INTRODUCTION

Top Abstract Introduction Materials and Methods Results and Discussion References

The field of microbial physiology was launched in 1958 with the fundamental discovery that the macromolecule composition of the bacterial cell changes with the growth rate (58). Faster-growing cells contain proportionally more stable RNAsrRNA and tRNA. The reason for this increased abundance of stable RNA is simple: in order to grow faster, bacteria must synthesize protein faster. The growth rate of the bacterial cell increases in proportion to the quality of the growth medium (although not necessarily in proportion to its exact composition), and this increase in growth rate is accomplished by an increase in the number of ribosomes and the concentrations of translation accessory factors (8). It is now understood that the seven Escherichia coli rRNA operons are under the control of growth rate-dependent promoters and that expression of the ribosomal proteins, translation factors, and the transcription apparatus are all tied to the cellular concentration of rRNA (8, 27, 35). The rate of transcription initiation of the growth rate-dependent rrn promoters is physiologically connected to the metabolic state of the cell by the concentration of nucleoside triphosphatesefficient transcription initiation from these promoters requires a high concentration of the initiating nucleotide (22). The presence of high-quality nutrients in the growth medium results in high intracellular nucleoside triphosphate concentrations; hence, this model unifies the idea that the quality of the growth medium dictates the growth rate of the cell.

Growth rate-dependent changes in cell composition are realized at the level of gene expression; for example, transcript levels corresponding to the protein components of the protein synthesis apparatus change in proportion to the growth rate as the rates of transcription or mRNA turnover are modulated (27, 35). Other changes in cellular physiology can be more subtle, such as redirection of intermediary metabolism in response to changes in growth medium composition or the flow of carbon and electrons that is coupled to ATP generation, although many of these adjustments in metabolism are accompanied by changes in the concentrations of metabolic enzymes and electron transport chain components (40, 41, 56, 63, 64). The expression of numerous other genes is affected by environmental stresses (9, 17, 26, 29, 48, 60, 69, 71). Almost all aspects of microbial physiology, including the myriad adjustments made by the cell in response to changes in the environment, have been cataloged by the scientific community in the form of the book Escherichia coli and Salmonella: Cellular and Molecular Biology. Since the publication of this compendium, the sequence of the E. coli genome has been completed and the way that we look at gene expression is forever changed (6). The genome sequence provides the tools necessary to take a global view of E. coli physiology.

Genomic expression assays provide an unprecedented ability not only to look at a single aspect of physiology but also to see how a particular gene, regulon, or modulon interacts with every other aspect of physiology. Genomewide methods have been developed for a number of uses, including drug discovery (43), measurement of gene copy number (50), discovery of disease-related genes in humans (18, 28), gene mapping (12), and gene expression: in humans (73), in yeast (13, 19, 31, 37, 65), and in Arabadopsis (59).

From the E. coli MG1655 genome sequence (6), 4,290 open reading frame (ORF)-specific primer pairs were designed for PCR amplification of all E. coli ORFs, and this set of 4,290 PCR-amplified, ORF-specific DNA fragments was used to develop DNA arrays for gene expression profiling (54). A similar set of ORF-specific DNA fragments was used to generate commercially available DNA macroarrays (12 by 24 cm) on nylon membranes (Sigma-GenoSys Biotechnologies, Inc., Woodland, Tex.). The advantage of the commercial arrays is that they can be used with equipment found in typical molecular biology laboratories. For these utilitarian investigations of bacterial physiology to be successful, it will be necessary to determine if DNA macroarrays can reveal differences in gene expression across the genome. Here we report on the expression profiles of E. coli under two very different growth conditions, and from the data we provide insights into growth rate-dependent gene expression, global regulation of biosynthetic regulons, and stress responses that appear to be involved in growth on minimal glucose medium.

	MATERIALS AND METHODS

Top Abstract Introduction Materials and Methods Results and Discussion References

Growth conditions. E. coli MG1655 cultures were grown in 50-ml batch cultures in 250-ml Erlenmeyer flasks at 37°C with aeration by gyrotary shaking (300 rpm). The culture media used were M63 minimal medium (57) containing 0.2% glucose and a rich medium, Luria broth (39) containing 0.2% glucose. Growth was monitored spectrophotometrically at 600 nm on a Spectronic 601 (Milton Roy). Cells were harvested in late logarithmic growth phase (absorbance at 600 nm = 0.6) from cultures that had been inoculated at low density and had maintained a constant growth rate for at least 10 generations.

Handling of RNA. The ability to isolate pure, intact mRNA is critical to the success of genomic expression assays. Cells in growing cultures were pipetted directly into boiling lysis buffer. The lysed cells were extracted twice with phenol (pH 5.0) at 60°C and then with phenol-chloroform (66). The RNA was precipitated with isopropanol, redissolved in water, treated with DNase I, and applied to an RNeasy column. The purified RNA was redissolved in water and stored at 70°C in 2 volumes of ethanol.

Probe synthesis. Hybridization probes were generated by standard cDNA synthesis. The protocol supplied by the manufacturer of the DNA arrays was suitable for achieving >70% incorporation of the ³³P-labeled nucleotide. Since it is not possible to purify bacterial mRNA from total RNA (i.e., by purification of polyadenylated mRNA as in eukaryotes), the labeling protocol takes into account the presence of rRNA and tRNA, which constitute 85% of the total RNA. The C-terminal primer set (4,290 ORF-specific C-terminal primers [Sigma-GenoSys Biotechnologies, Inc.]) was used to generate the hybridization probe in a standard first-strand cDNA synthesis. Briefly, 1 µg of RNA was mixed with dATP, dGTP, and dTTP (final concentrations, 0.33 mM each), and cDNA-labeling primers (Sigma-GenoSys), in a volume of 25 µl of first-strand buffer, heated to 90°C for 2 min and cooled to 42°C in 20 min. Then 200 U of Superscript II, 10 U of RNase inhibitor, and 20 µCi of [-³²P]dCTP (2,000 to 3,000 Ci/mmol) were added, bringing the total volume to 30 µl, and the cDNA synthesis reaction mixture was incubated at 42°C for 2 h. Unincorporated nucleotides were removed by gel filtration through a G-50 Sephadex column (57).

Hybridization. The DNA arrays (Panorama E. coli gene arrays) used in the hybridization experiments were produced by Sigma-GenoSys Biotechnologies, Inc. Each DNA array consists of a 12- by 24-cm positively charged nylon membrane on which 10 ng each of all 4,290 PCR-amplified ORF-specific DNA fragments are robotically printed in duplicate. The hybridization and washing steps were carried out as described by the manufacturer. Briefly, the blots were prehybridized in hybridization solution (5× SSPE [1× SSPE is 0.18 M NaCl, 10 mM NaH₂PO₄, and 1 mM EDTA, pH 7.7], 2% sodium dodecyl sulfate [SDS], 1× Denhardt's reagent, 100 µg of sheared salmon sperm DNA per ml) at 65°C for 1 h in a 30- by 3.5-cm roller bottle in a hybridization oven. The entire cDNA probe, generated as described above, was added to 3 ml of hybridization buffer, and the blot was hybridized with this solution for 15 h at 65°C. The blots were washed with buffer (0.5× SSPE, 0.2% SDS) three times for 5 min each at room temperature and three times for 30 min each at 65°C. The blots were then wrapped in clear plastic food wrap and exposed to a PhosphorImager screen (Molecular Dynamics, Sunnyvale, Calif.) for 48 h. For each of the data sets used in this study, the same blot was consecutively hybridized, stripped, and rehybridized (this can be done up to four times). The blots were stripped at 100°C with 1% SDS in Tris-EDTA buffer as specified by the manufacturer.

Data analysis. The exposed PhosphorImager screens were scanned with a pixel size of 100 µm (10,000 dots/cm²) on a STORM 840 PhosphorImager (Molecular Dynamics). The resulting TIFF image files were analyzed by determining the pixel density (intensity) for each spot in the array by using ImageQuant (version 5.0) software (Molecular Dynamics). A grid of individual ellipses corresponding to each of the DNA spots on the blots was laid down on the image to designate each spot to be quantified. Background was subtracted automatically by the software by using the local median background subtraction method. The intensities for each spot were exported from ImageQuant into a Microsoft Excel spreadsheet. Each ORF-specific spot was present in duplicate, and the intensities were averaged for analysis. Each averaged spot intensity was expressed as a percentage of the total of intensities of all the spots on the DNA array, which allowed direct comparison of the two conditions by normalizing with regard to the specific activity of the probes used. The correlation coefficients of the percent intensities determined individually for the duplicate spots on a single blot ranged from 0.986 to 0.999, and the standard deviations for the log ratios of intensities of the duplicate spots (determined as described below) ranged from 0.073 to 0.095 for four different hybridizations, thus providing a measure of reproducibility.

Functional groups. Two schemes for functional grouping of genes have been applied to the expression data generated in these experiments. The first scheme assigns genes to groups in accordance with their cellular function, as described previously (6). The second scheme of functional assignments is that of Riley (55), version M54, submitted by Plunkett et al. (19a), as it appears on the E. coli K-12 MG1655 complete genome at the National Center for Biotechnology Information (43a).

Internet access to data. An Internet accessible version of the expression data and details of the protocols has been created (49a). The data can also be accessed from a database (19a).

Chemicals. SuperScript II, an RNase H reverse transcriptase used for cDNA synthesis, was purchased from Gibco BRL (Bethesda, Md.). RNase inhibitor and DNase I were also purchased from Gibco BRL. PCR grade deoxyribonucleoside triphosphates were purchased from Roche Molecular Biochemicals (Indianapolis, Ind.). RNeasy columns were purchased from Qiagen, Inc. (Valencia, Calif.). [-³³P]dCTP (2,000 to 3,000 Ci/mmol) was purchased from New England Nuclear (Wilmington, Del.). Biochemicals were purchased from Sigma (St. Louis, Mo.).

	RESULTS AND DISCUSSION

Top Abstract Introduction Materials and Methods Results and Discussion References

The genomic expression profiles of E. coli MG1655 growing on rich and on minimal culture media (Fig. 1) were determined. The rich medium (Luria broth) contained amino acids as the nitrogen source, a number of other preformed building blocks of macromolecule synthesis (e.g., nucleosides and vitamins, etc., provided by tryptone and yeast extract), and also glucose as a carbon and energy source. The minimal medium contained glucose as the sole carbon and energy source and ammonia as the nitrogen source. In glucose minimal medium, the carbon backbone of the glucose molecule was rearranged through the biosynthetic pathways to generate each of the building blocks de novo. In addition to having fundamentally different metabolisms, the two cultures grew at significantly different rates: G = 25 min on the rich medium and G = 57 min on minimal glucose medium. As a control, data are provided for a culture growing on minimal gluconate medium (G = 60 min).

View larger version (18K): [in this window] [in a new window]   FIG. 1.   Growth of E. coli MG1655 on Luria broth plus glucose (open squares), minimal glucose medium (open circles), and minimal gluconate medium (solid circles). Cells were harvested for genomic expression analysis at an absorbance at 600 nm (A600) of 0.6.

Whole-genome perspective. RNA isolated from the cultures in Fig. 1 were used to generate the probes used for hybridization of the DNA arrays shown in Fig. 2, and the data were quantified as described in Materials and Methods. Calculation of the log expression ratios of corresponding spots allowed pairwise comparisons of the relative transcript levels for each of the 4,290 E. coli protein-encoding genes under the different growth conditions. The log expression ratios indicate whether gene expression is higher under one condition or the other or remains unchanged. The results are summarized in Table 1 and presented in chart form in Fig. 3 to 6. It is important to keep in mind that in vivo transcript levels are dynamically balanced by the rates of transcription initiation and transcript turnover. Thus, the data presented here as expression ratios reflect the relative transcript levels for individual genes without providing any indication of the mechanism of regulation. Furthermore, some individual expression ratios may be in error, due to technical problems, including cross-hybridization, PCR failures, misapplied DNA spots on the arrays, or scatter in the data (see reference 54 for a more comprehensive review of the technical aspects of using E. coli DNA arrays). A few of the ratios are in conflict with published results, and it is possible that other ratios will not be validated in subsequent experiments. Thus, these data should not be taken as specific evidence for gene regulation and should be independently verified. Nevertheless, the general trends of the data are substantially clear and will be of value for generating experimental leads.

View larger version (194K): [in this window] [in a new window]   FIG. 2.   DNA arrays of the entire set of E. coli genes hybridized with probes generated from RNA extracted from cells growing in late logarithmic phase on minimal glucose medium (left) and on Luria broth (LB) containing glucose (right).

View larger version (26K): [in this window] [in a new window]   FIG. 3.   The log expression ratios of all E. coli genes were plotted for minimal glucose versus Luria broth plus glucose (top) and for minimal glucose versus minimal gluconate (bottom). The entire data sets were sorted in Excel spreadsheets by the log expression ratio values, and a bar chart was generated by the software, with individual genes plotted on the x axis and the log expression ratios plotted on the y axis. Genes more highly expressed under the first condition are positive, and genes more highly expressed under the second condition are negative. The horizontal divisions (dashed lines) represent 99% confidence levels, such that any gene with a value extending beyond the first horizontal division in either direction is significantly expressed at a higher level under that condition.

Translation apparatus. The culture containing rich medium plus glucose grew more than twice as fast as did the cultures on minimal media (Fig. 1). It is known that faster-growing cells synthesize protein faster and contain more ribosomes (27, 35). There are 128 known genes encoding the enzymes, factors, and structural components that make up the translation apparatus. Of these 128 genes of the translation apparatus, 53 (41.4%) were expressed at significantly higher levels in the cells growing on rich medium and none of them were expressed at significantly higher levels on the minimal medium. Of the 53 translation genes that were expressed at higher levels on rich medium, 42 encoded ribosomal proteins. These data are charted in Fig. 4 and can be compared to the data for the cultures on minimal glucose versus gluconate medium, which had nearly identical growth rates and showed very few significant differences in expression of the translation genes. A comparison of the general pattern of expression of the translation genes (Fig. 4) to that of the entire E. coli gene set (Fig. 3) further illustrates the dramatic increase in production of the translation apparatus in the faster-growing cells.

View larger version (35K): [in this window] [in a new window]   FIG. 4.   Log expression ratios of the translation apparatus genes sorted by value. The set of all translation apparatus genes is shown in the top two panels for the minimal glucose versus minimal gluconate and minimal glucose versus Luria broth plus glucose experiments (see the legend to Fig. 3). The bottom three panels show the results of the minimal glucose versus Luria broth plus glucose experiment for functionally grouped subsets of the translation apparatus genes.

(i) tRNA synthetase genes. There are 37 known genes encoding the tRNA synthetases and other enzymes involved in tRNA modification. While none of the expression ratios of the tRNA synthetase genes varied significantly, it is clear from the chart in Fig. 4 that the transcript levels for these genes followed the same general trend as the complete set of translation genes. This result is consistent with the notion that synthesis of the tRNA synthetases is coupled to the synthesis of other ribosomal components (27).

(ii) Translation factors. There are 17 known genes that encode factors involved in translation and ribosome modification, including the initiation and elongation factors, and 7 of these genes were expressed at significantly higher levels on rich medium (Fig. 4; Table 2). This result is generally consistent with the coupled synthesis of translation factors and ribosome components (27). The expression ratio of infB was significantly higher on rich medium. The regulation of infB, which is downstream of and cotranscribed with the transcription factor gene nusA, is complex and is thought to be the result of autoregulation of the extent of readthrough at upstream terminators by NusA (27). The expression ratio of infB was 1.8-fold higher than that of nusA (data not shown). The expression ratios of the translation elongation factor genes tsf, tufB, tufA, and fusA were all significantly higher, in that order, on rich medium, which is consistent with their coordinate regulation with the ribosomal protein genes (27). The growth rate-dependent regulation of tsf, tufA, and fusA, all of which are located in ribosomal protein operons, is the result of mRNA destabilization in slowly growing cells (27). Interestingly, regulation of tufB appears to be at least partially dependent upon Fis (68), and the fis gene had one of the highest expression ratios on rich medium, as described in more detail below. A fifth elongation factor encoded by efp has been shown to be essential in E. coli for protein synthesis and viability, although the details of efp regulation have not been published (2). The results of this study indicate that efp was expressed at a significantly higher level (log ratio = 0.425) in the faster-growing cells on rich medium, paralleling the expression of the other elongation factors.

(iii) Ribosomal proteins. Of the 55 genes encoding the ribosomal proteins, 42 were expressed at significantly higher levels in the more rapidly growing cells in rich medium (Fig. 4; Table 2). This result is consistent with the paradigm of growth rate-dependent regulation of ribosome number (35). Although the ribosomal S10 operon is at least partially regulated at the transcriptional level, it is generally accepted that regulation of the 21 ribosomal protein operons is not at the level of transcription initiation (23, 35). Rather, the regulation of ribosomal protein synthesis involves a combination of translational control and transcriptional control at the level of mRNA stability. In general, growth conditions which lead to a decreased rate of ribosome synthesis result in an excess of ribosomal proteins, with certain ones serving as autoregulators by binding to their transcript and decreasing the translation rate of the mRNA, thus leading to destabilization of the transcript (35). While not all of the ribosomal protein operons have been studied at this level of detail, the experiment presented here indicates that most of the operons are regulated in such a way that their transcript levels are higher in faster growing cells. Clearly, these data demonstrate that any regulatory mechanism that contributes to the dynamic control of a particular mRNA concentration, whether it be the rate of transcription or the rate of turnover, can be visualized in genomic expression assays. The global regulation and coordination of ribosome number and components of the translation apparatus was the most obvious result of this experiment.

Nitrogen metabolism. The minimal medium used in this study contained ammonia as the nitrogen source and the rich medium contained amino acids as the nitrogen source. In general, cells growing on minimal medium are limited for amino acids while cells growing on rich medium are limited for nucleotides (47, 52, 76). These differences were reflected in the transcript levels of the genes involved in nitrogen assimilation and biosynthesis of amino acids. The genes involved in assimilation of ammonia as the nitrogen source were expressed at significantly higher levels on minimal medium, including gdhA, which encodes glutamate dehydrogenase, and gltD, which encodes a subunit of glutamate synthase (Table 3). While it is known that gdhA is transcriptionally regulated by ammonia, next to nothing is known about the mechanism (53). The gltBD operon is subject to complex regulation by certain amino acids and in a concentration-dependent fashion by leucine-responsive protein (Lrp) (20); thus, the high induction ratio of gltBD on minimal medium (0.329 for gltB; 0.889 for gltD) can be explained by amino acid repression in rich medium and a high induction ratio of Lrp on minimal medium (see below). Conversely, glnA, which encodes glutamine synthase and is induced by nitrogen limitation (as indicated by a low ratio of intracellular glutamine to -ketoglutarate), had the highest (although not significantly so) expression ratio (0.316) of any of the amino acid biosynthetic genes in rich medium (52). In summary, the genes involved in ammonia assimilation were induced for growth on minimal medium where ammonia was the nitrogen source.

Biosynthesis of amino acids. The overall expression pattern of the genes encoding the enzymes of amino acid biosynthesis indicated that these were generally induced for growth on minimal medium (Fig. 5; Table 3). The argA gene, which encodes N-acetylglutamate synthase, the first enzyme of the pathway, and also ygjG, a probable ornithine aminotransferase, were expressed at significantly higher levels on minimal medium. Expression of the genes of the branched-chain amino acid biosynthetic pathways (67)was significantly elevated in minimal medium. The first gene of the ilvGMEDA operon, which encodes the enzymes of isoleucine and valine synthesis, was expressed at significantly higher levels on minimal medium. Interestingly, the monocistronic gene ilvC, which is derepressed exclusively by valine, had a log expression ratio of 0.977 on minimal medium, the highest of any of the amino acid biosynthesis genes. The leucine biosynthetic genes, encoded by the leuABCD operon, were all expressed at significantly higher levels on minimal medium. The high expression ratios of the leucine and valine biosynthetic genes are consistent with the relatively high abundance of these two amino acids (third and fourth most abundant, respectively) in E. coli cells (45). The genes encoding the first enzymes of the four branches of the aromatic amino acid biosynthetic pathways were all significantly elevated in cells grown on minimal medium (51). The first step of the "common pathway" of chorismate synthesis, encoded by aroF, and the first step of tyrosine biosynthesis, encoded by tyrA, form an operon, in that order, and had log expression ratios of 0.847 and 0.934, respectively. The pheA gene was significantly elevated, as were four of the five genes of the trpEDCBA operon; the trpD transcript level was high in both minimal and rich media. The gene encoding the first step in serine biosynthesis, serA, and the gene which codes for the enzyme that forms glycine from serine, glyA, were expressed at significantly higher levels on minimal medium. The cysK gene, which encodes cysteine synthase A, was expressed at significantly higher levels on minimal medium, while cysM, the gene encoding cysteine synthase B, was expressed at slightly higher levels on rich medium. The cysE product, serine transacetylase, forms a multifunctional complex with the cysK product, and the relative expression ratios of cysK and cysE (0.497 versus 0.024) are consistent with the cysE product being much less abundant in the enzyme complex (36). The uniquely MetR-regulated methionine synthase gene, metE, was expressed at a significantly higher level on minimal medium, in contrast to the remaining MetJ-regulated genes of methionine biosynthesis, which did not vary significantly (25). The cobalamin-dependent methionine synthase encoded by metH was not expressed on minimal or rich media (data not shown). Overall, 8 of the 22 amino acid biosynthesis genes which were significantly elevated on minimal medium corresponded to the first step in the biosynthetic pathway. Thus, significant elevation of the first step in the amino acid biosynthetic pathways in cells grown on minimal medium was a recurring regulatory theme, consistent with the roles of these steps in controlling the flow of precursor metabolites out of the central pathways and into biosynthesis. Increased expression of the amino acid biosynthetic genes on minimal medium was indicative of the need to generate these building blocks from the sole carbon source, glucose.

View larger version (35K): [in this window] [in a new window]   FIG. 5.   Log expression ratios of biosynthetic genes were sorted by value for the minimal glucose versus Luria broth plus glucose experiment and are grouped by related pathways.

Biosynthesis of vitamins, cofactors, prosthetic groups and carriers. Expression of the 106 genes involved in biosynthesis of vitamins, cofactors, prosthetic groups, and carriers followed the same trend as the genes of amino acid biosynthesis, although the expression ratios were generally not so large (Fig. 5; Table 3). Among the genes expressed at significantly higher levels on minimal medium were hemC, involved in porphyrin biosynthesis; the first three genes of the entCEBA operon and entF, involved in enterobactin biosynthesis; grxB, encoding glutaredoxin 2; gst, encoding glutathione S-transferase; folE, encoding the first step in tetrahydrofolate biosynthesis; and ggt, involved in glutathione biosynthesis. Only one gene, thiH, had an expression ratio that was significantly higher on rich medium, but this was in contrast to the remainder of the thi genes, which in general were expressed at modestly higher levels on minimal medium. Given that the vitamins and cofactors synthesized by the pathways in this functional group are needed in very small amounts, so small that they are rarely included in chemical composition tables (45), it is not surprising that most of these genes did not have significant expression ratios. Nevertheless, the general trend of higher expression in minimal medium is again indicative of the need to generate these building blocks de novo from glucose.

Nucleotide biosynthesis. While expression of the genes involved in biosynthesis of amino acids, vitamins, enzyme cofactors, and prosthetic groups, etc., was generally elevated on minimal glucose medium, expression of the genes involved in nucleotide salvage and biosynthesis was more evenly divided between the two growth conditions (Fig. 5; Table 3). The pyrBI genes, which form an operon encoding the first step of pyrimidine biosynthesis, were expressed at significantly higher levels on minimal medium, perhaps reflecting the presence of uridine in the rich medium, which would tend to repress these genes (47). There are three enzymes involved in conversion of ribonucleotides to deoxyribonucleotides (33, 47), but of the three corresponding genetic loci only the nrdHIEF operon was expressed at significantly higher levels on minimal glucose medium. In previous studies, hyperinduction of nrdEF by hydroxyurea was measured, but this was the first experiment comparing nrdEF transcript levels under normal growth conditions (33), and this was the first indication that the genes encoding the NrdEF accessory proteins, NrdH and NrdI (34), are coregulated with nrdEF. There were five genes that were expressed at significantly higher levels in rich medium: the prsA gene, which encodes an enzyme that forms the first precursor of purine biosynthesis, and upp, gmk, pfs, and ndk, all of which encode enzymes involved in nucleotide salvage or interconversion, consistent with the availability of nucleotides in the rich medium.

Fatty acid biosynthesis and degradation. The cfa gene, which encodes an enzyme responsible for postsynthetic formation of cyclopropane fatty acids from unsaturated fatty acids, had a significantly high expression ratio on minimal medium (Table 3; also see Table 6). Since cfa is transcribed from an RpoS-dependent promoter (16), this result is consistent with elevated expression of rpoS on minimal medium (see below). All of the genes of the fad regulon (13) of fatty acid degradation (except for fadA [possibly an erroneous result]) were significantly elevated on rich medium, including fadB, which is in the fadAB operon, and fadD, which together encode the fatty acid oxidation multienzyme complex (Fig. 5). Also significantly elevated on rich medium were fadR, the repressor of the fad genes, and fadL, which encodes a long-chain fatty acid transporter. These results tend to indicate that the cells grown on rich medium were exposed to exogenous long-chain fatty acids, leading to induction of the fad regulon (14). Interestingly, the ato genes, which are involved in degradation of four-carbon fatty acids, were modestly elevated on minimal medium, and the sensor of the two-component regulator of these genes, encoded by atoS, was significantly elevated on minimal medium. These results suggest that the cells grown on minimal glucose medium were exposed to acetoacetate, which is the inducer of the ato genes (14). E. coli is not known to form acetoacetate from glucose, and it is possible that some closely related compound such as acetolactate, which is formed by E. coli, serves as an inducer of the ato genes (46, 67).

Carbon and energy metabolism. The cells grown on rich medium showed nothing remarkable with respect to the expression pattern of genes involved in carbon and energy metabolism. Of the 409 genes of carbon catabolism, central metabolism, and energy metabolism, only 8 were expressed at significantly higher levels on rich medium (Table 4). These included nuoM and nuoN of the large operon encoding NADH dehydrogenase I and cyoA of the operon encoding cytochrome oxidase c (24), suggesting that aerobic respiration was elevated under this growth condition.

Cellular processes and global regulators. Growth on minimal medium with glucose as the sole carbon and energy source places a burden on the cell to synthesize its amino acids de novo or starve. Thus, cells growing on minimal glucose medium are partially starved for amino acids, certainly a stressful situation and potentially having dramatic consequences on global gene regulation, elevating transcript levels of stress-inducible genes, and invoking the stringent response (8, 9, 29, 32). Several of the genes known to be regulated by the stringent-response signal molecule, ppGpp, were found to be differentially regulated on minimal and rich media (Fig. 6; Table 5). Most notable of these genes was rpoS, encoding the stationary-phase sigma factor, which was significantly elevated on minimal medium. In fact it appeared that RpoS-dependent gene expression was a prominent feature of the genomic expression pattern of cells grown on minimal medium (Table 6). It is not clear from these genomic expression assays if the elevated level of the rpoS transcript was the result of regulation by ppGpp, although this would be consistent with the positive correlation between ppGpp concentration and RpoS levels (9), because it was also found that expression of nlpD (which encodes a lipoprotein and is operonic with rpoS) was significantly elevated on minimal medium (Table 2). Thus, these data do not distinguish between the possibilities that the higher level of rpoS transcription was driven by the nlpD promoter or by the rpoS promoters located within the upstream nlpD gene (29, 38). Production of RpoS is also subject to complex posttranscriptional and translational regulation, and therefore it cannot be presumed that rpoS transcript levels are correlated with RpoS activity (29). However, the number of RpoS-inducible genes that were observed to be expressed at significantly elevated levels on minimal medium (21 of them) argues strongly in this case that the rpoS transcript level correlated with RpoS function. Interestingly, of the 21 RpoS-dependent genes which were significantly elevated on glucose minimal medium, more than half are known to be involved in the physiological changes that highlight entry into stationary phase (32). However, the cells used in these experiments were in late logarithmic growth phase, still in steady-state growth. Although most studies have focused on the role of RpoS in preparing cells for entry into stationary phase, it has been suggested that RpoS may play a role in logarithmic phase as well (29), and the results presented here support this idea. Since this question is likely to receive further attention, a time course study of genomic expression in cells growing on minimal glucose medium in batch culture would be invaluable.

View larger version (25K): [in this window] [in a new window]   FIG. 6.   Log expression ratios of cell process genes and regulatory genes sorted by value for the minimal glucose versus Luria broth plus glucose experiment.

Conclusion. In the single experiment presented here, the hallmark features of growth on minimal and rich media were revealed. Across the genome, we observed differences in the expression of functionally grouped genes that paralleled the physiology of these two growth conditions. Cells grown in rich medium with a good carbon and energy source, glucose, grew rapidly, turning off the pathways of biosynthesis and elevating the expression of the genes involved in macromolecule synthesis, most prominently protein synthesis. Cells in minimal medium faced the need to synthesize all of their building blocks from a single carbon and energy source, again glucose, and this burden was reflected not just in the turning on of biosynthetic pathways but also in the elevated expression of regulators of cell processes and regulons involved in stress tolerance. The most prominent features of growth on glucose minimal medium were the formation of overflow metabolites, in particular acetate, and protection of the cell from the stress of living in a self-formed acidic environment. All of these aspects of physiology were revealed not by painstaking and careful analysis in the laboratory of each system but, rather, by deduction from the genomic expression patterns of cells grown under these two rather different conditions. These deductions would not have been possible were it not for countless microbial physiology experiments published over the past 50 years (44). On the other hand, from the one simple experiment reported here, the tremendous potential of functional genomics is obvious. As a result of this experiment, several genes were added to functional groups on the basis of coregulation with similar and related genes. Also, several testable hypotheses were generated, in particular those involving the flow of carbon to acetate, coupling of glucose and acetate cometabolism, and acid resistance, the importance of which has been previously pointed to in enteric bacteria (4).