Skip Navigation

This Article
Right arrow Abstract Freely available
Right arrow Print PDF (2409K) Freely available
Right arrow Alert me when this article is cited
Right arrow Alert me if a correction is posted
Services
Right arrow Email this article to a friend
Right arrow Similar articles in this journal
Right arrow Similar articles in ISI Web of Science
Right arrow Similar articles in PubMed
Right arrow Alert me to new issues of the journal
Right arrow Add to My Personal Archive
Right arrow Download to citation manager
Right arrow Search for citing articles in:
ISI Web of Science (284)
Right arrow Commercial Re-use Guidelines
for Open Access NAR Content
Google Scholar
Right arrow Articles by Richmond, C. S.
Right arrow Articles by Blattner, F. R.
Right arrow Search for Related Content
PubMed
Right arrow PubMed Citation
Right arrow Articles by Richmond, C. S.
Right arrow Articles by Blattner, F. R.
Social Bookmarking
Add to CiteULike   Add to Connotea   Add to Del.icio.us  
What's this?

Nucleic Acids Research 27:3821-3835 (1999)
© 1999 Oxford University Press

Article

Genome-wide expression profiling in Escherichia coli K-12

Craig S. Richmonda, Jeremy D. Glasner, Robert Mau, Hongfan Jin and Frederick R. Blattner

Laboratory of Genetics, University of Wisconsin, Madison, WI 53706, USA


   ABSTRACT
TOP
 ABSTRACT
INTRODUCTION
 MATERIALS AND METHODS
 RESULTS
 DISCUSSION
 REFERENCES
 
We have established high resolution methods for global monitoring of gene expression in Escherichia coli. Hybridization of radiolabeled cDNA to spot blots on nylon membranes was compared to hybridization of fluorescently-labeled cDNA to glass microarrays for efficiency and reproducibility. A complete set of PCR primers was created for all 4290 annotated open reading frames (ORFs) from the complete genome sequence of E.coli K-12 (MG1655). Glass- and nylon-based arrays of PCR products were prepared and used to assess global changes in gene expression. Full-length coding sequences for array printing were generated by two-step PCR amplification. In this study we measured changes in RNA levels after exposure to heat shock and following treatment with isopropyl-ß-D-thiogalactopyranoside (IPTG). Both radioactive and fluorescence-based methods showed comparable results. Treatment with IPTG resulted in high level induction of the lacZYA and melAB operons. Following heat shock treatment 119 genes were shown to have significantly altered expression levels, including 35 previously uncharacter­ized ORFs and most genes of the heat shock stimulon. Analysis of spot intensities from hybridization to replicate arrays identified sets of genes with signals consistently above background suggesting that at least 25% of genes were expressed at detectable levels during growth in rich media.


   INTRODUCTION
TOP
 ABSTRACT
 INTRODUCTION
MATERIALS AND METHODS
 RESULTS
 DISCUSSION
 REFERENCES
 
The complete genome sequences from more than 21 prokaryotes and two eukaryotes have been reported. Of the completed microbial genomes the best studied is the enteric, Gram-negative bacterium Escherichia coli. The 4.6 Mbp E.coli genome is predicted to encode 4290 open reading frames (ORFs) and at least 115 stable RNAs (1). Although determination of the sequence of every gene in an organism allows a better understanding of the organism’s physiological potential, it is just the beginning of a complete description of how the cell works. Despite over 40 years of intensive study, more than 30% of the ORFs which make up the E.coli chromosome have no known function. This theme is reiterated in all completely sequenced genomes published to date (for example, 2–6) reinforcing the idea that knowledge of the nucleotide sequence of the complete genome is just the first step to understanding the dynamic nature of gene function that allows the cell to grow, replicate and respond to its environ­ment.

To provide a means of expression analysis in E.coli at single gene resolution we have created high density arrays composed of full-length ORF-specific PCR products. This strategy of gene expression analysis was originated by Chuang et al. using {lambda} clone spot blots to monitor gene expression in E.coli at 10 gene resolution (7). Brown and colleagues have subsequently miniaturized this method and have also brought it to single gene resolution (8,9). This basic approach for large scale expression monitoring has been used in a number of different formats including spotted cDNA microarrays (9–14) and oligo­nucleotide arrays (15–17). Microarrays of PCR fragments corresponding to most of the ORFs from Saccharomyces cerevisiae have been used to study transcriptional changes throughout the cell cycle, during sporulation and in response to a diauxic shift (12,13,18). Methods for global genome comparisons using Mycobacterium tuberculosis microarrays have also been reported (19,20).

The ability to determine changes in RNA levels simul­taneously for all the genes in a cell is an extremely powerful tool. Regardless of the state of cell growth, one can measure the relative expression levels for each gene under various growth conditions, different genetic states or over a time course during environmental change. These types of studies have begun to identify new sets of genes involved in specific physiological responses. For example, DeRisi et al. identified 183 genes whose expression levels change by greater than 4-fold during the diauxic shift in yeast, about half of which have no previously determined function (12). Human cDNA microarrays have been used to study the transcriptional response of human fibroblasts to serum stimulation (21), revealing many similarities between serum stimulation and wound repair.

Analysis of global gene expression data has been difficult due to the large number of data points collected in a single experiment. Eisen et al. recently presented a clustering method that allows easy visualization of genes that may be co-regulated (22). Their method clusters genes based on similarities in expression patterns and graphically presents the data using dendograms. Comparison of global expression data from S.cerevisiae using their clustering method found that genes having similar function often cluster together, suggesting that ORFs of unknown function may be involved in similar physiological pathways as those genes of known function in the same cluster.

To demonstrate the ability of our system to measure changes in gene expression on a global scale and to compare radio­active methods with fluorescence-based methods we have used two classic systems which affect gene expression in E.coli. Induction with isopropyl-ß-D-thiogalactopyranoside (IPTG) provided a simple test of the methods since we expected only a few gene transcripts to change, while the effect of heat shock allowed global regulatory affects to be observed. Hybridization of genomic DNA probes to high density arrays allowed assessment of signal variation between spots due to factors such as size, amount of DNA spotted and cross-hybridization to members of gene families.


   MATERIALS AND METHODS
TOP
 ABSTRACT
 INTRODUCTION
 MATERIALS AND METHODS
RESULTS
 DISCUSSION
 REFERENCES
 
PCR primer design
PCR primer pairs were designed to amplify each of the 4290 ORFs in the E.coli K-12 genome. Primers were designed to amplify each ORF beginning at the putative start codon and ending at the stop codon. All primer pairs used ATG as a start codon and TAA as a stop codon regardless of the start and stop codons found in the genomic sequence. Each primer contains a 13-base non-variable ‘adaptamer’ sequence at its 5" end followed by 20–25 bases of ORF-specific sequence. The length of the ORF-specific region of each primer was adjusted to achieve a melting temperature of 68–70°C. Adaptamer sequences were included to facilitate directional cloning in later stages of functional analysis (J.D.Glasner, C.S.Richmond, G.Plunkett III, S.Hinsa, B.Bochner and F.R.Blattner, manuscript in preparation). The adaptamer sequences of all N-terminal primers include a SapI restriction endonuclease recognition site and have the sequence 5"-TTGCTCTTCCATG . . .-3". All C-terminal primers also include a SapI site and have the adaptamer sequence 5"-TTG-CTCTTCGTTA . . .-3". Primers were synthesized then arrayed in 96-well plates (Genosys Biotechnologies) for convenient parallel processing of amplification reactions. These primers are available from Sigma-Genosys Biotechnologies as a complete set or as ORF-specific primer pairs.

PCR amplification
Amplification reactions were performed in two rounds using 96-well plates. For the first round of PCR, 1 U of Pfu DNA polymerase (Stratagene) was used in a 25 µl reaction volume containing 20–30 ng E.coli MG1655 genomic DNA template, 0.5 µM each primer and 200 µM dNTPs. Reactions were cycled 25 times as follows: 95°C for 15 s, 64°C for 15 s, 72°C for 4 min, with a final cycle of 72°C for 5 min. To generate re-amplified PCR products used in production of high density arrays, the first round PCR products were diluted 500-fold and 1 µl used as template in 100 µl reactions containing 200 µM dNTPs, 2.5 U ExTaq (PanVera) and 0.5 µM each primer. Reactions were cycled 25 times as follows: 95°C for 15 s, 64°C for 15 s, 72°C for 2 min, with a final cycle of 72°C for 5 min.

All PCR products (2 µl of the total reaction volume) were analyzed by electrophoresis on 1.4% agarose gels in 0.5x TBE. Gels were stained with ethidium bromide and digitally imaged using a CCD camera. BioImageTM software was used to determine the size and purity of each PCR product. An amplification was scored as ‘successful’ if a single product was within 10% of the expected size of the ORF predicted from the genomic sequence.

Unsuccessful PCRs in first round amplification reactions were of three types: (i) no product observed by ethidium bromide staining (693); (ii) multiple products observed (149); and (iii) reactions resulting in single products of unexpected size (38). Reactions that failed were repeated using conditions expected to favor amplification of the single desired product. For example, those reactions yielding no product were amplified using lower annealing temperatures. Reactions that resulted in multiple products were amplified using a lower primer concentration and/or a higher annealing temperature. Approximately 97% of the ORFs in E.coli have been successfully amplified using this approach.

Bacterial growth and isolation of total RNA
For all experiments a single colony of E.coli K-12 (MG1655) was used to inoculate 60 ml of Luria-Bertini (LB) broth [1% Bacto­tryptone (Difco), 0.5% yeast extract (Difco) and 1% NaCl] in 250 ml Erlenmeyer flasks and grown to an OD A600 between 0.8 and 0.9 at 37°C with constant aeration. IPTG treatment was performed by splitting a 60 ml culture in half and adding IPTG (final concentration of 1 mM) to one 30 ml sample (experimental) while the other sample (control) was untreated. Incubation was continued at 37°C for 30 min prior to harvesting cells. Heat shock induction was carried out by splitting a mid-log culture into two 30 ml samples with the control culture kept at 37°C for 7 min while the experimental culture was shifted to a 50°C shaking water bath for 7 min.

Total RNA was isolated from cells using Qiagen RNeasy columns. Cultures were pelleted by brief centrifugation at 5000 g and cell pellets flash-frozen on dry ice/ethanol. Cells were resuspended by vortexing in 1 ml of TE (10 mM Tris pH 8.0, 1 mM EDTA) containing 500 µg/ml lysozyme (Boehringer Mannheim) and RNA isolation proceeded following the manufacturer’s protocol. To remove contaminating genomic DNA from purified RNA, samples were treated with RQ1 RNAse-free DNAse (Promega) followed by Proteinase K digestion, phenol–chloroform extraction and precipitation with ethanol. Pelleted RNA samples were resuspended in diethyl pyrocarbonate (DEPC)-treated H2O, quantitated by absorbance at 260 nm, and stored in ethanol at –20°C until further use. For time-course studies cells were grown in rich, defined media (23) to an OD A600 of 0.8 and 30 ml samples taken at time 0, 5, 10, 15 and 20 min following transfer from 37 to 50°C. RNA was isolated as described above.

Genomic DNA labeling
Genomic DNA isolated from E.coli K-12 was labeled radio­actively using nick translation, or fluorescently labeled by random priming using the Klenow fragment of DNA polymerase. Genomic DNA was purified from MG1655 using standard protocols (24) and fragmented by sonication to an average length of 1500 bp. Each radioactive labeling reaction contained 500 ng of fragmented genomic DNA, 0.02 mM nucleotide mix (dGTP, dTTP, dATP), 50 µCi 33P-labeled dCTP, 50 mM Tris pH 7.9, 5 mM MgCl2, 10 mM ß-mercaptoethanol, 10 µg/ml BSA and 1.5 U of DNA polymerase I/DNAse I (Life Technologies, Inc.). Reactions were carried out at 15°C for 1 h and stopped by addition of EDTA to 0.05 M. Labeled genomic DNA was purified from unincorporated nucleotides by gel filtration on Sephadex G-50 columns. For each hybridization 2 x 106 c.p.m./ml of probe was used.

Fluorescence labeling reactions for microarray analysis contained 1 µg of fragmented genomic DNA, 5 µg of random hexamers, 0.5 mM nucleotide mix (dGTP, dATP, dCTP), 0.1 mM Cy3 or Cy5 labeled dUTP (Amersham-Pharmacia), 50 mM Tris–HCl pH 7.4, 10 mM MgCl2, 50 µg/ml BSA, 3 mM DTT and 5 U Klenow (Promega). Reactions were carried out at 25°C for 2 h and stopped by addition of 25 mM EDTA, 0.25% SDS. Labeled DNA was purified and concentrated prior to hybridization using Microcon 30 concentrators (Amicon).

Labeled cDNA synthesis
Random primed cDNA synthesis of total E.coli RNA was used to prepare 33P-labeled and fluorescently-labeled probes for array hybridization. For radioactive labeling 5 µg of total RNA was pelleted from ethanol, washed once with 70% ethanol and briefly dried prior to resuspension in 12 µl DEPC-treated H2O containing 250 ng of random hexamer oligonucleotides (Amersham-Pharmacia). Samples were heated to 70°C for 10 min and chilled on ice prior to probe synthesis. Probe synthesis was carried out at 42°C for 2 h in a 50 µl reaction volume containing 50 mM Tris–HCl (pH 8.3), 75 mM KCl, 3 mM MgCl2, 10 mM DTT, 0.5 mM dATP, 0.5 mM dTTP, 0.5 mM dGTP, 0.05 mM dCTP, 100 µCi 33P-labeled dCTP (1–3000 Ci/mmol; Amersham), 40 U RNAsin (Promega) and 200 U Superscript II reverse transcriptase (Life Technologies, Inc.). Following cDNA reactions RNA template was degraded by incubating for 30 min at 65°C in 0.27 M NaOH and 20 mM EDTA followed by neutralization with HCl and Tris buffer. Unincorporated nucleo­tides were removed by Sephadex G50 gel filtration chromatography (24).

Fluorescence cDNA labeling was performed essentially as described above. Random hexamers (10 µg) and RNA (20 µg) were mixed in a final volume of 12 µl and treated as above. Probe synthesis was performed at 42°C for 2 h in a 50 µl reaction volume containing 50 mM Tris–HCl (pH 8.3), 75 mM KCl, 3 mM MgCl2, 10 mM DTT, 0.5 mM dATP, 0.5 mM dGTP, 0.5 mM dCTP, 0.2 mM dTTP, 0.1 mM Cy dye-labeled dUTP (Amersham), 40 U RNAsin (Promega) and 200 U Superscript II reverse transcriptase (Life Technologies, Inc.). For all experiments Cy3-dUTP was used to label control samples and Cy5-dUTP used to label experimental samples. Following cDNA reactions the RNA template was degraded by incubating for 30 min at 65°C in 0.27 M NaOH followed by neutralization with HCl and Tris buffer. Labeled cDNA was purified and concentrated prior to hybridization using Microcon 30 concentrators (Amicon).

Preparation of high density arrays
Spot blots of ORF-specific PCR products and controls were printed on 23 x 24 cm nylon membranes (GeneScreen Plus, New England Nuclear) in a 64 x 68 hexagonal grid using a Gilson Model 215 liquid handling robot. Before printing, 2 µl of each re-amplified PCR was diluted with 1 µl of an alkaline denaturing solution (0.75 M NaOH, 30 mM EDTA) and directly spotted on pre-wet membranes. Immediately following spot deposition, membranes were neutralized for 20 min in 20 mM Tris (pH 7.0) and baked for 45 min at 80°C. Membranes were stored dry at room temperature or used immediately for hybridization. Control spots on all arrays include genomic DNA and {lambda} DNA. As negative controls, mock amplification reactions and amplification products from several yeast ORFs having no significant match to E.coli genes were used. Whole genome spot blots printed in a similar fashion to those described here are currently available from Sigma-Genosys Biotechnologies.

Microarrays were prepared as described previously (12) using re-amplified PCR products (60 µl reactions) that were precipitated with isopropanol, pelleted and resuspended in 15 µl of 3x SSC prior to spotting.

Radioactive hybridization and raw data analysis
Prior to hybridization, high density arrays were pre-wet in 2x SSC, 0.5% SDS and pre-hybridized for 2 h at 64°C in roller bottles containing 15 ml of 1x hybridization buffer (0.9 M NaCl, 90 mM sodium citrate pH 7.0, 0.1% Ficoll, 0.1% polyvinylpyrro­lidone, 0.1% BSA, 0.5% SDS and 100 µg/ml sheared salmon sperm DNA). Pre-hybridization buffer was removed and replaced with 1x hybridization buffer containing 2 x 106 c.p.m./ml cDNA probe and hybridization continued at 64°C for 40 h. Following hybridization, arrays were washed twice at room temperature and once at 65°C in 0.3 M NaCl, 30 mM sodium citrate, 0.5% SDS followed by one wash at 65°C in 15 mM NaCl, 1.5 mM sodium citrate, 0.5% SDS. Arrays were then sealed in thin polypropylene bags to avoid drying and exposed to a phosphor screen (Molecular Dynamics) for 96 h. Phosphor screens were scanned using a STORM phosphorimaging instrument (Molecular Dynamics) at 100 micron pixel resolution. Images were stored electronically and analyzed using ImageQuant v.4.1 analysis software (Molecular Dynamics). Between successive hybridizations, membranes were stripped by two 30 min incubations in 1% SDS at 100°C and dried for 30 min at 65°C.

To calculate signal intensities a grid of ellipses was drawn and overlaid on the array image and signal intensities for each spot calculated using the volume quantitation method of ImageQuant. Total intensity of all pixels within each ellipse was determined and data saved in spreadsheets. Background correction was achieved by sampling regions outside the array grid and averaging their volumes.

We calculate expression ratios as follows: after correction for background, the percentage of total signal is calculated for each spot as a means of normalization. These values are used to determine the ratio of experimental to control signal. To allow easy comparison of induction and repression ratios, signals that were higher under the control condition (transcription repressed) were used in the numerator of the ratio calculation and then converted to negative values. A threshold of minimum acceptable signal was used to eliminate expression ratios that were extremely high or low due to undetectable signal in control or experimental samples. To determine consistency of ratios across replicate hybridizations a t-test was applied. We report only those ratios with values greater than or equal to 5 and having a 95% confidence interval as determined by the t-test. The entire data sets are available at http://www.genetics.wisc.edu

Microarray hybridization and raw data analysis
Escherichia coli microarrays were hybridized in a final volume of 13 µl containing 3x SSC, 0.8 mg/ml salmon sperm DNA and 0.2% SDS. Prior to hybridization samples were heated to 95°C for 2 min and pipetted directly onto microarrays. A cover slip was applied and the arrays hybridized overnight at 64°C in a humidified hybridization chamber. Following hybridization, slides were washed for 3 min in 0.2x SSC, 0.1% SDS, followed by two washes of 2 min each in 0.2x SSC and rinsed in 0.05x SSC. To remove residual salts, slides were spun at 500 r.p.m. for 5 min prior to scanning on a ScanArray 3000 confocal laser scanner (GSI-Lumonics). Signal intensities for each spot were determined using ScanAlyze software (available at http://rana.stanford.edu/software/ ). The average fluorescence intensity for each spot was calculated and local background determined as the median pixel intensity in a square surrounding each spot. Following background subtraction, signal intensities were calculated as the percent of total signal as a means of normalization. Ratios and t-tests were performed as described above.


   RESULTS
TOP
 ABSTRACT
 INTRODUCTION
 MATERIALS AND METHODS
 RESULTS
DISCUSSION
 REFERENCES
 
Creation of PCR gene set
To begin systematic functional and expression analysis in E.coli K-12 a complete set of ORF-specific PCR primers was created. The primer pairs were designed to amplify the complete coding sequences from start to stop codon and included a 13-base, non-variable adaptamer sequence. A high fidelity PCR system was chosen to obtain PCR products for downstream cloning and expression studies. We used Pfu polymerase which has 3"->5" proofreading activity and a low error rate (25–27). All PCR products were analyzed by agarose gel electrophoresis and digital image analysis. Amplification of all 4290 ORFs with Pfu polymerase was followed by a re-amplification step designed to minimize carry-over of genomic DNA template when printing high density arrays. The re-amplification step used a 500-fold dilution of the Pfu PCR product as template and a Taq/Pfu thermostable DNA polymerase mix to improve yield of large fragments (28). Using this two-step method we have successfully amplified 97% of the 4290 E.coli ORFs using the original set of PCR primers. Products printed on nylon membranes represent >90% of the E.coli genome while those printed on microarrays represent 95% of the genome. Those ORFs that failed to amplify successfully were eliminated from further analysis during expression profiling.

High density array analysis and validation
To ensure that DNA samples were successfully deposited on nylon membranes and to assess differential hybridization to target genes, we hybridized 33P-labeled MG1655 genomic DNA to each spot blot. The total intensity of all pixels within each spot (corrected for background signal) was determined after scanning of exposed phosphor screens. In a few cases (32 spots) signals close to average were detected for ORFs that were scored as unamplified following electrophoresis of PCR products. These products were likely produced at levels below the limit of ethidium bromide detection. Only nine spots, scored as successful PCRs, were consistently scored as undetectable (defined as having a signal strength less than three standard deviations above the average background) in all hybridizations.

Figure 1 (upper panel) shows the distribution of 33P-labeled genomic DNA hybridization signal intensities as a function of ORF length. We noticed that small genes (<300 bp) often showed lower intensity signals even when the PCR product was abundant. For the largest genes, the hybridization signal also tended to be lower than average. Inspection of the PCR electrophoresis results reveals that most ORFs >2 kb were amplified less efficiently than smaller genes.



View larger version (85K):
[in this window]
[in a new window]
  		Figure 1. Scatter plot showing the relationship between signal intensities from genomic DNA hybridizations and ORF length. (Upper panel) Radioactive hybridization of 33P-labeled E.coli genomic DNA to spot blot arrays. Average percent spot intensity values from three hybridizations are plotted against ORF length in bp. (Lower panel) Same as upper panel using microarrays and fluorescently labeled genomic DNA. The average percent intensities from two hybridizations are plotted against ORF length in bp. Paralogous ORFs are highlighted in blue and failed PCR products in red.

 
We observed 61 genes with genomic DNA hybridization signals much higher than average (Fig. 1). Since the majority of these spots (45 spots, marked blue in Fig. 1) correspond to ORFs that are members of gene families (paralogs) with high degrees of sequence identity (>70% nucleotide identity over >200 bp) we conclude that the effect is likely due to cross-hybridization between paralogs. There is a general trend for genes with greater numbers of paralogs to have higher signal strengths although their intensity also varies depending on the length and degree of similarity between the paralogous sequences. ORFs with sequences that align over >200 bp with >80% sequence identity show the strongest intensity signals. This result is also seen using DNA microarrays (Fig. 1, lower panel). Genomic DNA was labeled separately with Cy3 and Cy5 dUTP, the probes were mixed and hybridized to a single microarray followed by laser scanning and quantification. As with the radioactive hybridizations, signal intensities were higher than average for paralogous ORFs. We also find, as with radioactive methods, that small ORFs tend to have weaker signals than do larger ORFs.

IPTG expression profile
The lac operon (lacZYA) is one of the best-studied operons in E.coli. Addition of lactose or IPTG to E.coli cultures in exponential phase growth results in transcriptional induction of the lac genes caused by release of LacI-mediated transcriptional repression. Table 1 shows all genes found to be significantly affected by 30 min treatment with 1 mM IPTG as measured by radioactive methods. As expected, genes of the lac operon are the most highly induced transcripts. Induction ratios for lacZ and lacY were near 60-fold while lacA had an induction ratio of 79-fold. This result confirms the ability of this method to measure specific changes in gene expression using a complex mixture of labeled cDNA. In addition, melA and melB, which make up the melibiose operon, are induced 8- and 6-fold, respectively. This is consistent with previous data from our laboratory (7). The melA gene encodes {alpha}-galactosidase and melB encodes a melibiose transport system permease II, both of which are induced by melibiose under the positive control of MelR (29). It is interesting to note that melibiose is also able to induce the lac operon (30) although there have been no previous reports of IPTG positively affecting MelR.


View this table:
[in this window]
[in a new window]
  		Table 1. Genes affected by IPTG
 
Radioactive methods also identified the cryptic ß-D-galacto­sidase gene, ebgA, and two ORFs of unknown function, b1441 and b1297, as induced by IPTG. The ebgA gene is part of the ebgRAC operon thought to have evolved from the lac operon (31). We did not, however, see significant induction of ebgC, which is thought to be co-transcribed with ebgA. Comparison of the ebgA and lacZ DNA sequences using the Wilbur–Lipman alignment method (32) shows ~50% sequence identity over >2000 bp. To test if this level of sequence similarity was enough to produce cross-hybridization artifacts using radioactive hybridization methods, in vitro transcribed RNA from the lacZ gene was radioactively labeled using reverse transcriptase and hybridized in the presence of unlabeled total cDNA to an array of 200 spots including lacZ and ebgA. The results confirm that cross-hybridization may have led to the observed ‘induction’ of ebgA expression (data not shown).

We compared results from radioactive methods to those using E.coli microarrays (Fig. 4E and F). Control and experimental RNA samples from the same preparations used for radioactive studies were fluorescently labeled and hybridized in parallel to a single microarray. The averaged results from two hybridizations were compared to those from radioactive hybridizations. Using microarrays, genes of the lac and mel operons were induced to levels similar to those obtained with radioactive methods (Table 1); however, no significant effect was observed for the two unknown ORFs (b1297 and b1441) or ebgA. This suggests that hybridization and wash conditions were stringent enough to avoid the cross-hybridization artifact or that the sensitivity of microarrays is not high enough to detect cross-hybridization of ebgA. However, Heller et al. report that short regions of sequence identity over the length of the target sequence do result in cross-hybridization artifacts using cDNA microarrays, although the degree of sequence identity for these cross-hybridizing species was not reported (33).



View larger version (75K):
[in this window]
[in a new window]
  		Figure 4. Comparison of radioactive and fluorescence-based hybridization methods. Percent total signal intensity from radioactive and fluorescence hybridization studies were plotted on a logarithmic scale. (A) and (B) Identical cDNA probes hybridized sequentially to the same membrane (A) or in parallel to a single microarray (B). (C) and (D) Identical cDNA probes hybridized to separate spot blots (C) or hybridized to separate microarrays (D). (E) and (F) Average percent signal intensity from control (LB) and experimental (IPTG) samples hybridized to same membrane (E) or to a single microarray (F). (G) and (H) Average percent signal intensity from control (37°C) and experimental (50°C, 7 min) samples hybridized sequentially to the same membrane (G) or in parallel to a single microarray (H). Spots in red in (E)–(H) are those that meet our cutoff criteria (see Materials and Methods).

 
Heat shock response expression profile
The heat shock response is well studied (34–37) and is conserved among many different organisms (36,38). Early studies of heat shock in E.coli used two-dimensional gel electrophoresis to identify proteins whose expression was induced due to increased temperature (39–41). More recently, experiments with {lambda} clone spot blots have identified additional members of the heat shock stimulon (7).

The heat shock response in E.coli is controlled at the transcriptional level by the alternative {sigma} factors rpoH ({sigma}32) (42) and rpoE ({sigma}E) (43,44). At least 51 loci have been identified as members of the heat shock stimulon in E.coli (reviewed in 35) although the specific gene sequence affected is known for only 30 of these loci. The remaining 21 loci, although named as heat shock inducible genes in the literature, are as yet functionally or genetically uncharacterized. Heat shock proteins where the biochemical activity of the gene product is known include proteases, chaperones, tRNA synthetases and {sigma} factors.

We used spot blots to identify the genes transcriptionally regulated following a temperature shift from 37 to 50°C using radioactive cDNA probes prepared from total E.coli RNA (Figs 2 and 4G). Most genes are apparently unaffected while a specific subset of genes show increased or decreased expression following heat shock treatment (for example, see Fig. 2C and D, boxed regions). This experiment identified 77 genes that are induced and 42 genes repressed by heat shock (Table 2). For 35 ORFs of unknown function, this study provides the first biochemical evidence for their expression and biological role, although for 12 of these ORFs a putative function was pre­viously assigned based on similarities to sequences of known function (1). The up-regulated genes include 23 previously identified members of the heat shock stimulon (Table 2), with an additional three known heat shock genes showing induction just below our cutoff values.



View larger version (116K):
[in this window]
[in a new window]
  		Figure 2. Escherichia coli spot blot of heat shock induction. (A) Escherichia coli array hybridized with 33P-labeled cDNA from RNA isolated from MG1655 grown at 37°C. (B) Same blot as in (A) following stripping and re-hybridization with cDNA prepared from RNA isolated from a culture that was heat shock treated at 50°C for 7 min. (C) and (D) are enlargements from boxed regions in (A) and (B) respectively. Spots indicated by boxes in (C) and (D): upper box includes ibpAB and hslUV, lower box is miaA-hfq-hflX.

 

View this table:
[in this window]
[in a new window]
  		Table 2. Genes significantly affected by heat shock
 
The remaining 19 genes with known function have not previously been described as heat shock inducible. Some of these genes are known to be affected in response to other stress conditions, such as the cadAB operon, induced in response to low pH or high lysine concentrations (45); and cspD, induced under conditions of nutritional stress (46). Another group encode proteins with functions related to heat shock genes: rseA, a negative regulator of {sigma} E (47,48); clpA, the ATPase component of the ClpAP protease (49,50) [clpP is a heat shock gene (51)]; and prlC, a trypsin-like proteinase in E.coli (52,53).

Nine of the induced ORFs are members of paralogous gene families. For example, marA has a stretch of 369 bp that are 51% identical to rob. However, although the marA signal was induced 6-fold, rob was not, confirming that cross-hybridization at this level of sequence similarity was not significant. Spots corresponding to IS186 elements were the only spots showing evidence for cross-hybridization in this experiment with all members induced to similar levels.

Of the 42 genes repressed, all but six have known biochemical activities in a number of functional categories. Sixteen are involved in purine or pyrimidine ribonucleotide biosynthesis while four are involved in chemotaxis or motility. One very interesting member of this group is the glyA gene which encodes serine hydroxymethyltransferase and is responsible for converting serine into glycine with the formation of one-carbon units. We have not been able to reproduce repression of purine and pyrimidine biosynthetic genes due to heat shock using a variety of growth conditions.

In previous experiments using spot blots prepared from {lambda} clones, 26 new heat shock genes were found on 12 different clones (7,34). Due to the low resolution of this method many loci were simply given a generic name (hslA–Z). Ten of the 12 {lambda} clones identified in the previous study encode genes shown in this report to be significantly induced by heat shock (Table 3). The criteria for our tentative correlations are as follows: if only one ORF, identified as induced in the present study, was present on a {lambda} clone, the ORF identified here was given that clone’s hsl designation. If {lambda} clones were shown previously to have multiple genes induced by heat shock, the corresponding ORFs were compared based on predicted protein size to determine correspondence. In some cases the hsl genes have been subsequently characterized as heat shock inducible [miaA locus (hslY and Z) (54), ftsJ (hslN), hflB (hslL) (55)] or were recently named based on biochemical characterization [clpYQ (hslUV) (56)].


View this table:
[in this window]
[in a new window]
  		Table 3. Correlation of heat shock locus (hsl) ORFs with specific genes
 
We also compared heat shock data from radioactive spot blot analysis to those collected using microarrays (Figs 3 and 4; Table 2). cDNA probes were prepared from RNA isolated at 5 min intervals following temperature shift from 37 to 50°C. Fluorescent probes were hybridized to E.coli microarrays comparing time 0 to each of the four subsequent time points. Comparing data from the first 5 min time point using microarrays to data from radioactive hybridizations we see that the expression profiles are comparable for both (Fig. 4G and H). The majority of ORFs identified as induced by radioactive methods are also detected by microarray analysis (Table 2). With microarrays, 62 of the 76 genes identified by radioactive methods were seen to be induced to levels of 3-fold or greater. Using a 5-fold cutoff for microarray data identifies 45 of the 76 genes. We also found an additional subset of genes to be induced to levels greater than 5-fold using microarray methods. In general, we found that genes significantly induced by 5 min remained high throughout the time course although a few genes appeared to be temporally regulated (data not shown).



View larger version (112K):
[in this window]
[in a new window]
  		Figure 3. Escherichia coli genome microarray. The microarray was hybridized with fluorescently labeled cDNA (Cy3 labeled control and Cy5 labeled heat shock samples) from the first 5 min time point of a heat shock time-course experiment. The actual size of this microarray is 1.8 cm2. Genes that are induced or repressed appear in the image as red or green spots respectively. Genes that are expressed at similar levels in both samples appear as yellow or orange spots. Labeled arrows identify examples of genes affected by heat shock.

 
Transcriptome analysis
Most approaches to global transcription analysis rely on generating ratios of signal intensities between control and experimental samples. The ratio between signals for a particular spot provides a robust measure of change in expression level. However, experi­mental conditions often result in only a small number of genes with altered levels of expression. Induction and repression ratios identify dramatic changes in transcript abundance but ignore the variations in signal intensities between spots that do not significantly change. To a large degree, these variations in signal reflect the absolute abundance of different transcripts in the cell, but one should be cautious when interpreting absolute signal intensities. Two different genes on an array may have different signals because their transcripts are present at different levels or because they differ in size, base composition, or concentration of DNA in the spot. Multiple replicates of an experiment are necessary before we can be confident that the signal derives from a detectable level of RNA resulting from gene transcription. For example, in four replicates of the heat shock experiment we observed 1023 genes expressed above background level in four out of four replicates of the control sample. Likewise, for 1158 genes we failed to detect a signal above background in any replicate. Of the 1158 genes that were not detected, 75% are of unknown function; in contrast, genes of unknown function constitute only 37% of the genes that are always detected. The number of genes with significant signals in only one or two experiments is large, indicating the necessity to repeat experiments to definitively determine if a gene is expressed or not. In fact the proportion of expressed genes is likely to be an underestimate. In these experiments we required that the hybridization signal be above three standard deviations of the background signal in all trials. In a few individual trials the variation in background signal was substantially greater than others leading to exclusion of a large number of genes. Taking these factors into account we observe that at least 25% of the E.coli ORFs were expressed at detectable levels in batch culture at 37°C.


   DISCUSSION
TOP
 ABSTRACT
 INTRODUCTION
 MATERIALS AND METHODS
 RESULTS
 DISCUSSION
REFERENCES
 
We describe the use of high density arrays of amplimers from the complete E.coli genome sequence to determine global changes in transcript abundance under various experimental conditions. The changes observed in response to treatment with IPTG and heat shock validate use of this method to detect alterations in transcript abundance from samples of total cellular RNA. We have also compared hybridization of radioactive cDNAs to spot blots on nylon membranes with fluorescence-based hybridization to glass microarrays to demonstrate the reliability and reproducibility of the two methods.

Figure 4 shows multiple comparisons of data collected using radioactive hybridization to spot blots and fluorescence hybridization to microarrays. Hybridization of identical RNA samples (RNA from E.coli grown in LB broth to mid-log) allowed for measurement of experimental reproducibility between the two methods. Parallel hybridization of identical samples to the same microarray (Fig. 4B) shows a higher degree of correlation than sequential hybridization to the same spot blot (Fig. 4A). With microarrays only four spots showed ratios greater than 3 with only one of these having a ratio greater than 4. In comparison, spot blots showed 30 spots with ratios greater than 3 and 11 of these with ratios greater than 4. We find that the variation in signal intensities is more apparent for low intensity signals with both methods. Low level signals are more difficult to accurately measure and are impacted to a larger degree by use of average background subtraction (spot blots) as compared to local background subtraction (microarrays).

Hybridization of identical RNA samples to different arrays was used to assess reproducibility of expression analysis when comparing data collected from separate arrays (Fig. 4C and D). For both methods it is apparent that use of different arrays for experimental and control hybridizations is less reproducible than hybridization of both samples to the same array (Fig. 4, compare A and B with C and D). Hybridization to different glass microarrays shows a similar correlation to sequential hybridization of the same RNA to a single spot blot.

The same variation between methods is seen when different RNAs from an experimental condition, known to affect few genes, are compared (Fig. 4E and F). In this comparison RNA from the IPTG study was used. These RNA samples were isolated independently from a culture that was split for 30 min and maintained at 37°C and would therefore be expected to show more variation than comparison of identical RNA samples. We do see a larger degree of variation when comparing independent preparations of RNA although we only see six spots with ratios greater than 3 using microarrays and 39 spots with ratios greater than 3 with spot blots. A similar degree of variation was seen by Wodicka et al. when independently prepared RNA samples were hybridized to separate oligonucleotide arrays (16). It is apparent that all global expression monitoring methods will show more variation when independently isolated RNA samples are used. This will become an even greater problem when different RNA samples from normal and wild-type cells are compared.

Based on our results we conclude that the microarray approach is more reproducible than radioactive hybridization to spot blots. The relative ease and minimum time required to perform expression studies with microarrays make this the preferred method. Radioactive hybridization to spot blots produce similar results but appear to be more subject to variation. One explanation for the higher variability with hybridization to separate membranes is variability in spot deposition. We used a single channel robotic printing device which may be more prone to variation than printing methods for microarrays which rely on capillary action of small volumes and contact printing. The advantage of radioactive methods is their minimal cost, availability of whole genome arrays, and the fact that most laboratories are currently equipped to perform these studies. However, if radioactive methods are used for global expression monitoring the preferred method should be sequential hybridization to the same membrane (Fig. 4A) rather than hybridization of control and experimental samples to separate membranes (Fig. 4C).

In the IPTG experiment described here we observed induction of the well-characterized lactose operon as well as the melibiose operon. Two ORFs of unknown function were seen to be significantly affected by IPTG using spot blots while these same two ORFs show no affect using microarray assays. Induction of ebgA by IPTG, as seen on spot blots, contrasted with results reported by Hall and Clarke (57) who found that IPTG treatment does not lead to increased enzymatic activity of EbgA assayed in crude protein extracts. The nucleotide sequences of lacZ and ebgA share ~50% identity over most of their length and further investigation reveals that lacZ cDNA can indeed bind to the ebgA PCR product during hybridization to spot blots. Although we did not observe the same affect on ebgA using glass microarray methods, Heller et al. report that cross-hybridization artifacts using microarray methods can occur due to low levels of sequence identity when they occur over the entire length of a target (33). The potential for cross-hybridization between genes with significant levels of sequence identity must always be considered when evaluating the results of global expression studies. In fact, 556 E.coli ORFs contain regions of >200 bp that share a minimum of 50% identity to at least one other ORF in the genome. We currently maintain a database of potential cross-hybridizing gene sequences and correlate these with expression analysis to avoid misinterpretation of potentially artifactual hybridization results (available at http://www.genetics. wisc.edu ).

In the heat shock analysis we identified 119 genes significantly affected by growth at 50°C for 7 min when determined by spot blots. More than half of the induced genes (46 of 77) were observed to be similarly affected when assayed using microarray analysis. Genes found to be down regulated using spot blots do not correlate as well when compared to microarray analysis. One explanation of the variation in ratios for these comparisons may be the use of rich defined growth media (23) for RNA samples used in microarray studies compared to growth in LB broth for spot blot analysis. In fact, the observation of repression of nucleotide biosynthetic genes following heat shock was not confirmed in additional heat shock experiments using LB broth, minimal media or rich defined media (not shown).

Correlation analysis of averaged heat shock data showed a very interesting result. As in control studies (IPTG), RNA samples from control and heat shock treated cells were isolated from a single culture that was split and subject to an environmental stimulus prior to RNA isolation. The correlation observed for this study (Fig. 4G and H) was quite different from that observed with IPTG treatment (Fig. 4E and F). Heat shock is pleiotropic effector of gene expression which is apparent in the plot. We reproducibly see a larger amount of signal variation between heat shock and control samples using either monitoring method. Again we see that this variation is diminished when using microarrays but the trend is still there. These results suggest that exposure to heat shock has much greater global effects than we are measuring using our significance tests. There may be a large number of genes subtly affected (either specifically or non-specifically) that we cannot accurately measure within the noise inherent to the method.

We have observed that different sets of genes are affected, both positively and negatively, when cells are subject to the same environmental stimuli in different media or at different times. For example, heat shock in LB broth (spot blots) showed a number of differences compared to heat shock in minimal media (data not shown) or rich defined media (microarrays). The majority of genes previously assigned to the heat shock stimulon, as well as a subset of ORFs with unassigned function, were consistently affected in all heat shock conditions, but a subset of genes responded in one condition but not in the other. In addition, when we compared RNA samples from batch cultures of MG1655 grown in different preparations of LB broth and grown to similar optical density, significant differences in the signal intensities of particular genes were observed. These observations argue for special care in the design of global expression studies. Since many interesting expression experiments will depend on addition or removal of compounds from the growth media we favor the use of a chemically defined medium rather than broth. The medium used for heat shock time course studies is ideal in that it is simple to prepare in bulk and supplement as required (23). Use of such a defined media will also facilitate integration of metabolite abundance information with global transcription analysis.

The ORFs affected by heat shock provide a list of potential players in the cell’s response to this stimulus but do not describe how the cell integrates these functions in response to stress. In some cases, however, it is possible to infer a biological role for a gene product based on its inclusion in the set of affected genes. For example, glyA is significantly repressed during the heat shock response under all growth conditions tested. The glyA gene encodes the enzyme serine hydroxymethyltransferase which converts serine to glycine generating 5,10-methylenetetrahydrofolate, a major source of one-carbon (C1) units in E.coli. The next step in catabolism of glycine to CO2 and ammonia by the enzymes of the gvc operon and lpdA also results in production of C1 units. (The lpdA gene is also down-regulated by heat shock; Table 2.) C1 units are used in a variety of biosynthetic pathways including deoxy­pyrimidine nucleotide biosynthesis and formyl-methionine biosynthesis. Gold has proposed that a possible mechanism of global regulation of protein synthesis may occur in the initiation of protein synthesis (58) for which C1 units are required. It is possible that down-regulation of the glyA and lpdA genes could lead to a global down-regulation of translation by limiting the supply of C1 units. This would suggest that simple repression of a few genes could lead to the global down-regulation of protein synthesis observed following heat shock (36,39).

Our findings also suggest that there is little change in the overall expression level of most genes following a 50°C heat shock. This observation is in agreement with the findings of Henry et al. (59) who found little change in the level of mRNA turnover following heat shock and suggest that the decrease in protein expression is controlled, not at the level of transcriptional repression, but at the translational or post-translational level.

The results presented here underscore the power of genome-wide measurements of transcript abundance. Detecting changes in response to different stimuli will define hypotheses regarding mechanisms of gene regulation and cellular responses for further analyses.


   ACKNOWLEDGEMENTS
 
We wish to thank Valerie Burland and Guy Plunkett III for stimulating discussion during preparation of this manuscript, Brian Peters and Patrick Brown’s lab for help in preparing microarrays, Chris Herring for work on lacZ and ebgA analysis, Nicole Perna for help with paralog analysis and Cathryn Lundberg and Shannon Hinsa for excellent technical assistance. This is Laboratory of Genetics Paper 3542. This work was supported by NIH grant R01 GM35682.


   FOOTNOTES
 
a To whom correspondence should be addressed. Tel: +1 608 262 2534; Fax: +1 608 263 7459; Email: craig{at}genetics.wisc.edu Back


   REFERENCES
TOP
 ABSTRACT
 INTRODUCTION
 MATERIALS AND METHODS
 RESULTS
 DISCUSSION
 REFERENCES
 

  1. 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) Science, 277, 1453–1474.

  2. Bult,C.J., White,O., Olsen,G.J., Zhou,L., Fleischmann,R.D., Sutton,G.G., Blake,J.A., FitzGerald,L.M., Clayton,R.A., Gocayne,J.D., Kerlavage,A.R., Dougherty,B.A., Tomb,J.F., Adams,M.D., Reich,C.I., Overbeek,R., Kirkness,E.F., Weinstock,K.G., Merrick,J.M., Glodek,A., Scott,J.L., Geoghagen,N.S.M. and Venter,J.C. (1996) Science, 273, 1058–1073.

  3. Cole,S.T., Brosch,R., Parkhill,J., Garnier,T., Churcher,C., Harris,D., Gordon,S.V., Eiglmeier,K., Gas,S., Barry,C.E.III, Tekaia,F., Badcock,K., Basham,D., Brown,D., Chillingworth,T., Connor,R., Davies,R., Devlin,K., Feltwell,T., Gentles,S., Hamlin,N., Holroyd,S., Hornsby,T., Jagels,K., Barrell,B.G. et al. (1998) Nature, 393, 537–544.

  4. Fleischmann,R.D., Adams,M.D., White,O., Clayton,R.A., Kirkness,E.F., Kerlavage,A.R., Bult,C.J., Tomb,J.F., Dougherty,B.A., Merrick,J.M. et al. (1995) Science, 269, 496–512.

  5. Himmelreich,R., Hilbert,H., Plagens,H., Pirkl,E., Li,B.C. and Herrmann,R. (1996) Nucleic Acids Res., 24, 4420–4449.

  6. Kunst,F., Ogasawara,N., Moszer,I., Albertini,A.M., Alloni,G., Azevedo,V., Bertero,M.G., Bessieres,P., Bolotin,A., Borchert,S., Borriss,R., Boursier,L., Brans,A., Braun,M., Brignell,S.C., Bron,S., Brouillet,S., Bruschi,C.V., Caldwell,B., Capuano,V., Carter,N.M., Choi,S.K., Codani,J.J., Connerton,I.F., Danchin,A. et al. (1997) Nature, 390, 249–256.

  7. Chuang,S.-E., Daniels.,D.L. and Blattner,F.R. (1993) J. Bacteriol., 175, 2026–2036.

  8. Shalon,D., Smith,S.J. and Brown,P.O. (1996) Genome Res., 6, 639–645.

  9. Schena,M., Shalon,D., Davis,R.W. and Brown,P.O. (1995) Science, 270, 467–470.

  10. Welford,S.M., Gregg,J., Chen,E., Garrison,D., Sorensen,P.H., Denny,C.T. and Nelson,S.F. (1998) Nucleic Acids Res., 26, 3059–3065.

  11. Khan,J., Simon,R., Bittner,M., Chen,Y., Leighton,S.B., Pohida,T., Smith,P.D., Jiang,Y., Gooden,G.C., Trent,J.M. and Meltzer,P.S. (1998) Cancer Res., 58, 5009–5013.

  12. DeRisi,J.L., Iyer,V.R. and Brown,P.O. (1997) Science, 278, 680–686.

  13. Chu,S., DeRisi,J., Eisen,M., Mulholland,J., Botstein,D., Brown,P.O. and Herskowitz,I. (1998) Science, 282, 699–705.

  14. Schena,M., Shalon,D., Heller,R., Chai,A., Brown,P.O. and Davis,R.W. (1996) Proc. Natl Acad. Sci. USA, 93, 10614–10619.

  15. Zhu,H., Cong,J.P., Mamtora,G., Gingeras,T. and Shenk,T. (1998) Proc. Natl Acad. Sci. USA, 95, 14470–14475.

  16. Wodicka,L., Dong,H., Mittmann,M., Ho,M.H. and Lockhart,D.J. (1997) Nature Biotechnol., 15, 1359–1367.

  17. Holstege,F.C., Jennings,E.G., Wyrick,J.J., Lee,T.I., Hengartner,C.J., Green,M.R., Golub,T.R., Lander,E.S. and Young,R.A. (1998) Cell, 95, 717–728.

  18. Spellman,P.T., Sherlock,G., Zhang,M.Q., Iyer,V.R., Anders,K., Eisen,M.B., Brown,P.O., Botstein,D. and Futcher,B. (1998) Mol. Biol. Cell, 9, 3273–3297.

  19. Gingeras,T.R., Ghandour,G., Wang,E., Berno,A., Small,P.M., Drobniewski,F., Alland,D., Desmond,E., Holodniy,M. and Drenkow,J. (1998) Genome Res., 8, 435–448.

  20. Behr,M.A., Wilson,M.A., Gill,W.P., Salamon,H., Schoolnik,G.K., Rane,S. and Small,P.M. (1999) Science, 284, 1520–1523.

  21. Iyer,V.R., Eisen,M.B., Ross,D.T., Schuler,G., Moore,T., Lee,J.C.F., Trent,J.M., Staudt,L.M., Hudson,J.Jr, Boguski,M.S., Lashkari,D., Shalon,D., Botstein,D. and Brown,P.O. (1999) Science, 283, 83–87.

  22. Eisen,M.B., Spellman,P.T., Brown,P.O. and Botstein,D. (1998) Proc. Natl Acad. Sci. USA, 95, 14863–14868.

  23. Neidhardt,F.C., Bloch,P.L. and Smith,D.F. (1974) J. Bacteriol., 119, 736–747.

  24. Sambrook,J., Fritsch,E.F. and Maniatis,T. (eds), (1989) Molecular Cloning: A Laboratory Manual. 2nd Edn. Cold Spring Harbor Laboratory, Cold Spring Harbor, New York, NY.

  25. Andre,P., Kim,A., Khrapko,K. and Thilly,W.G. (1997) Genome Res., 7, 843–852.

  26. Lundberg,K.S., Shoemaker,D.D., Adams,M.W., Short,J.M., Sorge,J.A. and Mathur,E.J. (1991) Gene, 108, 1–6.

  27. Cline,J., Braman,J.C. and Hogrefe,H.H. (1996) Nucleic Acids Res., 24, 3546–3551.

  28. Barnes,W.M. (1994) Proc. Natl Acad. Sci. USA, 91, 2216–2220.

  29. Webster,C., Kempsell,K., Booth,I. and Busby,S. (1987) Gene, 59, 253–263.

  30. Miller,J.H. (1980) In Miller,J.H. and Reznikoff,W.S. (eds), The Operon. Cold Spring Harbor Laboratory, Cold Spring Harbor, New York, NY, pp. 31–88.

  31. Stokes,H.W. and Hall,B.G. (1985) Mol. Biol. Evol., 2, 478–483.

  32. Wilbur,W.J. and Lipman,D.J. (1983) Proc. Natl Acad. Sci. USA, 80, 726–730.

  33. Heller,R.A., Schena,M., Chai,A., Shalon,D., Bedilion,T., Gilmore,J., Woolley,D.E. and Davis,R.W. (1997) Proc. Natl Acad. Sci. USA, 94, 2150–2155.

  34. Chuang,S.-E. and Blattner,F.R. (1993) J. Bacteriol., 175, 5242–5252.

  35. Gross,C.A. (1996) In Neidhardt,F.C., Curtiss,R.III, Ingraham,J.L., Lin,E.C.C., Low,K.B., Magasanik,B., Reznikoff,W.S., Riley,M., Schaechter,M. and Umbarger,H.E. (eds), Escherichia coli and Salmonella Cellular and Molecular Biology. 2nd Edn. ASM Press, Washington, DC, Vol. 1, pp. 1382–1400.

  36. Neidhardt,F.C., VanBogelen,R.A. and Vaughn,V. (1984) Annu. Rev. Genet., 18, 295–329.

  37. Neidhardt,F.C. (1987) Genes Dev., 1, 109–110.

  38. Lindquist,S. (1986) Annu. Rev. Biochem., 55, 1151–1191.

  39. Lemaux,P.G., Herendeen,S.L., Bloch,P.L. and Neidhardt,F.C. (1978) Cell, 13, 427–434.

  40. Yamamori,T., Ito,K., Nakamura,Y. and Yura,T. (1978) J. Bacteriol., 134, 1133–1140.

  41. Herendeen,S.L., VanBogelen,R.A. and Neidhardt,F.C. (1979) J. Bacteriol., 139, 185–194.

  42. Grossman,A.D., Erickson,J.W. and Gross,C.A. (1984) Cell, 38, 383–390.

  43. Erickson,J.W. and Gross,C.A. (1989) Genes Dev., 3, 1462–1471.

  44. Wang,Q.P. and Kaguni,J.M. (1989) J. Bacteriol., 171, 4248–4253.

  45. Watson,N., Dunyak,D.S., Rosey,E.L., Slonczewski,J.L. and Olson,E.R. (1992) J. Bacteriol., 174, 530–540.

  46. Yamanaka,K. and Inouye,M. (1997) J. Bacteriol., 179, 5126–5130.

  47. Missiakas,D., Mayer,M.P., Lemaire,M., Georgopoulos,C. and Raina,S. (1997) Mol. Microbiol., 24, 355–371.

  48. De Las Penas,A., Connolly,L. and Gross,C.A. (1997) Mol. Microbiol., 24, 373–385.

  49. Katayama,Y., Gottesman,S., Pumphrey,J., Rudikoff,S., Clark,W.P. and Maurizi,M.R. (1988) J. Biol. Chem., 263, 15226–15236.

  50. Gottesman,S., Clark,W.P. and Maurizi,M.R. (1990) J. Biol. Chem., 265, 7886–7893.

  51. Kroh,H.E. and Simon,L.D. (1990) J. Bacteriol., 172, 6026–6034.

  52. Conlin,C.A., Trun,N.J., Silhavy,T.J. and Miller,C.G. (1992) J. Bacteriol., 174, 5881–5887.

  53. Jiang,X., Zhang,M., Ding,Y., Yao,J., Chen,H., Zhu,D. and Muramatu,M. (1998) J. Biochem. (Tokyo), 124, 980–985.

  54. Tsui,H.C., Feng,G. and Winkler,M.E. (1996) J. Bacteriol., 178, 5719–5731.

  55. Herman,C., Thevenet,D., D’Ari,R. and Bouloc,P. (1995) Proc. Natl Acad. Sci. USA, 92, 3516–3520.

  56. Missiakas,D., Schwager,F., Betton,J.M., Georgopoulos,C. and Raina,S. (1996) EMBO J., 15, 6899–6909.

  57. Hall,B.G. and Clarke,N.D. (1977) Genetics, 85, 193–201.

  58. Gold,L. (1988) Annu. Rev. Biochem., 57, 199–233.

  59. Henry,M.D., Yancey,S.D. and Kushner,S.R. (1992) J. Bacteriol., 174, 743–748.

  60. . Riley,M. (1998) Nucleic Acids Res., 26, 54.

Received June 15, 1999. Revised and Accepted July 23, 1999.


Add to CiteULike CiteULike   Add to Connotea Connotea   Add to Del.icio.us Del.icio.us    What's this?


This article has been cited by other articles:


Home page
 J. Biol. Chem.Home page
P. Bross, S. Naundrup, J. Hansen, M. N. Nielsen, J. H. Christensen, M. Kruhoffer, J. Palmfeldt, T. J. Corydon, N. Gregersen, D. Ang, et al.
The Hsp60-(p.V98I) Mutation Associated with Hereditary Spastic Paraplegia SPG13 Compromises Chaperonin Function Both in Vitro and in Vivo
J. Biol. Chem., June 6, 2008; 283(23): 15694 - 15700.
[Abstract] [Full Text] [PDF]

Home page
 J. Bacteriol.Home page
T. S. Gunasekera, L. N. Csonka, and O. Paliy
Genome-Wide Transcriptional Responses of Escherichia coli K-12 to Continuous Osmotic and Heat Stresses
J. Bacteriol., May 15, 2008; 190(10): 3712 - 3720.
[Abstract] [Full Text] [PDF]

Home page
 J. Bacteriol.Home page
E. M. Widenbrant and C. M. Kao
Introduction of the Foreign Transposon Tn4560 in Streptomyces coelicolor Leads to Genetic Instability near the Native Insertion Sequence IS1649
J. Bacteriol., December 15, 2007; 189(24): 9108 - 9116.
[Abstract] [Full Text] [PDF]

Home page
 J. Bacteriol.Home page
K. Inoue, S. Basu, and M. Inouye
Dissection of 16S rRNA Methyltransferase (KsgA) Function in Escherichia coli
J. Bacteriol., December 1, 2007; 189(23): 8510 - 8518.
[Abstract] [Full Text] [PDF]

Home page
 J. Bacteriol.Home page
A. Rasouly, Y. Shenhar, and E. Z. Ron
Thermoregulation of Escherichia coli hchA Transcript Stability
J. Bacteriol., August 1, 2007; 189(15): 5779 - 5781.
[Abstract] [Full Text] [PDF]

Home page
 J. Bacteriol.Home page
X.-S. Zhang, R. Garcia-Contreras, and T. K. Wood
YcfR (BhsA) Influences Escherichia coli Biofilm Formation through Stress Response and Surface Hydrophobicity
J. Bacteriol., April 15, 2007; 189(8): 3051 - 3062.
[Abstract] [Full Text] [PDF]

Home page
 RNAHome page
B. Wegscheid and R. K. Hartmann
The precursor tRNA 3'-CCA interaction with Escherichia coli RNase P RNA is essential for catalysis by RNase P in vivo
RNA, December 1, 2006; 12(12): 2135 - 2148.
[Abstract] [Full Text] [PDF]

Home page
 Nucleic Acids ResHome page
B. K. Mohanty and S. R. Kushner
The majority of Escherichia coli mRNAs undergo post-transcriptional modification in exponentially growing cells
Nucleic Acids Res., November 14, 2006; 34(19): 5695 - 5704.
[Abstract] [Full Text] [PDF]