Identification of the CRP regulon using in vitro and in vivo transcrip
http://www.100md.com
《核酸研究医学期刊》
School of Biosciences, University of Birmingham, Birmingham B15 2TT, UK
* To whom correspondence should be addressed. Tel: +44 121 414 5564; Fax: +44 121 414 5925; Email: C.Constantinidou@bham.ac.uk
ABSTRACT
The Escherichia coli cyclic AMP receptor protein (CRP) is a global regulator that controls transcription initiation from more than 100 promoters by binding to a specific DNA sequence within cognate promoters. Many genes in the CRP regulon have been predicted simply based on the presence of DNA-binding sites within gene promoters. In this study, we have exploited a newly developed technique, run-off transcription/microarray analysis (ROMA) to define CRP-regulated promoters. Using ROMA, we identified 176 operons that were activated by CRP in vitro and 16 operons that were repressed. Using positive control mutants in different regions of CRP, we were able to classify the different promoters into class I or class II/III. A total of 104 operons were predicted to contain Class II CRP-binding sites. Sequence analysis of the operons that were repressed by CRP revealed different mechanisms for CRP inhibition. In contrast, the in vivo transcriptional profiles failed to identify most CRP-dependent regulation because of the complexity of the regulatory network. Analysis of these operons supports the hypothesis that CRP is not only a regulator of genes required for catabolism of sugars other than glucose, but also regulates the expression of a large number of other genes in E.coli. ROMA has revealed 152 hitherto unknown CRP regulons.
INTRODUCTION
The Escherichia coli cyclic AMP receptor protein (CRP) is an important transcription factor that regulates transcription initiation for more than 100 genes mainly involved in catabolism of carbon sources other than glucose (1). E.coli preferentially utilizes glucose over other sugars and only catabolizes other sugars when the supply of glucose has become depleted . The presence of glucose prevents E.coli from catabolizing alternative sugars by several mechanisms, one of which is that glucose lowers the level of cAMP, the inducer for CRP.
CRP functions as a dimer in the form of a CRP–cAMP complex, and regulates transcription initiation by binding to a symmetrical DNA sequence (consensus sequence 5'-AAATGTGATCTAGATCACATTT-3'), located near or within the promoter regions. At CRP-dependent promoters, CRP activates transcription by making direct protein–protein contacts with RNA polymerase (RNAP). Class I CRP-dependent promoters, e.g. lac, contain a single CRP-binding site upstream of the DNA-binding sites for RNAP. At these promoters, CRP activates transcription by interacting with the C-terminal domain of the RNAP subunit (CTD) via a surface-exposed patch, known as activating region 1 (AR1) (residues 156–164). Class II CRP-dependent promoters, e.g. galP1 and melR, contain a CRP-binding site overlapping the –35 hexamer for RNAP. At these promoters, CRP makes multiple interactions with RNAP, including between AR1 and CTD, the activating region 2 (AR2) (residues 19, 21 and 101) and the N-terminal domain of subunit (NTD) and between AR3 (residues 52–58) and the RNAP 70 subunit region 4. Many promoters contain tandem CRP sites and are known as Class III promoters. CRP activation at these promoters involves a combination of the Class I and Class II mechanisms. Additionally, many CRP-dependent promoters are co-dependent on a second activator (3).
The recent completion of several bacterial genome sequences has facilitated the development of computer-based bioinformatic approaches and microarray techniques to study transcription regulation at a genome-wide level. Bioinformatic analysis has allowed the prediction of the regulon of a particular transcription factor by searching for the consensus sequences, or by using algorithms to search for sequence patterns within the genome. CRP-binding sites present within the E.coli genome have been predicted using different computational approaches (4,5). In the latter study, Tan et al. have predicted 161 strong (including known sites) and 285 weak candidate CRP-binding sites, using a comparative genomic approach. However, in silico-predicted DNA-binding sites may not be occupied in vivo, or may only function in the presence of other factors. Conversely, operons with weak, but functional, sites may not have been identified using these approaches. In this study, we have utilized oligonucleotide microarray technology to determine the CRP regulation experimentally, and compared these data with the predicted CRP-binding sites.
The DNA microarray technology allows thousands of genes to be studied simultaneously in a single experiment, and thus provides a powerful tool to investigate gene function at a genomic level. Transcription profiles of E.coli in different media (6,7), under various stress conditions (8,9) and between different strains (10,11), have revealed the roles of many regulatory factors. However, the interference of other regulatory networks makes it difficult to distinguish direct effects on transcription from indirect effects, and therefore it is hard to directly link the results to a specific transcriptional factor. Recently, Helmann and co-workers (12) have developed a novel technique, which combined in vitro run-off transcription with macroarray analysis (ROMA), to define the direct effects of W on Bacillus subtilis promoters. Run-off transcription reactions in this experimental system used genomic DNA as template. The resulting 33P-labelled RNA products, transcribed by either W holoenzyme or core enzyme, were hybridized to macroarray membranes. Direct visual comparison of the core versus holoenzyme experiments allows the ready identification of genes targeted by W. The result from ROMA for W from B.subtilis was consistent with those determined by promoter consensus searching and in vivo transcriptional profiling. In this study, the term ‘in vitro’ refers to experiments completed with purified DNA and protein components, whereas ‘in vivo’ refers to experiments in E.coli cells grown in planktonic culture.
In this study, the ROMA procedure was modified so as to exploit oligonucleotide arrays on glass slides and ROMA was then used to identify CRP-regulated operons in vitro.
MATERIALS AND METHODS
Strains and growth conditions
E.coli K-12 MG1655 (CGSC 7740) was used in this study (13). A crp derivative of MG1655 was constructed using the gene disruption method of Datsenko and Wanner (14). Primers: CRP P1: 5'-GCTCTGGAGAAAGCTTATAACAGAGG ATAACCGCGCGTGTAGGCTGGAGCTGCTTC-3' and CRP P2:5'-TGGCGCGCTACCAGGTAACGCGCCACTCCGACGGGACATATGAATATCCTCCTTAG-3' were used to amplify a chloramphenicol resistance gene cassette from plasmid pKD3 (14). The bases underlined in the primer sequences shown above represent the short regions of homology to genomic sequences flanking crp, which were used in Red-mediated recombination of the chloramphenicol resistance cassette into the chromosome. Transformants were grown on Luria–Bertani (LB) agar (15) containing 0.2% glucose and 50 μg/ml chloramphenicol. PCR was used to screen chloramphenicol resistant transformants, for replacement of crp with the chloramphenicol resistance cassette, using the primers CRPSCREEN 1: 5'-GGATGCTACAGTAATACATTGATG-3', and CRPSCREEN2: 5'-GACCGAATCGTAATTCGCCAAG-3'. Amplicons generated by the screening PCR were sequenced using a BigDyeTM version 3 sequencing kit (Applied Biosystems, Warrington, UK) and analysed on an ABI 3700 sequencer (Applied Biosystems). MG1655 crp strains were grown on Maconkey agar containing maltose (1% w/v) to confirm the phenotype.
For in vivo microarray experiments, both wild-type and crp strains were grown in M9 minimal media (14) containing 0.2% fructose at 37°C to OD600 0.8. Glucose (0.2%) was then added to the cultures, which were grown for a further 15 min, prior to harvesting. The cell samples for RNA preparations were collected before and after the addition of glucose. Two volumes of RNAProtect (Qiagen Ltd, Crawley, UK) were added per volume of bacterial culture, prior to centrifugation.
E.coli oligonucleotide array
E.coli oligonucleotide arrays were produced by the UBEC group (University of Birmingham E.coli group). The oligonucleotides were designed and synthesized by Qiagen Operon Ltd as the E.coli Array ready oligo set vs 1.0 (Qiagen) and represented 4289 E.coli K12 strain (MG1655) open reading frames (ORFs); 1416 ORFs designed to the O157:H7 (EDL933) strain; and 273 ORFs unique to the O157:H7 (Sakai) strain (http://oligos.qiagen.com/arrays/oligosets_ecoli.php). In addition, 110 oligonucleotides representing ORFs from the EDL933 and Sakai plasmids were added to the oligonucleotide array set. Also included within the array set were 12 positive and 12 negative control oligonucleotides, which were printed within each subarray. The average size of each oligonucleotide used in the array was a 70mer, with a Tm = 75 ± 5°C. The position of each oligonucleotide within the ORF was more than 40 bases away from the 3' end. The oligonucleotides were arrayed on Corning CMT-GAPS II slides in 48 blocks, each containing 324 spots (18 rows by 18 columns) using a MicroGrid II robot (BioRobotics, UK). Each oligonucleotide was printed in duplicate on the array, and Amersham LucideaTM Universal ScorecardTM (Amersham, UK) controls were printed within each subarray.
Genomic run-off transcription for ROMA
Genomic DNA was isolated from the E.coli MG1655 (CGSC 7740) wild-type strain using phenol/chloroform extraction (http://www.research.umbc.edu/~jwolf/m1.htm), digested with EcoRI overnight and purified by phenol/chloroform extraction, followed by isopropanol precipitation. RNAP holoenzyme was purified as described previously (16). Wild-type CRP, AR1-mutated CRP (HL159) and AR2-mutated CRP (KE101) were purified as described previously (17).
For a single ROMA experiment, two run-off transcription reactions were set up in parallel. The control reaction contained 4 μg of EcoRI digested genomic DNA, and 1 mM each of ATP, GTP, CTP and UTP, in transcription buffer (40 mM Tris/acetate, pH 7.9, 10 mM MgCl2, 1 mM DTT, 100 mM KCl, 0.1 mg/ml BSA), which contained RNase Out (Invitrogen, UK) at a concentration of 50 U/reaction. The test reaction contained 2 μl of 5 mM cAMP (final concentration 200 μM), and 4 μl of 10 μM CRP (40 pmol of monomer, at a final concentration of 800 nM) in addition to the components of the control mixture. Both run-off transcription reaction mixtures were incubated for 10 min at 37°C and transcription started by the addition of 20 pmol of RNA polymerase holoenzyme. Run-off transcription reactions were incubated at 37°C for 30 min. After incubation, the reactions were stopped by the addition of 5 μl of 250 mM EDTA and placed on ice.
RNA purification and labelling
Both RNA transcripts from in vitro transcription reactions and total RNA from cell cultures were extracted using the Qiagen RNeasy mini kit. On-column DNase I digestion (Qiagen Ltd) was used to remove contaminating genomic DNA from the RNA preparations. The concentrations of RNA or in vitro transcription products were determined at OD260 and OD280. An indirect labelling method was used to obtain fluorescence-labelled cDNA for any RNA sample. Briefly, all RNA transcripts from each reaction or 10–20 μg of total RNA were mixed with 6 μg of random hexamers (Amersham Biosciences, Little Chalfont, UK) to a final volume of 18.4 μl, incubated at 70°C for 10 min and snap-cooled in ice. Reverse transcription labelling mixture (11.6 μl) was then added to the RNA template and random hexamers, which contained 0.5 mM dATP, dCTP, dGTP, 0.2 mM dTTP, 0.3 mM aminoallyl-dUTP (aa-dUTP), RNase inhibitor (30 U), 400 U SuperScript II (Invitrogen, Paisley, UK), 10 mM DTT and 1x first strand buffer. The mixture was incubated at 42°C for 3 h or overnight to generate aminoallyl-labelled cDNA. To hydrolyse the RNA template, 10 μl of 0.5 M EDTA and 10 μl of 1 M NaOH were added to the reaction and incubated at 65°C for 15 min. The reaction was neutralized by the addition of 10 μl 1 M HCl. Unincorporated aa-dUTP and free amines were removed by washing the cDNA in a microconcentrator (Microcon YM-30, Millipore) and the sample was then vacuum dried. The aminoallyl-cDNA pellet was resuspended in 4.5 μl of 0.1 M sodium carbonate buffer (pH 9.0) and coupled with Cy3 or Cy5 monoreactive dye (Amersham), prepared in dimethyl sulfoxide for 2 h at room temperature in the dark. For the ROMA experiment, the aa-cDNA from the control reactions was coupled with Cy3 and aa-cDNA from the CRP reactions coupled with Cy5. For in vivo transcriptional profiling, the aa-cDNA obtained from either wild-type strain or CRP strains before the addition of glucose was coupled with Cy3 and after the addition of glucose was coupled with Cy5. Uncoupled dyes were removed using a QIAquick PCR purification kit (Qiagen). LucideaTM Scorecard (Amersham) mRNA spike mixes were added to labelling reactions for both in vivo and ROMA experiments.
For the in vivo microarray work, the quality and concentration of the RNA prepared was assessed using the Agilent 2100 Bioanalyser. Total RNA (10–20 μg) was labelled using the CyScribe Post-Labelling Kit (Amersham Biosciences) as described by the manufacturer.
Prehybridization and hybridization
Before hybridization, the arrayed slides were prehybridized in a buffer containing 25% formamide, 5x SSC, 0.1% SDS and 10 mg/ml BSA at 42°C for 2 h and washed by dipping twice in distilled water. The slides were then dipped in 95% ethanol for 1 sec and dried in a clean 50 ml centrifuge tube by centrifugation at 1500 g for 10 min. The Cy3 and Cy5 labelled cDNAs were mixed, vacuum dried and resuspended in 70 μl hybridization buffer containing 25% formamide, 5x SSC, 2 μl of 50x Denhardt's, 2 μl yeast tRNA (20 μg/ml) and 0.1% SDS. The labelled cDNAs were denatured by heating at 95°C for 5 min, and applied to the prehybridized slide in a CMT-Hybridization chamber (Corning Inc., Corning, NY). A HybriSlip (Sigma) was carefully lowered onto the slide. To maintain humidity inside the chamber, 10 μl of distilled water was added to the two reservoir wells. The chamber was then tightly sealed and incubated at 42°C for 16–20 h in the dark. The slide was then removed from the chamber, washed for 5 min sequentially in 2x SSC/0.1% SDS buffer, 0.1x SSC/0.1% SDS buffer and 0.1x SSC buffer, rinsed in distilled water for 5 s and dried by centrifugation at 2000 r.p.m. for 10 min. The hybridized slides were scanned with a confocal laser scanner (Axon GenePix 4000A) using appropriate gains on the photomultiplier tube to obtain the highest intensity without saturation. Three replicates were completed for each experiment.
Image extraction and data normalization
Scanned images for Cy3 and Cy5 were then overlaid with GenePix Pro 3.0 software. Only data generated from spots representing E.coli MG1655 genes were analysed in our studies. Spots with background-subtracted intensity lower than 100 in both Cy3 and Cy5 channels were filtered out. GeneSpring software (SiliconGenetics) was used for global normalization and density-dependent normalization (Lowess) to correct artefacts dependent on density or caused by different dye incorporation rates for the two dyes. The duplicate spots for each gene on a single slide were taken as two individual spots. Three independent experiments were performed for each comparison and, therefore, six replicate data sets were obtained for each gene. The normalized data from GeneSpring software were exported to a Microsoft Excel spreadsheet.
Data reproducibility
To assess data reproducibility, the spot-to-spot variation was calculated as described by Loos et al. (18) and correlation between data sets analysed by the Pearson correlation coefficient (r). The spot-to-spot variation was represented by the percent error calculated by dividing the SD by the average ratio for each gene (% = /μ). Average variation within a slide was calculated as the average of the gene spot-to-spot variations on that slide (two spots), and average variation across slides was calculated as the average of gene spot-to-spot variations based on the six spots of the data set. The latter variation included spot, slide and cDNA variability. The Pearson correlation coefficient (r) was calculated for Cy3 data sets from any two slides and for Cy5/Cy3 from any two replicate slides.
Identification of differentially transcribed genes
Differentially transcribed genes were selected using an outlier iteration method (18,19). The data for each gene were averaged and the geometric mean and SD were calculated for the entire population. Any gene with a log-ratio more than three SDs away from the mean was considered an outlier. Outliers were then removed from the population and retained within the differentially expressed subset, and the mean and SDs were recalculated for the rest of the data. The step was repeated until few or no outliers were detected. The 99% predictive interval (PI) was set for the final cut-off ratio to define the remaining differentially transcribed genes within the now symmetrical and well-defined distribution.
The in vitro transcription profiles of CRP derivatives were compared with wild-type CRP, and significant changes were identified by outlier iteration. The average of the log-ratios for each gene from the CRP derivative experiments was normalized to the wild-type CRP experiment. The geometric mean and SD were calculated for the entire population. The outliers were selected as stated above and the 95% PI was set as a final cut-off. The genes with a log-ratio change between two experiments falling in the outlier group or beyond the PI were regarded as differentially regulated by CRP mutants and wild-type CRP.
Sequence processing
The differentially transcribed genes were grouped into operons using the RegulonDB database at http://www.cifn.unam.mx/Computational_Genomics/regulondb. The current version of the RegulonDB database (RegulonDB 4.0) identifies 87 CRP-regulated operons that have been verified experimentally. In this study, 24 of these operons were shown in the ROMA experiments to be regulated by CRP and, therefore, were further analysed. CRP-binding site predictions were based on the previous study (5), where a Positional Weight Matrix was generated from alignment of known CRP-binding sites and used to determine the strength of a site. In this study, a strong site was defined as a sequence with a score more than 10 (consistent with the prediction by Tan et al.) and a weak site as a sequence with a score between 5 and 10. Sequences with a score below 5 were rejected.
RESULTS
Combination of run-off transcription and microarray analysis (ROMA) on glass slides
The original ROMA procedure used macroarray membranes to analyse radioactively labelled RNA transcripts (12). In our study, the procedure was modified by using a post-transcriptional fluorescence-labelling procedure to enable hybridization to a sense-oligonucleotide array. To gauge data reproducibility, data from each of three replicate slides were subjected to statistical analysis. Gene spot-to-spot variation was 5.0% (range, 4.1–6.3%) between replicate spots within a given slide, and 13.9% (range, 12.6–15.4%) among six spots across three slides. The latter number includes both the hybridization variation and biological variation. Comparison of the duplicated probe sets within a single hybridization showed good reproducibility of the fluorescence signals with a Pearson correlation coefficient (r) of >0.98. This result illustrates uniform hybridization of the microarray. The reproducibility of signal between replicate reactions was lower, with a Pearson correlation coefficient (r) of >0.80. This variability was considered to be mainly due to variations from independent reactions. In addition, the detail of signal intensities showed that 90% of all genes gave a detectable signal.
CRP regulatory profiles in the ROMA system
Comparison of RNA transcripts from reactions containing wild-type CRP to control reactions containing only RNAP revealed many differentially transcribed genes (Figure 1a). Data for the majority of genes fall within the threshold values, with a ratio between +2 and –2, and therefore indicates that these genes are not significantly regulated by CRP. As expected, transcripts of many genes were increased by the addition of CRP and transcripts of very few genes were reduced, which is consistent with the view that CRP activates most of the genes that it regulates. A total of 280 genes, present in 188 different operons, had significantly higher transcriptional levels and 20 genes, in 16 operons, had reduced transcriptional levels in response to CRP. Taken together, these data show that 7% of the genes in the E.coli genome were transcribed differentially upon the addition of CRP in the ROMA experiment. The highest activation, 13-fold induction, was observed at the sdaC gene encoding a protein that is a putative serine transporter. Of the 176 operons identified by ROMA in this study, 24 are known members of the CRP regulon (Table 2), i.e. the promoter regions and CRP sites have been mapped experimentally. For genes present within the same polycistronic operon, the signal for the promoter-proximal gene should be the same or higher than the signal for promoter distal genes. Of the 188 operons activated by CRP, the promoter-proximal gene was activated in the vast majority (176 operons). For 8 of these 176 operons, the promoter-proximal gene was subjected to only marginal activation (above 95% PI), while the promoter distal gene(s) were activated at a significant level. In only 12 operons was a promoter-distal gene up-regulated and not the promoter-proximal gene (Tables 1 and 2). These 12 promoters were not included in the CRP-regulated set. Up-regulation of promoter-distal genes in the absence of regulation of the promoter-proximal genes may represent false positives in the assay; alternatively it may be due to CRP-dependent promoters within operons that are active in vitro.
Figure 1. Logarithmic-scale scatter plots of spot intensities. The data were subjected to global and density-dependent normalization, and plotted on GeneSpring 4.2.1. Each data point corresponds to the average of two spots representing an individual MG1655 gene. The ratios were colour coded, with red for ratio above one and green for ratio below one. The middle line passes genes with no change at two conditions (ratio = 1) and the other two lines demarcate threshold values for genes with significant increase or decrease in transcription in response to CRP deletion (ratio = ±2). (a) Transcriptional profiles from reactions with wild-type CRP (wt. CRP) versus without CRP (–CRP). (b) Transcriptional profiles from reactions with AR1-mutated CRP (HL159) versus without CRP (–CRP). (c) Transcriptional profiles from reactions with AR2-mutated CRP (KE101) versus without CRP (–CRP).
Table 2. Operons which have been experimentally verified as being regulated by CRP
Table 1. Operons activated by wild-type CRP in ROMA experiment
Determination of CRP-binding sites
We scanned the upstream regions of the 176 up-regulated operons for putative CRP-binding sites to determine whether the activation observed is due to a site that matches the CRP-binding site consensus. A previous study using a comparative genomic approach (5) identified 447 operons on the E.coli genome that contain putative CRP sites upstream from the translation start site. Using the same approach as Tan et al., 55 of the 176 (31%) operons identified in our study were found to contain a CRP site. Tan et al. predicted CRP-binding sites with a good match to the consensus (10 as a cut-off score). However, some known CRP-dependent promoters contain weak CRP sites that are not identified when using a cut-off of 10 (see Table 2), e.g. the melR CRP site has a score of 5.5 and rpoH as two sites with scores of 2.2 and 7.4. In addition, they also identified many false positives.
We therefore scanned the regulatory regions of the remaining operons for sequences with a lower match to the consensus (5 as a cut-off score) and found an additional 70 operons that contain CRP sites with a score between 5 and 10 (Tables 1 and 2). Thus, a total of 125 operons identified in the ROMA assay contain CRP sites identified by either Tan et al. (5) or this study. The remaining 51 operons (29%) identified by ROMA do not contain a recognizable CRP site (score <5).
Effect of CRP AR1 and AR2 on activation
Previous studies have shown that CRP activation is dependent on the interactions of its activating regions with subunits of RNAP . In this study, the effects of two CRP derivatives containing a mutation at either AR1 or AR2 were analysed. The HL159-CRP carries a His-to-Leu mutation at position 159 within AR1 and has been shown to be defective at both Class I and Class II promoters. Whereas KE101-CRP, which contains a Lys-to-Glu mutation at position 101 within AR2, fails to activate transcription at Class II promoters, but still functions at Class I promoters (20).
The transcriptional level of each gene in the presence of either HL159-CRP or KE101-CRP was compared with those from the control reaction. As expected, the results showed that HL159-CRP failed to activate transcription of 86% (151/176) of the CRP-dependent genes identified in the initial ROMA experiment (Table 1, Figure 2 and Figure 1b). Therefore only 14% (25/176) of operons were activated by HL159-CRP to the same level as by wild-type CRP. The AR2 mutant KE101-CRP was defective for activation at 59% (104/176) of CRP-dependent operons (Table 1, Figure 2 and Figure 1c). The AR2 mutant, therefore, still activated transcription for many genes although the level of activation was reduced. We predict that the promoter regions of the operons whose transcription is not up-regulated by the AR2 mutant contain a CRP-binding site overlapping the –35 hexamer (Class II binding site).
Figure 2. Regulatory map of 50 highest CRP-regulated operons. The figure indicates the activation by HL159-CRP (middle) and KE101-CRP (right) relative to wild-type CRP (=1), generated by GeneSpring clustering in distance order, so that those with similar regulation pattern were grouped together. The relative activation changes are colour coded: yellow indicates that a gene is regulated at the same level as wild-type CRP, blue indicates that a gene is regulated at a much lower level than wild-type CRP and red indicates that a gene is regulated at a higher level.
To validate the ROMA data, we considered the 24 up-regulated operons whose promoter regions and CRP sites have been mapped experimentally. Of these 24 operons, 7 contain a single Class I CRP-dependent promoter (possess a CRP site far upstream of the –35 hexamer), 9 contain a single Class II promoter (possess a CRP site overlapping the –35 hexamer) and 8 contain a Class III promoter with tandem CRP-binding sites (Table 2). All Class III promoters contain a Class II binding site and a Class I binding site. In the ROMA assay, HL159-CRP failed to activate all promoters, except the two Class III promoters, deoCABD and rpoH. Both the deoCABD and the rpoH operons have several promoter regions that are under differential regulation by CRP (21,22) and therefore the effect of the AR1 mutation may be compromised by transcription from different promoters. In contrast to HL159-CRP, KE101-CRP could still activate class I promoters, but failed to activate all class II promoters and half of the class III promoters. It is therefore clear from the experiments with activating region mutants that the CRP regulation observed in ROMA is similar to that observed in vivo. In addition, it allows us to predict that 59% of the CRP-regulated operons identified by ROMA contain a class II CRP site and are therefore class II or class III promoters.
CRP-repressed operons in the ROMA system
Although CRP is predominantly an activator, we found 16 operons where the promoter-proximal gene had reduced levels of transcription in response to CRP in the ROMA experiments (Table 3). The strongest repression, 3-fold reduction, occurred at the yjcB gene. In contrast to activation, the repression levels of most operons were not significantly affected by either HL159-CRP or KE101-CRP, which indicates that most CRP repression is not dependent on either the AR1 or the AR2 determinant. However, repression at four operons was slightly affected by the AR1 mutation, suggesting some role for AR1 at these operons. We have identified possible CRP-binding sites within 11 of the 16 promoter regions, 8 operons contain a strong CRP-binding site (>10) and 3 operons contain a weak site (5–10). The yjcB and ycfR operons contain a CRP site almost identical to the consensus.
Table 3. Operons repressed by wild-type CRP in ROMA experiment
Of the 16 repressed operons, promoter regions of two, nirB and pncB, have been determined experimentally (23,24). Sequence analysis indicates that their CRP-binding sites overlap the binding sites for RNAP (Figure 3) and it is therefore probable that CRP represses transcription at the nirB and pncB promoters by either blocking the interaction between RNAP and the promoter DNA, or blocking transcription from an alternative upstream promoter. The position of putative CRP-binding sites relative to the RNAP-binding site was determined for the other repressed operons (Figure 3). The position of the CRP site is variable, but at most promoters the CRP site overlaps the region between the –35 and –10 hexamers. Therefore, repression at these operons might also involve a simple blocking mechanism. However, at the metK, ybiS and yjcB predicted promoter regions, the CRP sites are located 3 bp upstream of the –35 hexamer. The yjcB promoter contains a very strong CRP-binding site (score = 22). The promoter architecture might be such that the CRP-binding site still stops productive binding of RNAP.
Figure 3. Non-template strand sequences of promoters repressed by CRP. Except nirB and pncB, promoter regions are predicted by searching the –10 and –35 sequences upstream of 10 down-regulated genes with a CRP site. The predicted –10 and –35 hexamers are underlined and in boldface with the –10 regions aligned together. The potential CRP-binding sites are shown as shaded boxes. *ydfK and ynaE are both associated with different prophage (Quin and Rac, respectively) but have essentially identical sequences.
Identification of CRP regulon by in vivo transcriptional profiling
An in vivo microarray experiment was also designed to identify CRP-regulated genes. Both wild-type and crp strains were grown in minimal media containing fructose and then pulsed with glucose, and the transcriptome before the addition of glucose for both strains was compared with that after glucose addition. In the presence of glucose, the cAMP level is expected to be decreased, and hence the cAMP CRP level, thus the CRP-dependent regulation is repressed. However, glucose also affects gene expression through other mechanisms, thus glucose effects will be observed in the crp strain. The CRP-regulated genes in this study were defined as genes regulated by glucose in the wild-type strain, but not regulated in the crp strain. In total, we identified only 17 operons repressed by glucose (CRP-activated operons) and six glucose-activated operons (CRP-repressed operons), listed in Table 4. In the 17 CRP-activated operons, 9 operons are known members of the CRP regulon and 7 operons were activated by CRP in the ROMA experiment. Our in vivo experiments failed to identify most CRP-regulated genes.
Table 4. CRP-regulated genesa identified by in vivo transcription profiling
DISCUSSION
The combination of run-off transcription and microarray analysis (ROMA) has shown advantages in defining W-dependent promoters in B.subtilis (12). In this study, we have established ROMA exploiting microarray glass slides and further demonstrated its application to investigate a specific transcription factor. The microarray glass slide has several advantages over a macroarray membrane. First, the slide array allows more genes to be studied in a single experiment (up to 50 000 genes) compared with a few thousand genes that can be printed on a macroarray membrane. Second, RNA from two different sources of interest can be dual fluorescence labelled and simultaneously hybridized on a single slide, whereas nylon arrays are generally probed in serial or parallel hybridization reactions to allow comparison. Third, the utilization of fluorescence instead of radioactive material is safer, easier to handle and produces higher resolution and lower background. Therefore, the establishment of ROMA on glass slides provides a convenient system to obtain high quality information. Our results have shown the high sensitivity and reproducibility of the system. However, our procedure does have some disadvantages compared with the original macroarray procedure. First, several purification steps are required which might result in the loss of short transcripts. Second, a large amount of template DNA and protein is required. Third, any trace of template DNA remaining after RNA purification will result in high background hybridization that might mask the actual induction level.
Using this newly established ROMA system, we have identified 152 novel CRP-dependent operons and 24 operons corresponding to previously known CRP regulons. Regulation at these operons was further verified by using two CRP mutants. Our experiments indicate that ROMA is a very powerful technique to identify direct regulation by a transcription initiation factor. The ROMA system was, however, unable to identify some known CRP-regulated operons. There are several reasons that could account for these false negatives. First, some operons require other factors for CRP to activate transcription, e.g. AraC at the araBAD promoter (25), which are missing in the ROMA system. Second, transcripts that can initiate from different promoters upstream from the same operon cannot be resolved. So genes where CRP alters the transcription start point will not be identified. For example, the galP1 promoter is a well-known Class II CRP-dependent promoter. However, in the absence of CRP, transcription of the gal operon can initiate from the upstream galP2 promoter (26), which will mask the weak transcription from galP1. Third, certain promoter structures result in a situation where CRP can activate transcription by binding to an upstream site but can also repress transcription by binding to a downstream site, e.g. the crp promoter (27). In addition, some operons may not be transcribed efficiently due to DNA relaxation, or in the experimental system described here, the existence of an EcoRI restriction site within the promoter region or coding region. Finally, because of the necessity for data manipulation to identify only the most robust changes in transcription, promoters that are only weakly activated by CRP in vitro will not be detected. The CRP regulon may therefore contain more operons than were identified in this study.
In the W ROMA experiment, 50% of the genes identified using ROMA were also identified using consensus sequence searching and in vivo transcriptional profiling (12). However, in the case of CRP, these approaches have serious limitations. Comparison of the ability of each method to identify known members of the CRP regulon demonstrates that sequence searching generates the highest number of hits, and in vivo transcriptional profiling generates the fewest (Figure 4). A combination of the three methods can identify more than 80% of known operons, but failed to identify 14 operons known to be regulated by CRP. Although the sequence prediction generates the highest number of hits, it is probable that this includes many false positives and is therefore not necessarily the best approach. In addition, many weak sites are difficult to identify by sequence searching. Lowering the cut-off score would facilitate the identification of known sites but would inevitably lead to more false positives. Besides DNA sequence, CRP activation is dependent on the location of the binding site. Previous studies have indicated that at a Class I promoter, the CRP-binding site is normally positioned at –61.5, –71.5, –81.5 or –91.5, and no activation occurs when CRP sites are located further than –113.5, which might be beyond the reach of CTD . Therefore, many in silico predicted sites might be null sites that are not functional during transcription. For promoters containing tandem CRP sites, the space between the two sites is an important determinant for synergistic activation (29–32). Therefore, although some operons contain several predicted binding sites, we hypothesized that at least some sites are not functional due to improper spacing. This may also explain why some operons with strong predicted binding sites were not regulated by CRP in our experiments. It is also probable that sequence searching generates many false negatives due to the ability of CRP to bind variable sequences. For example, previous work has identified the CRP sites within the melR, rpoH and rbs promoters that play important roles in regulation of these promoters (22). These sites have scores below 10 and were, therefore, not identified in the sequence search. It is therefore possible that several promoters identified by ROMA that do not contain a putative CRP-binding site, e.g. inaA and yfaH, may still be regulated by CRP.
Figure 4. Venn diagram of the CRP regulon as identified by three genomic approaches: ROMA, in vivo transcriptional profiling, sequence search (5). The result for (a) 87 known (experimentally verified) CRP regulon collected in RegulonDB database and (b) all operons are shown. The number of operons identified by each approach is presented in a coloured circle. The numbers covered by more than one circles are operons identified by two or three methods.
Comparing activation levels of operons with strong CRP sites to those with weak sites indicated that there is no significant relationship between the degree of activation and the ‘quality’ of CRP-binding sites. For example, the focA-pflB operon which is activated 3.9-fold contains a relatively weak CRP site (score = 10.1), whereas the tnaL operon which is activated 2-fold contains a very strong CRP-binding site (score = 19.6) (Table 2). Therefore, conservation of CRP-binding sequence is not sufficient to predict CRP activation levels.
In our study, the in vivo transcriptional profiles failed to identify many CRP-regulated genes. The main reason for this is the complexity of the CRP regulon. CRP activation at some promoters, such as melAB and araBAD, is dependent on the presence of an additional regulator (MelR and AraC, respectively) (33,25) that is only induced by a specific substrate. It is impossible to include all inducers in the growth media and, moreover, the presence of extra inducers may trigger expression of other genes that are independent on CRP, further complicating the interpretation of the data. In addition, most CRP-dependent promoters are subjected to repression by other factors, which will mask the CRP activation in vivo. For example, the galE promoter is repressed by GalR, the lacZ promoter repressed by LacI and many CRP-activated promoters are repressed by CytR (Table 2). Therefore, the variability of CRP regulation at different operons makes it difficult to design an in vivo microarray experiment to distinguish direct effects from indirect effects. ROMA is better suited for identifying the regulon of a factor that is part of a complex regulatory network. In the six glucose-induced operons, two operons, gcd and proP, are known CRP-repressed operons, which contain a strong CRP-binding site between the –10 and –35 regions, thus blocking RNAP binding. As observed in previous studies (34,35), the ptsG gene encoding the major glucose transporter was highly induced by glucose in the wild-type stain. Previous studies have indicated that the ptsG promoter, which contains a CRP activation site (see Table 2) (36), is both activated by CRP and strongly repressed by the pleiotropic transcriptional repressor Mlc. The glucose induction of ptsG is through a complicate Mlc-dependent mechanism that involves several layers of regulation (34–39). The case of ptsG demonstrates that an in vivo indirect effect can mask the real regulation by CRP, sometimes leading to the opposite conclusion. The regulation observed in the trp operon might also have resulted from an indirect effect, as there is no evidence from previous studies to indicate that CRP regulates this operon.
Further evidence of the limitations of in vivo transcription profiling of a complex regulon has come from a recent study of CRP using an Affymetrix E.coli array (40). Gosset et al completed a similar in vivo transcriptomic study of CRP and identified 39 operons under CRP-dependent glucose repression and 19 operons under CRP-dependent glucose activation. Among these operons, only seven glucose-repressed operons and one glucose-activated operon were identified in our in vivo transcriptional profiling (Table 4). However, direct comparison of the two data sets is difficult because of differences in the strain and the growth conditions used. Gosset et al in their study used the E.coli strain BW25113 and an isogenic crp mutant derivative and grew their cells in LB media with or without glucose. The difference in composition between the LB media used in the Gosset study and minimal media used in this study could contribute to a different profile of induction of some genes. For example, the presence of fructose in our growth conditions meant that the fru operon was identified as CRP regulated in our study, but not regulated in the Gosset study due to the lack of fructose in LB media. Gosset et al. (40) used long-term growth in glucose plus media rather than a glucose shock which was used in this study. Long-term growth in glucose should result in higher levels of induction but can also lead to a larger number of indirect effects, such as regulation of ribosomal protein-encoding genes, RNA-encoding genes, stress-related genes and temperature shock genes, which were significantly affected in their study. One-third of operons subject to CRP-dependent glucose repression (13/39) in Gosset et al. were activated by CRP in ROMA experiment. This figure is similar to the data obtained by in vivo transcription profiling in this study (7/17). This indicates that ROMA has limitations to detect a certain set of operons that may be subjected to complex regulation or regulated indirectly in vivo. However, as seen in this study, Gosset et al. failed to identify many CRP-regulated genes, such as rpoH, melR and rbs, so neither study produced a definitive list of members of the CRP regulon. This further indicates that the study of complex regulons, such as CRP, in vivo has many limitations. However, combining several approaches, such as in vivo profiling, sequence analysis and ROMA increases the likelihood of obtaining a more accurate definition of a complex regulon.
The mechanism of CRP activation, in particular the role of different activating regions, has been studied using several well-characterized promoters. At Class I promoters, CRP binds to a DNA sequence upstream of the RNAP-binding site and makes direct protein–protein contact to CTD via AR1 of the downstream subunit of the CRP dimer, and this interaction recruits CTD to its DNA target immediately downstream of the CRP-binding site. At Class II promoters, CRP binds to a site overlapping the –35 hexamer and makes several contacts with RNAP: AR1 of the upstream subunit of the CRP dimer binds CTD, AR2 of the downstream subunit of the CRP dimer binds NTD and AR3 of the downstream subunit binds region 4 of 70. The CTD binds to its target upstream of the CRP site. At Class III promoters that contain tandem sites, CRP activation involves both Class I and Class II mechanisms . Nearly 90% of operons identified by ROMA showed dependence on the AR1 determinant of CRP for full activation, while AR2 is required for activation of 59% (104) of operons. We predicted that these 104 AR2-dependent operons (including 14 known operons) contain a Class II binding site in their promoter regions.
It is not surprising that transcription of some Class III operons was still activated by KE101-CRP though they all contain a Class II binding site. It has been previously shown that CRP activation at this kind of promoter involves a Class I mechanism (upstream site) and a Class II mechanism (downstream site) (30). Inactivation of AR2 only affects the proximal Class II activation, while the distal Class I site could still drive efficient activation. In some cases, where CRP activation is mainly dependent on the proximal site, the AR2 determinant will be crucial.
Interestingly, a few known FNR-regulated operons, adhE, sdhC, sodA, cyoA and focA, were also activated by CRP in the ROMA experiments. FNR is a CRP homologue and regulates genes during anaerobic growth. The consensus sequence for FNR is similar to that for CRP consensus, and previous studies have shown that FNR activates transcription initiation in a similar way to CRP and that both proteins can bind to the DNA site for the other protein (41,42). Therefore, we propose that some CRP-regulated operons identified in this study are actually regulated by FNR in vivo and contribute to anaerobic growth.
At some promoters, transcription activation by CRP can be repressed by a transcriptional regulator protein, CytR. The CytR-regulated promoters usually contain two DNA sites for CRP, centred at positions –41.5 and –93.5 with respect to the transcription start point, and a DNA site for CytR located between the two CRP-binding sites (43). Cytidine is an allosteric inducer of CytR. In the presence of cytidine, CytR is inactive, and thus CRP can activate transcription in a Class III dependent manner. In the absence of cytidine, CytR is active and binds between the two CRP dimers, repressing CRP-dependent transcription . CytR binds to two 5'-TTGCAA-3' motifs between two CRP-binding sites separated by 10–30 bp (44). The deoC, nupG and tsx operons that are under CytR regulation were activated by CRP in our study. Many operons identified by ROMA possess tandem CRP-binding sites separated by 30–60 bp (Table 1). Sequence analysis of the promoter sequence of these operons may identify more CytR-regulated promoters.
In addition to activation, our data also shows that CRP can serve as a repressor at some promoters. There are several different mechanisms by which repressors can inhibit transcription initiation. The simplest mechanism is by blocking the interaction between RNAP and a promoter. This can occur if a binding site for a repressor protein is located overlapping the binding site for RNAP at a promoter, e.g. over the transcription start site or the –10 hexamer. The blocking mechanism is common for most repressors, such as LacI at the lacUV5 promoter and IclR at the iclR promoter (45–47). CRP repression at nirB, pncB, yeeF, ycfR, yhfC, ydfK, ynaE and apbA might also involve this mechanism since at these promoter regions the CRP-binding site overlaps the sites for RNAP. Repressors also can inhibit transcription initiation by direct contact with RNAP, a process more commonly associated with activators. The P4 protein of the B.subtilis bacteriophage 29 represses the transcription of early promoter A2C by inhibiting promoter clearance (48). The binding of P4 overstabilizes the open complex formed by RNAP and DNA, thus impeding the following promoter escape step. The repression at metK, ybiS and yjcB might involve this mechanism since they all contain a strong CRP-binding site immediately upstream of a recognizable –35 hexamer. Interestingly, the CRP-binding site and the –35 region at three promoters are separated by 3 bp, which suggests that the structure of protein–protein interaction might play an important role.
CRP has been identified as a regulator of genes required for catabolism of sugars other than glucose and a large number of other genes . The CRP regulon defined in this study includes several operons that are involved in carbohydrate transport and metabolism, such as fucPI for fucose, kdgT for gluconate and fucAO for fuculose. CRP also regulates transcription of genes required for energy production, amino acid metabolism, nucleotide metabolism and ion transport systems. In addition, CRP can regulate transcription of other transcription factors, such as MelR, RpoH, BlgG, Fis and PdhR, which could further regulate transcription of genes involved in the above processes. We predict that there are several hundred genes that are directly or indirectly subjected to regulation by CRP. In our list, the functions of more than one-third of the gene products are unknown. Identification of function of the gene product of these genes will help to define the biological function of CRP in E.coli.
ACKNOWLEDGEMENTS
We thank Nigel J. Savery and Georgina S. Lloyd for the CRP protein used in this study. This work was supported by Biotechnology and Biological Sciences Research Council Grants EGA16107and JIF 13209, and by a Darwin Trust of Edinburgh studentship (to D.Z.). The authors thank Steve J.W. Busby for his interest in this work and stimulating discussions.
REFERENCES
Kolb,A., Busby,S., Buc,H., Garges,S. and Adhya,S. ( (1993) ) Transcriptional regulation by cAMP and its receptor protein. Annu. Rev. Biochem., , 62, , 749–795.
Postma,P. ( (1986) ) Catabolite repression and related phenomena. Symp. Soc. Gen. Microbiol., , 39, , 21–49.
Busby,S. and Ebright,R. ( (1999) ) Transcription activation by catabolite activator protein (CAP). J. Mol. Biol., , 293, , 199–213.
Robison,K., McGuire,A.M. and Church,G.M. ( (1998) ) A comprehensive library of DNA-binding site matrices for 55 proteins applied to the complete Escherichia coli K-12 genome. J. Mol. Biol., , 284, , 241–254.
Tan,K., Moreno-Hagelsieb,G., Collado-Vides,J. and Stormo,G.D. ( (2001) ) A comparative genomics approach to prediction of new members of regulons. Genome Res., , 11, , 566–584.
Tao,H., Bausch,C., Richmond,C., Blattner,F.R. and Conway,T. ( (1999) ) Functional genomics: expression analysis of Escherichia coli growing on minimal and rich media. J. Bacteriol., , 181, , 6425–6440.
Oh,M.K., Rohlin,L., Kao,K.C. and Liao,J.C. ( (2002) ) Global expression profiling of acetate-grown Escherichia coli. J. Biol. Chem., , 277, , 13175–13183.
Zheng,M., Wang,X., Templeton,L.J., Smulski,D.R., LaRossa,R.A. and Storz,G. ( (2001) ) DNA microarray-mediated transcriptional profiling of the Escherichia coli response to hydrogen peroxide. J. Bacteriol., , 183, , 4562–4570.
Courcelle,J., Khodursky,A., Peter,B., Brown,P.O. and Hanawalt,P.C. ( (2001) ) Comparative gene expression profiles following UV exposure in wild-type and SOS-deficient Escherichia coli. Genetics, , 158, , 41–64.
Arfin,S., Long,A., Ito,E., Riehle,M., Paegle,E. and Hatfield,G. ( (2000) ) Global gene expression profiling in Escherichia coli K12. The effects of integration host factor. J. Biol. Chem., , 275, , 29672–29684.
Hommais,F., Krin,E., Laurent-Winter,C., Soutourina,O., Malpertuy,A., Le Caer,J.P., Danchin,A. and Bertin,P. ( (2001) ) Large-scale monitoring of pleiotropic regulation of gene expression by the prokaryotic nucleoid-associated protein, H-NS. Mol. Microbiol., , 40, , 20–36.
Cao,M., Kobel,P.A., Morshedi,M.M., Wu,M.F., Paddon,C. and Helmann,J.D. ( (2002) ) Defining the Bacillus subtilis W regulon: a comparative analysis of promoter consensus search, run-off transcription/macroarray analysis (ROMA), and transcriptional profiling approaches. J. Mol. Biol., , 316, , 443–457.
Blattner,F.R., Plunkett,G.,III, Bloch,C.A., Perna,N.T., Burland,V., Riley,M., Collado-Vides,J., Glasner,J.D., Rode,C.K., Mayhew,G.F., Gregor,J., Davis,N.W., Kirkpatrick,H.A., Goeden,M.A., Rose,D.J., Mau,B. and Shao,Y. ( (1997) ) The complete genome sequence of Escherichia coli K-12. Science, , 277, , 1453–1474.
Datsenko,K.A. and Wanner,B.L. ( (2000) ) One-step inactivation of chromosomal genes in Escherichia coli K-12 using PCR products. Proc. Natl Acad. Sci. USA, , 97, , 6640–6645.
Sambrook,J., Fritsch,E.F. and Maniatis,T. ( (1989) ) Molecular cloning: A Laboratory Manual. Cold Spring Harbor Laboratory Press, Cold Spring Harbor, NY.
Tang,H., Severinov,K., Goldfarb,A. and Ebright,R.H. ( (1995) ) Rapid RNA polymerase genetics: one-day, no-column preparation of reconstituted recombinant Escherichia coli RNA polymerase. Proc. Natl Acad. Sci. USA, , 92, , 4902–4906.
Rhodius,V.A., West,D.M., Webster,C.L., Busby,S.J. and Savery,N.J. ( (1997) ) Transcription activation at class II CRP-dependent promoters: the role of different activating regions. Nucleic Acids Res., , 25, , 326–332.
Loos,A., Glanemann,C., Willis,L.B., O'Brien,X.M., Lessard,P.A., Gerstmeir,R., Guillouet,S. and Sinskey,A.J. ( (2001) ) Development and validation of Corynebacterium DNA microarrays. Appl. Environ. Microbiol., , 67, , 2310–2318.
Britton,R.A., Eichenberger,P., Gonzalez-Pastor,J.E., Fawcett,P., Monson,R., Losick,R. and Grossman,A.D. ( (2002) ) Genome-wide analysis of the stationary-phase sigma factor (sigma-H) regulon of Bacillus subtilis. J. Bacteriol., , 184, , 4881–4890.
Rhodius,V. and Busby,S. ( (2000) ) Transcription activation by the Escherichia coli cyclic AMP receptor protein: determinants within activating region 3. J. Mol. Biol., , 299, , 295–310.
Valentin-Hansen,P., Holst,B., Josephsen,J., Hammer,K. and Albrechtsen,B. ( (1989) ) CRP/cAMP- and CytR-regulated promoters in Escherichia coli K12: the cdd promoter. Mol. Microbiol., , 3, , 1385–1390.
Kallipolitis,B.H. and Valentin-Hansen,P. ( (1998) ) Transcription of rpoH, encoding the Escherichia coli heat-shock regulator 32, is negatively controlled by the cAMP-CRP/CytR nucleoprotein complex. Mol. Microbiol., , 29, , 1091–1099.
Jayaraman,P.S., Peakman,T.C., Busby,S.J., Quincey,R.V. and Cole,J.A. ( (1987) ) Location and sequence of the promoter of the gene for the NADH-dependent nitrite reductase of Escherichia coli and its regulation by oxygen, the Fnr protein and nitrite. J. Mol. Biol., , 196, , 781–788.
Wubbolts,M.G., Terpstra,P., van Beilen,J.B., Kingma,J., Meesters,H.A. and Witholt,B. ( (1990) ) Variation of cofactor levels in Escherichia coli. Sequence analysis and expression of the pncB gene encoding nicotinic acid phosphoribosyltransferase. J. Biol. Chem., , 265, , 17665–17672.
Lobell,R. and Schleif,R. ( (1991) ) AraC-DNA looping: orientation of distance-dependent loop breaking by the cyclic AMP receptor protein. J. Mol. Biol., , 218, , 45–54.
Busby,S., Truelle,N., Spassky,A., Dreyfus,M. and Buc,H. ( (1984) ) The selection and characterization of two novel mutation in overlapping promoters of the Escherichia coli galactose operon. Gene, , 28, , 201–209.
Ishizuka,H., Hanamura,A., Inada,T. and Aiba,H. ( (1994) ) Mechanism of the down-regulation of cAMP receptor protein by glucose in Escherichia coli: role of autoregulation of the crp gene. EMBO J., , 13, , 3077–3082.
Ebright,R. ( (1993) ) Transcription activation at Class I Cap-dependent promoters. Mol. Microbiol., , 8, , 797–802.
Ushida,C. and Aiba,H. ( (1990) ) Helical phase dependent action of CRP: effect of the distance between the CRP site and the -35 region on promoter activity. Nucleic Acids Res., , 18, , 6325–6330.
Belyaeva,T., Rhodius,V., Webster,C. and Busby,S. ( (1998) ) Transcription activation at promoters carrying tandem DNA sites for the Escherichia coli cyclic AMP receptor protein: organisation of the RNA polymerase subunits. J. Mol. Biol., , 277, , 789–804.
Law,E., Savery,N. and Busby,S. ( (1999) ) Interactions between the Escherichia coli cAMP receptor protein and the C-terminal domain of the subunit of RNA polymerase at class I promoters. Biochem. J., , 337, , 415–423.
Tebbutt,J., Rhodius,V.A., Webster,C.L. and Busby,S.J. ( (2002) ) Architectural requirements for optimal activation by tandem CRP molecules at a class I CRP-dependent promoter. FEMS Microbiol. Lett., , 210, , 55–60.
Belyaeva,T., Wade,J., Webster,C., Howard,V., Thomas,M., Hyde,E. and Busby,S. ( (2000) ) Transcription activation at the Escherichia coli melAB promoter: the role of MelR and the cyclic AMP receptor protein. Mol. Microbiol., , 36, , 211–222.
Kimata,K., Inada,T., Tagami,H. and Aiba,H. ( (1998) ) A global repressor (Mlc) is involved in glucose induction of the ptsG gene encoding major glucose transporter in Escherichia coli. Mol. Microbiol., , 29, , 1509–1519.
Plumbridge,J. ( (1998b) ) Expression of ptsG, the gene for the major glucose PTS transporter in Escherichia coli, is repressed by Mlc and induced by growth on glucose. Mol. Microbiol., , 29, , 1053–1063.
Plumbridge,J. ( (1998a) ) Control of the expression of the manXYZ operon in Escherichia coli: Mlc is a negative regulator of the mannose PTS. Mol. Microbiol., , 27, , 369–380.
Decker,K., Plumbridge,J. and Boos,W. ( (1998) ) Negative transcriptional regulation of a positive regulator: the expression of malT, encoding the transcriptional activator of the maltose regulon of Escherichia coli, is negatively controlled by Mlc. Mol. Microbiol., , 27, , 381–390.
Lee,S.J., Boos,W., Bouche,J.P. and Plumbridge,J. ( (2000) ) Signal transduction between a membrane-bound transporter, PtsG, and a soluble transcription factor, Mlc, of Escherichia coli. EMBO J., , 19, , 5353–5361.
Shin,D., Cho,N., Heu,S. and Ryu,S. ( (2003) ) Selective regulation of ptsG expression by Fis. Formation of either activating or repressing nucleoprotein complex in response to glucose. J. Biol. Chem., , 278, , 14776–14781.
Gosset,G., Zhang,Z., Nayyar,S., Cuevas,W.A. and Saier,M.H. ( (2004) ) Transcriptome analysis of crp-dependent catabolite control of gene expression in Escherichia coli. J. Bacteriol., , 186, , 3516–3524.
Li,B., Wing,H., Lee,D., Wu,H.C. and Busby,S. ( (1998) ) Transcription activation by Escherichia coli FNR protein: similarities to, and differences from, the CRP paradigm. Nucleic Acids Res., , 26, , 2075–2081.
Sawers,G., Kaiser,M., Sirko,A. and Freundlich,M. ( (1997) ) Transcriptional activation by FNR and CRP: reciprocity of binding-site recognition. Mol. Microbiol., , 23, , 835–845.
Valentin-Hansen,P., S?gaard-Anderson,L. and Petersen,H. ( (1996) ) A flexible partnership: the CytR ant-activator and the cAMP-CRP activator protein, comrades in transcriptional control. Mol. Microbiol., , 20, , 461–466.
Perini,L.T., Doherty,E.A., Werner,E. and Senear,D.F. ( (1996) ) Multiple specific CytR binding sites at the Escherichia coli deoP2 promoter mediate both cooperative and competitive interactions between CytR and cAMP receptor protein. J. Biol. Chem., , 271, , 33242–33255.
Yamamoto,K. and Ishihama,A. ( (2003) ) Two different modes of transcription repression of the Escherichia coli acetate operon by IclR. Mol. Microbiol., , 47, , 183–194.
Müller-Hill,B. ( (1998) ) Some repressors of bacterial transcription. Curr. Opin. Microbiol., , 1, , 145–151.
Roy,S., Garges,S. and Adhya,S. ( (1998) ) Activation and repression of transcription by different contact: two sides of a coin. J. Biol. Chem., , 273, , 14059–14062.
Monsalve,M., Calles,B., Mencia,M., Salas,M. and Rojo,F. ( (1997) ) Transcription activation or repression by phage 29 protein p4 depends on the strength of the RNA polymerase-promoter interactions. Mol. Cell, , 1, , 99–107.
Gulati,A. and Mahadevan,S. ( (2000) ) Mechanism of catabolite repression in the bgl operon of Escherichia coli: involvement of the anti-terminator BglG, CRP-cAMP and EIIAGlc in mediating glucose effect downstream of transcription initiation. Genes Cells, , 5, , 239–250.
Davies,S.J., Golby,P., Omrani,D., Broad,S.A., Harrington,V.L., Guest,J.R., Kelly,D.J. and Andrews,S.C. ( (1999) ) Inactivation and regulation of the aerobic C(4)-dicarboxylate transport (dctA) gene of Escherichia coli. J. Bacteriol., , 181, , 5624–5635.
Bell,A.W., Buckel,S.D., Groarke,J.M., Hope,J.N., Kingsley,D.H. and Hermodson,M.A. ( (1986) ) The nucleotide sequences of the rbsD, rbsA, and rbsC genes of Escherichia coli K12. J. Biol. Chem., , 261, , 7652–7658.
Holcroft,C.C. and Egan,S.M. ( (2000) ) Interdependence of activation at rhaSR by cyclic AMP receptor protein, the RNA polymerase subunit C-terminal domain, and rhaR. J. Bacteriol., , 182, , 6774–6782.
Cunningham,L. and Guest,J.R. ( (1998) ) Transcription and transcript processing in the sdhCDAB-sucABCD operon of Escherichia coli. Microbiology, , 144, , 2113–2123.
Deeley,M.C. and Yanofsky,C. ( (1982) ) Transcription initiation at the tryptophanase promoter of Escherichia coli K-12. J. Bacteriol., , 151, , 942–951.
Klein,W., Horlacher,R. and Boos,W. ( (1995) ) Molecular analysis of treB encoding the Escherichia coli enzyme II specific for trehalose. J. Bacteriol., , 177, , 4043–4052.
Weickert,M.J. and Adhya,S. ( (1993) ) Control of transcription of gal repressor and isorepressor genes in Escherichia coli. J. Bacteriol., , 175, , 251–258.
Larson,T.J., Cantwell,J.S. and van Loo-Bhattacharya,A.T. ( (1992) ) Interaction at a distance between multiple operators controls the adjacent, divergently transcribed glpTQ-glpACB operons of Escherichia coli K-12. J. Biol. Chem., , 267, , 6114–6121.
Otsuka,J., Watanabe,H. and Mori,K.T. ( (1996) ) Evolution of transcriptional regulation system through promiscuous coupling of regulatory proteins with operons; suggestion from protein sequence similarities in Escherichia coli. J. Theor. Biol., , 178, , 183–204.
Webster,C., Gaston,K. and Busby,S. ( (1988) ) Transcription from the Escherichia coli melR promoter is dependent on the cyclic AMP receptor protein. Gene, , 68, , 297–305.
Ryu,S., Ramseier,T.M., Michotey,V., Saier,M.H.,Jr and Garges,S. ( (1995) ) Effect of the FruR regulator on transcription of the pts operon in Escherichia coli. J. Biol. Chem., , 270, , 2489–2496.
Hanamura,A. and Aiba,H. ( (1991) ) Molecular mechanism of negative autoregulation of Escherichia coli crp gene. Nucleic Acids Res., , 19, , 4413–4419.
Golby,P., Kelly,D.J., Guest,J.R. and Andrews,S.C. ( (1998) ) Transcriptional regulation and organization of the dcuA and dcuB genes, encoding homologous anaerobic C4-dicarboxylate transporters in Escherichia coli. J. Bacteriol., , 180, , 6586–6596.
Plumbridge,J. and Kolb,A. ( (1991) ) CAP and Nag repressor binding to the regulatory regions of the nagE-B and manX genes of Escherichia coli. J. Mol. Biol., , 217, , 661–679.
Craig,J.E., Zhang,Y. and Gallagher,M.P. ( (1994) ) Cloning of the nupC gene of Escherichia coli encoding a nucleoside transport system, and identification of an adjacent insertion element, IS 186. Mol. Microbiol., , 11, , 1159–1168.
Munch-Petersen,A. and Jensen,N. ( (1990) ) Analysis of the regulatory region of the Escherichia coli nupG gene, encoding a nucleoside-transport protein. Eur. J. Biochem., , 190, , 547–551.
Gerlach,P., S?gaard-Andersen,L., Pedersen,H., Martinussen,J., Valentin-Hansen,P. and Bremer,E. ( (1991) ) The cyclic AMP (cAMP)-cAMP receptor protein complex functions both as an activator and as a corepressor at the tsx-p2 promoter of Escherichia coli K-12. J. Bacteriol., , 173, , 5419–5430.(Dongling Zheng, Chrystala Constantinidou)
* To whom correspondence should be addressed. Tel: +44 121 414 5564; Fax: +44 121 414 5925; Email: C.Constantinidou@bham.ac.uk
ABSTRACT
The Escherichia coli cyclic AMP receptor protein (CRP) is a global regulator that controls transcription initiation from more than 100 promoters by binding to a specific DNA sequence within cognate promoters. Many genes in the CRP regulon have been predicted simply based on the presence of DNA-binding sites within gene promoters. In this study, we have exploited a newly developed technique, run-off transcription/microarray analysis (ROMA) to define CRP-regulated promoters. Using ROMA, we identified 176 operons that were activated by CRP in vitro and 16 operons that were repressed. Using positive control mutants in different regions of CRP, we were able to classify the different promoters into class I or class II/III. A total of 104 operons were predicted to contain Class II CRP-binding sites. Sequence analysis of the operons that were repressed by CRP revealed different mechanisms for CRP inhibition. In contrast, the in vivo transcriptional profiles failed to identify most CRP-dependent regulation because of the complexity of the regulatory network. Analysis of these operons supports the hypothesis that CRP is not only a regulator of genes required for catabolism of sugars other than glucose, but also regulates the expression of a large number of other genes in E.coli. ROMA has revealed 152 hitherto unknown CRP regulons.
INTRODUCTION
The Escherichia coli cyclic AMP receptor protein (CRP) is an important transcription factor that regulates transcription initiation for more than 100 genes mainly involved in catabolism of carbon sources other than glucose (1). E.coli preferentially utilizes glucose over other sugars and only catabolizes other sugars when the supply of glucose has become depleted . The presence of glucose prevents E.coli from catabolizing alternative sugars by several mechanisms, one of which is that glucose lowers the level of cAMP, the inducer for CRP.
CRP functions as a dimer in the form of a CRP–cAMP complex, and regulates transcription initiation by binding to a symmetrical DNA sequence (consensus sequence 5'-AAATGTGATCTAGATCACATTT-3'), located near or within the promoter regions. At CRP-dependent promoters, CRP activates transcription by making direct protein–protein contacts with RNA polymerase (RNAP). Class I CRP-dependent promoters, e.g. lac, contain a single CRP-binding site upstream of the DNA-binding sites for RNAP. At these promoters, CRP activates transcription by interacting with the C-terminal domain of the RNAP subunit (CTD) via a surface-exposed patch, known as activating region 1 (AR1) (residues 156–164). Class II CRP-dependent promoters, e.g. galP1 and melR, contain a CRP-binding site overlapping the –35 hexamer for RNAP. At these promoters, CRP makes multiple interactions with RNAP, including between AR1 and CTD, the activating region 2 (AR2) (residues 19, 21 and 101) and the N-terminal domain of subunit (NTD) and between AR3 (residues 52–58) and the RNAP 70 subunit region 4. Many promoters contain tandem CRP sites and are known as Class III promoters. CRP activation at these promoters involves a combination of the Class I and Class II mechanisms. Additionally, many CRP-dependent promoters are co-dependent on a second activator (3).
The recent completion of several bacterial genome sequences has facilitated the development of computer-based bioinformatic approaches and microarray techniques to study transcription regulation at a genome-wide level. Bioinformatic analysis has allowed the prediction of the regulon of a particular transcription factor by searching for the consensus sequences, or by using algorithms to search for sequence patterns within the genome. CRP-binding sites present within the E.coli genome have been predicted using different computational approaches (4,5). In the latter study, Tan et al. have predicted 161 strong (including known sites) and 285 weak candidate CRP-binding sites, using a comparative genomic approach. However, in silico-predicted DNA-binding sites may not be occupied in vivo, or may only function in the presence of other factors. Conversely, operons with weak, but functional, sites may not have been identified using these approaches. In this study, we have utilized oligonucleotide microarray technology to determine the CRP regulation experimentally, and compared these data with the predicted CRP-binding sites.
The DNA microarray technology allows thousands of genes to be studied simultaneously in a single experiment, and thus provides a powerful tool to investigate gene function at a genomic level. Transcription profiles of E.coli in different media (6,7), under various stress conditions (8,9) and between different strains (10,11), have revealed the roles of many regulatory factors. However, the interference of other regulatory networks makes it difficult to distinguish direct effects on transcription from indirect effects, and therefore it is hard to directly link the results to a specific transcriptional factor. Recently, Helmann and co-workers (12) have developed a novel technique, which combined in vitro run-off transcription with macroarray analysis (ROMA), to define the direct effects of W on Bacillus subtilis promoters. Run-off transcription reactions in this experimental system used genomic DNA as template. The resulting 33P-labelled RNA products, transcribed by either W holoenzyme or core enzyme, were hybridized to macroarray membranes. Direct visual comparison of the core versus holoenzyme experiments allows the ready identification of genes targeted by W. The result from ROMA for W from B.subtilis was consistent with those determined by promoter consensus searching and in vivo transcriptional profiling. In this study, the term ‘in vitro’ refers to experiments completed with purified DNA and protein components, whereas ‘in vivo’ refers to experiments in E.coli cells grown in planktonic culture.
In this study, the ROMA procedure was modified so as to exploit oligonucleotide arrays on glass slides and ROMA was then used to identify CRP-regulated operons in vitro.
MATERIALS AND METHODS
Strains and growth conditions
E.coli K-12 MG1655 (CGSC 7740) was used in this study (13). A crp derivative of MG1655 was constructed using the gene disruption method of Datsenko and Wanner (14). Primers: CRP P1: 5'-GCTCTGGAGAAAGCTTATAACAGAGG ATAACCGCGCGTGTAGGCTGGAGCTGCTTC-3' and CRP P2:5'-TGGCGCGCTACCAGGTAACGCGCCACTCCGACGGGACATATGAATATCCTCCTTAG-3' were used to amplify a chloramphenicol resistance gene cassette from plasmid pKD3 (14). The bases underlined in the primer sequences shown above represent the short regions of homology to genomic sequences flanking crp, which were used in Red-mediated recombination of the chloramphenicol resistance cassette into the chromosome. Transformants were grown on Luria–Bertani (LB) agar (15) containing 0.2% glucose and 50 μg/ml chloramphenicol. PCR was used to screen chloramphenicol resistant transformants, for replacement of crp with the chloramphenicol resistance cassette, using the primers CRPSCREEN 1: 5'-GGATGCTACAGTAATACATTGATG-3', and CRPSCREEN2: 5'-GACCGAATCGTAATTCGCCAAG-3'. Amplicons generated by the screening PCR were sequenced using a BigDyeTM version 3 sequencing kit (Applied Biosystems, Warrington, UK) and analysed on an ABI 3700 sequencer (Applied Biosystems). MG1655 crp strains were grown on Maconkey agar containing maltose (1% w/v) to confirm the phenotype.
For in vivo microarray experiments, both wild-type and crp strains were grown in M9 minimal media (14) containing 0.2% fructose at 37°C to OD600 0.8. Glucose (0.2%) was then added to the cultures, which were grown for a further 15 min, prior to harvesting. The cell samples for RNA preparations were collected before and after the addition of glucose. Two volumes of RNAProtect (Qiagen Ltd, Crawley, UK) were added per volume of bacterial culture, prior to centrifugation.
E.coli oligonucleotide array
E.coli oligonucleotide arrays were produced by the UBEC group (University of Birmingham E.coli group). The oligonucleotides were designed and synthesized by Qiagen Operon Ltd as the E.coli Array ready oligo set vs 1.0 (Qiagen) and represented 4289 E.coli K12 strain (MG1655) open reading frames (ORFs); 1416 ORFs designed to the O157:H7 (EDL933) strain; and 273 ORFs unique to the O157:H7 (Sakai) strain (http://oligos.qiagen.com/arrays/oligosets_ecoli.php). In addition, 110 oligonucleotides representing ORFs from the EDL933 and Sakai plasmids were added to the oligonucleotide array set. Also included within the array set were 12 positive and 12 negative control oligonucleotides, which were printed within each subarray. The average size of each oligonucleotide used in the array was a 70mer, with a Tm = 75 ± 5°C. The position of each oligonucleotide within the ORF was more than 40 bases away from the 3' end. The oligonucleotides were arrayed on Corning CMT-GAPS II slides in 48 blocks, each containing 324 spots (18 rows by 18 columns) using a MicroGrid II robot (BioRobotics, UK). Each oligonucleotide was printed in duplicate on the array, and Amersham LucideaTM Universal ScorecardTM (Amersham, UK) controls were printed within each subarray.
Genomic run-off transcription for ROMA
Genomic DNA was isolated from the E.coli MG1655 (CGSC 7740) wild-type strain using phenol/chloroform extraction (http://www.research.umbc.edu/~jwolf/m1.htm), digested with EcoRI overnight and purified by phenol/chloroform extraction, followed by isopropanol precipitation. RNAP holoenzyme was purified as described previously (16). Wild-type CRP, AR1-mutated CRP (HL159) and AR2-mutated CRP (KE101) were purified as described previously (17).
For a single ROMA experiment, two run-off transcription reactions were set up in parallel. The control reaction contained 4 μg of EcoRI digested genomic DNA, and 1 mM each of ATP, GTP, CTP and UTP, in transcription buffer (40 mM Tris/acetate, pH 7.9, 10 mM MgCl2, 1 mM DTT, 100 mM KCl, 0.1 mg/ml BSA), which contained RNase Out (Invitrogen, UK) at a concentration of 50 U/reaction. The test reaction contained 2 μl of 5 mM cAMP (final concentration 200 μM), and 4 μl of 10 μM CRP (40 pmol of monomer, at a final concentration of 800 nM) in addition to the components of the control mixture. Both run-off transcription reaction mixtures were incubated for 10 min at 37°C and transcription started by the addition of 20 pmol of RNA polymerase holoenzyme. Run-off transcription reactions were incubated at 37°C for 30 min. After incubation, the reactions were stopped by the addition of 5 μl of 250 mM EDTA and placed on ice.
RNA purification and labelling
Both RNA transcripts from in vitro transcription reactions and total RNA from cell cultures were extracted using the Qiagen RNeasy mini kit. On-column DNase I digestion (Qiagen Ltd) was used to remove contaminating genomic DNA from the RNA preparations. The concentrations of RNA or in vitro transcription products were determined at OD260 and OD280. An indirect labelling method was used to obtain fluorescence-labelled cDNA for any RNA sample. Briefly, all RNA transcripts from each reaction or 10–20 μg of total RNA were mixed with 6 μg of random hexamers (Amersham Biosciences, Little Chalfont, UK) to a final volume of 18.4 μl, incubated at 70°C for 10 min and snap-cooled in ice. Reverse transcription labelling mixture (11.6 μl) was then added to the RNA template and random hexamers, which contained 0.5 mM dATP, dCTP, dGTP, 0.2 mM dTTP, 0.3 mM aminoallyl-dUTP (aa-dUTP), RNase inhibitor (30 U), 400 U SuperScript II (Invitrogen, Paisley, UK), 10 mM DTT and 1x first strand buffer. The mixture was incubated at 42°C for 3 h or overnight to generate aminoallyl-labelled cDNA. To hydrolyse the RNA template, 10 μl of 0.5 M EDTA and 10 μl of 1 M NaOH were added to the reaction and incubated at 65°C for 15 min. The reaction was neutralized by the addition of 10 μl 1 M HCl. Unincorporated aa-dUTP and free amines were removed by washing the cDNA in a microconcentrator (Microcon YM-30, Millipore) and the sample was then vacuum dried. The aminoallyl-cDNA pellet was resuspended in 4.5 μl of 0.1 M sodium carbonate buffer (pH 9.0) and coupled with Cy3 or Cy5 monoreactive dye (Amersham), prepared in dimethyl sulfoxide for 2 h at room temperature in the dark. For the ROMA experiment, the aa-cDNA from the control reactions was coupled with Cy3 and aa-cDNA from the CRP reactions coupled with Cy5. For in vivo transcriptional profiling, the aa-cDNA obtained from either wild-type strain or CRP strains before the addition of glucose was coupled with Cy3 and after the addition of glucose was coupled with Cy5. Uncoupled dyes were removed using a QIAquick PCR purification kit (Qiagen). LucideaTM Scorecard (Amersham) mRNA spike mixes were added to labelling reactions for both in vivo and ROMA experiments.
For the in vivo microarray work, the quality and concentration of the RNA prepared was assessed using the Agilent 2100 Bioanalyser. Total RNA (10–20 μg) was labelled using the CyScribe Post-Labelling Kit (Amersham Biosciences) as described by the manufacturer.
Prehybridization and hybridization
Before hybridization, the arrayed slides were prehybridized in a buffer containing 25% formamide, 5x SSC, 0.1% SDS and 10 mg/ml BSA at 42°C for 2 h and washed by dipping twice in distilled water. The slides were then dipped in 95% ethanol for 1 sec and dried in a clean 50 ml centrifuge tube by centrifugation at 1500 g for 10 min. The Cy3 and Cy5 labelled cDNAs were mixed, vacuum dried and resuspended in 70 μl hybridization buffer containing 25% formamide, 5x SSC, 2 μl of 50x Denhardt's, 2 μl yeast tRNA (20 μg/ml) and 0.1% SDS. The labelled cDNAs were denatured by heating at 95°C for 5 min, and applied to the prehybridized slide in a CMT-Hybridization chamber (Corning Inc., Corning, NY). A HybriSlip (Sigma) was carefully lowered onto the slide. To maintain humidity inside the chamber, 10 μl of distilled water was added to the two reservoir wells. The chamber was then tightly sealed and incubated at 42°C for 16–20 h in the dark. The slide was then removed from the chamber, washed for 5 min sequentially in 2x SSC/0.1% SDS buffer, 0.1x SSC/0.1% SDS buffer and 0.1x SSC buffer, rinsed in distilled water for 5 s and dried by centrifugation at 2000 r.p.m. for 10 min. The hybridized slides were scanned with a confocal laser scanner (Axon GenePix 4000A) using appropriate gains on the photomultiplier tube to obtain the highest intensity without saturation. Three replicates were completed for each experiment.
Image extraction and data normalization
Scanned images for Cy3 and Cy5 were then overlaid with GenePix Pro 3.0 software. Only data generated from spots representing E.coli MG1655 genes were analysed in our studies. Spots with background-subtracted intensity lower than 100 in both Cy3 and Cy5 channels were filtered out. GeneSpring software (SiliconGenetics) was used for global normalization and density-dependent normalization (Lowess) to correct artefacts dependent on density or caused by different dye incorporation rates for the two dyes. The duplicate spots for each gene on a single slide were taken as two individual spots. Three independent experiments were performed for each comparison and, therefore, six replicate data sets were obtained for each gene. The normalized data from GeneSpring software were exported to a Microsoft Excel spreadsheet.
Data reproducibility
To assess data reproducibility, the spot-to-spot variation was calculated as described by Loos et al. (18) and correlation between data sets analysed by the Pearson correlation coefficient (r). The spot-to-spot variation was represented by the percent error calculated by dividing the SD by the average ratio for each gene (% = /μ). Average variation within a slide was calculated as the average of the gene spot-to-spot variations on that slide (two spots), and average variation across slides was calculated as the average of gene spot-to-spot variations based on the six spots of the data set. The latter variation included spot, slide and cDNA variability. The Pearson correlation coefficient (r) was calculated for Cy3 data sets from any two slides and for Cy5/Cy3 from any two replicate slides.
Identification of differentially transcribed genes
Differentially transcribed genes were selected using an outlier iteration method (18,19). The data for each gene were averaged and the geometric mean and SD were calculated for the entire population. Any gene with a log-ratio more than three SDs away from the mean was considered an outlier. Outliers were then removed from the population and retained within the differentially expressed subset, and the mean and SDs were recalculated for the rest of the data. The step was repeated until few or no outliers were detected. The 99% predictive interval (PI) was set for the final cut-off ratio to define the remaining differentially transcribed genes within the now symmetrical and well-defined distribution.
The in vitro transcription profiles of CRP derivatives were compared with wild-type CRP, and significant changes were identified by outlier iteration. The average of the log-ratios for each gene from the CRP derivative experiments was normalized to the wild-type CRP experiment. The geometric mean and SD were calculated for the entire population. The outliers were selected as stated above and the 95% PI was set as a final cut-off. The genes with a log-ratio change between two experiments falling in the outlier group or beyond the PI were regarded as differentially regulated by CRP mutants and wild-type CRP.
Sequence processing
The differentially transcribed genes were grouped into operons using the RegulonDB database at http://www.cifn.unam.mx/Computational_Genomics/regulondb. The current version of the RegulonDB database (RegulonDB 4.0) identifies 87 CRP-regulated operons that have been verified experimentally. In this study, 24 of these operons were shown in the ROMA experiments to be regulated by CRP and, therefore, were further analysed. CRP-binding site predictions were based on the previous study (5), where a Positional Weight Matrix was generated from alignment of known CRP-binding sites and used to determine the strength of a site. In this study, a strong site was defined as a sequence with a score more than 10 (consistent with the prediction by Tan et al.) and a weak site as a sequence with a score between 5 and 10. Sequences with a score below 5 were rejected.
RESULTS
Combination of run-off transcription and microarray analysis (ROMA) on glass slides
The original ROMA procedure used macroarray membranes to analyse radioactively labelled RNA transcripts (12). In our study, the procedure was modified by using a post-transcriptional fluorescence-labelling procedure to enable hybridization to a sense-oligonucleotide array. To gauge data reproducibility, data from each of three replicate slides were subjected to statistical analysis. Gene spot-to-spot variation was 5.0% (range, 4.1–6.3%) between replicate spots within a given slide, and 13.9% (range, 12.6–15.4%) among six spots across three slides. The latter number includes both the hybridization variation and biological variation. Comparison of the duplicated probe sets within a single hybridization showed good reproducibility of the fluorescence signals with a Pearson correlation coefficient (r) of >0.98. This result illustrates uniform hybridization of the microarray. The reproducibility of signal between replicate reactions was lower, with a Pearson correlation coefficient (r) of >0.80. This variability was considered to be mainly due to variations from independent reactions. In addition, the detail of signal intensities showed that 90% of all genes gave a detectable signal.
CRP regulatory profiles in the ROMA system
Comparison of RNA transcripts from reactions containing wild-type CRP to control reactions containing only RNAP revealed many differentially transcribed genes (Figure 1a). Data for the majority of genes fall within the threshold values, with a ratio between +2 and –2, and therefore indicates that these genes are not significantly regulated by CRP. As expected, transcripts of many genes were increased by the addition of CRP and transcripts of very few genes were reduced, which is consistent with the view that CRP activates most of the genes that it regulates. A total of 280 genes, present in 188 different operons, had significantly higher transcriptional levels and 20 genes, in 16 operons, had reduced transcriptional levels in response to CRP. Taken together, these data show that 7% of the genes in the E.coli genome were transcribed differentially upon the addition of CRP in the ROMA experiment. The highest activation, 13-fold induction, was observed at the sdaC gene encoding a protein that is a putative serine transporter. Of the 176 operons identified by ROMA in this study, 24 are known members of the CRP regulon (Table 2), i.e. the promoter regions and CRP sites have been mapped experimentally. For genes present within the same polycistronic operon, the signal for the promoter-proximal gene should be the same or higher than the signal for promoter distal genes. Of the 188 operons activated by CRP, the promoter-proximal gene was activated in the vast majority (176 operons). For 8 of these 176 operons, the promoter-proximal gene was subjected to only marginal activation (above 95% PI), while the promoter distal gene(s) were activated at a significant level. In only 12 operons was a promoter-distal gene up-regulated and not the promoter-proximal gene (Tables 1 and 2). These 12 promoters were not included in the CRP-regulated set. Up-regulation of promoter-distal genes in the absence of regulation of the promoter-proximal genes may represent false positives in the assay; alternatively it may be due to CRP-dependent promoters within operons that are active in vitro.
Figure 1. Logarithmic-scale scatter plots of spot intensities. The data were subjected to global and density-dependent normalization, and plotted on GeneSpring 4.2.1. Each data point corresponds to the average of two spots representing an individual MG1655 gene. The ratios were colour coded, with red for ratio above one and green for ratio below one. The middle line passes genes with no change at two conditions (ratio = 1) and the other two lines demarcate threshold values for genes with significant increase or decrease in transcription in response to CRP deletion (ratio = ±2). (a) Transcriptional profiles from reactions with wild-type CRP (wt. CRP) versus without CRP (–CRP). (b) Transcriptional profiles from reactions with AR1-mutated CRP (HL159) versus without CRP (–CRP). (c) Transcriptional profiles from reactions with AR2-mutated CRP (KE101) versus without CRP (–CRP).
Table 2. Operons which have been experimentally verified as being regulated by CRP
Table 1. Operons activated by wild-type CRP in ROMA experiment
Determination of CRP-binding sites
We scanned the upstream regions of the 176 up-regulated operons for putative CRP-binding sites to determine whether the activation observed is due to a site that matches the CRP-binding site consensus. A previous study using a comparative genomic approach (5) identified 447 operons on the E.coli genome that contain putative CRP sites upstream from the translation start site. Using the same approach as Tan et al., 55 of the 176 (31%) operons identified in our study were found to contain a CRP site. Tan et al. predicted CRP-binding sites with a good match to the consensus (10 as a cut-off score). However, some known CRP-dependent promoters contain weak CRP sites that are not identified when using a cut-off of 10 (see Table 2), e.g. the melR CRP site has a score of 5.5 and rpoH as two sites with scores of 2.2 and 7.4. In addition, they also identified many false positives.
We therefore scanned the regulatory regions of the remaining operons for sequences with a lower match to the consensus (5 as a cut-off score) and found an additional 70 operons that contain CRP sites with a score between 5 and 10 (Tables 1 and 2). Thus, a total of 125 operons identified in the ROMA assay contain CRP sites identified by either Tan et al. (5) or this study. The remaining 51 operons (29%) identified by ROMA do not contain a recognizable CRP site (score <5).
Effect of CRP AR1 and AR2 on activation
Previous studies have shown that CRP activation is dependent on the interactions of its activating regions with subunits of RNAP . In this study, the effects of two CRP derivatives containing a mutation at either AR1 or AR2 were analysed. The HL159-CRP carries a His-to-Leu mutation at position 159 within AR1 and has been shown to be defective at both Class I and Class II promoters. Whereas KE101-CRP, which contains a Lys-to-Glu mutation at position 101 within AR2, fails to activate transcription at Class II promoters, but still functions at Class I promoters (20).
The transcriptional level of each gene in the presence of either HL159-CRP or KE101-CRP was compared with those from the control reaction. As expected, the results showed that HL159-CRP failed to activate transcription of 86% (151/176) of the CRP-dependent genes identified in the initial ROMA experiment (Table 1, Figure 2 and Figure 1b). Therefore only 14% (25/176) of operons were activated by HL159-CRP to the same level as by wild-type CRP. The AR2 mutant KE101-CRP was defective for activation at 59% (104/176) of CRP-dependent operons (Table 1, Figure 2 and Figure 1c). The AR2 mutant, therefore, still activated transcription for many genes although the level of activation was reduced. We predict that the promoter regions of the operons whose transcription is not up-regulated by the AR2 mutant contain a CRP-binding site overlapping the –35 hexamer (Class II binding site).
Figure 2. Regulatory map of 50 highest CRP-regulated operons. The figure indicates the activation by HL159-CRP (middle) and KE101-CRP (right) relative to wild-type CRP (=1), generated by GeneSpring clustering in distance order, so that those with similar regulation pattern were grouped together. The relative activation changes are colour coded: yellow indicates that a gene is regulated at the same level as wild-type CRP, blue indicates that a gene is regulated at a much lower level than wild-type CRP and red indicates that a gene is regulated at a higher level.
To validate the ROMA data, we considered the 24 up-regulated operons whose promoter regions and CRP sites have been mapped experimentally. Of these 24 operons, 7 contain a single Class I CRP-dependent promoter (possess a CRP site far upstream of the –35 hexamer), 9 contain a single Class II promoter (possess a CRP site overlapping the –35 hexamer) and 8 contain a Class III promoter with tandem CRP-binding sites (Table 2). All Class III promoters contain a Class II binding site and a Class I binding site. In the ROMA assay, HL159-CRP failed to activate all promoters, except the two Class III promoters, deoCABD and rpoH. Both the deoCABD and the rpoH operons have several promoter regions that are under differential regulation by CRP (21,22) and therefore the effect of the AR1 mutation may be compromised by transcription from different promoters. In contrast to HL159-CRP, KE101-CRP could still activate class I promoters, but failed to activate all class II promoters and half of the class III promoters. It is therefore clear from the experiments with activating region mutants that the CRP regulation observed in ROMA is similar to that observed in vivo. In addition, it allows us to predict that 59% of the CRP-regulated operons identified by ROMA contain a class II CRP site and are therefore class II or class III promoters.
CRP-repressed operons in the ROMA system
Although CRP is predominantly an activator, we found 16 operons where the promoter-proximal gene had reduced levels of transcription in response to CRP in the ROMA experiments (Table 3). The strongest repression, 3-fold reduction, occurred at the yjcB gene. In contrast to activation, the repression levels of most operons were not significantly affected by either HL159-CRP or KE101-CRP, which indicates that most CRP repression is not dependent on either the AR1 or the AR2 determinant. However, repression at four operons was slightly affected by the AR1 mutation, suggesting some role for AR1 at these operons. We have identified possible CRP-binding sites within 11 of the 16 promoter regions, 8 operons contain a strong CRP-binding site (>10) and 3 operons contain a weak site (5–10). The yjcB and ycfR operons contain a CRP site almost identical to the consensus.
Table 3. Operons repressed by wild-type CRP in ROMA experiment
Of the 16 repressed operons, promoter regions of two, nirB and pncB, have been determined experimentally (23,24). Sequence analysis indicates that their CRP-binding sites overlap the binding sites for RNAP (Figure 3) and it is therefore probable that CRP represses transcription at the nirB and pncB promoters by either blocking the interaction between RNAP and the promoter DNA, or blocking transcription from an alternative upstream promoter. The position of putative CRP-binding sites relative to the RNAP-binding site was determined for the other repressed operons (Figure 3). The position of the CRP site is variable, but at most promoters the CRP site overlaps the region between the –35 and –10 hexamers. Therefore, repression at these operons might also involve a simple blocking mechanism. However, at the metK, ybiS and yjcB predicted promoter regions, the CRP sites are located 3 bp upstream of the –35 hexamer. The yjcB promoter contains a very strong CRP-binding site (score = 22). The promoter architecture might be such that the CRP-binding site still stops productive binding of RNAP.
Figure 3. Non-template strand sequences of promoters repressed by CRP. Except nirB and pncB, promoter regions are predicted by searching the –10 and –35 sequences upstream of 10 down-regulated genes with a CRP site. The predicted –10 and –35 hexamers are underlined and in boldface with the –10 regions aligned together. The potential CRP-binding sites are shown as shaded boxes. *ydfK and ynaE are both associated with different prophage (Quin and Rac, respectively) but have essentially identical sequences.
Identification of CRP regulon by in vivo transcriptional profiling
An in vivo microarray experiment was also designed to identify CRP-regulated genes. Both wild-type and crp strains were grown in minimal media containing fructose and then pulsed with glucose, and the transcriptome before the addition of glucose for both strains was compared with that after glucose addition. In the presence of glucose, the cAMP level is expected to be decreased, and hence the cAMP CRP level, thus the CRP-dependent regulation is repressed. However, glucose also affects gene expression through other mechanisms, thus glucose effects will be observed in the crp strain. The CRP-regulated genes in this study were defined as genes regulated by glucose in the wild-type strain, but not regulated in the crp strain. In total, we identified only 17 operons repressed by glucose (CRP-activated operons) and six glucose-activated operons (CRP-repressed operons), listed in Table 4. In the 17 CRP-activated operons, 9 operons are known members of the CRP regulon and 7 operons were activated by CRP in the ROMA experiment. Our in vivo experiments failed to identify most CRP-regulated genes.
Table 4. CRP-regulated genesa identified by in vivo transcription profiling
DISCUSSION
The combination of run-off transcription and microarray analysis (ROMA) has shown advantages in defining W-dependent promoters in B.subtilis (12). In this study, we have established ROMA exploiting microarray glass slides and further demonstrated its application to investigate a specific transcription factor. The microarray glass slide has several advantages over a macroarray membrane. First, the slide array allows more genes to be studied in a single experiment (up to 50 000 genes) compared with a few thousand genes that can be printed on a macroarray membrane. Second, RNA from two different sources of interest can be dual fluorescence labelled and simultaneously hybridized on a single slide, whereas nylon arrays are generally probed in serial or parallel hybridization reactions to allow comparison. Third, the utilization of fluorescence instead of radioactive material is safer, easier to handle and produces higher resolution and lower background. Therefore, the establishment of ROMA on glass slides provides a convenient system to obtain high quality information. Our results have shown the high sensitivity and reproducibility of the system. However, our procedure does have some disadvantages compared with the original macroarray procedure. First, several purification steps are required which might result in the loss of short transcripts. Second, a large amount of template DNA and protein is required. Third, any trace of template DNA remaining after RNA purification will result in high background hybridization that might mask the actual induction level.
Using this newly established ROMA system, we have identified 152 novel CRP-dependent operons and 24 operons corresponding to previously known CRP regulons. Regulation at these operons was further verified by using two CRP mutants. Our experiments indicate that ROMA is a very powerful technique to identify direct regulation by a transcription initiation factor. The ROMA system was, however, unable to identify some known CRP-regulated operons. There are several reasons that could account for these false negatives. First, some operons require other factors for CRP to activate transcription, e.g. AraC at the araBAD promoter (25), which are missing in the ROMA system. Second, transcripts that can initiate from different promoters upstream from the same operon cannot be resolved. So genes where CRP alters the transcription start point will not be identified. For example, the galP1 promoter is a well-known Class II CRP-dependent promoter. However, in the absence of CRP, transcription of the gal operon can initiate from the upstream galP2 promoter (26), which will mask the weak transcription from galP1. Third, certain promoter structures result in a situation where CRP can activate transcription by binding to an upstream site but can also repress transcription by binding to a downstream site, e.g. the crp promoter (27). In addition, some operons may not be transcribed efficiently due to DNA relaxation, or in the experimental system described here, the existence of an EcoRI restriction site within the promoter region or coding region. Finally, because of the necessity for data manipulation to identify only the most robust changes in transcription, promoters that are only weakly activated by CRP in vitro will not be detected. The CRP regulon may therefore contain more operons than were identified in this study.
In the W ROMA experiment, 50% of the genes identified using ROMA were also identified using consensus sequence searching and in vivo transcriptional profiling (12). However, in the case of CRP, these approaches have serious limitations. Comparison of the ability of each method to identify known members of the CRP regulon demonstrates that sequence searching generates the highest number of hits, and in vivo transcriptional profiling generates the fewest (Figure 4). A combination of the three methods can identify more than 80% of known operons, but failed to identify 14 operons known to be regulated by CRP. Although the sequence prediction generates the highest number of hits, it is probable that this includes many false positives and is therefore not necessarily the best approach. In addition, many weak sites are difficult to identify by sequence searching. Lowering the cut-off score would facilitate the identification of known sites but would inevitably lead to more false positives. Besides DNA sequence, CRP activation is dependent on the location of the binding site. Previous studies have indicated that at a Class I promoter, the CRP-binding site is normally positioned at –61.5, –71.5, –81.5 or –91.5, and no activation occurs when CRP sites are located further than –113.5, which might be beyond the reach of CTD . Therefore, many in silico predicted sites might be null sites that are not functional during transcription. For promoters containing tandem CRP sites, the space between the two sites is an important determinant for synergistic activation (29–32). Therefore, although some operons contain several predicted binding sites, we hypothesized that at least some sites are not functional due to improper spacing. This may also explain why some operons with strong predicted binding sites were not regulated by CRP in our experiments. It is also probable that sequence searching generates many false negatives due to the ability of CRP to bind variable sequences. For example, previous work has identified the CRP sites within the melR, rpoH and rbs promoters that play important roles in regulation of these promoters (22). These sites have scores below 10 and were, therefore, not identified in the sequence search. It is therefore possible that several promoters identified by ROMA that do not contain a putative CRP-binding site, e.g. inaA and yfaH, may still be regulated by CRP.
Figure 4. Venn diagram of the CRP regulon as identified by three genomic approaches: ROMA, in vivo transcriptional profiling, sequence search (5). The result for (a) 87 known (experimentally verified) CRP regulon collected in RegulonDB database and (b) all operons are shown. The number of operons identified by each approach is presented in a coloured circle. The numbers covered by more than one circles are operons identified by two or three methods.
Comparing activation levels of operons with strong CRP sites to those with weak sites indicated that there is no significant relationship between the degree of activation and the ‘quality’ of CRP-binding sites. For example, the focA-pflB operon which is activated 3.9-fold contains a relatively weak CRP site (score = 10.1), whereas the tnaL operon which is activated 2-fold contains a very strong CRP-binding site (score = 19.6) (Table 2). Therefore, conservation of CRP-binding sequence is not sufficient to predict CRP activation levels.
In our study, the in vivo transcriptional profiles failed to identify many CRP-regulated genes. The main reason for this is the complexity of the CRP regulon. CRP activation at some promoters, such as melAB and araBAD, is dependent on the presence of an additional regulator (MelR and AraC, respectively) (33,25) that is only induced by a specific substrate. It is impossible to include all inducers in the growth media and, moreover, the presence of extra inducers may trigger expression of other genes that are independent on CRP, further complicating the interpretation of the data. In addition, most CRP-dependent promoters are subjected to repression by other factors, which will mask the CRP activation in vivo. For example, the galE promoter is repressed by GalR, the lacZ promoter repressed by LacI and many CRP-activated promoters are repressed by CytR (Table 2). Therefore, the variability of CRP regulation at different operons makes it difficult to design an in vivo microarray experiment to distinguish direct effects from indirect effects. ROMA is better suited for identifying the regulon of a factor that is part of a complex regulatory network. In the six glucose-induced operons, two operons, gcd and proP, are known CRP-repressed operons, which contain a strong CRP-binding site between the –10 and –35 regions, thus blocking RNAP binding. As observed in previous studies (34,35), the ptsG gene encoding the major glucose transporter was highly induced by glucose in the wild-type stain. Previous studies have indicated that the ptsG promoter, which contains a CRP activation site (see Table 2) (36), is both activated by CRP and strongly repressed by the pleiotropic transcriptional repressor Mlc. The glucose induction of ptsG is through a complicate Mlc-dependent mechanism that involves several layers of regulation (34–39). The case of ptsG demonstrates that an in vivo indirect effect can mask the real regulation by CRP, sometimes leading to the opposite conclusion. The regulation observed in the trp operon might also have resulted from an indirect effect, as there is no evidence from previous studies to indicate that CRP regulates this operon.
Further evidence of the limitations of in vivo transcription profiling of a complex regulon has come from a recent study of CRP using an Affymetrix E.coli array (40). Gosset et al completed a similar in vivo transcriptomic study of CRP and identified 39 operons under CRP-dependent glucose repression and 19 operons under CRP-dependent glucose activation. Among these operons, only seven glucose-repressed operons and one glucose-activated operon were identified in our in vivo transcriptional profiling (Table 4). However, direct comparison of the two data sets is difficult because of differences in the strain and the growth conditions used. Gosset et al in their study used the E.coli strain BW25113 and an isogenic crp mutant derivative and grew their cells in LB media with or without glucose. The difference in composition between the LB media used in the Gosset study and minimal media used in this study could contribute to a different profile of induction of some genes. For example, the presence of fructose in our growth conditions meant that the fru operon was identified as CRP regulated in our study, but not regulated in the Gosset study due to the lack of fructose in LB media. Gosset et al. (40) used long-term growth in glucose plus media rather than a glucose shock which was used in this study. Long-term growth in glucose should result in higher levels of induction but can also lead to a larger number of indirect effects, such as regulation of ribosomal protein-encoding genes, RNA-encoding genes, stress-related genes and temperature shock genes, which were significantly affected in their study. One-third of operons subject to CRP-dependent glucose repression (13/39) in Gosset et al. were activated by CRP in ROMA experiment. This figure is similar to the data obtained by in vivo transcription profiling in this study (7/17). This indicates that ROMA has limitations to detect a certain set of operons that may be subjected to complex regulation or regulated indirectly in vivo. However, as seen in this study, Gosset et al. failed to identify many CRP-regulated genes, such as rpoH, melR and rbs, so neither study produced a definitive list of members of the CRP regulon. This further indicates that the study of complex regulons, such as CRP, in vivo has many limitations. However, combining several approaches, such as in vivo profiling, sequence analysis and ROMA increases the likelihood of obtaining a more accurate definition of a complex regulon.
The mechanism of CRP activation, in particular the role of different activating regions, has been studied using several well-characterized promoters. At Class I promoters, CRP binds to a DNA sequence upstream of the RNAP-binding site and makes direct protein–protein contact to CTD via AR1 of the downstream subunit of the CRP dimer, and this interaction recruits CTD to its DNA target immediately downstream of the CRP-binding site. At Class II promoters, CRP binds to a site overlapping the –35 hexamer and makes several contacts with RNAP: AR1 of the upstream subunit of the CRP dimer binds CTD, AR2 of the downstream subunit of the CRP dimer binds NTD and AR3 of the downstream subunit binds region 4 of 70. The CTD binds to its target upstream of the CRP site. At Class III promoters that contain tandem sites, CRP activation involves both Class I and Class II mechanisms . Nearly 90% of operons identified by ROMA showed dependence on the AR1 determinant of CRP for full activation, while AR2 is required for activation of 59% (104) of operons. We predicted that these 104 AR2-dependent operons (including 14 known operons) contain a Class II binding site in their promoter regions.
It is not surprising that transcription of some Class III operons was still activated by KE101-CRP though they all contain a Class II binding site. It has been previously shown that CRP activation at this kind of promoter involves a Class I mechanism (upstream site) and a Class II mechanism (downstream site) (30). Inactivation of AR2 only affects the proximal Class II activation, while the distal Class I site could still drive efficient activation. In some cases, where CRP activation is mainly dependent on the proximal site, the AR2 determinant will be crucial.
Interestingly, a few known FNR-regulated operons, adhE, sdhC, sodA, cyoA and focA, were also activated by CRP in the ROMA experiments. FNR is a CRP homologue and regulates genes during anaerobic growth. The consensus sequence for FNR is similar to that for CRP consensus, and previous studies have shown that FNR activates transcription initiation in a similar way to CRP and that both proteins can bind to the DNA site for the other protein (41,42). Therefore, we propose that some CRP-regulated operons identified in this study are actually regulated by FNR in vivo and contribute to anaerobic growth.
At some promoters, transcription activation by CRP can be repressed by a transcriptional regulator protein, CytR. The CytR-regulated promoters usually contain two DNA sites for CRP, centred at positions –41.5 and –93.5 with respect to the transcription start point, and a DNA site for CytR located between the two CRP-binding sites (43). Cytidine is an allosteric inducer of CytR. In the presence of cytidine, CytR is inactive, and thus CRP can activate transcription in a Class III dependent manner. In the absence of cytidine, CytR is active and binds between the two CRP dimers, repressing CRP-dependent transcription . CytR binds to two 5'-TTGCAA-3' motifs between two CRP-binding sites separated by 10–30 bp (44). The deoC, nupG and tsx operons that are under CytR regulation were activated by CRP in our study. Many operons identified by ROMA possess tandem CRP-binding sites separated by 30–60 bp (Table 1). Sequence analysis of the promoter sequence of these operons may identify more CytR-regulated promoters.
In addition to activation, our data also shows that CRP can serve as a repressor at some promoters. There are several different mechanisms by which repressors can inhibit transcription initiation. The simplest mechanism is by blocking the interaction between RNAP and a promoter. This can occur if a binding site for a repressor protein is located overlapping the binding site for RNAP at a promoter, e.g. over the transcription start site or the –10 hexamer. The blocking mechanism is common for most repressors, such as LacI at the lacUV5 promoter and IclR at the iclR promoter (45–47). CRP repression at nirB, pncB, yeeF, ycfR, yhfC, ydfK, ynaE and apbA might also involve this mechanism since at these promoter regions the CRP-binding site overlaps the sites for RNAP. Repressors also can inhibit transcription initiation by direct contact with RNAP, a process more commonly associated with activators. The P4 protein of the B.subtilis bacteriophage 29 represses the transcription of early promoter A2C by inhibiting promoter clearance (48). The binding of P4 overstabilizes the open complex formed by RNAP and DNA, thus impeding the following promoter escape step. The repression at metK, ybiS and yjcB might involve this mechanism since they all contain a strong CRP-binding site immediately upstream of a recognizable –35 hexamer. Interestingly, the CRP-binding site and the –35 region at three promoters are separated by 3 bp, which suggests that the structure of protein–protein interaction might play an important role.
CRP has been identified as a regulator of genes required for catabolism of sugars other than glucose and a large number of other genes . The CRP regulon defined in this study includes several operons that are involved in carbohydrate transport and metabolism, such as fucPI for fucose, kdgT for gluconate and fucAO for fuculose. CRP also regulates transcription of genes required for energy production, amino acid metabolism, nucleotide metabolism and ion transport systems. In addition, CRP can regulate transcription of other transcription factors, such as MelR, RpoH, BlgG, Fis and PdhR, which could further regulate transcription of genes involved in the above processes. We predict that there are several hundred genes that are directly or indirectly subjected to regulation by CRP. In our list, the functions of more than one-third of the gene products are unknown. Identification of function of the gene product of these genes will help to define the biological function of CRP in E.coli.
ACKNOWLEDGEMENTS
We thank Nigel J. Savery and Georgina S. Lloyd for the CRP protein used in this study. This work was supported by Biotechnology and Biological Sciences Research Council Grants EGA16107and JIF 13209, and by a Darwin Trust of Edinburgh studentship (to D.Z.). The authors thank Steve J.W. Busby for his interest in this work and stimulating discussions.
REFERENCES
Kolb,A., Busby,S., Buc,H., Garges,S. and Adhya,S. ( (1993) ) Transcriptional regulation by cAMP and its receptor protein. Annu. Rev. Biochem., , 62, , 749–795.
Postma,P. ( (1986) ) Catabolite repression and related phenomena. Symp. Soc. Gen. Microbiol., , 39, , 21–49.
Busby,S. and Ebright,R. ( (1999) ) Transcription activation by catabolite activator protein (CAP). J. Mol. Biol., , 293, , 199–213.
Robison,K., McGuire,A.M. and Church,G.M. ( (1998) ) A comprehensive library of DNA-binding site matrices for 55 proteins applied to the complete Escherichia coli K-12 genome. J. Mol. Biol., , 284, , 241–254.
Tan,K., Moreno-Hagelsieb,G., Collado-Vides,J. and Stormo,G.D. ( (2001) ) A comparative genomics approach to prediction of new members of regulons. Genome Res., , 11, , 566–584.
Tao,H., Bausch,C., Richmond,C., Blattner,F.R. and Conway,T. ( (1999) ) Functional genomics: expression analysis of Escherichia coli growing on minimal and rich media. J. Bacteriol., , 181, , 6425–6440.
Oh,M.K., Rohlin,L., Kao,K.C. and Liao,J.C. ( (2002) ) Global expression profiling of acetate-grown Escherichia coli. J. Biol. Chem., , 277, , 13175–13183.
Zheng,M., Wang,X., Templeton,L.J., Smulski,D.R., LaRossa,R.A. and Storz,G. ( (2001) ) DNA microarray-mediated transcriptional profiling of the Escherichia coli response to hydrogen peroxide. J. Bacteriol., , 183, , 4562–4570.
Courcelle,J., Khodursky,A., Peter,B., Brown,P.O. and Hanawalt,P.C. ( (2001) ) Comparative gene expression profiles following UV exposure in wild-type and SOS-deficient Escherichia coli. Genetics, , 158, , 41–64.
Arfin,S., Long,A., Ito,E., Riehle,M., Paegle,E. and Hatfield,G. ( (2000) ) Global gene expression profiling in Escherichia coli K12. The effects of integration host factor. J. Biol. Chem., , 275, , 29672–29684.
Hommais,F., Krin,E., Laurent-Winter,C., Soutourina,O., Malpertuy,A., Le Caer,J.P., Danchin,A. and Bertin,P. ( (2001) ) Large-scale monitoring of pleiotropic regulation of gene expression by the prokaryotic nucleoid-associated protein, H-NS. Mol. Microbiol., , 40, , 20–36.
Cao,M., Kobel,P.A., Morshedi,M.M., Wu,M.F., Paddon,C. and Helmann,J.D. ( (2002) ) Defining the Bacillus subtilis W regulon: a comparative analysis of promoter consensus search, run-off transcription/macroarray analysis (ROMA), and transcriptional profiling approaches. J. Mol. Biol., , 316, , 443–457.
Blattner,F.R., Plunkett,G.,III, Bloch,C.A., Perna,N.T., Burland,V., Riley,M., Collado-Vides,J., Glasner,J.D., Rode,C.K., Mayhew,G.F., Gregor,J., Davis,N.W., Kirkpatrick,H.A., Goeden,M.A., Rose,D.J., Mau,B. and Shao,Y. ( (1997) ) The complete genome sequence of Escherichia coli K-12. Science, , 277, , 1453–1474.
Datsenko,K.A. and Wanner,B.L. ( (2000) ) One-step inactivation of chromosomal genes in Escherichia coli K-12 using PCR products. Proc. Natl Acad. Sci. USA, , 97, , 6640–6645.
Sambrook,J., Fritsch,E.F. and Maniatis,T. ( (1989) ) Molecular cloning: A Laboratory Manual. Cold Spring Harbor Laboratory Press, Cold Spring Harbor, NY.
Tang,H., Severinov,K., Goldfarb,A. and Ebright,R.H. ( (1995) ) Rapid RNA polymerase genetics: one-day, no-column preparation of reconstituted recombinant Escherichia coli RNA polymerase. Proc. Natl Acad. Sci. USA, , 92, , 4902–4906.
Rhodius,V.A., West,D.M., Webster,C.L., Busby,S.J. and Savery,N.J. ( (1997) ) Transcription activation at class II CRP-dependent promoters: the role of different activating regions. Nucleic Acids Res., , 25, , 326–332.
Loos,A., Glanemann,C., Willis,L.B., O'Brien,X.M., Lessard,P.A., Gerstmeir,R., Guillouet,S. and Sinskey,A.J. ( (2001) ) Development and validation of Corynebacterium DNA microarrays. Appl. Environ. Microbiol., , 67, , 2310–2318.
Britton,R.A., Eichenberger,P., Gonzalez-Pastor,J.E., Fawcett,P., Monson,R., Losick,R. and Grossman,A.D. ( (2002) ) Genome-wide analysis of the stationary-phase sigma factor (sigma-H) regulon of Bacillus subtilis. J. Bacteriol., , 184, , 4881–4890.
Rhodius,V. and Busby,S. ( (2000) ) Transcription activation by the Escherichia coli cyclic AMP receptor protein: determinants within activating region 3. J. Mol. Biol., , 299, , 295–310.
Valentin-Hansen,P., Holst,B., Josephsen,J., Hammer,K. and Albrechtsen,B. ( (1989) ) CRP/cAMP- and CytR-regulated promoters in Escherichia coli K12: the cdd promoter. Mol. Microbiol., , 3, , 1385–1390.
Kallipolitis,B.H. and Valentin-Hansen,P. ( (1998) ) Transcription of rpoH, encoding the Escherichia coli heat-shock regulator 32, is negatively controlled by the cAMP-CRP/CytR nucleoprotein complex. Mol. Microbiol., , 29, , 1091–1099.
Jayaraman,P.S., Peakman,T.C., Busby,S.J., Quincey,R.V. and Cole,J.A. ( (1987) ) Location and sequence of the promoter of the gene for the NADH-dependent nitrite reductase of Escherichia coli and its regulation by oxygen, the Fnr protein and nitrite. J. Mol. Biol., , 196, , 781–788.
Wubbolts,M.G., Terpstra,P., van Beilen,J.B., Kingma,J., Meesters,H.A. and Witholt,B. ( (1990) ) Variation of cofactor levels in Escherichia coli. Sequence analysis and expression of the pncB gene encoding nicotinic acid phosphoribosyltransferase. J. Biol. Chem., , 265, , 17665–17672.
Lobell,R. and Schleif,R. ( (1991) ) AraC-DNA looping: orientation of distance-dependent loop breaking by the cyclic AMP receptor protein. J. Mol. Biol., , 218, , 45–54.
Busby,S., Truelle,N., Spassky,A., Dreyfus,M. and Buc,H. ( (1984) ) The selection and characterization of two novel mutation in overlapping promoters of the Escherichia coli galactose operon. Gene, , 28, , 201–209.
Ishizuka,H., Hanamura,A., Inada,T. and Aiba,H. ( (1994) ) Mechanism of the down-regulation of cAMP receptor protein by glucose in Escherichia coli: role of autoregulation of the crp gene. EMBO J., , 13, , 3077–3082.
Ebright,R. ( (1993) ) Transcription activation at Class I Cap-dependent promoters. Mol. Microbiol., , 8, , 797–802.
Ushida,C. and Aiba,H. ( (1990) ) Helical phase dependent action of CRP: effect of the distance between the CRP site and the -35 region on promoter activity. Nucleic Acids Res., , 18, , 6325–6330.
Belyaeva,T., Rhodius,V., Webster,C. and Busby,S. ( (1998) ) Transcription activation at promoters carrying tandem DNA sites for the Escherichia coli cyclic AMP receptor protein: organisation of the RNA polymerase subunits. J. Mol. Biol., , 277, , 789–804.
Law,E., Savery,N. and Busby,S. ( (1999) ) Interactions between the Escherichia coli cAMP receptor protein and the C-terminal domain of the subunit of RNA polymerase at class I promoters. Biochem. J., , 337, , 415–423.
Tebbutt,J., Rhodius,V.A., Webster,C.L. and Busby,S.J. ( (2002) ) Architectural requirements for optimal activation by tandem CRP molecules at a class I CRP-dependent promoter. FEMS Microbiol. Lett., , 210, , 55–60.
Belyaeva,T., Wade,J., Webster,C., Howard,V., Thomas,M., Hyde,E. and Busby,S. ( (2000) ) Transcription activation at the Escherichia coli melAB promoter: the role of MelR and the cyclic AMP receptor protein. Mol. Microbiol., , 36, , 211–222.
Kimata,K., Inada,T., Tagami,H. and Aiba,H. ( (1998) ) A global repressor (Mlc) is involved in glucose induction of the ptsG gene encoding major glucose transporter in Escherichia coli. Mol. Microbiol., , 29, , 1509–1519.
Plumbridge,J. ( (1998b) ) Expression of ptsG, the gene for the major glucose PTS transporter in Escherichia coli, is repressed by Mlc and induced by growth on glucose. Mol. Microbiol., , 29, , 1053–1063.
Plumbridge,J. ( (1998a) ) Control of the expression of the manXYZ operon in Escherichia coli: Mlc is a negative regulator of the mannose PTS. Mol. Microbiol., , 27, , 369–380.
Decker,K., Plumbridge,J. and Boos,W. ( (1998) ) Negative transcriptional regulation of a positive regulator: the expression of malT, encoding the transcriptional activator of the maltose regulon of Escherichia coli, is negatively controlled by Mlc. Mol. Microbiol., , 27, , 381–390.
Lee,S.J., Boos,W., Bouche,J.P. and Plumbridge,J. ( (2000) ) Signal transduction between a membrane-bound transporter, PtsG, and a soluble transcription factor, Mlc, of Escherichia coli. EMBO J., , 19, , 5353–5361.
Shin,D., Cho,N., Heu,S. and Ryu,S. ( (2003) ) Selective regulation of ptsG expression by Fis. Formation of either activating or repressing nucleoprotein complex in response to glucose. J. Biol. Chem., , 278, , 14776–14781.
Gosset,G., Zhang,Z., Nayyar,S., Cuevas,W.A. and Saier,M.H. ( (2004) ) Transcriptome analysis of crp-dependent catabolite control of gene expression in Escherichia coli. J. Bacteriol., , 186, , 3516–3524.
Li,B., Wing,H., Lee,D., Wu,H.C. and Busby,S. ( (1998) ) Transcription activation by Escherichia coli FNR protein: similarities to, and differences from, the CRP paradigm. Nucleic Acids Res., , 26, , 2075–2081.
Sawers,G., Kaiser,M., Sirko,A. and Freundlich,M. ( (1997) ) Transcriptional activation by FNR and CRP: reciprocity of binding-site recognition. Mol. Microbiol., , 23, , 835–845.
Valentin-Hansen,P., S?gaard-Anderson,L. and Petersen,H. ( (1996) ) A flexible partnership: the CytR ant-activator and the cAMP-CRP activator protein, comrades in transcriptional control. Mol. Microbiol., , 20, , 461–466.
Perini,L.T., Doherty,E.A., Werner,E. and Senear,D.F. ( (1996) ) Multiple specific CytR binding sites at the Escherichia coli deoP2 promoter mediate both cooperative and competitive interactions between CytR and cAMP receptor protein. J. Biol. Chem., , 271, , 33242–33255.
Yamamoto,K. and Ishihama,A. ( (2003) ) Two different modes of transcription repression of the Escherichia coli acetate operon by IclR. Mol. Microbiol., , 47, , 183–194.
Müller-Hill,B. ( (1998) ) Some repressors of bacterial transcription. Curr. Opin. Microbiol., , 1, , 145–151.
Roy,S., Garges,S. and Adhya,S. ( (1998) ) Activation and repression of transcription by different contact: two sides of a coin. J. Biol. Chem., , 273, , 14059–14062.
Monsalve,M., Calles,B., Mencia,M., Salas,M. and Rojo,F. ( (1997) ) Transcription activation or repression by phage 29 protein p4 depends on the strength of the RNA polymerase-promoter interactions. Mol. Cell, , 1, , 99–107.
Gulati,A. and Mahadevan,S. ( (2000) ) Mechanism of catabolite repression in the bgl operon of Escherichia coli: involvement of the anti-terminator BglG, CRP-cAMP and EIIAGlc in mediating glucose effect downstream of transcription initiation. Genes Cells, , 5, , 239–250.
Davies,S.J., Golby,P., Omrani,D., Broad,S.A., Harrington,V.L., Guest,J.R., Kelly,D.J. and Andrews,S.C. ( (1999) ) Inactivation and regulation of the aerobic C(4)-dicarboxylate transport (dctA) gene of Escherichia coli. J. Bacteriol., , 181, , 5624–5635.
Bell,A.W., Buckel,S.D., Groarke,J.M., Hope,J.N., Kingsley,D.H. and Hermodson,M.A. ( (1986) ) The nucleotide sequences of the rbsD, rbsA, and rbsC genes of Escherichia coli K12. J. Biol. Chem., , 261, , 7652–7658.
Holcroft,C.C. and Egan,S.M. ( (2000) ) Interdependence of activation at rhaSR by cyclic AMP receptor protein, the RNA polymerase subunit C-terminal domain, and rhaR. J. Bacteriol., , 182, , 6774–6782.
Cunningham,L. and Guest,J.R. ( (1998) ) Transcription and transcript processing in the sdhCDAB-sucABCD operon of Escherichia coli. Microbiology, , 144, , 2113–2123.
Deeley,M.C. and Yanofsky,C. ( (1982) ) Transcription initiation at the tryptophanase promoter of Escherichia coli K-12. J. Bacteriol., , 151, , 942–951.
Klein,W., Horlacher,R. and Boos,W. ( (1995) ) Molecular analysis of treB encoding the Escherichia coli enzyme II specific for trehalose. J. Bacteriol., , 177, , 4043–4052.
Weickert,M.J. and Adhya,S. ( (1993) ) Control of transcription of gal repressor and isorepressor genes in Escherichia coli. J. Bacteriol., , 175, , 251–258.
Larson,T.J., Cantwell,J.S. and van Loo-Bhattacharya,A.T. ( (1992) ) Interaction at a distance between multiple operators controls the adjacent, divergently transcribed glpTQ-glpACB operons of Escherichia coli K-12. J. Biol. Chem., , 267, , 6114–6121.
Otsuka,J., Watanabe,H. and Mori,K.T. ( (1996) ) Evolution of transcriptional regulation system through promiscuous coupling of regulatory proteins with operons; suggestion from protein sequence similarities in Escherichia coli. J. Theor. Biol., , 178, , 183–204.
Webster,C., Gaston,K. and Busby,S. ( (1988) ) Transcription from the Escherichia coli melR promoter is dependent on the cyclic AMP receptor protein. Gene, , 68, , 297–305.
Ryu,S., Ramseier,T.M., Michotey,V., Saier,M.H.,Jr and Garges,S. ( (1995) ) Effect of the FruR regulator on transcription of the pts operon in Escherichia coli. J. Biol. Chem., , 270, , 2489–2496.
Hanamura,A. and Aiba,H. ( (1991) ) Molecular mechanism of negative autoregulation of Escherichia coli crp gene. Nucleic Acids Res., , 19, , 4413–4419.
Golby,P., Kelly,D.J., Guest,J.R. and Andrews,S.C. ( (1998) ) Transcriptional regulation and organization of the dcuA and dcuB genes, encoding homologous anaerobic C4-dicarboxylate transporters in Escherichia coli. J. Bacteriol., , 180, , 6586–6596.
Plumbridge,J. and Kolb,A. ( (1991) ) CAP and Nag repressor binding to the regulatory regions of the nagE-B and manX genes of Escherichia coli. J. Mol. Biol., , 217, , 661–679.
Craig,J.E., Zhang,Y. and Gallagher,M.P. ( (1994) ) Cloning of the nupC gene of Escherichia coli encoding a nucleoside transport system, and identification of an adjacent insertion element, IS 186. Mol. Microbiol., , 11, , 1159–1168.
Munch-Petersen,A. and Jensen,N. ( (1990) ) Analysis of the regulatory region of the Escherichia coli nupG gene, encoding a nucleoside-transport protein. Eur. J. Biochem., , 190, , 547–551.
Gerlach,P., S?gaard-Andersen,L., Pedersen,H., Martinussen,J., Valentin-Hansen,P. and Bremer,E. ( (1991) ) The cyclic AMP (cAMP)-cAMP receptor protein complex functions both as an activator and as a corepressor at the tsx-p2 promoter of Escherichia coli K-12. J. Bacteriol., , 173, , 5419–5430.(Dongling Zheng, Chrystala Constantinidou)