse transcription (RT)-PCR was performed for serC, stxA2, stxB2, stxA1, stxB1, and recA with serial dilutions of the cDNA (Table (Table2).2). We defined a copy number for each dilution. Real-time PCR was also performed for experimental and control cDNA from E. coli O157:H7 strain EDL933 with an initial concentration of 2,800 pg of cDNA for each sample.
In order to study the effects of a low norfloxacin concentration on the transcriptome of EHEC O157:H7 strain EDL933, the following strategy was adopted: (i) use of a concentration with measurable effects on the EDL933 transcriptome (i.e., induction of Stx production) and (ii) use of a concentration that should not strongly inhibit bacterial growth during the experiment.
At first, growth curve experiments (measurement of turbidity) were performed with 100, 200, 500, and 1,000 ng of norfloxacin per ml added 90 min (OD600 = 0.3) after inoculation of the culture and with a control without norfloxacin. Using norfloxacin concentrations of 500 and 1,000 ng/ml, we observed strong reductions in the OD600 30 min after the addition of the antibiotic (Fig. (Fig.1).1). With 100 and 200 ng of norfloxacin per ml, only a slight growth reduction was detected within 120 min, and this turned into a strong OD600 reduction after that time (Fig. (Fig.1).1). Using an induction time of 120 min and a norfloxacin concentration of 200 ng/ml, we noticed an increase in the level of Stx production (Fig. (Fig.2).2). Since we needed a marker for conditions under which particular EDL933 genes were expressed, we performed further experiments with 200 ng of norfloxacin per ml and an induction time of 120 min.
To better characterize the experimental conditions, we performed a time course of induction with 200 ng of norfloxacin per ml and a control without norfloxacin, including the parameters OD600, viable cell counts, Stx production, and phage titer (in PFU). The results are depicted in Fig. Fig.2.2. It could be confirmed that the OD600 decreased only slightly during the induction time period (Fig. (Fig.2,2, ODs for the control and experimental samples). Interestingly, the viable cell count increased during the first 30 min of induction, decreased about 1.5 log units during the following 90 min (Fig. (Fig.2,2, viable cell counts for the control and the experimental samples), and then further decreased only slightly. The fact that the viable cell counts decreased strongly during induction but the OD600 decreased only slightly could be due to filamentation of the cells with norfloxacin treatment. Similar effects of ciprofloxacin on gram-negative and gram-positive bacteria have been shown by Voigt and Zeiler (50). To look at this effect in more detail, the norfloxacin MIC was determined and scanning electron microscopy was performed. The MIC was 250 ng/ml in repeated experiments, so the induction experiments were conducted with concentrations marginally below the MIC. Control experiments with glacial acetic acid without norfloxacin showed that the amount of acid in the test did not influence the MIC. Scanning electron microscopy performed 120 min after induction revealed a distinct filamentation of cells in the presence of norfloxacin (Fig. (Fig.3).3). Whereas the normal logarithmically growing E. coli cells appeared as 1- to 2-μm-long rods (Fig. (Fig.3B),3B), the norfloxacin-treated cells filamented up to a length of 13 to 15 μm (Fig. (Fig.3A).3A). Therefore, the reduction in viable cell counts during the initial induction period, when the OD600 was still increasing, may be explained by a reduced generation time, an inhibitory effect of norfloxacin, and extended filamentation of cells, which appeared as single colonies on the agar plate.
The level of Stx production started to increase 90 min after induction, and 120 min after induction it increased exponentially. The phage titer started to increase 110 min after induction and showed a maximum of 4.5 × 104 PFU/ml 130 min after induction. Soon after it reached this peak, the phage titer dropped during the following 30 min.
RNA was prepared from three different norfloxacin-induced cultures (200 ng/ml, 120 min) and three cultures incubated without norfloxacin and subsequently reverse transcribed with incorporation of fluorescent cyanine dyes, as described below. The whole procedure was performed two times on two different days, and the two procedures were termed trial 2 and trial 3, respectively. For trial 1, the approach was conducted on a third day under the same conditions but without norfloxacin (Table (Table3).3). This additional trial was carried out to evaluate the expression profile under control conditions. The labeled cDNAs were hybridized with the MWG Biotech AG O157 arrays and analyzed as described below. In addition, cDNA from trials 1 and 2 was reverse labeled to control for the different incorporation efficiencies of the cyanine dyes.
All spots on all three experimental slides with expression ratios above 2.0 (upregulation) or below 0.5 (downregulation) were included in the analysis of the array data. Simultaneously, the ratios of the corresponding spots of control slides 18 and 19 had to be between 2.0 and 0.5. Consequently, 240 spots were detected on the array that displayed regulated genes of E. coli O157:H7 strain EDL933 in response to 200 ng of norfloxacin per ml. The ratios for 118 spots (P < 0.05) indicated upregulation, whereas the ratios for 122 spots (P < 0.05) indicated downregulation.
The three strains represented on the E. coli O157 array together contained 14,982 genes, which were specified by 6,167 oligonucleotides. A large part of these genes consisted of highly similar sequences, meaning that a given oligonucleotide may be specific for more than one gene. Therefore, we had to analyze a huge amount of raw data. Since we were primarily interested in EDL933 genes and to simplify the tables, we mention only genes with the prefix Z (EDL933) and omit those with the prefixes Ecs (RIMD0509952) and B (MG1655).
Due to the presence of a number of homologous phage sequences, even in the EDL933 genome, some problems arose with the assignment of spots to genes of particular phages. A reason for that is that those genes (e.g., lysis genes S, R, and Rz) consist of highly similar sequences in the different prophages present in EDL933. The oligonucleotides used for a given spot, e.g., for gene S, may be homologous to eight different S genes present in EDL933. Therefore, we used the designation “unequivocal” for the specific assignment of spots to a particular phage and “ambiguous” when the spots could be ascribed to homologous genes of distinct prophages.
In order to investigate the reproducibilities of trials 2 and 3, the ratios for slide 43 were plotted versus the ratios for slide 45 in a double-logarithmic plot (Fig. (Fig.4).4). The same was performed for slide 42 and slide 45 and for slide 42 and slide 43. These data demonstrate that from 84.6 to 96.7% of the ratios were in the twofold range, which indicates a high degree of reproducibility of the experiments described here (Table (Table4).4). The correlation coefficients show the variability between two experiments (slides), which ranged from 0.7 to 0.8 (Table (Table44).
Eighty-five of the 118 oligonucleotides representing upregulated genes were specific for EDL933-associated phages (Table (Table55 to Table Table7).7). Among these, 33 oligonucleotides were specific for Stx2 phage BP-933W and 9 were specific for Stx1 phage CP-933V. Ten oligonucleotides were ambiguously assigned to both phages BP-933W and CP-933V (Table (Table55).
Furthermore, 10 oligonucleotides were specific for prophages CP-933H (n = 1), CP-933K (n = 2), CP-933N (n = 1), CP-933X (n = 1), CP-933O (n = 1), CP-933R (n = 1), CP-933T (n = 1), CP-933U (n = 1), and an additional prophage-associated element, as shown in Table Table66.
Twenty-three additional oligonucleotides could not been assigned unequivocally to single phages (Table (Table7).7). Twenty-two oligonucleotides were specific for phages of E. coli O157:H7 strain RIMD0509952 (Table (Table8),8), and 11 oligonucleotides represent genes that were not phage associated (Table (Table99).
According to the genome sequence of E. coli O157:H7 strain EDL933 (NCBI database accession no. NC_002655), prophage BP-933W contains 71 ORFs, which are depicted in Fig. Fig.5.5. Thirty-three of the genes upregulated in our experiments could be unequivocally assigned to BP-933W, 18 upregulated genes could be ambiguously assigned only to BP-933W, 7 genes appeared not to be regulated, 3 genes were not represented on the slide, and 9 BP-933W genes had to be excluded from the analysis because the slide data did not match our quality criteria (e.g., P > 0.05) (Fig. (Fig.5).5). Interestingly, one BP-933W gene (orf1498) appeared to be downregulated.
By considering these findings in detail, the immediate-early region spanning from the integrase gene int to the antiterminator gene N contained six upregulated genes unequivocally assigned to BP-933W; six genes were ambiguously upregulated; four genes, including int, appeared not to be regulated; and one gene had to be excluded from the analysis (Fig. (Fig.5).5). In the region between the antiterminator genes N and Q, four genes were unequivocally allocated to BP-933W, six genes were ambiguously upregulated, one gene appeared not to be regulated, and five genes had to be excluded from the analysis. One of the unequivocally upregulated genes is serine-threonine protein kinase gene stk (Z1444), which was upregulated 21.9-fold and which does not belong to the basic phage gene equipment.
Twenty-three upregulated genes were unequivocally assigned to the late region of BP-933W, starting downstream of gene Q and ending with the structural (head-tail) phage genes. The 99.0-fold upregulated gene Z1466 is homologous to the yihS gene of E. coli (37), the function of which is not known. The genes belonging to the lysis region, i.e., orf1467, S, R, and Rz, were upregulated but could be only ambiguously ascribed to the BP-933W genome. The tRNA genes ileZ, argN, and argO were not present on the array. Late-phase genes lom and bor appeared not to be regulated (Fig. (Fig.5),5), whereas the data for the late antiterminator gene Q, orf1490, orf1494, and orf1503 had to be excluded. Since bor is transcribed in the opposite direction from the other late-phase genes, this may be plausible. For lom, the situation is more difficult. The results of real-time RT-PCR with primers directed to lom were negative. Comparison of the restriction fragment length polymorphism of BP-933W derived from the EDL933 derivative used in our study and that of BP-933W from strain C600/933W revealed a slightly different pattern in this region, indicating that parts of lom have been deleted (data not shown). Since we used an E. coli EDL933 derivative that has been stored in Germany for >10 years, it may be possible that genetic rearrangements in the stab culture were responsible for this phenomenon. Such a phenomenon has been described by Papadopoulos et al. (34).
By taking a closer look at the ratios for the upregulated BP-933W genes, it is conspicuous that a number of ratios are approximately 10-fold higher than the others (Table (Table5).5). The genes with higher ratios are located mainly in the late region downstream of gene Q (Table (Table5).5). Interestingly, the ratios for the upregulated genes, which were ambiguously allocated to this region, i.e., lysis genes, were also in this range. This observation allow us to suggest that those genes could probably also be allocated to BP-933W.
The Stx1 prophage CP-933V contains 64 genes (NCBI database accession no. NC_002655). Eight upregulated genes could be unequivocally allocated to CP-933V, 19 upregulated genes appeared to be ambiguously located in the CP-933V genome, 36 genes were not regulated as determined by our approach, and 1 gene was excluded from analysis (P > 0,05). Most of the nonregulated genes occurred in the early and the beginning of the late-phase region (Fig. (Fig.6;6; Table Table5).5). The upregulated genes of the lysis gene cassette appeared to be ambiguously assigned to CP-933V. In this case, however, the transcription ratios do not fit into the general ratio pattern for CP-933V, so we suggest that upregulation of the lysis gene spots was specified by BP-933W cDNA. Interestingly, the stx1 genes appeared not to be induced by norfloxacin under the conditions applied.
Twenty-three oligonucleotides reported upregulated genes that possess more than two homologous relatives in the EDL933 genome (Table (Table7).7). These genes are mainly associated with replication, recombination, morphogenesis, and lysis. For example, a putative holin protein gene (Table (Table7)7) is highly homologous to eight other EDL933 phage-related holin genes in the EDL933 genome and is covered by a single oligonucleotide. Therefore, the spot could not be exactly assigned to a particular phage, and we have assembled the affected genes in their own category (Table (Table7).7). Some of these genes demonstrated high ratios of up to 25, indicating intensive transcription activity. From our data, we cannot exactly conclude whether the upregulation was caused by RNA from a single gene or RNA from a number of homologous genes present in the EDL933 genome.
Twenty-two upregulated genes were assigned to prophages present in the genome of the Sakai outbreak strain RIMD0509952 (Table (Table8).8). We suggest that upregulation of RIMD0509952 phage-specific genes is caused by cross-hybridization of RIMD0509952 phage-specific oligonucleotides on the slide with EDL933 cDNA. One hint for this suggestion is the homology of EDL933 and RIMD0509952 phages (15). Most of the genes detected are yet uncharacterized (Table (Table88).
We also found a category of regulated genes that are apparently not phage regulated (Table (Table9).9). These genes display ratios from 2.7 to 5.8 and were considered to be slightly induced. Among these were a putative phage integrase not present in any of the prophages, some stress-induced proteins, and DNA damage-inducible proteins associated with the SOS response. It is noteworthy that recA and uvrD appeared not to be upregulated under these conditions on the array.
According to the conditions applied, 122 genes were downregulated. Only two of these genes belonged to prophages. Most of the downregulated genes are responsible for metabolic functions such as energy, amino acid, fatty acid, protein, and cell wall metabolism (Table (Table10).10). We have omitted a detailed description of these genes because we turned our attention to prophage genes. Beside this basic cellular machinery, the seven locus of enterocyte effacement genes, espA, espB, espD, cesT, sepZ, sepL, and escV, were downregulated (Table (Table10).10). In addition, the genes for an EHEC factor for adherence cytotoxin (Z4332) and a Shet2-like toxin (Z4326) were also downregulated.
In order to verify the results obtained with our array, we performed real-time RT-PCR with stx1 and stx2 genes. As a reference, we wanted to use a chromosomal E. coli gene that is usually constitutively expressed. Therefore, we selected the housekeeping gene serC.
For the generation of standard curves, we analyzed the LightCycler runs by the fit-point method, as described in the manual accompanying the LightCycler instrument (LightCycler operator's manual, version 3.0, May 1999). Crossing points (CPs) were plotted against the input copy number to obtain standard curves and calculate the slopes with the software accompanying the LightCycler instrument. The corresponding real-time efficiency (E) of each standard curve was calculated by the equation 10(−1/slope). To calculate the relative expression ratio of the regulated gene of interest (target [T]) versus that of the unregulated housekeeping gene (reference [R]) we used the formula described by Pfaffl et al. (36): EΔCPT/EΔCPR, where ΔCP is the change in CP (control CP − sample CP). We provide a comparison of particular microarray data and real-time RT-PCR data in Table Table11.11. The microarray-generated ratios for the control with the housekeeping gene serC and the stx genes are about 1, indicating nonregulated genes as expected. The means of the ratios in trial 2 and trial 3 also indicate nonregulation for serC. Array analysis has shown norfloxacin induced the stxA2 gene >157-fold but induced stxB2 only 40.7-fold. In general, the data obtained by real-time RT-PCR with the LightCycler instrument confirm these results (Table (Table11).11). However, the numerical values were higher than those obtained with the arrays, which is probably a result of the higher efficiency of the RT step (no cyanine dye incorporation) and the higher sensitivity of the real-time PCR. The same is true for stx1. Whereas the ratios for the stx1 genes were close to being below the threshold values for regulation on the array (1.8 for stxA1) and 1.5 (stxB1), real-time PCR indicated upregulation, with higher ratios of 7.0 (stxA1) and 5.9 (stxB1).
Real-time RT-PCR data for recA indicated slight upregulation under the experimental conditions used.
The results of our experiments have shown that DNA microarray analysis is a valuable tool for analysis of global gene expression in response to environmental stimuli. Whereas E. coli K-12 arrays have been used in a number of investigations (13, 19, 39, 42), the E. coli O157 array opens the door for transcriptome analysis of this serious bacterial pathogen.
Little is known about the global genetic regulation of E. coli O157:H7 genes. Sperandio et al. (45) have shown by DNA microarray analysis that quorum sensing is a global regulatory mechanism for basic physiological functions as well as for the production of virulence factors in E. coli O157:H7. They found that 404 genes were upregulated by luxS at least fivefold. Moreover, an involvement of the SOS response was indicated in their study. Whereas the transcription of stx2 was 3-fold higher in strain 86-24 than in a luxS mutant, SOS genes such as recA, uvrA, and sulA were upregulated more than 20-fold. Dahan et al. (8) investigated the changes in gene expression upon binding of E. coli O157:H7 to plasma membranes. They found increased levels of stress-associated mRNA and decreased levels of mRNA encoding proteins involved in translation and type III secretion.
Bacteria are able to rapidly respond to environmental signals (24, 47). Besides signal transduction systems that forward stimuli from the outside of the cell through the cell wall to the inside, chemical substances may enter the cells and directly or indirectly affect gene transcription. Bacterial pathogens interact with a specific environment, the human host. Besides interaction with epithelial surfaces and competitive growth with the commensal flora, they come into contact with chemicals that are used for therapy. One group of chemicals used most frequently for the treatment of infectious diseases are antibiotics. Antibiotics are thought to interact specifically with bacterial metabolism and target specific structures. Targeting of these structures with therapeutic concentrations of antibacterial chemicals inhibits bacterial growth by interfering with cell wall biosynthesis, nucleotide synthesis, or protein biosynthesis. Previous studies from a number of investigators have shown that the mechanisms of antibiotic action are more complex and that subinhibitory concentrations of antibiotics may alter the bacterium's properties (2, 23). Since the use of antibiotics is common in veterinary and human medicine, as well as in animal husbandry and the food industry worldwide, bacteria frequently come into contact with antibiotics. Certain antibiotics may cause induction of bacteriophages, i.e., Stx-encoding bacteriophages (21, 57), and even subinhibitory concentrations may modulate Stx production (11, 21, 54, 56).
Norfloxacin is a member of the group of 4-quinolone antibiotics that interact with both DNA gyrase and topoisomerase IV (10). For E. coli, DNA gyrase is the primary target. Quinolones inhibit DNA synthesis by triggering the formation of stable complexes of DNA and topoisomerase IV as well as DNA gyrase. The quinolone-mediated cell death is thought to be initiated by the generation of double-stranded breaks (10). Norfloxacin is a good inducer of the SOS response, and it has been shown that genes of the SOS regulon are induced on E. coli K-12 microarrays upon incubation with norfloxacin in a dose-dependent manner (44). Those investigators used concentrations ranging from 0.03 to 8.0 μg of norfloxacin per ml for 30 min. The highest extent of gene regulation was observed with the highest norfloxacin concentration, which caused the upregulation of 22 genes involved in DNA metabolism. With a norfloxacin concentration comparable to that which we have used, only four genes of the SOS response were upregulated after 30 min of induction.
We have used 200 ng of norfloxacin per ml over an induction period of 120 min and detected by array analysis only two upregulated genes that were associated with the SOS response. One of these encodes a protein homologous to DinI of Serratia marcescens, and the other one encodes a DinD homologue (Table (Table9)9) (53). The ratios for both genes were 2.7 and 5.8, respectively, indicating moderate induction (Table (Table9).9). The real-time RT-PCR performed for determination of the expression of recA and uvrD confirmed the array data for uvrD and indicated a moderately increased level of recA expression. In general, RT-PCR is more sensitive because RT can be performed without incorporation of cyanine dyes. It is not clear why only three SOS-associated genes are switched on, but this may depend on the fact that particular DNA-damaging agents can preferentially induce particular genes of the SOS regulon (3), on the prolonged induction time, or on the low norfloxacin concentration.
The physiological response to norfloxacin probably cannot be compared between E. coli K-12 and EDL933, since EDL933 probably does not respond with a similar strength as E. coli K-12. However, 200 ng of norfloxacin per ml induced phage genes; and the genes most strongly affected were those of BP-933W in the late region, including stx2, which was induced more than 150-fold. It is intriguing that three SOS-related genes were induced. However, a low concentration of norfloxacin could probably switch on alternate DNA repair pathways. E. coli possesses multiple inducible DNA repair pathways; among these are recA-independent pathways, such as the UVM response (1, 17). However, the most plausible explanation would be that the low norfloxacin concentration is not able to fully induce the SOS regulon genes.
Regulation of gene transcription is an important response to environmental changes and is stringently regulated in bacteria. Our present understanding indicates that the coordination of gene transcription in E. coli involves a number of hierarchical levels, such as the local control of individual operons, the regional control of multiple operons within a regulon, or the regional control of multiple regulons in regulatory networks that are termed stimulons or modulons (29).
A high hierarchical level of gene regulation is represented by changes in the DNA conformation, such as DNA supercoiling or the presence of intrinsically curved DNA (14, 33). In studies with mutants it has been shown that the maximal abundance of proteins in a set of 88 proteins investigated occurred at supercoiling levels below that of the wild type, and others were most abundant with elevated levels of negative superhelices (46). Since the superhelical status is caused and influenced by the activities of DNA gyrase and topoisomerase, DNA gyrase inhibitors also influence supercoiling and gene transcription (10). It has been demonstrated that DNA gyrase inhibitors also lower the levels of DNA supercoiling. It may be hypothesized that norfloxacin at higher concentrations inhibits bacterial growth and probably modulates phage gene expression by causing changes in the DNA tertiary structure at lower levels. However, genes that have already been shown by other investigators to be regulated by supercoiling were not upregulated in our study. Future experiments will be needed to investigate this phenomenon.
It is conspicuous that the genes of prophage BP-933W are transcribed most strongly. In concordance with a temporal transcription scheme (48), genes of the late-phase region are upregulated manyfold (Table (Table5).5). However, some genes, such as int, cI, lom, bor, and some ORFs, do not fit in that scheme (Fig. (Fig.4).4). In order to obtain reproducible results, we used a strict system for the inclusion of spots in our analyses, as described above. For example, when the ratio for a spot on one of the three experimental slides was less than 2.0, the spot was precluded from further analyses. For the int gene, we obtained values of 4.0, 5.0, and 1.16 (data not shown). Although int appeared to be upregulated on two of the slides, it had to excluded since the ratio for one spot was less than 2.0. Therefore, from the slide data we cannot suggest exactly whether this gene is regulated or not. A similar situation was found for some other genes, such as cI. The bor gene is transcribed in the opposite direction and is not thought to be cotranscribed with the late phage genes, and lom was not present in our isolate in a fully intact form and therefore had to be excluded from the analysis.
Another interesting question to be addressed is the high level of expression of stx2 and the difference between the transcription efficiencies of stxA2 and stxB2. Up to now it has not been clear whether the relation of the A and B subunits of 1:5 is regulated on the transcriptional level or the translational level (30). By taking into account the fact that slide ratios cannot be considered absolute values, we have validated these data with real-time RT-PCR experiments. The A-subunit genes are transcribed more efficiently than the B-subunit genes. This was not expected and may be interpreted as evidence for regulation at the posttranscriptional level. This suggestion needs further research.
A number of phage genes are demonstratively regulated very strongly. These are mainly BP-933W genes. Whereas the stx2 gene is upregulated more than 150-fold, most genes of the late region displayed ratios from 10- to 99-fold. This is in contrast to CP-933V and non-Stx phages (Tables (Tables55 to to7),7), which show lower regulation ratios, with a maximum of 20-fold. This points to the question of whether the different prophages possess different mechanisms of regulation of their growth cycle.
In older studies it was observed that toxin production is affected by the presence of a second toxin-encoding phage. The same could be true in the context of our study. Repressors of the BP-933W phage may alter gene transcription in CP-933V or the other prophages. Alternatively, these phages could contain other mechanisms of regulation. An interesting experiment would be investigation of whether the DNA supercoiling status of BP-933W may be responsible for that phenomenon.
Microarray analysis opens new dimensions in E. coli O157 research. The fact, however, that E. coli O157:H7 contains so many homologous DNA sequences aggravates the handling of E. coli O157:H7 arrays in this specific field of phage research. Prophage genes, present as functional or defective phage fragments, show gene transcription activity upon induction by environmental stimuli. Future research will be needed to clarify the impact of these observations for population biology and to determine regulation networks that enable the prophages to respond to and communicate with each other and alter transcription activity.
This work was supported by Deutsche Forschungsgemeinschaft priority program SSP1407, grant Schm1360/1-3.
We thank Stefanie Müksch for skillful technical assistance. We are grateful to Matthias Frosch, Guido Dietrich, and Stefanie Theiss, University of Würzburg, who supported us with helpful discussions and technology transfer in the initial phase of the experiments. Andrea Hörster, MWG Biotech AG, supported us with valuable information on slide technology and the design of oligonucleotides for the real-time PCR.