RpoS-Regulated Genes of Escherichia coli Identified by Random lacZ Fusion Mutagenesis

Abstract

RpoS is a conserved alternative sigma factor that regulates the expression of many stress response genes in Escherichia coli. The RpoS regulon is large but has not yet been completely characterized. In this study, we report the identification of over 100 RpoS-dependent fusions in a genetic screen based on the differential expression of an operon-lacZ fusion bank in rpoS mutant and wild-type backgrounds. Forty-eight independent gene fusions were identified, including several in well-characterized RpoS-regulated genes, such as osmY, katE, and otsA. Many of the other fusions mapped to genes of unknown function or to genes that were not previously known to be under RpoS control. Based on the homology to other known bacterial genes, some of the RpoS-regulated genes of unknown functions are likely important in nutrient scavenging.

RpoS, an alternative sigma factor, controls a large regulon that is specifically expressed when the cell is nutrient deprived or is subjected to external stress, such as osmotic shock (32). Genes that are dependent on RpoS have been identified in different contexts from many studies of regulation in various gram-negative bacteria (38). From these studies it is clear that the regulon encodes many proteins that help bacteria adapt to adverse conditions, including those that are involved in nutrient scavenging, DNA repair, protein turnover, and protection from external environmental insult (31).

Characterization of the RpoS regulon has relied on several experimental approaches, including two-dimensional gel electrophoresis (50), identification of genes that are induced by RpoS-related stimuli such as carbon starvation (85), and by identifying gene fusions that depend on RpoS for expression (66). Use of bacterial cDNA microarrays or macroarrays, while potentially useful in identifying additional members of the RpoS regulon, have been employed in only a few studies of Escherichia coli and have not yet been used to directly assess RpoS dependence of genes in isogenic wild-type and rpoS mutant strains. In a previous study (66), our lab introduced an RpoS-null allele into a bank of random operon-lacZ fusions to facilitate the identification of RpoS-dependent genes based solely on RpoS requirement. This resulted in the identification of several new highly RpoS-dependent genes in E. coli (66), and similar approaches have been employed with Salmonella spp. where other RpoS-dependent genes have been identified (37). Lac gene fusions are useful measures of gene expression, because the product, β-galactosidase, is stable and easily assayed (16). This makes lacZ fusions particularly suitable for the study of genes that are expressed in stationary phase, where the stability of the RNA and protein products of the affected gene is not known.

In this study, we report the completion of the identification of reporter fusions identified in a previous genetic screen of lacZ-expressing fusion mutants. In addition, we examined the expression of the RpoS-dependent operon fusions in rich and minimal media and classified the RpoS-dependent genes according to their function.

MATERIALS AND METHODS

Growth conditions.

Expression assays were performed using derivatives of E. coli K-12 (Table 1). Cultures were grown overnight from single colony isolates in M9 minimal medium or Luria-Bertani (LB) medium containing appropriate antibiotics. The rich medium used was LB broth (53). Cell growth was monitored by measuring optical density at 600 nm (OD₆₀₀) (Novospec II; Pharmacia LKB, Cambridge, United Kingdom). For β-galactosidase expression studies, the cultures were maintained in early exponential phase (OD₆₀₀ < 0.2) for at least eight generations prior to the start of each experiment. Bacterial cultures were grown at 37°C and shaken at 200 rpm, sampled, and assayed for β-galactosidase activity.

TABLE 1.

E. coli strains and plasmid used in this study

Strain or plasmid	Genotype or characteristics	Reference
Strain
GC4468	ΔlacU169 rpsL	68
GC122	Like GC4468, but φ(rpoS::Tn10)	67
MC4100	Δ(argF-lacZ)205 araD139 fibB5301 relA1 RpsL150 thi ptsF25	68
HS1002-HS1100	Like GC4468, but carries RpoS-dependent operon-lacZ fusions	66
Plasmid
pMMkatF3	Carries the rpoS (katF) gene	54

Chemicals and media.

Chemicals were supplied by Fisher Scientific, Ltd. (Toronto, Ontario, Canada), Sigma Chemicals Co. (St. Louis, Mo.), and Gibco BRL (Burlington, Ontario, Canada). Antibiotics, amino acids, sugars, and other nonautoclavable solutions were sterilized by filtration. When required, medium was supplemented with kanamycin, 50 μg/ml; streptomycin, 50 μg/ml; tetracycline, 15 μg/ml; and/or X-Gal (5-bromo-4-chloro-3-indolyl-β-d-galactopyranoside), 50 μg/ml.

Identification of RpoS-dependent promoters.

To identify RpoS-dependent lacZ fusions, mutants containing individual random λplacMu insertions were conjugated with an Hfr strain containing an rpoS::Tn10 mutation that is transferred a few minutes after the point of origin during conjugation (66). LacZ⁺ mutant strains that yielded transconjugants having reduced blue color on X-Gal-containing plates were considered to contain presumptive RpoS-dependent operon fusions. Of several hundred identified mutants, 105 could be efficiently complemented when transformed with an RpoS-expressing plasmid (determined by testing catalase activity in colonies grown on LB plates and flooded with 30% hydrogen peroxide). The fusion mutations were then transduced (53) into GC4468 for further study. A small proportion of the strains in the operon fusion mutant bank carried double fusions (∼2%). During conjugation, loss of a high-expressing lacZ⁺ allele by recombination occasionally produced transconjugants that were kanamycin resistant (due to retention of a second operon fusion) and possessed reduced lacZ activity. These strains were easily identified by the catalase complementation test and were not studied further. A diagrammatic representation of the protocol used to isolate mutants is shown in Fig. 1.

FIG. 1.

Genetic screening method used to identify RpoS-dependent genes in E. coli.

Enzyme assays.

β-Galactosidase activity was assayed using whole cells with ONPG (o-nitrophenyl-β-d-galactopyranoside) as the substrate (53). Units of activity were calculated as (1,000 × OD₄₂₀)/(time of incubation [in minutes] × volume [in milliliters] × OD₆₀₀) and are expressed as Miller units (53).

DNA template preparation and sequencing of λ fusion junctions.

High-titer plate lysates were prepared by UV induction of lysogens (16, 62). DNA was extracted from the phage by phenol:chloroform extraction and ethanol precipitation (49). Bacterial DNA proximal to the λ fusion junction was sequenced by MOBIXlab (McMaster University, Hamilton, Ontario) using a phage Mu c end primer (5′-CCCGAATAATCCAATGTCCTCCCGG-3′) (62). The identities of the sequences obtained were determined by performing alignments using BLASTN (3). The predicted functions of the genes thus identified were either based on SWISSPROT (9) or ECOCYC (40) databases as indicated in the text. To confirm that the 5′ end of the fusion was correct, we designed primers for the 5′ area of the genes identified and for the common lacZ gene. These were used to PCR amplify the 5′ fusion junctions of the integrated mutator phage. All identified genes yielded products consistent with the sequenced 3′ junctions, indicating that no large rearrangements occurred during mutagenesis.

RESULTS

Identification of RpoS-dependent genes.

Previously, several hundred operon lacZ fusion mutants were isolated, of which 105 contained independent RpoS-dependent fusions identified on the basis of reduced β-galactosidase activity in strains carrying an rpoS-null allele (67). RpoS dependence was confirmed by restoration of gene activity following complementation with a functional rpoS gene (66). Duplicate strains that contained RpoS-dependent lacZ fusions in the same gene (albeit in different positions within the gene) were not considered further, and 48 resulting strains with mutations in unique genes were selected for further study. Many of the new lacZ fusions mapped to genes located in polycistronic operons. Although we did not isolate fusions to all RpoS-regulated promoters, the identification of a single RpoS-dependent gene within a predicted or known operon is suggestive that an entire given operon is under RpoS control. When this point is taken into consideration, the number of new RpoS-dependent genes identified in this study is almost 80. As the number of previously known RpoS regulon members is 100 genes (38), the total size of the regulon is approximately 200 members.

Quantitative expression of lacZ fusions.

To quantify the relative expression levels and RpoS dependence of the 48 lacZ fusion strains, β-galactosidase activity was assayed in rpoS⁺ and rpoS mutant strains grown in rich and minimal media (Table 2). All fusion strains were confirmed to be RpoS dependent in rich and minimal media in stationary phase, with the exception of the strain carrying a fusion in the argH-oxyR intergenic region, which was RpoS dependent only in rich media. RpoS dependence of the fusions in rich media ranged from 2- to 40-fold, with 80% of the strains showing greater than 5-fold dependence (Table 2). Similarly, in minimal media, 80% of the strains showed greater than fivefold dependence in stationary phase (Table 2).

TABLE 2.

Growth-phase-dependent expression of RpoS-dependent operon-lacZ fusions in strains grown in rich and minimal media^a

Strain	Gene	β-Galactosidase activity (Miller units) in:
		LB						M9 glucose
		Exponential phase			Stationary phase			Exponential phase			Stationary phase
		WT	rpoS mutant	RpoS dependent	WT	rpoS mutant	RpoS dependent	WT	rpoS mutant	RpoS dependent	WT	rpoS mutant	RpoS dependent
HS1002	yjbJ	6.2	3.2	2	74.4	3.4	22	20.2	3.6	6	53.7	0.3	180
HS1006	yjbE	4.2	2.1	2	45.2	2.1	22	23.9	1.0	24	63.3	1.1	58
HS1008	aldB	5.0	1.8	3	91.4	4.3	22	4.3	0.3	14	8.2	0.4	21
HS1010	gabP	5.1	2.8	2	111	4.7	24	6.9	0.5	14	26.8	1.5	18
HS1011	ygaU	15.4	18.9	1	144	33.5	4	58.9	6.9	9	198	16.0	12
HS1012	ygdI	13.8	3.9	4	54.5	7.6	7	8.4	4.7	2	21.3	4.0	5
HS1014	katE	12.7	1.9	7	80.3	2.0	40	23.4	0.3	78	79.4	1.2	66
HS1019	yqhE	7.4	10.5	2	91.4	12.2	8	20.2	3.4	6	48.3	4.6	11
HS1020	mltB	12.3	38.1	0.3	90.1	52.5	2	3.5	5.3	1	26.1	16.3	2
HS1022	narY	3.4	2.5	1	28.3	3.0	10	0.9	1.2	1	5.4	2.4	2
HS1024	ygaF	3.1	10.2	0.3	49.7	18.8	3	2.3	1.0	2	16.9	2.5	7
HS1026	ybaY	29.8	10.1	3	243	16.4	15	16.1	1.5	11	132	11.0	12
HS1028	yjgR	14.4	7.4	2	63.0	11.5	6	3.5	1.5	2	27.6	8.1	3
HS1033	ydaM	8.6	0.9	10	35.2	2.2	16	20.8	2.1	10	38.2	1.5	25
HS1035	ydcS	6.7	2.4	3	21.7	3.9	6	2.0	1.1	2	18.9	12.7	2
HS1036	ydcK	0.7	1.8	0.4	46.2	3.0	15	5.2	1.9	3	26.7	3.1	9
HS1037	gabD-gabT	0.7	0.6	1	11.0	1.3	8	1.9	0.3	6	3.7	0.6	6
HS1042	yhjY	11.7	3.6	3	66.8	4.5	15	3.0	1.2	3	29.1	3.2	9
HS1045	yhiN	16.1	11.1	1	60.1	13.0	5	12.7	1.6	8	16.3	2.5	7
HS1049	yeaG	11.7	2.2	5	42.2	3.3	13	12.3	0.5	25	13.7	1.4	10
HS1050	phnP	17.4	5.4	3	51.1	6.5	8	2.7	1.0	3	2.2	1.4	2
HS1054	yebF	23.0	2.7	9	225	5.5	41	44.6	3.6	13	87.7	3.6	25
HS1056	yodC	7.1	1.9	4	166	4.0	41	9.78	0.9	11	23.6	1.4	17
HS1057	gabD	4.6	3.7	1	245	15.7	16	12.4	1.3	10	27.1	2.1	13
HS1059	talA	3.2	3.3	1	106	5.3	20	13.3	0.9	15	42.6	1.6	26
HS1061	aroM	2.1	2.1	1	74.8	5.0	15	1.81	0.3	6	31.9	0.4	78
HS1063	ugpE	2.7	3.5	1	47.0	8.1	6	7.3	0.6	12	10.8	1.9	6
HS1064	nlpA-yicM	5.9	7.8	1	30.7	11.1	3	8.8	6.1	1	23.1	6.0	4
HS1066	argH-oxyR	12.3	14.8	1	30.0	18.9	2	88.3	81.4	1	102	102	1
HS1067	ybiO	2.4	1.0	2	11.9	1.0	13	7.0	0.4	18	15.8	0.6	26
HS1071	yhjG	12.8	6.0	2	159	9.4	17	27.7	1.8	15	95.7	6.3	15
HS1072	argH	9.2	12.4	1	33.9	18.3	2	42.8	24.5	2	90.8	17.9	5
HS1073	yliI	2.5	1.0	3	20.3	1.2	17	13.4	0.4	34	22.0	0.6	37
HS1075	ugpC	2.7	1.3	2	26.0	1.3	20	3.2	0.7	5	9.9	1.1	9
HS1077	yhiV	2.1	1.9	1	101	3.3	31	13.2	1.2	11	86	22.5	4
HS1078	uspB	4.5	5.6	1	30.0	7.1	4	21.1	2.0	11	66.6	8.2	8
HS1079	ldcC	4.9	3.4	1	29.8	4.5	7	18.1	1.3	14	46.6	4.3	11
HS1080	appB	2.7	0.9	3	112	4.6	25	8.4	1.6	5	9.3	1.3	7
HS1081	yhiU	6.0	2.1	3	112	2.9	39	34.0	2.5	14	117	14.0	8
HS1082	aidB	8.8	3.0	3	138	3.8	36	22.2	1.0	20	111	3.4	33
HS1084	yehX	3.6	1.5	2	58.3	2.1	28	23.1	1.0	23	55.5	2.6	21
HS1090	yphA	6.0	2.7	2	64.5	2.7	24	25.2	1.0	25	67.5	3.7	18
HS1091	osmY	15.9	6.0	3	264	7.5	35	110	2.3	48	210	3.3	64
HS1092	yfcG	3.3	1.2	3	62.3	1.7	37	16.3	0.6	27	67.9	1.3	52
HS1094	otsA	8.5	3.1	3	165	5.4	31	80.1	2.3	35	159.8	2.8	57
HS1095	ecnB	15.0	7.3	2	144	6.3	23	48.9	1.9	26	134.0	1.8	74
HS1099	yhjD	29.2	37.9	1	172	33.6	5	47.3	20.9	2	87.4	9.5	34
HS1100	yjiN	8.6	6.2	1	66.8	15.6	4	30.2	4.0	8	52.2	9.5	6

^aStrains were grown overnight in LB broth, subcultured, and maintained in the appropriate media for eight generations prior to sampling. Cultures were sampled at exponential phase (OD₆₀₀ was 0.3 for rich media and 0.2 for minimal media) and stationary phase (OD₆₀₀ was 1.5 for rich media and 1.3 for minimal media). Cultures of strain HS1072 grown in M9 glucose medium were supplemented with 20 μg of arginine/ml. Values are the averages of three independent experiments, and the coefficient of variation was less than 10%.

As expected for RpoS-dependent genes, in rich media, gene fusions in rpoS⁺ strains were expressed to much higher levels in stationary phase than in exponential phase (Table 2). β-Galactosidase activity was 2- to 66-fold higher in stationary phase. Expression levels for the fusions in an rpoS mutant background were generally low in exponential and stationary phases, confirming that the stationary phase dependence exhibited in the rpoS⁺ strains is due to the presence of RpoS rather than other regulatory factors.

Residual expression levels for several identified genes remained high in an rpoS mutant background. This was the case for rpoS mutants carrying fusions in ygaU, mltB, and yhjD (Table 2) and in the argH-oxyR intergenic region (Table 2) suggesting that these genes are not solely RpoS-dependent. Other regulators, in particular other sigma factors, may recognize the promoters of these genes and activate their transcription. The nucleotide sequence of promoters recognized by RpoS are similar to those recognized by RpoD, the housekeeping sigma factor for RNA polymerase, and in fact some RpoS-dependent promoters can also be activated by RpoD in vitro (11). In support of the hypothesis that these genes are regulated by other factors, gene fusions in rpoS mutant strains that had a high level of expression in stationary phase also tended to have a similarly high level of expression in exponential phase (Table 2). Many RpoS-dependent genes are regulated by other transcription factors (Table 3); for example, the promoters of genes ygaU and yhjD are known to be activated by Lrp (77).

TABLE 3.

RpoS-dependent genes and their known or predicted functions^d

Genea	Blattner no.	Functionb	Other regulator(s)	Previously reported as RpoS dependentc	RpoS dependence in:		Reference(s)
					Rich Medium	Minimal Medium
ybaY	b0453	Unknown		No	15	12
gabP	b2663	GABA transporter protein	GabC(−), Nac(+)	Yes	24	18	55, 66, 93
ugpE	b3451	G-3-P transporter membrane bound component	PhoBR(+)	No	6	6	15
ugpC	b3450	ATP binding protein of G-3-P transporter	PhoBR(+)	No	20	9	15
ydcS	b1440	Putative binding protein of ABC transporter	Nac(+)	Yes	6	2	93
yehX	b2129	Hypothetical ATP-binding protein of ABC transporter		No	28	21
narY	b1467	Beta subunit of nitrate reductase-II		Yes	10	2	13
appB	b0979	Subunit of cytochrome bd-II oxidase	AppR(+)	Yes	25	7	23
IdcC	b0186	Lysine decarboxylase		Yes	7	11	47, 79
aidB	b4187	Induced by alkylating agents; protects DNA from alkylation damage		Yes	36	33	45, 46
katE	b1732	Hydroperoxidase II	Lrp(+)	Yes	40	66	74
uspB	b3494	Universal stress protein B; required for ethanol resistance		Yes	4	8	26
osmY	b4376	Periplasmic protein induced by high osmolarity; function unknown	Lrp(−), Crp(−), IHF(−)	Yes	35	64	20, 90, 91
yhiV	b3514	Multidrug efflux pump	EvgA(+), Lrp(+)	No	31	4	56, 57, 74, 78
yhiU	b3513	Multidrug efflux pump	EvgA(+)	Yes	39	8	56, 57, 78
mltB	b2701	Membrane-bound lytic transglycosylase B		No	2	2	25
otsA	b1896	Trehalose phosphate synthase	Lrp(+)	Yes	31	57	72, 74
ecnB		Bacteriolytic lipoprotein entericidin B	EnvZ/OmpR(−)	Yes	23	74	12
argH	b3960	Argino-succinate lyase	ArgR(−)	No	2	5	77
aroM	b0390	Unknown; may play a role in aromatic amino acid biosynthesis	TyrR(+)	No	15	78	24
yhjY	b3548	Putative lipase		No	15	9
yqhE	b3012	2,5-diketo-d-gluconic acid reductase A; catalyzes the conversion of 2,5-KGDR to 2-KLG		No	8	11	92
aldB	b3588	Aldehyde dehydrogenase B; converts lactaldehyde and NAD+ to lactate and NADH		Yes	22	21	74, 87
gabD	B2661	Succinate-semialdehyde dehydrogenase	GabC(−), Nac(+)	No	16	13	55, 93
phnP	B4092	Membrane-bound subunit of carbon-phosphorus lyase complex	PhnF and PhnO(+)	No	8	2	51
talA	B2464	Transaldolase A; converts Sed-7-P and Gly-3-P to Fru-6-P and Ery-4-P	CreBC(+)	No	20	26	7, 70
yfcG	B2302	Putative glutathione S-transferase	SdiA(+)	No	37	52	84
yliI	B0837	Putative dehydrogenase		Yes	17	37	66
yjbJ	B4045	Unknown		No	22	179
yjbE	B4026	Unknown		No	22	58
ygaU	B2665	Unknown	Lrp(+)	No	4	12	74
ygdI	B2809	Unknown		No	7	5
ygaF	B2660	Unknown		No	3	7
yjgR	B4263	Unknown		No	6	3
ydaM	B1341	Unknown		No	16	23
ydcK	B1428	Unknown		No	15	9
yhiN	B3492	Unknown		No	5	7
yebF	B1847	Unknown		No	41	25
yjiN	B4336	Unknown		No	4	6
yeaG	B1783	Unknown	Lrp(+)	Yes	13	10	63, 74
yodC	B1957	Unknown		No	41	17
ybiO	B0808	Unknown		No	13	26
yhjG	B3524	Unknown		No	17	15
yphA	B2543	Unknown		No	24	18
yhjD	B3522	Unknown	Lrp(+)	Yes	5	34	63

^aGenes were identified using the BLAST algorithm (3).

^bFunctions were assigned to each gene based on either SWISSPROT or ECOCYC databases.

^cIn E. coli.

^dG-3-P, glyceraldehyde-3-phosphate; Sed-7-P, sedoheptulose-7-phosphate; Fru-6-P, fructose-6-phosphate; Ery-4-P, erythrose-4-phosphate.

Classification of RpoS-dependent genes.

All proteins encoded by the E. coli genome have been grouped into 22 classes based on their known or predicted functions (14). The 48 RpoS-regulated genes identified in this study encode proteins that fall into 12 of these functional classes (Table 4), many of them (17 genes) into the class of hypothetical, unclassified, and unknown proteins.

TABLE 4.

Classification of proteins encoded by the RpoS-dependent genes identified based on Blattner functional categories (14)

Functional category(ies)	No. of proteins in each category	Protein(s)
Cell structure	1	YbaY
Transport and binding proteins	3	GabP, UgpE, UgpC
Putative transport proteins	2	YdcS, YehX
Energy metabolism	3	NarY, AppB, LdcC
DNA replication, recombination, modification, and repair	1	AidB
Cell processes (including adaptation and protection)	8	KatE, UspB, OsmY, YhiV, YhiU, MltB, OtsA, EcnB
Amino acid biosynthesis and metabolism	2	ArgH, AroM
Fatty acid and phospholipid metabolism	1	YhjY
Carbon compound catabolism	2	YqhE, AldB
Central intermediary metabolism	4	GabD, PhnP, TalA, YfcG
Putative enzymes	1	Ylil
Hypothetical, unclassified, unknown	17	YjbJ, YjbE, YgaU, Ygdl, YgaF, YjgR, YdaM, YdcK, YhiN, YebF, YjiN, YeaG, YodC, YbiO, YhjG, YphA, YhjD

Cell processes (including adaptation and protection).

Of the 48 genes identified in this study, eight are in this class (cell processes), underscoring the importance of RpoS-dependent genes in adaptation. From the expression data (Table 2), it is clear that genes required for stress protection, such as otsA, osmY, and katE, are expressed highly in both minimal and rich media, especially in stationary phase, and are among the most highly RpoS dependent.

Two of the eight genes, yhiV and mltB, are newly identified as RpoS dependent in this study (Table 4). The yhiV gene along with yhiU (previously reported to be RpoS dependent [4]) codes for a multidrug efflux pump, overexpression of which increases resistance to a variety of antimicrobial compounds (56). In addition to RpoS, these are positively regulated by the response regulator EvgA of the EvgSA two-component regulatory system (57). The mltB gene codes for a membrane-bound lytic transglycosylase required for mucopeptide recycling during cell wall growth in E. coli (25).

Six genes of this class are known to be RpoS controlled. The katE gene encodes catalase (hydroperoxidase II), which converts highly reactive hydrogen peroxide into oxygen and water (81). The uspB gene codes for a 14-kDa protein that may play a role in sensing or changing membrane composition during stationary phase (26). Mutations in uspB render the cell sensitive to ethanol in stationary phase, and overexpression causes cell death in stationary phase (26). Mutations in the osmY gene, which encodes a periplasmic protein induced during high osmolarity, increase the susceptibility of E. coli to osmotic stress (91). The otsA gene forms a two-member trehalose biosynthetic operon with otsB (72). Biosynthesis of trehalose is activated in response to high osmolarity and upon entry into stationary phase (72). Trehalose can serve as both an osmoprotectant and as a carbon source (72).

The ecnB gene codes for a bacteriolytic lipoprotein known as entericidin B (12). Along with ecnA, ecnB forms an antidote/toxin gene pair termed an addiction module that induces programmed cell death of bacterial populations in stationary phase (12). Expression of the ecnB gene is induced by high osmolarity in stationary phase, positively regulated by RpoS (66) and negatively regulated by the EnvZ/OmpR two-component regulator (12).

Cell structure.

In stationary phase, changes occur in cellular morphology as well as in membrane and cell wall composition of E. coli (36). A distinctive feature of stationary-phase cells is an oval cell morphology, a consequence of RpoS-mediated induction of bolA (65). RpoS also regulates other cell-structure-related genes, such as ftsQAZ, required for cell septum formation (69); csgAB, required for the formation of curli fimbriae (59); and cfa, required for cyclopropane fatty acid synthesis, which accumulates in the cell membrane during stationary phase and has been implicated in acid resistance (19).

Expression of ybaY, encoding a cell-structure-related protein required for glycolipid/polysaccharide metabolism (40), was also found to be RpoS dependent in this study.

Transport and binding proteins.

Operon fusions to two new operons belonging to the transport and binding class were identified, ugpBAECQ and ydcSTUV. ugpC and ugpE are members of the ugpBAECQ operon, which codes for a binding protein-dependent ATP-binding cassette (ABC) transporter (15). The ugp operon is also a member of the Pho regulon that is regulated by PhoBR in response to external phosphate limitation (83). UgpB to UgpQ are required for phosphate scavenging and for utilization of glycerol-3-phosphate as a sole carbon source (15), suggestive of a role for RpoS in nutrient scavenging.

The ydcS gene encodes a putative binding protein-dependent ABC transporter (40) that is similar in amino acid sequence to polyamine transporters PotD and PotF. Although the function of this gene has not been confirmed, microarray analysis has revealed that ydcS, and the entire ydcSTUV operon, is positively regulated by the nitrogen assimilation control protein Nac (93), which induces genes required for nitrogen scavenging under nitrogen-limited conditions. The gabP gene, previously reported to be RpoS controlled (66), codes for a γ-aminobutyric acid (GABA) permease (55).

DNA replication, recombination, modification, and repair.

In stationary phase, the E. coli chromosome is stabilized by nucleoid-associated proteins, including Dps (86) and H-NS (5). Dps is one of the most abundant proteins during stationary phase, and the dps gene is positively controlled by RpoS (2). Dps cocrystallizes with DNA, forming a ferritin-like structure which effectively deprives the cell of iron required for the Fenton reaction, which can lead to oxidative damage (86). H-NS, another nucleoid-associated protein that is controlled by RpoS (10), compacts DNA (22) and can modulate gene expression (5). Under starvation conditions, endogenous DNA-methylating agents are produced in E. coli by nitrosation reactions that lead to methylation of DNA (76). Methylated DNA mispairs with thymine during DNA replication, leading to GC-to-AC transition mutations (76). There are various mechanisms to counteract DNA damage during stationary phase. One of these is by the induction of the adaptive response by the Ada protein that acts as a methyl transferase; in its methylated form, Ada acts as a transcriptional regulator of the Ada operon genes ada, aidB, alkA, and alkB (44). The aidB gene is known to be RpoS dependent (80). Expression of the aidB gene is also induced by alkylating agents, anaerobiosis, and acetate at acidic pH (45). Though the exact function of AidB is not known, it may neutralize nitrosoguanidines or their intermediates (46). Microarray analysis of mitomycin C-treated E. coli cells has revealed that numerous stationary-phase genes, including that encoding RpoS, are induced by mitomycin C treatment, confirming the role of RpoS in alleviating DNA damage (42).

Energy metabolism.

E. coli, a facultative anaerobe, has two terminal oxidases, cytochrome bo, encoded by the cyo operon, and cytochrome bd, encoded by the cyd operon (6). The cyo operon operates under conditions of high oxygen concentration and has low affinity to oxygen, while cyd has high affinity to oxygen and operates under microaerobic conditions (28). A third cytochrome oxidase encoded by appB has been identified (73) and was found (this study and reference 23) to be RpoS dependent. The appB gene is in an operon with appA (encoding pH 2.5 acid phosphatase) and encodes a cytochrome bd-type oxidase (73). In addition to RpoS, appB is regulated by AppY (6), which itself is subject to regulation by RpoS (6). What might be the physiological significance of RpoS upregulation of cytochrome oxidases during stationary phase? In stationary phase, the density of cells in a culture increases, and, as a consequence, the amount of oxygen available to the cells is reduced. Under these microaerobic conditions, upregulation of cytochrome bd II and AppB, which have a high affinity for oxygen, may be required for efficient electron transport.

In the absence of oxygen as an electron acceptor, E. coli preferentially uses nitrate as an electron acceptor (28). E. coli has three functional nitrate reductases, one periplasmic nitrate reductase (NapA) and two membrane-bound nitrate reductases, nitrate reductase A (NRA) and nitrate reductase Z (NRZ) (71). Nitrate reductase Z is encoded by members of the narZYWV operon (71). The RpoS-dependent narY gene codes for the beta subunit of nitrate reductase Z, which functions as an electron acceptor during anaerobic growth on nitrate in E. coli (18).

The ldcC gene, encoding lysine decarboxylase C (47), which is classified as an energy metabolism function in E. coli (14), was previously shown to be RpoS dependent (89). LdcC catalyzes the decarboxylation of lysine to cadaverine (47) and may be necessary for utilization of lysine as a nutrient source in stationary phase.

Amino acid biosynthesis and metabolism.

Two amino acid biosynthetic genes, argH and aroM, were found to be RpoS regulated. The argH gene, the terminal member of the argCBH operon, codes for arginino-succinate lyase and is required for arginine biosynthesis in E. coli (21). Arginine biosynthesis is likely important during stationary phase, because this amino acid is a precursor for synthesis of polyamines (29), which protect DNA from oxidative damage by scavenging free radicals (30). Other arginine metabolic functions are controlled by RpoS, including the catabolic astCADBE operon, which is required for growth in arginine as a sole nitrogen source (43). This operon is also controlled by RpoN (43). Some of the well-known amino acid catabolic operons that have protective functions in E. coli are arginine decarboxylase (encoded by adiA) and glutamate decarboxylase (encoded by gadABC) (17). These operons are part of the acid resistance (AR) system in E. coli and are regulated by RpoS (17).

The aroM gene, which is highly RpoS dependent, particularly in minimal medium (Table 2), forms an operon with aroL that codes for shikimate kinase II (24). This enzyme is required for the conversion of shikimate to shikimate-3-phosphate, which is ultimately converted to chorismate by the aroA and aroC gene products (61). Chorismate is the branching point for the synthesis of three aromatic amino acids: tyrosine, phenylalanine, and tryptophan (61). Though aroM is cotranscribed with aroL, the function of AroM is not unknown (24).

Carbon compound catabolism.

Two RpoS-dependent genes identified in this study, yqhE and aldB, are in the carbon compound catabolism class. The aldB gene, previously identified as RpoS dependent (87), codes for a broad-spectrum aldehyde dehydrogenase that converts aldehydes to their corresponding acids. AldB may thus detoxify aldehydes produced during stationary phase (87). The yqhE gene, found to be RpoS dependent in this study, encodes 2,5-diketo-d-gluconic acid reductase A (92). This enzyme is part of the ketogluconate utilization pathway that converts 2,5-diketo-d-gluconate into 2-keto-l-gluconate, which can then be converted to d-gluconate and further catabolized through the Entner-Doudoroff and pentose-phosphate pathways (92).

Other catabolic genes regulated by RpoS include treA (34) and glgS (33). In addition to its role as an osmoprotectant, trehalose, synthesized by the products of otsBA (39), is converted to glucose by trehalose, which is encoded by treA (72), and thus can be used as a carbon source. The glgS gene is required for glycogen synthesis and is under the positive control of RpoS (33). Carbon-starved E. coli cells accumulate glycogen, which acts as a carbon reserve during starvation (36). Regulation by RpoS has been proposed to provide an alternative mechanism of carbon catabolite control in E. coli (64), especially in stationary-phase metabolism.

Central intermediary metabolism.

We found four genes, gabD, phnP, talA, and yfcG, which fall into the central intermediary metabolism functional group, none of which have been previously reported to be RpoS dependent. The gab operon (gabDTP) is required for conversion of gamma-amino butyric acid (GABA) into glutamate and succinate (55). Succinate and glutamate thus produced can be utilized as a carbon source via the trichloroacetic acid (TCA) cycle. Glutamate protects cells against acid stress, and this pathway may be the only mechanism by which internal glutamate pools are generated (35). The phnP gene, along with 13 other genes, codes for phosphonate transport and catabolism (88). Although E. coli cannot utilize phosphonates as carbon sources, they can be used as a sole phosphorus source (88).

The talA gene codes for transaldolase A of the pentose phosphate pathway and forms an operon with tktB (70). Intermediates of the pentose phosphate pathway serve as precursors for the biosynthesis of aromatic amino acids and also for heptoses found in lipopolysaccharides (70). The reason why talA is regulated by RpoS is not clear; however, talA may replenish the glycolytic pathway during stationary phase by providing the necessary intermediates from the pentose phosphate pathway. The yfcG gene codes for a putative glutathione S-transferase (GST) (40) and is highly RpoS dependent, especially in minimal medium (Table 2). In eukaryotes, GSTs function as detoxifying enzymes (82). The role of GST in prokaryotes is not well characterized, but it has been implicated in the degradation of xenobiotics (82).

Hypothetical, unclassified, unknown.

Of the 48 RpoS-dependent genes identified in this study, 17 have unknown or hypothetical function. It is not surprising to find a majority of genes in this class, because 30% of E. coli genes do not yet have an assigned function (14). The yjbJ gene has a very high RpoS dependence in stationary phase, especially when grown in minimal media (Table 2), and is one of the most abundant proteins produced in stationary phase (48). Moreover, this gene has homology similarities to csbD of Bacillus subtilis, which is also regulated by a stress sigma factor (1). Some of the RpoS-dependent genes of this class have recently been shown to be upregulated under different growth conditions. The yeaG gene is induced in artificial seawater medium (63), while ybaF is induced under in vivo conditions in mice (41). Though the physiological functions of these genes were not revealed by these studies, the growth conditions mentioned above are stressful conditions, emphasizing that some of these RpoS-regulated genes may have a role in adaptation or in counteracting stress.

DISCUSSION

In this study, we employed a mutational screen to identify genes that require RpoS for expression. Because this screen did not rely on a phenotypic property of the identified gene, the class of genes identified is likely less biased than might be obtained by using other screens dependent on phenotypic differences (such as carbon starvation or osmotic shock). This was found to be important, as many of the genes identified have paralogous functions whose activity may be masked by the presence of another gene product (e.g., NarG or NarZ). There are several published RpoS-dependent paralogs, including NarZ (nitrate reductase) (18) and Ldc (lysine decarboxylase) (89), that comprise only a small percentage of the cell's total activity. There are many other paralogous functions that are controlled by RpoS, and collectively this suggests that expressing a subset of key duplicated metabolic functions is an important physiological imperative in stationary-phase cultures. These functions include polyamine transport (through PotGHIJ), balancing the activities of glycolytic and oxidative pentose pathways (through TalA/TktB), acid resistance (through lysine decarboxylase), drug efflux (YhiUV), and the use of alternative electron acceptors (NarZ).

In some cases, RpoS-regulated functions may form redundant or alternate metabolite cycles. For example, both arginine biosynthetic (argH, this study) and catabolic genes (astCADBE [8, 27]) are RpoS dependent (albeit weakly). Similarly, the RpoS-dependent gabDTP operon (52) with glutamate decarboxylases (gadAB [17]) could comprise a cycle that produces fumarate from succinate, similar to the succinate dehydrogenase step of the TCA cycle but with less energy conserved as chemical-reducing power. In both cases, concomitant decarboxylation likely confers acid resistance. Trehalose synthesis and catabolism, encoded by RpoS-dependent treA and otsAB, respectively, may allow the cell to accumulate a compatible solute for catabolic purposes when the cell is starved of preferred carbon sources.

The relatively large size of the RpoS regulon has made its complete characterization difficult. There are several approaches by which members of a regulon can be characterized. Two-dimensional gel electrophoresis can be used to identify differentially expressed proteins based on mass and charge (58). Limitations to this approach include difficulty in identifying proteins that are expressed at very low levels, masking of one protein spot by another, leading to an underestimation of the total number of proteins separated, and loss of detectable protein spots from treatment of the gels under denaturing conditions (58).

Microarray and cDNA macroarray analyses can also potentially be used to identify members of a large regulon like RpoS. To date, however, there have been only a few reports where these tools have been used to identify regulon members by comparing RNA expression profiles of wild-type and isogenic regulator-minus mutants (60, 74). Such methodology has identified several genes that are dependent on Lrp in stationary phase (74). Interestingly, the majority of these were previously identified as RpoS dependent. Here, we extend the number of Lrp-regulated genes identified that are also regulated by RpoS to include talA, ydcS, and yjbJ. Array methodology can be employed to examine gene expression under different growth conditions, and this may also indirectly provide information about a regulon. Macroarrays have been used (75) to identify many genes that are upregulated in minimal medium as opposed to rich media. As expected, many of these, especially those that are involved in adaptation and protection, are RpoS dependent (75). Array analysis, though exhaustive, yields expression results that are more conservative than either Northern analyses or reporter fusions (unpublished data). The results of this study should be a useful independent measure to validate future array studies.

Unlike microarray analysis, the genetic screening technique used in this study provides a direct estimate of gene expression levels, because the expression of the lacZ reporter gene is primarily a function of the promoter activity of the gene into which lacZ is inserted (16). Moreover, as a reporter, lacZ is a sensitive indicator of gene expression. Genes with low expression levels can be detected easily by β-galactosidase assays, and direct sequencing of the fusion junctions makes this technique useful (16). In addition, insertion of lacZ into a gene generally inactivates the gene, which can then be exploited to study gene function. Many E. coli strains used in expression studies, including GC4468 and its derivatives, carry a very large deletion (approximately 80 kb) in the region of the lac operon, beginning at 6.2 min. Thus, genes in this region, including any that may be regulated by RpoS, cannot be studied in these strains. These drawbacks notwithstanding, we have employed this technique successfully to characterize genes of the RpoS regulon and have identified 48 genes under RpoS control.

In summary, we have identified a large number of genes that are dependent on RpoS for expression, particularly in stationary phase in cultures grown in rich media. The functions of the identified genes are consistent with previous suggestions that the RpoS regulon is required for nutrient scavenging, pH homeostasis, and protection from oxidative stress. Further insight into the mechanisms by which bacterial cells cope with suboptimal conditions may be gained from learning the functions of the genes identified here whose physiological role is not yet known. RpoS and the genes that it controls are conserved among many gram-negative bacteria; therefore, such studies are likely to reveal fundamental information regarding the adaptive physiology of other bacteria in addition to E. coli.