Global transcriptional response of Escherichia coli O157:H7 to growth transitions in glucose minimal medium.

BACKGROUND: Global patterns of gene expression of Escherichia coli K-12 during growth transitions have been deeply investigated, however, comparable studies of E. coli O157:H7 have not been explored, particularly with respect to factors regulating virulence genes and genomic islands specific to this pathogen. To examine the impact of growth phase on the dynamics of the transcriptome, O157:H7 Sakai strain was cultured in MOPS minimal media (0.1% glucose), RNA harvested at 10 time points from early exponential to full stationary phase, and relative gene expression was measured by co-hybridization on high-density DNA microarrays. Expression levels of 14 genes, including those encoding Shiga toxins and other virulence factors associated with the locus of enterocyte effacement (LEE), were confirmed by Q-PCR.
RESULTS: Analysis of variance (R/MAANOVA, Fs test) identified 442 (36%) of 1239 O157-specific ORFs and 2110 (59%) of 3647 backbone ORFs that changed in expression significantly over time. QT cluster analysis placed 2468 of the 2552 significant ORFs into 12 groups; each group representing a distinct expression pattern. ORFs from the largest cluster (n = 1078) decreased in expression from late exponential to early stationary phase: most of these ORFs are involved in functions associated with steady state growth. Also represented in this cluster are ORFs of the TAI island, encoding tellurite resistance and urease activity, which decreased approximately 4-fold. Most ORFs of the LEE pathogenicity island also decreased approximately 2-fold by early stationary phase. The ORFs encoding proteins secreted via the LEE encoded type III secretion system, such as tccP and espJ, also decreased in expression from exponential to stationary phase. Three of the clusters (n = 154) comprised genes that are transiently upregulated at the transition into stationary phase and included genes involved in nutrient scavenging. Upregulated genes with an increase in mRNA levels from late exponential to early stationary phase belonged to one cluster (n = 923) which includes genes involved in stress responses (e.g. gadAB, osmBC, and dps). These transcript levels remained relatively high for > 3 h in stationary phase. The Shiga toxin genes (stx1AB and stx2B) were significantly induced after transition into stationary phase.
CONCLUSION: Expression of more than 300 O157-specific ORFs, many implicated in virulence of the O157 pathogen, was modulated in a growth dependent manner. These results provide a baseline transcriptional profile that can be compared to patterns of gene expression of this important foodborne pathogen under adverse environmental conditions.

Background

Enterohemorrhagic Escherichia coli (EHEC), a food and water- borne pathogen of zoonotic origin, are an important cause of acute gastroenteritis in humans. O157:H7 is the predominant serotype of EHEC causing illness in the United States [1]. One of the hallmarks of EHEC pathogenesis is the formation of attaching and effacing lesions on intestinal epithelial cells [2,3], resulting in intimate adherence of the bacterial cell to the intestinal mucosa. The ability to form A/E lesions is encoded by the Locus of Enterocyte Effacement (LEE), a 35 kb pathogenicity island [4]. Another principal virulence characteristic of EHEC is the production of Shiga toxins, two component cytotoxins that inhibits protein synthesis in eukaryotic cells [5]. The genes for production of Shiga toxin (stx1 and stx2) are located in intact or partial genomes of lambda prophages that are inserted into the chromosome [6,7]. Together, the LEE and Shiga toxins are considered to be two of the principal virulence determinants that mediate development of hemorrhagic colitis and the life-threatening hemolytic uremic syndrome [8].

Regulation of these virulence factors is influenced by global regulators as well as regulators specific to the virulence factor. Many of the virulence factors in EHEC are carried on 'foreign' DNA, DNA which is not present in E. coli K-12, and their expression is typically regulated by transcriptional regulators also carried on these mobile elements. EHEC can utilize global regulators that are common to all E. coli as well as O157-specific regulators to control expression of virulence factors [8,9]. For example, the expression of LEE is dependent on growth phase and responds to nutrient downshifts via ppGpp signaling through DksA to Ler (LEE encoded regulator) and PchA [10]. Also, LEE expression can be upregulated by activation of Ler in the presence of bicarbonate ions [11], and by quorum sensing, in which autoinducer 3 activates LEE expression through Ler [12]. Shiga toxin expression is linked to expression of the phage lysis genes [13,14] and can be controlled by the iron responsive regulator Fur [15]. In addition to the known virulence determinants, E. coli O157:H7 Sakai and EDL-933 contain ~1.3 Mb of sequence not found in E. coli K-12; these sequences occur as part of the genetic material of 18 lysogenic phages and 6 phage-like elements that have integrated into the Sakai genome [16] and as 177 O157 strain-specific islands (OI number) in the EDL-933 strain [17]. Recently it was identified that many effector proteins that are secreted via the LEE-encoded type three secretion system are specified by genes distributed throughout the genome on lambdoid phages [18].

In general, E. coli populations are ecologically and metabolically versatile and can adapt to growth under a wide range of conditions, adaptations essential to a bimodal lifestyle either in the primary habitat within animal hosts or in secondary habitats as free living cells in the natural environment [19]. Slowing metabolic activity during stationary phase reflects a survival mechanism in nutrient poor environments in which bacteria undergo a variety of morphological and physiological changes. In stationary phase, E. coli are more resistant to a number of stresses, including pH stress [20,21] and osmotic stress [22], both of which occur during transit through the colonization of the host intestinal tract. Important to the transmission route of E. coli O157:H7 is the ability to persist in adverse environments until entering a new host. Hence studying growth transitions and how they relate to the lifestyle of E. coli O157:H7 is essential to understanding the persistence and spread of this pathogen from the bovine reservoir to foods and to humans.

Microarrays have been utilized to observe changes induced in the transcriptome of E. coli during growth transitions and in response to different environmental stresses. Global transcription profiling of E. coli K-12 indicates a coordinated response to growth arrest [23]. The sigma factor RpoS plays a crucial role in transcription of stationary phase and stress response genes [24-26]. Microarray studies have been conducted to understand how E. coli K-12 responds to growth at low pH [27,28], on rich and minimal media [29], during anaerobic growth [30-32], and growth transitions in rich medium [33]. Transcriptional responses of O157:H7 during growth on plasma membranes [34] and upon exposure to norfloxacin [35], have been investigated, and microarrays have been used to identify genes regulated by luxS in E. coli O157:H7 [36], but a full picture of global gene expression during growth transitions, although well-studied in E. coli K-12, has not been investigated in the food borne pathogen E. coli O157:H7.

A key component to understanding the adaptations and ecology of E. coli O157:H7 is to determine how the O157-specific genomic elements respond to one of the critical stages of growth, stationary phase. It has been documented that some virulence factors are regulated by environmental cues that signal the presence of the host environment, but what is less understood is how these virulence factors are expressed in a non-host environment. Here we address the question of how O157:H7 responds to growth transitions by determining the changes in global gene expression patterns in O157:H7 during the transition from exponential to stationary phase in minimal medium, with particular interest in expression patterns of O157-specific genes. The dynamics of the transcriptome of E. coli O157:H7 Sakai was investigated by comparative microarray hybridizations at 10 different time points during exponential and stationary phase and compared to obtain a temporal pattern of gene expression. Gene expression patterns were also verified by Q-PCR of 14 ORFs including those encoding principal virulence factors (e.g. LEE-encoded proteins and Shiga toxins).

Results

Analysis of expression ratios

The results are based on O157:H7 Sakai cultures collected from 10 time points of growth in minimal media with 4 replicates (i.e. 4 independent cultures that were each sampled 10 times) for a total of 40 samples. The resulting array data have been deposited in the NCBI Gene Expression Omnibus, accession GSE7477. The samples collected from the mid-exponential phase 3-hour time point (Fig. 1) were used as a common reference for hybridization and analysis for all subsequent time points in a replicated reference design where all the samples are biologically replicated, including the reference [37]. This design allowed for indirect comparison between all time points via the 3-hour samples [37,38]. We incorporated the dye-swaps among the four biological replicates which can confound dye effects and biological replicate effects. However, if there is significant variation in the rate of dye incorporation from one labeling reaction to another, this would result in large dye-effects compared to biological replicate effect [39], an effect not found in our experiments. In general, the variance estimates for the biological replicate effects were small (median 0.0008), compared to the variance estimates (<0.5%) for the array effects (median 0.16). Most ORFs (4362/5886, 90%) had similar signal intensities for the mid- and late- exponential phase samples, indicating that there were few overall differences in expression between these time points (Fig. 2A and 2B). In contrast, there were many differences in expression found between mid-exponential phase samples and samples from the transition point (Fig. 2C) and early stationary phase (Fig. 2D).

Figure 1

Average growth of . Increase in cell density is measured at OD600 at 10 time periods (hr) of growth into stationary phase. Error bars represent the standard deviation of four culture replicates. Samples were taken from the culture at the time points plotted and RNA was extracted. The 5 growth phases, separated by dotted lines, are defined as exponential growth, late exponential growth, transition to stationary phase, early stationary phase, and stationary phase.

Figure 2

Average signal intensity plots for four of the time points of growth. The normalized signals of 4,886 ORFs for E. coli O157:H7 Sakai were averaged from four hybridizations, representing four biological replicates: A. Average signal intensity for exponential phase (3 h) vs late exponential phase (4 h). B. Average signal intensity for exponential phase (3 h) vs late exponential phase (4.6 h). C. Average signal intensity for exponential phase (3 h) vs transition to stationary phase (5 h). D. Average signal intensity for exponential phase (3 h) vs. stationary phase (6 h. The Lowess line is plotted for each graph in red.

Summary of significant changes in gene expression over time

Significant changes in expression through time were determined using ANOVA and the Fs-test [40] for the 4,886 O157:H7 Sakai ORFs targeted on the microarray. ANOVA does not capture how time affects gene expression, but how expression changes over time. The p values were adjusted to correct for type I error using the Benjamini-Hochberg (B-H) false discovery rate approach with the linear step-up correction implemented in R/MAANOVA. The B-H correction for multiple testing applies a step-wise correction to the p values, and allows selection of a suitable p value cutoff based on practical considerations. The 2,552 ORFs with the lowest adjusted p values (< 0.0000001) were considered for further analysis (Additional File 1). Fold change in expression between time points was determined by calculating the difference in log2 expression between the time points of interest (Additional Files 1 and 2).

Trends in gene expression identified by significant contrasts between subsequent time points and QT clustering

Expression profiles of the 2,552 ORFs with significant differences were clustered into 12 groups using Quality Threshold (QT) clustering (Fig. 3), which groups similar expression profiles based on jackknife correlations [41]. The majority of ORFs were classified into 12 groups; only 84 were not placed into a cluster. ORFs were also classified based on the time intervals at which a significant change in gene expression occurred. A subset of 524 ORFs had differences in expression from mid- to late-exponential phase (3 h to 4 h). The majority of ORFs, 1,668 of 4,886 (34%), had significantly changed in transcript level at the stationary phase transition point (4.6 to 5 h). Additionally, 1,045 of 4,886 (20%) ORFs significantly changed during the transition point to early stationary phase (5 to 5.5 h).

Figure 3

Quality Threshold (QT) clusters of the 2,552 ORFs with significant changes in gene expression. The number of ORFs in each cluster is listed at the top of each plot. Average expression profiles were determined for the ORFs in each cluster and plotted with the standard deviation for each time point.

Significant changes in exponential phase – response to limiting oxygen

There were 524 ORFs with significant differences in transcript level during exponential phase – a total of 113 genes changed only in exponential phase, whereas the remainder had altered levels of expression in exponential as well as stationary phase. ORFs that decreased in expression included those encoding enzymes of the TCA cycle, such as sucA, and sdhABCD (Table 1, Fig. 3; QTC 10). Many of the ORFs (94/524, 18%) that changed in transcript level are regulated by the transcription factor FNR [31,32], whereas a smaller set of ORFs (29/524, 5%) are known to be regulated by ArcA [42]. Genes that increased included those encoding anaerobic electron acceptors, fumarate reductase, frdAB, and DMSO reductase, dmsA and dmsC, as well as a DMSO reductase paralog, ynfE (Table 1, Fig. 3; QTC 6). Genes encoding the cytochrome c biogenesis system (ccmABCDEFG) and the cytochrome c-like protein (napB and napC) also had a significant increase from 3 h to 4 h. The operon encoding the low affinity cytochrome oxidase bo (cyoABCD) decreased 9 to15- fold in expression from 3 h to 4 h (Table 1). Concomitantly, expression of the high affinity cytochrome oxidase bd, cydAB, increased 4–5 fold. Expression of the high and low affinity cytochrome oxidases is directly influenced by the level of available oxygen in the media [43,44]. Gene set enrichment analysis (GSEA) found that genes involved in anaerobic energy metabolism were significantly enriched in the 4 h sample compared to the 3 h sample, and that genes involved in aerobic energy metabolism and the TCA cycle were significantly enriched in the 3 h compared to the 4 h culture (Additional File 3). The altered expression of the cytochrome oxidases as well as the increase in FNR-controlled, anaerobic associated genes and decrease in TCA cycle enzymes is consistent with the hypothesis that oxygen became limiting under the culture conditions that were used. Subsequent measurement of the residual dissolved oxygen tension in the culture during growth indicates that residual dissolved O2 levels decreased almost to zero (95% decrease in O2) during exponential phase (Fig. 4). As cultures entered stationary phase, the O2 levels began to increase, as the demand for O2 by the culture decreased.

Table 1

Significant ORFs (p value < 1 × 107) with greater than 4-fold significant change between 3 h and 4 h.

Ecs number	Genea	function	log2 change in expression	QT clusterb
ECs0236	O157	Unknown function	-2.89	10
ECs0482	cyoE	protoheme IX farnesyltransferase	-2.08	10
ECs0483	cyoD	cytochrome o ubiquinol oxidase subunit IV	-3.18	10
ECs0484	cyoC	cytochrome o ubiquinol oxidase subunit III	-3.32	10
ECs0485	cyoB	cytochrome o ubiquinol oxidase subunit I	-3.87	10
ECs0486	cyoA	cytochrome o ubiquinol oxidase subunit II	-3.89	10
ECs0660	dcuC	transport of dicarboxylates	2.20	6
ECs0746	sdhC	succinate dehydrogenase, cytochrome b556	-3.15	10
ECs0747	sdhD	succinate dehydrogenase, hydrophobic subunit	-2.66	10
ECs0748	sdhA	succinate dehydrogenase	-3.06	10
ECs0749	sdhB	succinate dehydrogenase	-2.79	10
ECs0750	-	orf, hypothetical protein	-2.24	10
ECs0751	sucA	2-oxoglutarate dehydrogenase	-2.20	10
ECs0768	cydA	cytochrome d terminal oxidase	2.56	4
ECs0769	cydB	cytochrome d terminal oxidase	2.58	4
ECs0916	yliH	putative receptor	3.84	2
ECs0979	dmsA	anaerobic dimethyl sulfoxide reductase	3.71	4
ECs0981	dmsC	anaerobic dimethyl sulfoxide reductase	2.08	4
ECs0986	pflB	formate acetyltransferase 1	2.04	6
ECs1572	pepT	putative peptidase T	2.64	2
ECs1728	narK	nitrite extrusion protein	2.98	6
ECs1729	narG	nitrate reductase 1, alpha subunit	2.25	4
ECs1741	adhE	CoA-linked acetaldehyde dehydrogenase	2.50	6
ECs1756	yciD	putative outer membrane protein	4.15	4
ECs1875	ycjL	probable amidotransferase subunit	-2.28	2
ECs2027	ydcI	putative transcriptional regulator LYSR-type	-2.41	2
ECs2078	fdnG	formate dehydrogenase-N, nitrate-inducible	2.30	6
ECs2150	ydfZ	orf, hypothetical protein	4.70	4
ECs2293	ynfE	putative oxidoreductase, major subunit	2.83	6
ECs2457	ydjY	orf, hypothetical protein	2.03	6
ECs2463	ynjE	putative thiosulfate sulfur transferase	3.47	1
ECs2614	yecH	orf, hypothetical protein	3.09	6
ECs3061	fruB	PTS system, fructose-specific IIA/fpr component	2.02	1
ECs3084	dsbE	disulfide oxidoreductase	2.04	1
ECs3088	ccmC	heme exporter protein C	2.19	1
ECs3090	ccmA	ATP binding protein of heme exporter A	2.64	1
ECs3091	napC	cytochrome c-type protein	3.55	1
ECs3092	napB	cytochrome c-type protein	3.30	1
ECs3093	napH	ferredoxin-type protein: electron transfer	3.04	un
ECs3180	ackA	acetate kinase	2.28	6
ECs3226	yfcZ	orf, hypothetical protein	3.02	6
ECs3445	yfiD	putative formate acetyltransferase	4.48	6
ECs3460	yfiA	putative yhbH sigma 54 modulator	2.12	2
ECs3582	hypA	pleiotrophic effects on 3 hydrogenase isozymes	2.92	6
ECs3583	hypB	guanine-nucleotide binding protein	2.10	6
ECs3586	hypE	plays structural role in maturation of all 3 hydrogenases	2.10	6
ECs3799	O157	orf; hypothetical protein	2.92	6
ECs3800	O157	orf; hypothetical protein	2.33	6
ECs3802	O157	putative ATP-binding protein of ABC transport system	2.40	6
ECs3833	ansB	periplasmic L-asparaginase II	3.41	4
ECs3880	hybB	probable cytochrome Ni/Fe component of hydrogenase-2	2.05	4
ECs4040	yhbV	orf, hypothetical protein	3.11	1
ECs4216	nirB	nitrite reductase (NAD(P)H) subunit	3.55	1
ECs4343	nikA	periplasmic binding protein for nickel	3.46	1
ECs4344	nikB	transport of nickel, membrane protein	2.10	6
ECs4347	nikE	ATP-binding protein of nickel transport system	2.85	1
ECs4398	yhjA	putative cytochrome C peroxidase	2.86	6
ECs4456	yiaI	orf, hypothetical protein	2.40	4
ECs4750	yigI	orf, hypothetical protein	-2.55	10
ECs4834	sodA	superoxide dismutase, manganese	-2.40	7
ECs4874	gldA	glycerol dehydrogenase, (NAD)	2.86	6
ECs5052	nrfA	periplasmic cytochrome c(552)	3.08	un
ECs5053	nrfB	formate-dependent nitrite reductase	2.86	6
ECs5054	nrfC	formate-dependent nitrite reductase; Fe-S centers	2.96	un
ECs5134	frdB	fumarate reductase, anaerobic, iron-sulfur protein subunit	2.50	8
ECs5135	frdA	fumarate reductase, anaerobic, flavoprotein subunit	2.11	8
ECs5214	nrdG	anaerobic ribonucleotide reductase activating protein	2.02	6
ECs5215	nrdD	anaerobic ribonucleoside-triphosphate reductase	3.54	6
ECs5298	yjiM	orf, hypothetical protein	2.76	8

aGenes in bold have been identified as being under the control of the transcriptional regulator FNR [29, 30]. Underlined genes have been identified as being controlled by the transcriptional regulator ArcA [40].

bQT cluster un. means unassigned

Figure 4

Residual dissolved oxygen (O. The mean rDOT was determined for three independent cultures of E. coli O157:H7 Sakai in glucose limited MOPS minimal medium and plotted with the error bars indicating the standard deviation.

Significant change in O157-specific genes during exponential phase

In contrast to the backbone genes, only a small proportion (6%) of the O157-specific ORFs changed significantly in expression measured during exponential phase. While statistically significant, most of the differences in expression level for these 61 genes were not substantial, ranging from 1.2- to 2- fold (Additional File 2). Many of these ORFs are of unknown function, but of the 45 O157-specific ORFs that increased from 3–4 h, 9 of them were from a single Sakai phage (Sp) 15 region which encodes the Shiga toxin 1 phage, and contains 5 ORFs from Sp 5 which encodes the Shiga toxin 2 phage. Interestingly, ECs2381 a locus whose expression has been shown to contribute to colonization in the bovine gastrointestinal tract [45], was also significantly induced 1.8 fold from 3 to 4 h. There were 4 additional O157-specific ORFs that had a greater than 4-fold change in expression from 3–4 h (Table 1). These 4 ORFs were not associated with any of the Sakai phages, and have unknown or putative functions.

Genes with significant changes in expression at the stationary phase transition point

The majority of significant changes in transcript level occurred at the stationary phase transition point. There were 176 ORFs with a 4-fold or greater change in expression at the stationary phase transition point (Table 2). A total of 363 ORFs that significantly increased or decreased during the transition to stationary phase (4.6–5 h) also increased or decreased in early stationary phase (5–5.5 h) (Table 3). A reduced level of expression of ribosomal genes and genes involved in nucleotide and amino acid synthesis was observed from 4.6–5 h and 5–5.5 h (Tables 2 and 3, Fig. 3; QTC 1). Many of the genes that were significantly down-regulated upon entry into stationary phase have been also observed in E. coli K-12 [23], including decreased expression of the molecular chaperones encoded by groEL, htpG, and dnaK (Tables 2 and 3, Fig. 3; QTC 1, Additional File 1).

Table 2

Significant ORFs (p value < 1 × 107) with greater than 4-fold significant changes in expression at the stationary phase transition point (4.6 to 5 h).

Ecs no.	Gene	log2 change in expression	QT cluster	Ecs no.	Gene	log2 change in expression	QT cluster
ECs0034	dapB	-2.63	1	ECs1286	O157	-2.09	1
ECs0035	carA	-4.23	1	ECs1426	mdoG	-2.46	5
ECs0068	araC	2.57	2	ECs1440	pyrC	-2.65	1
ECs0072	tbpA	-2.25	1	ECs1468	plsX	-2.16	1
ECs0083	fruL	-2.06	1	ECs1479	ptsG	-3.09	1
ECs0124	speD	-2.74	1	ECs1492	mfd	-2.33	1
ECs0149	dksA	2.12	2	ECs1604	ycfC	-2.19	1
ECs0150	sfsA	2.83	2	ECs1662	O157	-2.53	1
ECs0191	yaeO	2.42	2	ECs1684	dadA	2.34	2
ECs0207	dniR	-2.11	1	ECs1710	ychH	4.61	2
ECs0334	O157	-2.07	1	ECs1712	prsA	-2.68	1
ECs0374	O157	2.42	2	ECs1859	rnb	-2.70	1
ECs0401	mhpR	4.33	3	ECs1954	O157	-2.61	1
ECs0416	O157	3.55	2	ECs1955	O157	-3.22	1
ECs0468	ribH	-2.27	1	ECs2011	ynbB	2.34	2
ECs0483	cyoD	2.06	10	ECs2021	aldA	3.60	2
ECs0485	cyoB	2.01	10	ECs2027	ydcI	3.35	2
ECs0489	bolA	2.10	2	ECs2031	ydcH	2.99	2
ECs0584	purK	-2.27	1	ECs2037	ydcN	2.71	2
ECs0585	purE	-2.66	1	ECs2083	sfcA	-2.18	1
ECs0610	cusC	-2.31	1	ECs2182	O157	3.21	2
ECs0611	cusF	-2.82	1	ECs2293	ynfE	-2.97	6
ECs0612	cusD	-2.62	1	ECs2296	ynfH	-2.03	6
ECs0613	ybdE	-2.70	1	ECs2362	lhr	-2.16	1
ECs0670	dacA	-2.32	1	ECs2367	purR	-2.51	1
ECs0681	ybeL	2.78	2	ECs2431	yniA	3.52	2
ECs0773	tolR	-2.29	1	ECs2433	yniC	2.44	3
ECs0782	aroG	-3.09	1	ECs2445	osmE	2.33	2
ECs0800	ybhC	-2.13	1	ECs2510	yeaV	2.06	3
ECs0814	O157	-2.15	1	ECs2692	yodD	2.96	2
ECs0869	ybhQ	3.14	2	ECs2714	espJ	-2.01	1
ECs0890	dps	2.64	2	ECs2737	pchC	3.43	2
ECs0943	artJ	-2.81	1	ECs2814	yeeD	-2.09	1
ECs0966	cspD	2.75	2	ECs2816	yeeF	-2.55	1
ECs0968	clpA	2.89	3	ECs2839	O157	-2.23	1
ECs0986	pflB	-2.61	6	ECs2840	wbdP	-2.14	1
ECs1008	ycbB	2.15	2	ECs2847	O157	-2.25	1
ECs1029	pyrD	-2.88	1	ECs3038	yeiT	2.14	3
ECs1037	rmf	2.24	2	ECs3060	fruK	-2.13	1
ECs1072	O157	-2.13	5	ECs3061	fruB	-2.38	1
ECs1091	O157	3.24	2	ECs3136	yfaX	-2.04	1
ECs1137	yccC	-2.27	1	ECs3149	menC	-2.06	1
ECs1138	yccY	-2.78	1	ECs3180	ackA	-2.33	6
ECs1139	yccZ	-2.74	1	ECs3194	argT	4.08	2
ECs3196	purF	-2.94	1	ECs4188	hopD	2.38	2
ECs3197	cvpA	-3.35	1	ECs4192	rpsG	-2.57	1
ECs3212	mepA	-2.29	1	ECs4193	rpsL	-2.67	1
ECs3287	ptsH	-2.66	1	ECs4204	yheT	-2.00	1
ECs3296	cysP	-2.17	1	ECs4249	yhgF	-2.53	1
ECs3361	purM	-4.24	1	ECs4294	yhhA	2.48	2
ECs3387	sseA	2.18	2	ECs4343	nikA	-3.22	1
ECs3403	hcaR	2.50	3	ECs4347	nikE	-2.38	1
ECs3423	purL	-2.45	1	ECs4366	yhiO	4.26	2
ECs3448	trxC	2.24	2	ECs4481	lldP	2.39	2
ECs3463	tyrA	3.67	3	ECs4490	gpmI	-3.35	1
ECs3464	aroF	3.30	3	ECs4512	rpmB	-2.38	1
ECs3471	yfjA	-3.58	1	ECs4553	O157	-2.10	1
ECs3546	emrR	-3.47	1	ECs4575	escC	-2.00	1
ECs3559	srlA_1	2.08	3	ECs4632	yidA	-2.02	1
ECs3595	rpoS	2.80	2	ECs4638	rpmH	-2.47	1
ECs3606	cysD	-2.59	1	ECs4639	rnpA	-2.92	1
ECs3640	pyrG	-2.98	1	ECs4679	atpE	-2.33	1
ECs3659	fucO	2.17	3	ECs4716	rho	-2.56	5
ECs3675	argA	-4.09	1	ECs4720	wecC	-2.34	1
ECs3677	recB	-2.17	1	ECs4741	xerC	-2.08	1
ECs3701	yqeF	2.50	3	ECs4756	yigL	-2.18	1
ECs3743	ygeW	2.08	2	ECs4759	metE	-2.76	1
ECs3749	yqeC	2.45	2	ECs4791	glnL	-2.80	1
ECs3799	O157	-2.69	6	ECs4870	metF	-3.65	1
ECs3800	O157	-2.43	6	ECs4885	ppc	-3.62	1
ECs3802	O157	-2.43	6	ECs4887	argC	-3.99	1
ECs3811	yggG	2.28	2	ECs4888	argB	-3.62	1
ECs3818	metK	-3.03	1	ECs4893	O157	2.09	2
ECs3840	nupG	2.44	2	ECs4907	rplA	-3.02	1
ECs3931	glgS	2.94	2	ECs4908	rplJ	-2.64	1
ECs4052	argG	-2.68	1	ECs4909	rplL	-2.60	1
ECs4103	rpsI	-2.28	1	ECs4929	purH	-4.14	1
ECs4132	yhdG	-2.54	1	ECs4931	metA	-3.78	1
ECs4150	smg	2.24	2	ECs4932	aceB	5.35	2
ECs4156	mscL	2.01	2	ECs5004	O157	2.14	2
ECs4161	rpsD	-2.17	1	ECs5046	yjcD	-3.81	1
ECs4162	rpsK	-2.36	1	ECs5051	acs	5.55	2
ECs4164	rpmJ	-2.03	1	ECs5067	O157	2.21	2
ECs4165	prlA	-2.53	1	ECs5153	purA	-2.06	1
ECs5164	yjfN	4.54	2	ECs5240	yjgR	2.24	2
ECs5192	cysQ	2.01	2	ECs5259	O157	-2.09	1
ECs5231	argI	-3.29	1	ECs5354	rob	2.15	2
ECs5235	valS	-2.58	1	ECs5360	yjjY	2.03	2

Table 3

Significant ORFs (p value < 1 × 107) with greater than 4-fold significant changes in expression from late exponential to early stationary phase (4.6 to 5.5 h)

Ecs no.	Genea	Log2 change in expression	QT cluster	Ecs no.	Genea	Log2 change in expression	QT cluster
ECs0009	mogA	2.27	2	ECs0526	htpG	-2.70	1
ECs0015	dnaJ	-2.32	1	ECs0527	adk	-3.22	1
ECs0026	rpsT	-4.14	1	ECs0540	O157	3.52	2
ECs0028	ribF	-2.36	1	ECs0541	O157	4.11	2
ECs0029	ileS	-2.81	1	ECs0552	ybbK	3.00	2
ECs0030	lspA	-2.38	1	ECs0561	ybbD	2.16	2
ECs0032	lytB	-2.20	1	ECs0615	ybdG	-2.36	1
ECs0036	carB	-3.28	1	ECs0619	ybdK	2.98	2
ECs0052	ccdA	2.41	2	ECs0645	ahpF	-2.11	1
ECs0078	leuA	2.55	un	ECs0646	ybdQ	3.35	3
ECs0108	guaC	-2.45	1	ECs0656	citD	2.79	2
ECs0118	aceE	-3.27	1	ECs0864	ybhL	2.61	2
ECs0119	aceF	-2.55	1	ECs0866	ybhN	2.27	2
ECs0123	yacL	3.17	2	ECs0881	ybiI	3.35	2
ECs0128	gcd	2.69	2	ECs0896	ybiS	-2.07	1
ECs0168	dapD	-3.16	1	ECs0915	yliG	-2.77	1
ECs0172	tsf	-4.58	1	ECs0919	dacC	2.40	2
ECs0173	pyrH	-3.15	1	ECs0934	potF	2.12	2
ECs0182	fabZ	-2.20	1	ECs0944	artM	-2.40	1
ECs0199	yaeC	-3.14	1	ECs0963	ybjX	-2.08	1
ECs0200	yaeE	-2.63	1	ECs0974	lrp	-3.19	1
ECs0201	abc	-3.32	1	ECs0979	dmsA	-3.37	4
ECs0216	O157	2.02	2	ECs0989	ycaP	2.56	2
ECs0248	fadE	4.57	2	ECs0990	serC	-3.29	1
ECs0253	dinJ	2.46	2	ECs0991	aroA	-4.24	1
ECs0269	proB	-2.11	1	ECs1007	mukB	-2.12	1
ECs0383	yahO	4.53	2	ECs1013	asnS	-2.74	1
ECs0384	prpR	3.63	2	ECs1014	pncB	-2.37	1
ECs0390	codA	-2.78	1	ECs1022	ycbR	-3.16	1
ECs0415	afuA	5.21	2	ECs1048	-	2.44	2
ECs0434	psiF	3.83	2	ECs1050	yccV	3.88	2
ECs0458	yajC	-2.33	1	ECs1158	agp	3.47	2
ECs0459	secD	-2.40	1	ECs1159	yccJ	2.12	2
ECs0490	tig	-2.96	1	ECs1211	O157	-3.29	1
ECs0497	ybaW	2.93	2	ECs1266	phoH	7.69	2
ECs0502	mdlA	-3.80	1	ECs1351	terZ	-2.58	1
ECs0503	mdlB	2.30	2	ECs1355	terD	-2.53	1
ECs0504	glnK	-3.73	1	ECs1356	terE	-2.90	1
ECs0507	ybaY	3.19	2	ECs1423	-	2.34	2
ECs0509	ybaA	2.56	2	ECs1438	yceP	4.78	2
ECs0518	aefA	2.12	2	ECs1466	yceD	-2.86	1
ECs1467	rpmF	-2.36	1	ECs2307	ydgG	-2.69	1
ECs1500	potC	-2.14	1	ECs2308	pntB	-2.80	1
ECs1603	purB	-3.58	1	ECs2314	rstA	-2.63	1
ECs1605	ycfB	-2.43	1	ECs2319	manA	-2.47	1
ECs1663	ompT	-2.72	1	ECs2346	tyrS	-2.97	1
ECs1683	ycgB	5.36	2	ECs2385	ynhG	3.80	2
ECs1692	ymgE	2.35	2	ECs2387	sufS	2.88	2
ECs1705	-	2.55	2	ECs2389	sufC	3.58	2
ECs1722	chaB	2.85	2	ECs2390	sufB	2.93	2
ECs1729	narG	-2.70	4	ECs2391	sufA	3.68	2
ECs1743	oppA	-2.75	1	ECs2397	-	2.44	2
ECs1747	oppF	-2.11	1	ECs2423	rplT	-3.11	1
ECs1831	yciG	2.91	2	ECs2430	ydiZ	2.09	2
ECs1848	-	2.73	2	ECs2450	ydjS	3.94	2
ECs1849	acnA	4.07	2	ECs2451	astB	4.11	2
ECs1858	yciR	3.14	2	ECs2452	astD	5.88	2
ECs1874	ycjK	4.73	2	ECs2453	astA	4.62	2
ECs1875	ycjL	5.14	2	ECs2454	astC	6.97	2
ECs1914	ydaA	2.88	2	ECs2460	-	-2.03	1
ECs1997	ynaF	3.42	2	ECs2463	ynjE	-4.04	1
ECs2028	ydcJ	3.64	2	ECs2467	gdhA	-5.15	1
ECs2036	ydcL	3.78	2	ECs2491	yeaF	-2.61	1
ECs2044	ydcS	5.22	2	ECs2492	yeaG	4.08	2
ECs2045	ydcT	5.23	2	ECs2511	yeaW	2.43	2
ECs2054	ydcU	2.59	2	ECs2514	fadD	2.50	2
ECs2082	adhP	3.62	2	ECs2519	yoaC	2.05	2
ECs2084	rpsV	2.45	2	ECs2527	manX	-2.95	1
ECs2086	osmC	2.01	2	ECs2528	manY	-3.21	1
ECs2091	-	3.38	2	ECs2546	ydjX	4.09	2
ECs2092	-	3.70	2	ECs2560	eda	-2.65	1
ECs2113	O157	-3.24	1	ECs2574	yebC	-2.87	1
ECs2118	ydeV	3.60	2	ECs2575	ntpA	-2.15	1
ECs2119	ydeW	2.54	2	ECs2586	argS	-2.77	1
ECs2120	-	4.17	2	ECs2609	araF	3.61	2
ECs2126	tam	2.83	2	ECs2613	ftn	-3.88	1
ECs2145	ydeI	3.74	2	ECs2614	yecH	-2.94	6
ECs2146	ydeJ	2.05	2	ECs2668	yedE	-2.15	1
ECs2150	ydfZ	-5.10	4	ECs2669	yedF	-2.78	1
ECs2281	O157	2.30	2	ECs2670	yedK	2.55	2
ECs2292	-	3.22	2	ECs2693	-	2.62	2
ECs2295	ynfG	-2.81	6	ECs2785	erfK	2.37	2
ECs2820	hisG	-2.80	1	ECs3458	yfiO	-2.61	1
ECs2821	hisD	-2.57	1	ECs3465	yfiL	3.81	2
ECs2822	hisC	-2.52	1	ECs3469	rplS	-4.28	1
ECs2824	hisH	-2.22	1	ECs3470	trmD	-5.46	1
ECs2837	wbdQ	-2.35	1	ECs3472	rpsP	-5.57	1
ECs2888	yegP	4.47	2	ECs3520	ygaT	4.88	2
ECs2892	yegS	3.38	2	ECs3522	gabD	2.53	2
ECs2900	fbaB	2.45	2	ECs3523	gabT	3.14	2
ECs2920	metG	-2.59	1	ECs3533	ygaM	3.32	2
ECs3022	yohC	6.12	2	ECs3554	alaS	-2.20	1
ECs3029	O157	4.04	2	ECs3605	cysN	-3.19	1
ECs3030	O157	3.38	2	ECs3617	cysH	-2.76	1
ECs3053	yeiJ	-2.67	1	ECs3619	cysJ	-4.11	1
ECs3055	yeiL	2.38	2	ECs3655	ygdH	2.14	2
ECs3073	yejG	2.91	2	ECs3669	-	2.28	2
ECs3077	rplY	-3.64	1	ECs3742	ygeV	3.64	2
ECs3078	yejK	-2.55	1	ECs3750	ygfJ	4.14	2
ECs3102	ada	2.24	2	ECs3761	-	2.76	2
ECs3114	gyrA	-3.38	1	ECs3762	lysS	-2.92	1
ECs3181	pta	-3.50	1	ECs3771	-	3.52	2
ECs3224	fadJ	4.26	2	ECs3772	yqfB	-2.02	1
ECs3225	fadI	4.50	2	ECs3784	serA	-3.66	1
ECs3227	fadL	5.39	2	ECs3810	tktA	-4.16	1
ECs3230	yfdC	2.55	2	ECs3815	yqgB	-2.40	1
ECs3259	yfdZ	3.57	2	ECs3833	ansB	-2.31	4
ECs3271	-	3.06	2	ECs3896	yqhE	2.86	2
ECs3272	nupC	-2.08	1	ECs3904	ygiS	-2.33	1
ECs3278	gltX	-2.47	1	ECs3948	rpsU	-3.14	1
ECs3286	cysK	-3.68	1	ECs3955	ygjG	5.55	2
ECs3293	cysA	-2.71	1	ECs3963	ygjL	4.06	2
ECs3302	yfeX	-2.24	1	ECs3979	yqjC	2.12	2
ECs3322	-	2.10	2	ECs3980	yqjD	2.28	2
ECs3323	-	2.15	2	ECs4033	yraR	2.71	2
ECs3338	purC	-5.04	1	ECs4034	yhbO	3.67	2
ECs3339	nlpB	-2.13	1	ECs4040	yhbV	-3.40	1
ECs3369	guaA	-2.99	1	ECs4064	rpmA	-3.29	1
ECs3397	-	2.55	2	ECs4068	murA	-3.01	1
ECs3401	csiE	3.77	2	ECs4069	yrbA	-2.73	1
ECs3409	yphA	3.07	2	ECs4091	gltB	-3.19	1
ECs3434	lepB	-2.08	1	ECs4092	gltD	-3.80	1
ECs3445	yfiD	-6.36	6	ECs4104	rplM	-4.05	1
ECs4112	yhcO	3.56	2	ECs4404	yhjG	2.50	2
ECs4127	accB	-2.57	1	ECs4427	O157	2.20	2
ECs4128	accC	-3.51	1	ECs4433	yhjY	3.88	2
ECs4133	fis	-3.76	1	ECs4438	yiaE	-2.49	1
ECs4159	rplQ	-3.99	1	ECs4440	yiaG	2.73	2
ECs4160	rpoA	-4.11	1	ECs4449	xylF	2.34	2
ECs4163	rpsM	-3.05	1	ECs4463	yiaW	2.75	2
ECs4166	rplO	-3.63	1	ECs4464	aldB	3.02	2
ECs4167	rpmD	-3.70	1	ECs4473	yibH	3.78	2
ECs4168	rpsE	-3.75	1	ECs4475	mtlA	2.84	2
ECs4169	rplR	-3.75	1	ECs4511	rpmG	-3.21	1
ECs4170	rplF	-3.62	1	ECs4514	dfp	-2.28	1
ECs4171	rpsH	-4.00	1	ECs4554	espB	-2.41	1
ECs4172	rpsN	-3.13	1	ECs4555	espD	-2.60	1
ECs4173	rplE	-4.26	1	ECs4611	ilvN	4.84	2
ECs4174	rplX	-4.14	1	ECs4612	ilvB	4.41	2
ECs4175	rplN	-4.46	1	ECs4633	yidB	3.85	2
ECs4176	rpsQ	-4.30	1	ECs4672	glmU	-2.20	1
ECs4177	rpmC	-4.12	1	ECs4674	atpD	-2.97	1
ECs4178	rplP	-3.60	1	ECs4675	atpG	-3.77	1
ECs4179	rpsC	-5.07	1	ECs4676	atpA	-4.02	1
ECs4180	rplV	-4.74	1	ECs4677	atpH	-4.50	1
ECs4181	rpsS	-4.14	1	ECs4678	atpF	-4.08	1
ECs4182	rplB	-4.83	1	ECs4680	atpB	-3.27	1
ECs4183	rplW	-5.75	1	ECs4684	mioC	-2.10	1
ECs4184	rplD	-6.50	1	ECs4686	asnA	-3.68	1
ECs4185	rplC	-4.69	1	ECs4708	ilvC	-3.31	1
ECs4186	rpsJ	-5.93	1	ECs4757	yigM	-2.00	1
ECs4189	-	3.24	2	ECs4758	metR	-3.26	1
ECs4191	fusA	-3.27	1	ECs4760	-	3.29	3
ECs4210	argD	-2.36	1	ECs4774	fadB	6.58	2
ECs4212	fic	4.44	2	ECs4783	dsbA	-2.15	1
ECs4213	yhfG	4.13	2	ECs4785	yihG	2.17	2
ECs4232	aroK	-3.21	1	ECs4792	glnA	-5.34	1
ECs4244	yhgE	-3.33	1	ECs4793	yihK	-2.43	1
ECs4278	asd	-2.55	1	ECs4847	yiiS	2.66	2
ECs4299	ugpB	6.46	2	ECs4869	metL	-2.30	1
ECs4348	yhhG	-2.67	1	ECs4884	yijP	-3.62	1
ECs4398	yhjA	-2.20	6	ECs4889	argH	-4.04	1
ECs4399	treF	3.03	2	ECs4898	murI	-2.51	1
ECs4402	yhjD	2.51	2	ECs4906	rplK	-4.80	1
ECs4910	rpoB	-2.90	1	ECs5123	groES	-3.66	1
ECs4911	rpoC	-2.86	1	ECs5130	blc	3.64	2
ECs4914	thiG	-2.22	1	ECs5176	rpsF	-4.84	1
ECs4915	thiF	-2.80	1	ECs5177	priB	-5.00	1
ECs4916	thiE	-2.34	1	ECs5178	rpsR	-4.46	1
ECs4917	thiC	-2.35	1	ECs5179	rplI	-3.90	1
ECs4928	purD	-4.58	1	ECs5185	fklB	-2.32	1
ECs4933	aceA	4.98	2	ECs5194	ytfJ	4.59	2
ECs4934	aceK	2.90	2	ECs5205	ytfQ	6.29	2
ECs4990	O157	2.13	2	ECs5206	-	2.57	3
ECs4991	O157	2.12	2	ECs5207	ytfT	3.32	2
ECs5007	lysC	-3.43	1	ECs5208	yjfF	2.59	2
ECs5038	pfkA	2.14	2	ECs5211	yjgA	2.20	2
ECs5039	yjbR	2.21	2	ECs5213	cybC	2.20	2
ECs5049	actP	4.17	2	ECs5222	pyrB	-2.95	1
ECs5050	yjcH	3.78	2	ECs5253	O157	2.70	2
ECs5089	phnB	4.22	2	ECs5332	yjjG	3.09	2
ECs5109	yjdJ	3.07	2

aGenes with "O157" indicate that these are O157-specific ORFs.

Expression of rpoS, encoding the stationary phase sigma factor, increased 7 fold during the transition point and the RpoS activator, dksA, increased 4.3 fold at the same interval (Table 2). Increased expression of genes known to be regulated by RpoS, such as bolA, dps, osmE, osmC, adhP and ugpB occurred at the stationary phase transition point (Tables 2 and 3, Fig. 3; QTC 2). Gene set enrichment analysis also indicated that these genes (members of the adaptation to atypical conditions group) were significantly enriched in the stationary phase time points compared to exponential phase (Additional File 4). ORFs encoding toxin-antitoxin systems, such as dinJ-yafQ (Table 3, Fig. 3; QTC 2) and chpB-chpS (Fig. 3; QTC 3), were upregulated at the stationary phase transition point. The greatest increase in transcriptional level for an upregulated gene (45-fold) during the stationary phase transition (4.6 to 5 h interval) was acs, encoding acetyl CoA synthetase (Table 2). Acs expression is controlled in part by RpoS [46]. Also induced at a very high level during the transition point into stationary phase were aceB (40 fold) and aceA (32 fold), which encode components of the glyoxalate shunt. Increased expression of acetyl CoA synthetase as well as genes of the glyoxalate shunt is consistent with the hypothesis that cells are scavenging acetate produced during rapid growth on glucose in exponential phase.

There are a variety of O157-specific genes that were downregulated in the transition to stationary phase include those encoding the O157 LPS antigen (ECs2839, ECs2840, and ECs2847) with a greater than 4 fold decrease in expression (Table 3, Fig. 3; QTC 1) as well as ECs2835, ECs2836, ECs2838, ECs2841, ECs2844, and ECs2845 (1.6 to 3 fold decrease) from that genomic region. ECs2113, part of the F9 fimbrial operon [47], decreased in expression from exponential to stationary phase (Table 3). Other ORFs in this operon, ECs2107-2112, also decreased significantly (~1.5 fold) upon entry into stationary phase. O157-specific ORFs that increased in stationary phase included ECs2182, ECs2737, and ECs5067, which encode putative transcriptional regulators. ECs2737 (pchC) encodes a perC-like homolog reported to regulate LEE transcription [48]. The recently identified chromosomal CcdAB toxin-antitoxin system [49], encoded by ECs0052 and ECs0053, increased 5.3-fold from the transition point to early stationary phase (Table 3, Fig. 3; QTC 2). ECs0415, afuA, encoding a periplasmic ferric iron binding protein, had the greatest increase in expression of the O157-specific ORFs, with a 37-fold increase during the transition point (Table 3, Fig. 3; QTC 2).

Genes with transient expression at the stationary phase transition point

A subset of 120 ORFs were expressed transiently from late exponential to early stationary phase. These ORFs changed significantly during the 4.6 to 5 h time interval, and then again during the 5 to 5.5 h interval, but in the opposite direction (Fig. 3; QTC 3, 11, and 12). Genes with this pattern of transcription included many involved in nutrient scavenging and turnover. Genes encoding both the anaerobic (glpABC) and aerobic (glpD) sn-glycerol 3 phosphate dehydrogenase increased >4 fold during the transition into stationary phase, and then decreased significantly in early stationary phase (Table 4). This pattern is consistent with the general model that the glp metabolic system serves as a salvage pathway for glycerol derived from degradation of phospholipids and triacylglycerol. The murPQ operon was expressed transiently, with a 5.3 fold increase followed by a 2 fold decrease in early stationary phase. The products of murPQ are involved in peptidoglycan turnover [50] suggesting that membrane components are being recycled and utilized for energy. A dicarboxylate transporter, dctA, was induced 24-fold upon entry into stationary phase, then decreased 4-fold in early stationary phase (Table 4). Genes encoding transport and binding proteins for carbohydrates, sugar alcohols and acids were significantly enriched at the stationary phase transition point compared to mid-exponential phase, also indicating the release of glucose-dependent catabolite repression (Additional File 4). Genes encoding salvage and scavenging pathways are also induced during glucose-limited growth [51,52] (Additional File 1).

Table 4

Significant ORFs (p value < 1 × 107) with transient expression at the stationary phase transition point with expression changes greater than 4-fold.

Ecs no.	gene	log2 change in expression 4.6–5 h	log2 change in expression 5–5.5 h	QT cluster
ECs0057	pdxA	-2.43	0.66	1
ECs0287	O157	-1.09	2.91	2
ECs0358	betB	3.07	-1.25	3
ECs0359	betI	2.99	-1.48	3
ECs0418	O157	2.19	-0.64	3
ECs0486	cyoA	2.77	-1.08	10
ECs0746	sdhC	2.93	-1.74	10
ECs0747	sdhD	2.55	-1.59	10
ECs0748	sdhA	2.76	-1.38	10
ECs0778	nadA	-2.57	0.76	1
ECs0946	artI	-2.66	0.40	5
ECs0987	focA	-3.38	1.11	6
ECs0994	rpsA	-3.78	0.66	1
ECs1261	putP	2.80	-0.92	10
ECs1741	adhE	-4.01	1.17	6
ECs1744	oppB	-2.05	1.36	7
ECs1772	O157	-0.68	2.22	2
ECs2383	pykF	-2.42	1.49	un
ECs2387	sufS	0.71	2.17	2
ECs2481	ydjH	-2.44	0.84	6
ECs2482	ydjI	-2.18	0.81	6
ECs2486	yeaC	3.00	-0.57	2
ECs2487	yeaA	3.08	-0.60	3
ECs3042	mglB	3.96	-1.13	3
ECs3043	galS	2.14	-0.87	3
ECs3117	nrdA	1.83	-3.64	11
ECs3125	glpT	1.65	-2.22	4
ECs3126	glpA	3.61	-2.78	3
ECs3127	glpB	2.59	-2.15	3
ECs3128	glpC	2.92	-2.94	un
ECs3142	yfbF	-2.54	1.59	12
ECs3143	yfbG	-2.50	1.03	5
ECs3144	yfbH	-2.66	0.98	5
ECs3299	yfeU	2.39	-0.89	3
ECs3300	yfeV	2.28	-0.99	3
ECs3305	ypeA	2.05	-0.81	3
ECs3393	hscB	-1.37	2.51	7
ECs3748	yqeB	2.23	-0.78	3
ECs4109	mdh	2.05	-1.40	un
ECs4269	glpD	4.17	-1.82	3
ECs4408	dctA	4.59	-1.93	3
ECs4498	rfaF	-2.10	0.85	12
ECs4705	ilvD	-2.03	1.03	12
ECs4726	yifM	3.51	-3.62	un
ECs4750	yigI	2.77	-1.14	10
ECs4840	fieF	3.76	-3.29	3
ECs4851	glpK	3.54	-3.73	4
ECs4852	glpF	5.22	-4.10	3
ECs4918	yjaE	4.21	-1.58	3
ECs5298	yjiM	0.71	-2.14	8
ECs5313	yjiY	1.38	-3.28	un

Only 12 O157-specifc ORFs were expressed transiently upon entry into stationary phase. ECs1772, a putative intestinal colonization factor, decreased upon entry into stationary phase, then significantly increased in expression (>4 fold increase from 5 to 5.5 h) (Table 4).

Genes with a significant change in expression in early stationary phase

The interval with the second largest number of significantly modulated ORFs was in early stationary phase, during the 5 to 5.5 h interval. A number of these ORFs that had significant increases in expression levels during this interval are transcriptional regulators, such as the AraC-like regulators gadW and gadX, involved in regulation of the glutamate decarboxylase system of acid resistance (Table 5). Genes encoding DNA binding proteins, such as hha and cbpA, also increased in transcription in early stationary phase. Components of the glycerol metabolism pathway were activated in early stationary phase, including ugpC and ugpE, components of the glycerol 3 phosphate transporter, and ugpQ, the cytosolic glycerophosphoryl diester phosphodiesterase (Table 5).

Table 5

Significant ORFs (p value < 1 × 107) with greater than 4-fold expression changes in early stationary phase.

Ecs no.	gene	log2 change in expression 5–5.5 h	QT cluster	Ecs no.	gene	log2 change in expression 5–6 h	QT cluster
ECs0342	ykgC	2.30	2	ECs0568	gcl	4.04	2
ECs0513	hha	2.55	2	ECs0569	gip	3.56	2
ECs0514	ybaJ	2.16	2	ECs0570	ybbQ	4.89	2
ECs0732	ybgA	2.27	2	ECs0571	ybbV	3.27	2
ECs0733	phrB	2.15	2	ECs2095	yddV	4.27	2
ECs0769	cydB	-2.03	4	ECs2122	ydeZ	4.88	2
ECs0957	poxB	2.47	2	ECs2123	yneA	5.14	2
ECs0973	trxB	-2.25	1	ECs2124	yneB	4.46	2
ECs1155	cbpA	2.26	2	ECs2705	yedU	4.54	2
ECs1236	lomW	2.21	2
ECs1756	yciD	-3.21	4
ECs2047	ydcV	3.59	2
ECs2048	ydcW	2.35	2
ECs2386	ynhA	2.75	2
ECs2388	ynhC	2.89	2
ECs2493	yeaH	4.26	2
ECs2529	manZ	-2.65	1
ECs3395	iscU	2.61	2
ECs3396	yfhO	2.04	2
ECs4046	rpsO	-2.20	1
ECs4295	ugpQ	2.20	2
ECs4296	ugpC	2.60	2
ECs4297	ugpE	2.12	2
ECs4305	livK	-2.23	1
ECs4395	gadW	2.10	2
ECs4396	gadX	2.76	2
ECs4534	intL	2.50	2
ECs4673	atpC	-2.24	1
ECs5108	yjdI	2.20	2
ECs5269	yjhT	3.09	2

Expression of LEE island genes upon entry to stationary phase

Of the 41 LEE island ORFs, 23 had a significant (p < 0.00001) differences in expression in growth into stationary phase under minimal media (Fig. 5). Most of these ORFs were downregulated in transition to stationary phase (between 4.6 and 5 h). Three of these ORFs, espB, espD, and espZ, also had a significant change in expression in early stationary phase (5 h to 5.5 h). The greatest reduction in expression was observed in espB and espD, which both decreased 5.5- fold from late log to early stationary phase. Expression of espA also decreased > 5- fold over this interval, but the error variance for the espA probe was large and thus was not deemed significant. While the overall trend for the LEE island genes was a decline in expression during the transition to stationary phase, the exception was espZ which had a 3.2 fold increase in expression from late log to early stationary phase, indicating that this ORF may be activated by stationary phase associated regulators. The LEE ORFs as a group were found to be significantly enriched in the 3 h log-phase sample compared to the 5 h, 5.5 h, and 6 h stationary phase samples (Additional File 4).

Figure 5

Heatmap of gene expression over time for the ORFs of the LEE Pathogenecity Island. Expression values determined from the ANOVA analysis are represented colorimetrically, with dark red representing expression = 1.8 and dark green representing expression = -1.8 on a log2 scale. ORFs marked with an asterisk denotes that a statistically significant change in expression levels of that ORF was detected for at least one time point.

Based on the array data, a significant change in expression could not be detected for espA or eae as a result of the high variability in signal for probes targeting those ORFs. Because espB, espD and tir significantly decreased in expression, we used Q-PCR to determine if expression levels changed for espA and eae. Q-PCR confirmed the array data for espB, and detected a 32-fold decrease in espA and a 16-fold decrease in eae transcript levels from 4.6 to 5.5 h (Fig. 6A). Overall, the Q-PCR data detected greater amounts of decrease in espA, espB (22- to 35-fold), tir, and eae (7- to 16-fold) expression compared to the array data (3.2-fold decrease for tir and eae, 5- to 7-fold decrease for espA and espB) for the 4.6 to 5.5 h interval. Transcript levels for espZ were estimated to increase 3.2-fold for the 4.6 to 5.5 h interval from the array data. Q-PCR for espZ found only a slight change (~1.2-fold) in expression over the time intervals tested (Fig. 6A). Whereas the Q-PCR measurements did not confirm the increase in expression determined from the array data for espZ, it is clear that espZ expression does not follow the same pattern as other genes on the LEE island.

Figure 6

Relative changes in expression for virulence-associated genes from the LEE Pathogenecity Island. A. Expression differences of 5 genes transported through or adherence components of the LEE island as determined by Q-PCR. B. Expression differences of 3 genes encoding effectors (Esp) molecules transported through the LEE Type III Secretion System (TTSS) as determined by Q-PCR results. Average changes in expression for 2 independent replicate cultures were determined relative to the transcript level of each ORF at the 3 h mid-exponential phase time point.

Expression changes of LEE effectors

More than 60 putative effectors that are translocated through the LEE-encoded TTSS were recently identified [53] and of these TTSS effectors, 33 were found here to have significant (p < 0.00001) changes in expression over time. A total of 7 TTSS effector genes (nleA, nleG, espM1, espM2, nleG2-2, espZ, and nleG2-3) amplified in expression between mid- and late- exponential phase (3 h to 4 h). At the stationary phase transition point, 11 ORFs increased in expression and 7 ORFs had a significant reduction in expression (espJ, tccP, tir, espF, espB, espM2, and nleD). In early stationary phase, eleven ORFs were upregulated, with the exception of espB (2.5 fold decrease). From late exponential to early stationary phase (4.6 h to 5.5 h), 8 ORFs had an increase in expression > 2- fold, with espY1 increasing 9 fold. Over this same time interval, 5 TTSS effector ORFs had a 2-fold or greater decrease in expression, with espB decreasing 5.3 fold (Table 6). Array data for the TTSS effectors espY1 and espM1 indicated an increase in expression from exponential to stationary phase, which was reflected in the Q-PCR data for these ORFs (Fig. 6B). Expression of the TTSS espJ decreased in stationary phase and a 21 -fold decrease in expression during the transition from log to stationary phase (4.6 to 5 h) was determined by Q-PCR (Fig. 6B).

Table 6

Significant changes in transcript levels of Type III secretion system (TTSS) effectors.

ECs number	Gene name	Log2 change in expression from 4.6 h to 5.5 h
ECs0061	espY1	3.03
ECs0073	espY2	1.24
ECs1825	espM1	1.47
ECs1994	nleG2-2	1.41
ECs2156	nleG2-3	1.63
ECs2229	nleG2-4	-1.15
ECs2714	espJ	-2.18
ECs2715	tccP	-1.39
ECs3488	nleG6-3	1.11
ECSs4554	espB	-2.41
ECs4561	tir	-1.72
ECs4571	espZ	1.69
ECs4643	espL3	1.77

Significant increase of genes from the acid fitness region in stationary phase

A cluster of 12 genes located in the E. coli K-12 genome at position 3652706 to 3665603 bp has recently been termed the acid fitness island (AFI) [54]. In E. coli O157:H7 Sakai genome, the homologous region contains a 9 kb insertion between yhiF (ECs4378) and yhiD (ECs4388). This AFI genomic region contains multiple transcriptional regulators that control expression of the glutamate dependent acid resistance (GDAR) system, such as GadE, GadX, and GadW, that are also known to influence LEE expression [55,56]. The microarray comparisons demonstrate that the expression of the AFI genes increased significantly from exponential to stationary phase, and included some of the greatest increases in transcript level among the whole genome (Fig. 7). Transcript levels of gadA increased 111 fold and levels of gadE increased 85 fold from mid-exponential (3 h) to stationary phase (6 h). Over half of the AFI genes (7/12) had a significant increase in expression from mid- to late- exponential phase (Fig. 7). Transcript levels of gadA, gadE, and slp also increased significantly from 4.3 to 4.6 h, which is different from the expression profiles of the other known stationary phase activated genes that increase from 4.6–5 h. GSEA found the AFI ORFs as a group to be significantly enriched in all of the time points subsequent to the mid-exponential phase sample (Additional File 4).

Figure 7

Expression heat map of 12 genes from the acid fitness region (AFI). Expression values determined from the ANOVA analysis are represented colorimetrically, with dark red representing expression = 3 and dark green representing expression = -4 on a log2 scale. Over half of the AFI genes (7/12) had a significant increase in expression from mid- to late- exponential phase

Increased expression of Shiga toxin genes upon entry into stationary phase

More than 50% of Sp 15 ORFs and 60% of Sp 5 ORFs had significant (p < 0.00001) changes in expression over time into stationary growth phase. The genes encoding Shiga toxin 1, stx1A and stx1B, as well as those encoding Shiga toxin 2, stx2A and stx2B, had a significant difference in expression between mid- and late- log phase (3 to 4 h). Stx1 genes decreased ~1.8 fold over this interval, while Stx2 genes increased ~1.7 fold. Stx1 genes increased significantly in early stationary phase, with a ~4- fold increase from 5 to 6 h. A significant 2.5- fold increase for stx2B occurred at the transition point into stationary phase, but a concomitant increase was not observed for the stx2A ORF on the microarray. Q-PCR was used to monitor expression of stx1A, stx2A, and stx2B over time (Fig. 8). Expression of stx1A decreased 4-fold from mid-exponential phase to the stationary phase transition point, and then increased 2-fold during early stationary phase (5 to 6 h interval), which is similar to the array data. A 2-fold increase in expression from mid- to late- log phase was observed for stx2B using Q-PCR. While stx2B transcript levels remained higher in stationary phase compared to exponential phase, a subsequent increase during the stationary phase transition was not observed. Transcript levels of stx2A increased 1.5-fold from mid- to late- exponential phase, and decreased 1.5-fold from late exponential to stationary phase (Fig. 8). The ORFs encoding the phage replication factors O and P, as well as the CII regulatory protein and the Q antiterminator protein, increased in expression upon entry into stationary phase, but it is difficult to attribute these ORFs to a specific phage, as there are multiple copies of these ORFs throughout the genome.

Figure 8

Relative changes in expression of Shiga toxin genes. Changes in expression of the phage-encoded toxin genes stx1A, stx2A, and stx2B as determined by Q-PCR. Average changes in expression for 2 independent replicate cultures were determined relative to the transcript level of each ORF at the 3 h mid-exponential phase time point.

Expression changes in tellurite resistance genes and urease genes

Proteins involved in tellurite resistance are encoded on a pathogenicity island (TAI) along with genes encoding urease and are present in a single copy in E. coli O157:H7 Sakai [16]. Significant changes in expression over time were detected only for terZ, terD, and terE (Fig. 9A). Probes targeting terB and terC are not present on the array. Expression of these three genes increased significantly between mid- and late- exponential phase (~1.5 fold) and then decreased 5.9- to 7.4- fold during the transition into stationary phase and in early stationary phase (4.6 h to 5.5 h interval). These three ORFs were grouped into QT cluster 1. The ureD, ureA, ureC, ureE, and ureG ORFs had a significant reduction in expression from late exponential to early stationary phase (1.5 to 3 fold) and were grouped into QT cluster 1 (Fig. 9B). Also located on this genomic island is the adherence factor iha [57], but there is no significant change in iha mRNA over time. Quantitative real-time PCR targeting terZ and ureD confirmed the reduced expression of these ORFs from exponential to stationary phase, detecting an 8-fold decrease for terZ and ureD (Fig. 9C).

Figure 9

Expression profiles of 12 ORFs encoded on the tellurite resistance island (TAI). Expression values determined from the ANOVA analysis are plotted over time for the A. tellurite resistance and B. the urease-encoding ORFs. C. Average relative expression of terZ and ureD was determined by Q-PCR of two independent replicate cultures.

Discussion

By monitoring global transcript levels over time, we have measured the numerous alterations in gene expression patterns of E. coli O157:H7 when growing exponentially and then entering stationary phase in minimal medium. E. coli O157:H7 responded to low O2 by increasing expression of genes encoding anaerobic electron acceptors and decreasing expression of genes encoding the TCA cycle. O157-specific ORFs, including nleA and nleG, which encode TTS effectors, were upregulated during this interval. As cells entered stationary phase, ~50% of the ORFs in the genome responded with significant changes in transcript level. The global trends include down regulation of some of virulence factors, such as the components of the TTSS and some of the TTSS effectors, in stationary phase. Other TTSS effectors and the genes encoding Shiga toxin increased in transcript level upon entry to stationary phase. Entry into stationary phase was characterized by the increase in expression of known stress and survival related genes, such as the osmotic stress proteins encoded by osmB and osmC, and the glutamate dependent acid resistance system encoded by gadA and gadBC.

As E. coli O157:H7 cells grew in the minimal medium, O2 levels in the medium decreased and were not replenished from the atmosphere (Fig. 4). The decrease in available oxygen was reflected in the transcriptome, with significant increases in genes encoding anaerobic electron acceptors, decrease in transcript levels of TCA cycle genes, and increase in expression of pyruvate formate lyase and concomitant decrease in expression of the pyruvate dehydrogenase complex. Decreased expression of cyo and sdh operons, with concurrent increased expression of the cyd operon was also observed during anaerobic growth of E. coli K-12 at pH 4.7 compared to aerobic growth at acidic pH [28] (Additional File 1). 5% (73/1239) of O157-specific ORFs also had significant changes in expression over the same time interval, which leads one to speculate that these O157-specific ORFs may be regulated by transcription factors that influence expression of respiratory genes, such as FNR and ArcA. Virulence factor expression in enteric pathogens, such as Salmonella enterica serovar Typhimurium and Vibrio cholerae, can be controlled by the redox state regulators FNR and ArcA [58-60]. A search for the FNR and ArcA binding sequences 500 bp upstream of all ORFs in the Sakai genome (using Pattern Search on coliBASE [61]) did not identify any of the ORFs that were significantly expressed or repressed in exponential phase. It may be that these O157-specific ORFs are regulated indirectly through FNR or ArcA, or are modulated in a completely different manner. Recently Ando et al. [62] demonstrated that anaerobic growth in the presence of the electron acceptors nitrate or TMAO accelerates maturation of the TTSS, independently of new protein synthesis. They found that the narGHIJ operon must be present for the maturation of the TTSS. Our findings indicate that the expression of this operon increases significantly in late exponential phase, which is consistent with the narGHIJ operon contribution to the maturation of the TTSS in this environment.

Much is known about regulation of LEE expression; multiple circuits of activation and repression have been identified. LEE expression can be induced in response to a number of environmental stimuli, and while common transcriptional regulators and signaling systems can activate LEE expression, they all typically act on Ler or the recently described GrlA and GrlR regulators located on the LEE island [63], which then alter expression of other LEE genes. As LEE expression is thought to be maximal under conditions similar to the intestinal environment [11], we would expect to see increased expression of LEE under these conditions compared to the minimal medium used here. We did observe similar growth-phase dependent changes in expression in the LEE genes that has also been observed for O157:H7 grown in rich medium. LEE expression is known to be maximal in late exponential phase [12,64]. Nakanishi et al. determined that LEE can be stimulated during nutrient downshifts through activation of ppGpp and DksA [10]. The pattern of espB expression observed during induction of LEE with ppGpp is similar to our observation, in that espB and espA expression increases from mid- to late-exponential phase, and then decreases in stationary phase (Fig. 5 and Fig. 6A). LEE transcription can also be stimulated by quorum sensing, via the auto-inducer 3 signaling system that activates ler [12]. The quorum sensing activated regulator QseA which activates LEE transcription [65] had a decrease in transcript levels from exponential to stationary phase (Fig. 3; QTC 1). Many of the common E. coli regulators that have been shown to repress LEE expression, such as Hha [66], SdiA [67], GadE [56], YhiF [56], and IHF [68,69], all increased in mRNA levels from exponential to stationary phase (Table 2, Fig. 3; QTC 2). Other LEE activators, such as PchC [48], had a significant increase in mRNA in stationary phase.

The decrease in expression of LEE genes from exponential to stationary phase that we observed is similar to the expression changes of LEE genes when O157:H7 is attached to red blood cells [34]. In comparing our dataset to the genes that were found to be significantly up- or down-regulated in O157:H7 attached to red blood cells, we found that of the 299 ORFs reported to be down regulated in attached cells, 185 had significant changes in expression over time, and 73% of these ORFs were classified in QT cluster 1 (Fig. 3), which includes the LEE genes. Of the 105 ORFs reported to be upregulated in attached cells, 89 had significant changes in expression over time, and 56% of these significant ORFs were placed in QT cluster 2 (Fig. 3), suggesting that many of the ORFs expressed or repressed during attachment to plasma membranes are also regulated by growth phase.

Not as much known about regulation of non-LEE encoded TTSS effector genes. The presence of some of the TTSS effectors in the genome is variable among O157 strains [70]. The TTSS effectors tccP and espJ are located on Sakai phage 14 (EDL-933 OI#79); tccP (espFu) is necessary for actin recruitment and A/E lesion formation in E. coli O157:H7, while espJ influences colonization dynamics [71-73]. Reading and colleagues recently determined that the AI-3, epinephrine, and norepinephrine response regulator QseF (yfhA) is an indirect transcriptional activator of tccP [74]. Our data indicate that as Sakai O157:H7 entered stationary phase, qseF mRNA levels decrease significantly (2 fold) (QTC 1) and tccP-espJ levels decreased significantly as well. This decrease in tccP and espJ transcript levels in stationary phase is in contrast to a previous study, where the activity of the tccP-espJ promoter was found to be the same in exponential vs. stationary phase cultures of EDL-933 in Dulbecco's Modified Eagle Medium as monitored by a gfp fusion [75]. Some of the TTSS effectors appear to respond to low O2 conditions, as 7 effectors increased significantly from mid- to late-exponential phase when oxygen levels decreased. Of the TTSS effectors with significant changes in expression over time, ~50% have decreased expression from exponential to stationary phase, while the other half have increased expression in stationary phase. Several of these TTSS effectors are predicted to be pseudogenes; one example is espL3, an ORF that was found to have a significant increase in expression at the stationary phase transition point. Many of the TTSS effectors with increased expression levels in stationary phase are members of the NleG family (Table 6), one of the largest families of TTS effectors that contain many duplications [53]. The utility of the increased expression of some of the TTS effectors in stationary phase is unclear and warrants further investigation.

The ability to resist low pH is a crucial adaptation and critical component of the low infectious dose of E. coli O157:H7. The GDAR system provides the greatest protection against low pH, allowing survival at levels of gastric acidify (e.g. pH 2.0) as long as glutamate is present in the environment [76]. Genes encoding the glutamate decarboxylase isozymes and the glutamate-gamma-aminobutyric acid antiporter are under tight control and can be activated via a number of environmental cues and transcriptional regulators [77], and many components of the system are located in the AFI region [54]. Also located on the AFI are hdeA, slp, and yhiF, which encode proteins involved in protecting cells from their metabolic by-products when the medium is acidified, and yhiD and hdeD, which are required for density-dependent acid resistance [78]. ORFs of the AFI are induced by a variety of stressful environments, including growth in the presence of acetate [79], anaerobic, low pH growth [28], during attachment to red blood cells [34], and in stationary phase [23]. Induction of the AFI ORFs as a response to multiple types of stresses highlights the importance of acid stress survival to the route of transmission for E. coli. Transcription factors located on the AFI, such as GadE and GadX, regulate a large number (>40) of ORFs outside of the GAD system in E. coli K-12 [54,80], and it is known that these factors serve as repressors of LEE transcription [55,56], but it is unknown what the effect of these regulators are on a genome-wide scale in E. coli O157:H7.

Chromosomal Toxin-antitoxin (TA) loci encode TA systems which are thought to induce reversible stasis, allowing survival during episodes of extreme nutritional stress [81,82]. Increased expression of TA loci in E. coli K-12 during growth transitions led to the development of a model integrating inhibition of all macromolecular synthesis within the stringent response model [23], while recent findings suggest that the chromosomal TA systems do not significantly influence bacterial fitness during nutrient limitation [83]. E. coli O157:H7 has some of the same TA loci as E. coli K-12; dinJ-yafQ [84], chpAR, chpBS [85], as well as the recently described chromosomal ccdAB system that is present in E. coli O157:H7 and not in E. coli K-12 [49]. CcdB encodes a DNA gyrase toxin [49], chpB encodes a translation inhibitor [86], and yafQ encodes an mRNA endoribonuclease [84]. Here we observed increased expression of the dinJ-yafQ, chpBS, and ccdAB TA loci upon entry to stationary phase, indicating that these systems are actively transcribed in stationary phase, but their utility remains to be elucidated.

Important questions remain about the factors regulating the Shiga toxins which are encoded by lambda-like bacteriophages [7]. These phages can be transferred between many E. coli strains in nature and were the principal virulence factors acquired as a key step in the evolution of E. coli O157:H7 [87]. How they become influenced in expression and regulated when acquired into a new genome is not fully understood. Early work described two types of Shiga toxin [88] encoded by Stx1 and Stx2, both located within the late operons of the Stx-encoding phages [89,90]. Stx1 expression is regulated in part by the iron-dependent transcriptional repressor, Fur [15]. Exposure of E. coli O157:H7 to low iron conditions result in an increase of stx1 transcription [14]. Our growth experiments indicate that the stx1 genes decrease in expression from mid- to late-exponential phase, at the same interval when O2 levels decrease (Fig. 8). Iron becomes more soluble as O2 concentration decreases, which could lead to the decreased stx1 expression. Stx1 expression can be induced at 2 other promoters in addition to the Fur binding site, those that require the phage anti-terminator Q [91] and the phage anti-terminator N [14]. Stx2 transcription is initiated at the late phage promoter and is dependent on the anti-repressor Q protein [13]. The production of Stx2 is part of the lytic cycle of the bacteriophage, and can be induced by antimicrobials [92,93]. Herold et al. found that phage genes were induced upon exposure of E. coli O157:H7 EDL-933 to norfloxacin; phage late genes and stx2 were induced 150-fold [35]. Levels of stx2 mRNA are higher in starved O157:H7 compared to exponentially growing cells [94], indicating an overall trend of stx2 expression induction by stress. Here we observed that stx1 expression increased in stationary phase and stx2 expression was higher in stationary phase compared to exponential phase (Fig. 8). Herold et al. also reported that stx2A is transcribed more efficiently than stx2B in O157:H7 EDL-933, and may be regulated at the posttranscriptional level [35]. This difference in transcription efficiency between stx2A and stx2B was not reflected in our data, as stx2A levels did not appear to increase as the stx2B levels did in stationary phase.

The genomic island termed the Tellurite and Adherence Island (TAI) contains genes encoding tellurite resistance, urease, and an adherence factor, iha [57]. The urease operon and the tellurite genes are thought to be ubiquitously distributed in EHEC strains [95,96] but is not present in O157:NM or O55:H7 strains [97]. Urease activity has been detected in relatively few EHEC strains. In O157:H7 Sakai, the urease genes have been detected at the transcript level, but a functional protein is not produced due to a stop codon present in ureD [98]. Saridakis et al. surveyed 25 O157:H7 strains that were isolated from humans, cattle, and pigs for ureC and the ability to resist low pH using the urease enzyme. UreC was detected in all of the isolates but there was no evidence for urease-mediated acid resistance [99]. Here we see that the genes encoding urease are expressed during exponential growth and transcript levels decrease significantly as cells enter stationary phase. The genes encoding tellurite resistance are functional in EHEC, and can be separated into two classes – those that are expressed constitutively during growth (terDEZ) and those that are expressed only during growth in the presence of tellurite (terBCF) [96]. Our data show this as well, with terD, E, and Z all expressed during exponential growth, with a significant decrease in transcript levels upon entry to stationary phase, and no significant changes in transcript levels for terA and F.

As E. coli O157:H7 cells transitioned to stationary phase, we observed a transient increase in expression of ORFs encoding proteins involved in acquisition and utilization of alternative carbon sources. ORFs involved in acetate uptake and utilization increased in expression, and some of these genes, such as acs, are known to be controlled, in part, by RpoS [46]. RpoS is activated during slow growth on glucose [52], and growth on less desirable carbon sources [100]. RpoS is also activated during the switch from one carbon source to another, as observed in E. coli K-12 growing in a rich medium [33] and in minimal medium [23]. The glyoxalate shunt is thought to play an important role under glucose-limited conditions, where the absence of catabolite repression is an important factor for activity of the cycle [101]; here we observed that ORFs encoding glyoxalate shunt enzymes increased in expression as E. coli O157:H7 entered stationary phase (Tables 3 and 4). Over expression of high affinity sugar transporters occurs during the short-term adaptation of E. coli K-12 to carbon-limited growth [52,102]. Sugar transporters and carbon uptake systems are also activated in E. coli K-12 during growth on less desirable carbon sources [100]. In E. coli O157:H7, we observed increased expression of ORFs encoding transport systems for carbohydrates, sugar alcohols and acids during the transition to stationary phase. These data suggest that as glucose was depleted, other carbon sources were sought, but were not available in the minimal medium. In a rich medium, we would expect cells to resume growth, albeit at a lower rate, once glucose was consumed.

Transition to stationary phase is characterized by complex physiological changes to the bacterial cell. Many of the changes in expression patterns observed for O157:H7 in this study have also been observed for E. coli K-12 entering stationary phase [23]. Growth under stressful conditions, such as in the presence of acetate [79] or at low pH under anaerobic conditions [28], also induce expression of stationary phase associated genes. The sigma factor RpoS plays a critical role in transcribing genes associated with stationary phase and stress response [24-26], many of these genes, such as bolA, dksA, osmC, and osmB, have been identified in this study as well. We still do not know about most of the O157-specific ORFs, and what regulators control their expression. Here we have classified some of the well-known O157 genes, as well as O157 genes of unknown function, as being modulated during growth transitions. Global transcriptional profiling of isogenic mutants for a specific regulator have been conducted for a number of regulators in E. coli K-12. As we know more about the regulation of virulence loci, such as the LEE, by global regulators, it would be very useful to perform these profiling experiments in E. coli O157:H7 and other pathogenic E. coli.

Conclusion

The principal contribution of this study is the first complete description of global transcriptome profiling of the human pathogenic strain, E. coli O157:H7, during growth transitions. Many O157-specific ORFs responded to the growth transition with significant changes in gene expression including multiple genes that are encoded on pathogenicity islands or toxin-converting bacteriophages. Because a majority of the O157-specific ORFs have unknown functions, these findings will have practical use for comparing transcriptional responses to a variety of growth conditions and elucidating the function and factors regulating these unknown ORFs, some of which may contribute to disease processes. These results also generate new hypotheses and provide opportunities to identify ORFs of interest, based on expression profiles, and compare expression within E. coli O157:H7 populations or with other EHEC populations. It is useful to understand what transcriptional changes occur from exponential to stationary phase, as these transitions are important adaptations to the growth and survival of E. coli O157:H7 in natural conditions.

Methods

Growth conditions

E. coli O157:H7 RIMD0509952 (Sakai) involved in a radish sprout outbreak [103] was stored at -70°C in LB broth and 10% glycerol. The Sakai strain was inoculated into 10 mL of LB broth from freezer stocks and grown to OD600 = 0.1. This ~4 h period of growth in LB allowed cells to recover before transfer to MOPS minimal media. MOPS minimal media (10×) was prepared as described in Neidhardt et al. [104] with the addition of 20 mL of micronutrient solution instead of 10 mL. MOPS minimal media contained 100 mL of 10× MOPS minimal media, 10 mL of 0.132 M K2HPO4, and 5 mL of 20% D-glucose per liter. Cultures were inoculated into 50 mL of MOPS minimal media with 0.1% glucose at a ratio of 1:200 and grown to stationary phase at 37°C with shaking. Cultures in MOPS were transferred to 100 mL MOPS in 250 mL plastic culture flasks (Nalgene, Rochester, N.Y.) at a ratio of 1:75 and grown for 9 h before transfer again to 100 mL MOPS in 250 mL plastic culture flasks at a ratio of 1:30 and this culture was used to sample RNA during growth (Fig. 1). Four independent cultures were sampled at 3 h (OD600 ≈ 0.22), 4 h (OD600 ≈ 0.45), 4.33 h (OD600 ≈ 0.73), 4.66 h (OD600 ≈ 0.93), 5 h (OD600 ≈ 1), 5.5 h, 6 h, 7 h, 7.5 h, and 8 h. After 5 h, the culture density remained at OD600 ≈ 1. Residual dissolved oxygen tension was monitored using a dO2 probe and Consort C535 multimeter (Topac Instrumentation, Hingham, Mass.) for three independent cultures in MOPS minimal medium.

RNA isolation

At each time point, 4 mL of culture was mixed with 8 mL of RNAProtect (Qiagen, Valencia, Calif.), vortexed, and centrifuged at 4°C, 7500 rpm for 10 min to pellet cells. The supernatant was removed and cell pellets stored at -70°C for 1 week or less before RNA extraction. Cell pellets were suspended in 700 uL of 95C lysis buffer (20 mM sodium acetate pH 5.2, 2 mM EDTA pH 8.0, and 0.5% SDS), held at 95°C for 30s then mixed with 700 uL of 65°C Acid-Phenol: Chloroform, pH 4.5 (with IAA, 125:25:1) (Ambion, Austin, Tex.). Samples were held at 65°C with periodic shaking for at least 6 minutes before centrifuging at 12000 rpm for 10 min. Supernatant was extracted again with acid-phenol:chloroform and then with chloroform:isoamyl alcohol (24.1). RNA was precipitated for at least 1 hour at -20°C in 2.5 V 100% ethanol and 1/10 V 3 M sodium acetate pH 5.2. RNA was suspended in 1 mM sodium citrate pH 6.5 and stored at -70°C. RNA samples were purified and treated with Dnase using the Rneasy kit (Qiagen). RNA quality was assessed by electrophoresis on formaldehyde-agarose gel. RNA was checked for complete DNA digestion by PCR.

cDNA synthesis and hybridizations

Reverse transcription reactions contained 6 ug RNA, 2 ug random primers (Invitrogen, Carlsbad, Calif.), 1× first strand buffer (Invitrogen), 10 mM DTT, 400 U Superscript II (Invitrogen), 0.5 mM each dATP, dCTP, and dGTP, 0.3 mM dTTP, and 0.2 mM amino-allyl dUTP. 30 uL reactions were incubated overnight at 42°C. cDNA was purified using PCR cleanup columns (Qiagen) with Phosphate wash buffer (5 mM K2HPO4, pH 8.0, 80%EtOH) and phosphate elution buffer (4 mM K2HPO4, pH 8.5). Amino-allyl labeled cDNA was dried and suspended in 0.1 M sodium carbonate pH 9.3 and coupled with either Cy3 or Cy5 (Amersham Biosciences, Piscataway, N.J.). Uncoupled dye was removed by another purfication using the PCR cleanup kit. Concentration of cDNA and amount of incoporated dye was measured for each sample using a Nanodrop spectrophotometer (Ambion).

Oligo microarray

The E. coli oligo set version 1.0.1 (Operon) was printed onto Corning UltraGaps (Corning Incorporated, Acton, Mass.) coated slides at the Research Technology Support Facility at Michigan State University. The Qiagen oligo set contained 5,978 probes specific for three E. coli strains, K-12 (MG1655) [105], O157:H7 Sakai [16], and O157:H7 EDL-933 [17], in which 5,943 probes were 70-mer oligonucleotides, and 35 probes had lengths that ranged from 41–69 bp. All of these probes are spotted in duplicate on each array. The oligo set also contained 12 randomized negative control 70-mer oligonucleotides. There are a common set of 3807 probes that target 3807 ORFs present based on the genome sequences of E. coli K-12 MG1655, E. coli O157:H7 EDL-933, and O157:H7 Sakai, which we refer to as the backbone ORFs. There are also 1741 probes targeting ORFs that are specific to one or two of the genomes, typically to O157:H7 EDL-933 and Sakai. These probes are referred to as the O157-specific ORFs. Probes designed to target K-12-specific or EDL-933 specific ORFs were not considered after the data analysis stage because these genes are not present in the Sakai genome. We used the Qiagen annotation for the probe set, and analyzed only the 4886 probes that were assigned a ECs number.

Hybridization conditions

Arrays were cross-linked by exposure to 600 mJ UV before blocking in 1% SDS, 5× SSC, and 1 mg/mL BSA at 42°C for 1 hour. After blocking, arrays were washed 2× 5 min in 0.1× SSC and 2× 30 s in H2O. Dried arrays were placed into hybridization cassettes (TeleChem International, Sunnyvale, Calif.) and the cDNA samples were suspended in 10 mM EDTA, denatured at 95°C for 5 min and then mixed with 40 uL of SlideHyb buffer 1 (Ambion) and loaded under a coverslip onto the array. Hybridizations were carried out at 47°C for 16–18 h. After hybridization, arrays were washed in 2× SSC, 0.5% SDS 37C for 5 min, followed by 2× 5 min in 0.1× SSC, 0.1% SDS 37°C, and then 2 × 2.5 min in room temperature 0.1× SSC. Arrays were scanned using an Axon 4000b scanner (Molecular Devices, Sunnyvale, Calif) and images were analyzed using GenePix 6.0 (Molecular Devices). The 3 h sample served as a common reference and was hybridized with all subsequent samples. Array data have been deposited at the NCBI Gene Expression Omnibus (Accession GSE7477).

Quantitative real-time PCR

Expression levels of 14 ORFs determined to be differentially expressed (p < 0.00001) were verified by quantitative real-time PCR (Q-PCR). Primer pairs were designed based on the published reference genome sequence of E. coli O157:H7 strain Sakai using the Primer3 server, and secondary structure scrutinized using the Mfold algorithm [106,107] (Additional File 4). cDNA was synthesized from 1 ug of total RNA using the iScript Select cDNA synthesis kit (BioRad, Hercules, Calif.) and random hexamers supplied with the kit. Template cDNA was diluted to 10-1 to 10-3 for use in Q-PCR. Q-PCR reactions contained 12.5 uL 2× iQ SYBR green supermix (BioRad), 0.63 uL of each primer (10 uM stock), 9.24 uL H2O, and 2 uL cDNA and were conducted under the following conditions: 2 min at 95C, followed by 40 cycles of 10 sec at 95C then 20 sec at the specific annealing temperature. Q-PCR reactions were performed in triplicate for each cDNA sample tested. Relative expression was determined using the method described by Pfaffl et al. [108] and the fluorescence data from the 16S rRNA target were used for normalization within samples. All samples were then compared to the expression levels of the mid-exponential (3 hour) samples. The average log2 expression and standard deviation from two independent RNA samples are reported for each time point tested.

Data analysis

Raw intensity values for all probes on each array were normalized using pin-tip LOWESS [109] in R v.2.2.1 [110] with the MAANOVA (v. 0.98–8) package [111]. Signals from two replicate probes on each array were averaged and log2 transformation applied. Differences in transcription levels over time were determined using a mixed model ANOVA in R/MAANOVA, where the log transformed intensity data, Y = A (array) + D (dye) + T (time) + S (sample = biological replicate) + E (error). The ANOVA modeling allows for consideration of appropriate error structures for experiments with multiple sources of variation in microarray measurements [112]. The random effects of the model were biological replicate and array effects, whereas the fixed effects were time point and dye effects [113]. The Fs statistic, a shrinkage estimator for gene-specific variance components that makes no assumptions about the distribution of variances across genes, was estimated [40]. Significant changes in expression over time were determined by calculating the p values for the Fs statistic for each gene using 1000 random permutations. The p-values were adjusted to correct for type I error with the Benjamini-Hochberg (B-H) linear step-up correction implemented in R/MAANOVA and a cutoff adjusted p-value of 0.0000001. Pair wise contrasts of time points were estimated by the t-test in R/MAANOVA. Contrast p-values were corrected for multiple testing by using the B-H step-up correction. Probes with significant changes in expression over time were grouped by the time points at which the significant change occurred, as determined by the contrast analysis (adjusted p value < 0.05). QT clustering of significant ORFs was conducted in MeV v. 3.1 [116] with a diameter of 0.4 and a minimum cluster size of 15.

Gene Set Enrichment Analysis (GSEA) [114] was conducted using GSEA v. 2.0 [115] on a ranked list of log2 expression ratios (determined from the ANOVA estimates) for 5 of the time points relative to the mid-exponential time point. The following parameters were used for the analysis: 1000 permutations and exclusion of gene sets with less than 5 members and greater than 700 members. Gene sets were defined based on the E. coli O157:H7 Sakai role categories obtained from the Comprehensive Microbial Resource at The Institute for Genomic Research [116]. The GSEA results are summarized in Additional File 4. All gene sets determined to be enriched with a FDR (q-value) less than 0.25 are reported.

Authors' contributions

TMB designed and carried out all the microarray experiments, data analysis, and composed the manuscript. LMW and WQ assisted in the experimental design and statistical data analyses. JTR designed and optimized the Q-PCR assays and LMO conducted the experimental assays. TSW oversaw the project, contributed to the design and final analyses, and manuscript completion. All authors have read and approved the final version of the manuscript.

Additional File 1

Entire data set of O157 ORFs sorted by significance. Log2 expression indices for each time point for 4886 ORFs in the multigenome microarray

Click here for file

Additional File 2

Significant O157-specific ORFs. Log2 expression indices for each time point

Click here for file

Additional File 4

GSEA results. Gene Set Enrichment Analysis (GSEA) conducted using GSEA v. 2.0 on a ranked list of log2 expression ratios (determined from the ANOVA estimates) for 5 of the time points relative to the mid-exponential time point.

Click here for file

Additional File 3

QPCR primers. PCR primers designed for QPCR analysis.

Click here for file