Metabolic Context and Possible Physiological Themes of ς⁵⁴-Dependent Genes in Escherichia coli

Abstract

ς⁵⁴ has several features that distinguish it from other sigma factors in Escherichia coli: it is not homologous to other ς subunits, ς⁵⁴-dependent expression absolutely requires an activator, and the activator binding sites can be far from the transcription start site. A rationale for these properties has not been readily apparent, in part because of an inability to assign a common physiological function for ς⁵⁴-dependent genes. Surveys of ς⁵⁴-dependent genes from a variety of organisms suggest that the products of these genes are often involved in nitrogen assimilation; however, many are not. Such broad surveys inevitably remove the ς⁵⁴-dependent genes from a potentially coherent metabolic context. To address this concern, we consider the function and metabolic context of ς⁵⁴-dependent genes primarily from a single organism, Escherichia coli, in which a reasonably complete list of ς⁵⁴-dependent genes has been identified by computer analysis combined with a DNA microarray analysis of nitrogen limitation-induced genes. E. coli appears to have approximately 30 ς⁵⁴-dependent operons, and about half are involved in nitrogen assimilation and metabolism. A possible physiological relationship between ς⁵⁴-dependent genes may be based on the fact that nitrogen assimilation consumes energy and intermediates of central metabolism. The products of the ς⁵⁴-dependent genes that are not involved in nitrogen metabolism may prevent depletion of metabolites and energy resources in certain environments or partially neutralize adverse conditions. Such a relationship may limit the number of physiological themes of ς⁵⁴-dependent genes within a single organism and may partially account for the unique features of ς⁵⁴ and ς⁵⁴-dependent gene expression.

INTRODUCTION AND OVERVIEW

Scope

The dissociable sigma subunits of RNA polymerase are responsible for specific binding to DNA and are therefore important determinants of differential gene expression (68, 170). ς⁵⁴ was discovered during an analysis of glutamine synthetase and nitrogen assimilation in enteric bacteria (71, 77). Subsequent studies have confirmed its role in nitrogen assimilation but have also shown that it is involved in a variety of seemingly unrelated functions, such as carbon source utilization, certain fermentation pathways, flagellar synthesis, and bacterial virulence (97, 151). There have been several reviews about ς⁵⁴ and ς⁵⁴-dependent activation, including recent ones (14, 28, 97, 110, 113, 148, 151). However, discussions of ς⁵⁴ function usually consider genes from several organisms and therefore remove ς⁵⁴ from the metabolic or physiological context of a single organism. The primary purpose of this review is to redress this imbalance and to discuss the functions of the known ς⁵⁴-dependent genes from a single organism, Escherichia coli. The most meaningful discussion of ς⁵⁴ function requires a complete set of ς⁵⁴-dependent genes. Two sources provide information on likely ς⁵⁴-dependent promoters: a recent DNA microarray analysis of transcripts present during nitrogen-limited growth (176) and computer analysis of potential ς⁵⁴-dependent promoters, which are readily identified from the completed nucleotide sequence of the E. coli genome (23). Results of these analyses are integrated into the discussion of the function of the known ς⁵⁴-dependent genes. The central thesis of this review is that the ς⁵⁴-dependent genes of E. coli have only a few metabolic themes and that these themes may be related.

Sigma Subunits and Their Function in E. coli

E. coli has seven ς subunits, and each has a distinct function. ς⁷⁰ is considered the primary sigma factor. Core RNA polymerase (E) associated with ς⁷⁰ initiates transcription of housekeeping genes (68). Eς⁷⁰ also initiates the transcription of nonessential genes that are induced in specialized environments. ς^S has been called either an alternative sigma factor or a second primary sigma factor (69). Although Eς^S binds the same sequences as Eς⁷⁰ (unpublished results cited in reference 69), ς^S is considered a general stress factor since it is associated with a variety of growth-impairing stresses: nutrient depletion, oxidative stress, high temperature, high osmolarity, acidic pH, or exposure to ethanol (69). ς³² and ς^E are also associated with stress. ς³² is required for the response to damage of cytoplasmic proteins, which is most commonly associated with heat shock, and ς^E controls the response to extracytoplasmic or extreme heat stress (174). ς^FecI and ς²⁸ (FliA) are required for synthesis of the ferric citrate transporter and flagella, respectively (8, 115). As mentioned above, ς⁵⁴ is usually associated with nitrogen assimilation.

Unique Features of ς⁵⁴-Dependent Transcription

The ς factors in E. coli are homologous to ς⁷⁰, except for ς⁵⁴ (110). Not surprisingly, ς⁵⁴-dependent transcription has several distinctive features (reviewed in reference 28). Core RNA polymerase (E) complexed to a ς⁷⁰-like factor can be sufficient for open promoter complex formation. In contrast, Eς⁵⁴ catalyzes strand separation only with the help of a distinct class of transcriptional activators. A consequence of this property is that transcription can be completely turned off. Such absolute control may account for the evolutionary persistence of ς⁵⁴ (see “Physiological function of ς⁵⁴” below).

The activators of Eς⁵⁴-dependent genes are unusual. (The individual E. coli activators are discussed in a separate section.) Unlike most eubacterial transcriptional activators, the ς⁵⁴-dependent activators bind to sites that are effective regardless of distance and orientation (28, 29, 131). In this respect, the activator binding sites are analogous to eukaryotic enhancers, and the activators are often called enhancer binding proteins. However, the analogy to eukaryotic enhancer binding proteins is not precise, since the bacterial proteins do not enhance basal transcription but, instead, are absolutely required for any transcription. The activators interact with Eς⁵⁴ from these binding sites. This interaction sometimes requires a DNA bending protein. The DNA bending proteins that participate in ς⁵⁴-dependent gene expression in E. coli are integration host factor (IHF) and ArgR, the arginine repressor (72, 103). When transcription from a ς⁵⁴-dependent promoter does not require a DNA bending protein, DNA curvature facilitates the interaction between the activator and RNA polymerase (32). Another distinctive feature of ς⁵⁴-dependent activators is an essential ATPase activity (166).

Control of ς⁵⁴-Dependent Promoters

The most important control of ς⁵⁴-dependent genes is through modulation of the activator's ATPase activity. The ς⁵⁴-dependent activators usually contain a regulatory domain that controls ATPase activity. Several mechanisms control the interaction of the regulatory domain with the ATP binding domain: phosphorylation, interaction with a low-molecular-weight ligand, or interaction with one or several regulatory proteins (148). Some control mechanisms are extremely complex, such as that for PspF (discussed below). Variations in ς⁵⁴ activity, either by ligand binding or by covalent modification, are not known in E. coli. Furthermore, the intracellular level of ς⁵⁴ is apparently constant (87), and expression of rpoN, which specifies ς⁵⁴, appears to be constitutive (33). Strain W3110 contains about 700 molecules of ς⁷⁰ per cell and 110 molecules of ς⁵⁴, whereas strain MC4100 may contain only about 285 molecules of ς⁷⁰ and as few as 13 to 22 molecules of ς⁵⁴ (87). The low level of ς⁵⁴ could have regulatory implications. For example, it is possible that different ς⁵⁴-dependent operons compete for limiting ς⁵⁴. This possibility is even more plausible if there is some physiological relationship between the ς⁵⁴-dependent genes. One purpose of this review is to explore this issue.

COMMON FEATURES OF ς⁵⁴-DEPENDENT PROMOTERS

There are 11 ς⁵⁴-dependent promoters for which the transcription start site has been determined in vivo, in vitro, or both. These promoters precede astCADBE, fdhF, glnALG, glnHPQ, hycABCDEFGHI, hydN-hypF, hypABCDE-fhlA, prpBCDE, pspABCDE, rtcBA, and ygjG. It is virtually a certainty that atoDAEB, glnK-amtB, nac, zraP, and zraSR (hydHG) have a ς⁵⁴-dependent promoter, since their expression absolutely requires ς⁵⁴ and their promoter regions contain an easily recognizable site for Eς⁵⁴. Also, argT-hisJQMP has a verified ς⁵⁴-dependent promoter in Salmonella enterica serovar Typhimurium, and nitrogen limitation induces these genes in E. coli, which suggests that this operon possesses a ς⁵⁴-dependent promoter.

We will use the promoters for these 17 operons to characterize the ς⁵⁴-dependent promoters in E. coli (Table 1). These promoters suggest an apparent consensus of aaN₃TGGCAcN₆TGCNNt, where small letters indicate two to five mismatches, capital letters denote zero or one mismatch, and underlined bases have no mismatches. By similar criteria, the consensus derived from 186 promoters from a variety of organisms is N₅tGGcacN₅ttGC (14), which is similar but not identical to the apparent E. coli consensus. Figure 1 shows the locations of the ς⁵⁴-dependent operons within the context of neighboring genes, the size of the transcripts, and the direction of expression. Figure 2 shows the binding sites for Eς⁵⁴ in relation to the binding sites for the ς⁵⁴-dependent activators (when known) and the nearest upstream structural gene.

TABLE 1.

E. coli operons with confirmed ς⁵⁴-dependent promoters

Operon	Regulator	Function	Promoter sequencea
argT-hisJQMP	NRI-NRII	Arginine, histidine transport	CAAGA TGGCA TAAGA CCTGC ATGAA AGAG
astCADBE	NRI-NRII-ArgR	Arginine catabolism	CTGGC TGGCA CGAAC CCTGC AATCT ACAT
atoDAEB	AtoC-AtoS	Acetoacetate catabolism	CATTC TGGCA CTCCC CTTGC TATTG CCTG
fdhF	FhlA	FDH	AAATG TGGCA TAAAA GATGC ATACT GTAG
glnALG	NRI-NRII	Nitrogen assimilation	AAAGT TGGCA CAGAT TTCGC TTTAT CTTT
glnHPQ	NRI-NRII	Glutamine transport	AAAAC TGGCA CGATT TTTTC ATATA TGTG
glnK-amtB	NRI-NRII	PII paralog, NH3 transport	ATTTC TGGCA CACCG CTTGC AATAC CTTC
hycABCDEFGHI	FhlA	Hydrogenase	AAAGT TGGCA CAAAA AATGC TTAAA GCTG
hydN-hypF	FhlA	Hydrogenase	AAAGA TGGCA TGATT TCTGC TGTCA GAAA
hypABCDE-fhlA	FhlA	Hydrogenase	AACAC TGGCA CAATT ATTGC TTGTA GCTG
nac	NRI-NRII	Nitrogen assimilation	AAAAC TGGCA AGCAT CTTGC AATCT GGTT
prpBCDE	PrpR	Propionate catabolism	AATTG TGGCA CACCC CTTGC TTTGT CTTT
pspABCDE	PspF	Phage shock and other stresses	AAAAT TGGCA CGCAA ATTGT ATTAA CAGT
rtcBA	RtcR	RNA terminal cyclase	TTTTC TGGCA CGACG GTTGC AATTA TCAG
ygjG	NRI-NRII?	Uncharacterized transaminase	CGGAG TGGCG CAATC CCTGC AATAC TTAA
zraP	ZraR-ZraS	Zinc tolerance	ATCGT TGGCA CGGAA GATGC AATAC CCGA
zraSR (hydGH)	ZraR-ZraS	Regulators of zinc tolerance	AAAGA TGGCA TGATT TCTGC TGTCA CCGA

^aResidues that are conserved 100% are shown in bold type. 

FIG. 1.

ς⁵⁴-dependent genes in E. coli. The known ς⁵⁴-dependent operons and their transcripts are shown in the context of neighboring genes. The open and solid boxes indicate counterclockwise and clockwise transcription, respectively. If the gene has been assigned a function, this is indicated by the gene name underneath. Boxes without gene names indicate that the gene has an unknown function. The sizes of the genes and their intergenic regions are to scale.

FIG. 2.

Regulatory regions of the characterized ς⁵⁴-dependent genes. The following features are shown for each of the ς⁵⁴-dependent promoters: binding sites for activators, DNA bending proteins (when required), and RNA polymerase. Solid boxes indicate that the binding site has been demonstrated. A hatched box indicates a confirmed binding site, but binding is weak. An open box signifies a proposed binding site. The relative location and orientation of the nearest upstream gene are also shown. All diagrams are to the same scale, except for the hyp operon.

Several generalizations can be made concerning these promoters. First, all the known ς⁵⁴-dependent promoters are located outside the structural genes. This does not necessarily mean that an authentic binding site for Eς⁵⁴ will not be found within a gene, but it does mean that such binding sites will be the exception. Second, the activator binding sites are also outside the structural genes. The most spectacular example is the FhlA binding site for the hyp operon in the hyp-hyc regulatory region. The entire hycA gene is located between the binding sites for FhlA and RNA polymerase (Fig. 2). Even though the FhlA binding site for the hyp operon is within the hyc operon, it is not within a structural gene. Instead, it is completely located within the 130-base intergenic region between hycA and hycB. Third, the activator binding sites are almost always a significant distance from the adjacent upstream gene and therefore do not apparently interfere with expression of these genes (Fig. 2). The only exception is the binding site for PspF, which activates the pspABCDE operon. The PspF sites for activation overlap with the promoter for the pspF gene (Fig. 2), and it has been proposed that this has physiological significance (see “psp operon and phage shock response” below). The average size of the intergenic region that contains a known ς⁵⁴-dependent promoter is 267 ± 106 bases, with a range from 148 to 507 bases. This distance is apparently large enough for binding sites for both Eς⁵⁴ and an activator. Fourth, the distance from the 3′ end of the Eς⁵⁴ binding site to the nucleotides coding for the initiation codon is, on average, 50 bases. Such a short distance reduces the potential for RNA secondary structures or protein binding regions near the translational start site and therefore reduces the potential for translational control. Finally, the average A+T content of the 50 bases just upstream from the Eς⁵⁴ binding site is 70%, and no upstream region has an AT content less than 50%. Two rationales for such a bias can be suggested. Some ς⁵⁴-dependent promoters require a DNA bending protein. Both of the DNA bending proteins that facilitate the activation of ς⁵⁴-dependent genes in E. coli, IHF and ArgR, have an AT-rich consensus sequence. Alternately, ς⁵⁴-dependent promoters that do not require a DNA bending protein often have an intrinsic bend between the activator and RNA polymerase binding sites (32), and this may favor AT-rich regions. In either case, the requirement for long-range protein-protein interactions appears to bias the base composition of DNA just upstream from the Eς⁵⁴ binding site.

Despite the reasonably uniform properties of the known ς⁵⁴-dependent promoters, there is little useful information for promoter prediction. For uncharacterized promoters, the activator binding sites will probably not be known, and there may be uncertainty in the size of intergenic regions if the protein coding regions have been misidentified (which is not uncommon). A more useful predictor of potential promoters is based on computer identification of binding sites for Eς⁵⁴, which is described in the next section.

COMPUTER IDENTIFICATION OF POTENTIAL ς⁵⁴-DEPENDENT PROMOTERS

Site Identification and the Problem of False Positives

The most meaningful analysis of the physiological function of ς⁵⁴ requires a comprehensive set of ς⁵⁴-dependent genes. One method to help determine the complete set of such promoters is a computer analysis of potential ς⁵⁴-dependent promoters from the completed E. coli genome sequence. We used the SeqScan program (B. T. Nixon, Department of Biochemistry and Molecular Biology, Pennsylvania State University [http://www.bmb.psu.edu/seqscan]) for this analysis. This program uses 86 ς⁵⁴-dependent promoters from several organisms to define a consensus sequence that does not differ substantially from that derived from the 17 known E. coli promoters (Table 2). The program uses a weighted matrix to give sites a score from 0 to 100, and then reports sites with a score higher than 60.

TABLE 2.

Comparison of consensus sequence for ς⁵⁴-dependent promoters from known E. coli promoters and the SeqScan program^a

Source	Sequence
E. coli promoters	A71 A71 NNN T100 G100 G100 C100 A94 C71(A/G)94 NNNN T52T94 G94 C94 (A/T)100 (A/T)88 T71
SeqScan program	T90  G100 G100 C81  A84 C70 (A/G)87 NNNN T72 T86 G99C90 (A/T)86

^aA base-by-base comparison of the frequency distributions of the consensus sequences is shown. N indicates any base. Residues with greater than 90% conservation are shown in bold type. 

The SeqScan program identified approximately 8,000 potential ς⁵⁴-dependent promoters from the E. coli genome, or about 1 for every 600 bases. Clearly, there are a number of false positives. We can use the properties of the known ς⁵⁴-dependent promoters to eliminate many of these false positives. We eliminated all the sites within structural genes, which reduced the number of potential sites over 30-fold. The 17 known ς⁵⁴-dependent promoters were in the set of 213 intergenic sites in which the potential promoter transcribed a gene in the correct direction. We found that 121 potential intergenic Eς⁵⁴ binding sites transcribe genes in the wrong orientation. Table 3 and its footnotes list all of the intergenic sites that could potentially transcribe a gene in the correct direction.

TABLE 3.

Computer ranking of intergenic ς⁵⁴ sites^a

Typeb	Score	Gene	Ntr inductionc	Distanced		Function
				Down	Up
X	95.4	rtcBA	—	38	134p	RNA terminal phosphate cyclase
	92.8	rpoH	—	40	188t	Heat shock sigma factor
X	90.2	glnK-amtB	10–20	52	112t	Alternate Ntr regulator and ammonia transport
X	89.6	prpBCDE	—	46	176p	Propionate catabolism
(X)	88.6	yhdWXYZ	2–9	187	334t	Putative amino acid transport operon
X	88.6	zraP	2	34	196p	Zinc tolerance
(X)	88.0	gltlJKL	3–6	?	99t	Interrupted glutamate-aspartate transport genes
X	88.0	hypABCDE-fhlA	—	37	165p	Formate hydrogen lyase
	86.0	b2710-ygbD	—	47	48p	Flavodoxin, oxidoreductase
X	85.9	atoDAEB	?	45	134t	Acetoacetate metabolism
X	85.5	astCADBE	7–11	72	357p	Arginine degradation
	85.4	kch	—	264	20t	Potassium channel
X	85.2	nac	15–50	55	255p	Regulator of nitrogen assimilation genes
	83.9	yfhKGA	—	190	288t	Two-component regulatory system
(X)	83.8	yeaGH	2–4	103	316p	Possible operon, unknown function
X	83.3	glnALG	11–24	83	273p	Nitrogen assimilation and regulation of Ntr response
X	80.7	glnHPQ	5–9	54	333t	Glutamine transport
X	80.0	pspABCDE	1.5–2.5	51	84p	Phage shock
X	79.1	hycABCDEFGHI	—	36	159p	Formate hydrogen lyase
	78.8	ybhK	—	55	325p	RocR-like protein
	77.3	yaiS	—	33	656p	Hypothetical protein, 136 residues
(X)	76.8	b1012-1006	24–47	28	201p	Possible pyrimidine catabolic operon
(X)	76.6	ddpXABCDE	52–60	44	198t	d-Ala–d-Ala dipeptide transport and dipeptidase
	76.5	topA	—	201	162p	Topoisomerase I
X	76.2	zraSR (hydHG)	—	40	190p	Two-component regulatory system for zinc tolerance
X	72.6	fdhF	—	51	130t	Formate hydrogen lyase
X	72.3	argT-hisJQMP	10–18	68	181t	Arginine and histidine transport
X	70.7	ygjG	3–6	136	355p	An omega-transaminase
(X)	68.1	chaC	4–5	74	46t	Ca/H antiporter, in chaBC intercistronic region?
X	64.5	hydN-hypF	—	37	95t	Formate hydrogen lyase
(X)	63.9	potFGHI	5–6	59	275t	Putrescine transport
(X)	62.2	ycjJ	3	192	39t	Putative amino acid transporter

^aThere are over 200 sites with a computer score over 60. This table provides details of all sites with (i) verified ς⁵⁴-dependent promoters, (ii) potential sites before genes that are induced by nitrogen limitation, and (iii) sites with a score of 76 or higher (there are only seven such sites that fail to meet either criterion a or b). These criteria might exclude some authentic ς⁵⁴-dependent promoters. The following is a list of all the other intergenic sites. The following genes have sites with a score from 70.0 to 75.9 (from highest to lowest): yigL, ygeW, b2343, b2878, yjcB-vacB, ytfJ, crl, acrD, ptrA, clpPX, mutH, csgBA, b2878, pyrG, sohA, yehR, rmf, mscL, mdoB, gcvTHP, intF, and yfaO. The following sites have a ranking between 65.0 and 69.9: b1983, b2670, yhcC, yabI, malK-lamB-malM, hyfABCDEFGHIR, sgcCQ, yehZYXW, yejH, metZ, yhfA, ybcLM, proP, purC, ygiF, b1440-1444, yeaQ, yigM; b0833-0834, yefM, gdhA, rfe-wzzE, emrD, cysDNC, gcvA, yhfZY, ychH, b2380-2382, caiF, xseB-ispA-dxs, hcaA1, xdhABC (b2866-2868), ndh, pgpB, b1722, tolC, galETKM, ymcC, degQ, yheB, b0540, aer, glyQS, cadCBA, b2374, arsR, mviN, yieP, b2420, and ygdP-ptsP. The following sites have a ranking between 60.1 and 64.9: ybgE, fumB, b0805, ompF, ygdH, accBC, yciK-btuR, brnQ, acs, yceC-yceF, purM, ybcLM, b1017, kbl-tdh, b1625, yjhQP, aroG, zipA, yiiP, cybC, yjjJ, dinG, prsA, ydcD, emrD, secE, mog, ybiH, b0836, ymdD, dnaA, sapABCDF, slyA, b1826, sanA, yjdE, intB, hyaABCDEF, yhcL, yjhA, b2444-2445, tktA, bglX, yafW, dsbB, yfjZ-ypjF, yicK, panD, ycbG, rplY, yigK, yafK, yahA, araE, dfp, yfcD, feaB, gnd, yhhI, ybiA, dicF, b3000-b2999, yhaM, ydiQ, yagA, yhfA, ilvGMEDA, adiY, yjgL, trpR, osmB, sgaT, yjhQ, ydiQ, yecH, upp-uraA, rfe-wzzE (again), map, ybiT, trpEDCBA, b1592, fepB, yrbG, cydA, ssrA, yhcC, yhgN, yhjR, metA, talB, lpxC, b1432, hisS, mreB, rrfF, xylB, pfkA, hslVU, yjhR, ahpF, b1506, yihG, sbp, nhaA, proBA, hrpA, nirC-cysG, and yigK. 

^bX, the site is a verified ς⁵⁴-dependent promoter; (X), a probable ς⁵⁴-dependent promoter as suggested by computer identification of an appropriately located site and induction by nitrogen limitation. 

^cNtr induction refers to the elevation of expression induced by nitrogen limitation as determined by microarray analysis. Dashes indicate no induction. The number indicates the range of the increase for the most highly induced gene of the operon. 

^d“Down” indicates the distance from the 3′ end of the ς⁵⁴ site to the first base of the downstream structural gene. “Up” indicates the distance from the 5′ end of the ς⁵⁴ site to the most proximal base of the nearest upstream gene. The notation p or t indicates whether the promoter or the terminator, respectively, of the adjacent upstream gene is nearest the ς⁵⁴ site. The implication is that a promoter region may require a larger region. The size of the intergenic region is the sum of down plus up plus the 16 bases of the ς⁵⁴ site. 

Predictive Value of the Promoter Scores

The distribution of scores for properly and improperly oriented sites is shown in Fig. 3. There are only 2 incorrectly oriented sites (1.6%), and there are 25 correctly oriented sites (11.7%) with a score of at least 76. Of the 25 high-scoring correctly oriented sites, 18 (72%) are known ς⁵⁴-dependent promoters or are induced by nitrogen limitation. Together, these results imply that properly oriented high-scoring sites are not common and are likely to contain an authentic ς⁵⁴-dependent promoter. In contrast, only 7 (3%) of 213 sites with a score lower than 76 are known or likely (i.e., nitrogen limitation-induced) ς⁵⁴-dependent promoters. Therefore, scores below 76 are a reasonably good, but not infallible, indicator that the site is not a promoter. Despite the correlation with promoter scores, the promoter ranking is not an exact indicator of promoter strength. For example, the nac promoter has a higher score than the ς⁵⁴-dependent promoter of the glnALG operon, even though the latter is stronger (54). Nonetheless, and this point cannot be emphasized too much, the computer program recognizes all known Eς⁵⁴ binding sites, which implies that the failure to detect such a site is a reliable indicator that such a site does not exist.

FIG. 3.

Distribution of scores with properly and improperly oriented sites. The open bars indicate sites that are oriented toward the 3′ end of a gene; the solid bars indicate sites that are oriented toward the 5′ end.

Estimating the Number of ς⁵⁴-Dependent Promoters in E. coli

There are 17 ς⁵⁴-dependent promoters which either have been verified by direct evidence or whose expression requires ς⁵⁴; there are 7 other operons which are induced by nitrogen limitation (often associated with ς⁵⁴), appear to require a ς⁵⁴-dependent activator, and for which computer analysis suggests the presence of an appropriately located sequence for a ς⁵⁴-dependent promoter: b1012-b1006, chaC, ddpXABCDE, gltIJKL, potFGHI, yeaGH, and yhdWXYZ. Assuming that all of these genes have a functional ς⁵⁴-dependent promoter (which is unlikely) and that a few have been missed for various reasons (e.g., misidentified open reading frames), we estimate that E. coli contains about 30 ς⁵⁴-dependent promoters.

ς⁵⁴-DEPENDENT GENES OF NITROGEN METABOLISM

Of the 17 known ς⁵⁴-dependent promoters, 7 are involved in nitrogen metabolism. In addition, microarray analysis has identified 7 other genes that are induced by nitrogen limitation (176), and these have an appropriately placed potential ς⁵⁴-dependent promoter. This section discusses the 14 operons that are involved in nitrogen metabolism and the functions of their products. To understand the metabolic context of these proteins, it is first necessary to discuss nitrogen assimilation and the response to nitrogen deprivation. Because there have recently been some major changes in our understanding of these topics, the next sections summarize our current knowledge.

Nitrogen Assimilation and Its Control

Nitrogen assimilation.

Glutamate and glutamine are the major intracellular nitrogen donors, and they provide about 75 and 25% of the cell's nitrogen, respectively (calculated from numbers presented in reference 116). Nitrogen assimilation must therefore result in the synthesis of these two nitrogen donors.

Ammonia can be considered the focal point of nitrogen assimilation. There are two routes of ammonia assimilation (Fig. 4). For the first pathway, glutamate dehydrogenase assimilates ammonia and synthesizes glutamate. For the second pathway, glutamine synthetase (GS) assimilates ammonia, and glutamate synthase synthesizes glutamate. The former pathway is often associated with the presence of ammonia, and the latter pathway is associated with low ammonia levels or growth with a nitrogen source other than ammonia, since the K_m for ammonia for glutamate dehydrogenase is about 20-fold higher than that for GS. However, the most important difference between the two pathways appears to be that the former does not consume ATP but the latter does. Helling has shown that the glutamate dehydrogenase pathway is physiologically advantageous during carbon- and energy-limited growth while the GS-glutamate synthase pathway is used whenever energy is readily available (66, 67). (The energy difference between the ammonia assimilation pathways can be calculated, and it is significant. A 1-g amount of E. coli requires the synthesis of about 57,000 μmol of ATP and contains about 10,500 μmol of nitrogen. The ATP requirement for glutamine synthesis depends on the pathway of ammonia assimilation. If glutamate dehydrogenase assimilates ammonia, the cell requires about 2,300 μmol of glutamine and a corresponding amount of ATP for its synthesis. If the GS-glutamate synthase route assimilates ammonia, glutamine is also the precursor for glutamate and the cell must synthesize an additional 8,070 μmol of glutamine and ATP, or 14% more energy.)

FIG. 4.

Pathways of ammonia assimilation. GDH, glutamate dehydrogenase.

Control of ammonia assimilation and GS activity: role of glutamine.

Ammonia assimilation involves the regulation of three enzymes. Little is known about the regulation of glutamate dehydrogenase activity or synthesis in E. coli (127). The leucine-responsive regulatory protein (Lrp) controls the synthesis of glutamate synthase (53). A discussion of the function of Lrp and Lrp-dependent regulation, which does not require ς⁵⁴, is beyond the scope of the review. Instead, we will focus on the control of GS activity.

Two different but related mechanisms control GS activity: cumulative feedback inhibition by metabolites that require glutamine for their synthesis and covalent adenylylation (127). GS is a dodecamer, and adenylylation inactivates the modified subunit and renders the remaining subunits sensitive to feedback inhibition (127). A major function of adenylylation is to determine the function of GS. When GS is highly adenylylated and subject to cumulative feedback inhibition, its primary function is glutamine synthesis. In this situation, GS is just active enough to supply glutamine but not to supply glutamate. E. coli requires about 2,310 μmol of glutamine for biosyntheses per g (dry weight). In contrast, unadenylylated GS, which is not subject to feedback inhibition, can assimilate enough ammonia to meet all the cell's need for organic nitrogen. In this situation, E. coli needs to synthesize about 10,300 μmol of glutamine per g. A second function of adenylylation is to prevent the depletion of intracellular glutamate during the transition to a nitrogen-rich environment (96).

A cascade of three proteins controls GS adenylylation: the uridylyltransferase (UTase)-uridylyl removing (UR) enzyme, which in turn controls the activities of P_II and adenylyltransferase (ATase). It has been a long-standing paradigm that the ratio of α-ketoglutarate to glutamine (a sensor of relative carbon-to-nitrogen sufficiency) controls UTase-UR activity and therefore GS adenylylation. This conclusion was based on the properties of partially purified UTase-UR (2), which were not confirmed with purified UTase-UR (83). Furthermore, metabolite measurements suggested that low intracellular glutamine levels might be sufficient to control the response to nitrogen limitation (discussed below), which UTase-UR also controls (79). These results suggest that glutamine is the primary effector of UTase-UR and therefore of GS adenylylation (Fig. 5). Low glutamine levels stimulate UTase activity, which uridylylates P_II. P_II-UMP interacts with adenylyltransferase, which now removes adenylyl groups from GS and activates GS activity. High glutamine (nitrogen excess) stimulates UR activity, which results in the formation of unmodified P_II, whose interaction with adenylyltransferase stimulates adenylylation and reduces GS specific activity. Even though α-ketoglutarate does not affect UTase-UR, it does control the activity of unmodified P_II (discussed below).

FIG. 5.

Regulation of GS activity and the Ntr response. The pathways are shown for conditions of nitrogen excess (high glutamine) (top) and nitrogen limitation (low glutamine) with partial GlnK uridylylation (bottom). The open arrow in the bottom panel is meant to indicate that only partial uridylylation occurs. It is assumed that partial uridylylation occurs either during nitrogen limitation or during the transition to steady-state nitrogen-limited growth. The T-like symbol indicates an inhibition.

Nitrogen-Regulated (Ntr) Response

Nitrogen sources.

Ammonia is considered the preferred nitrogen source for E. coli grown in a minimal medium because ammonia supports the fastest growth and its presence prevents the synthesis of several proteins of nitrogen metabolism (reviewed in references 127 and 129). In place of ammonia, E. coli and related organisms can utilize a small number of nitrogen sources, usually amino acids, nucleosides, nucleobases, and a few inorganic nitrogen sources, e.g., nitrite and nitrate, which are reduced to ammonia. Steady-state growth on alternate nitrogen sources is slower and is said to be nitrogen limited. Catabolism of the alternate nitrogen sources must produce ammonia for the synthesis of glutamine, one of the intracellular nitrogen donors. For growth with a nitrogen source that cannot transfer its nitrogen to glutamate by transamination (e.g., adenosine), ammonia becomes an obligatory intermediate for all cellular nitrogen. In these situations, GS is the primary enzyme of ammonia and nitrogen assimilation (Fig. 4). Nitrogen-limited growth results in maximal synthesis of GS and also induces proteins that transport and catabolize several nitrogen sources. The coordinated response to nitrogen limitation is called the nitrogen-regulated (Ntr) response.

Control of the Ntr response by glutamine.

The two proteins that control GS adenylylation, UTase-UR and P_II, also control the Ntr response. High glutamine levels (nitrogen sufficiency) stimulate UR activity, which prevents uridylylation of P_II. Unmodified P_II interacts with nitrogen regulator II (NR_II, also called NtrB) and stimulates the dephosphorylation of nitrogen regulator I (NR₁, also called NtrC). The net effect is low expression of the glnALG operon and failure to activate Ntr genes. Low glutamine levels (nitrogen limitation) result in the formation of P_II-UMP, which is unable to interact with NR_II. In this situation, NR_II phosphorylates itself and transfers the activated phosphate to NR_I. NR_I-P then activates the expression of the glnALG operon and other Ntr genes.

α-Ketoglutarate counters the effects of unmodified P_II.

Even though recent studies have suggested that the ratio of glutamine to α-ketoglutarate does not regulate UTase-UR activity, α-ketoglutarate does affect P_II activity (83–85). α-Ketoglutarate counteracts the effects of unmodified P_II (present when glutamine levels are high and nitrogen is in excess) and therefore stimulates glnALG expression and increases GS activity. In other words, the ratio of glutamine (via UTase-UR) to α-ketoglutarate (via P_II) appears to control nitrogen assimilation during relative nitrogen sufficiency. This leaves the question whether there is a mechanism to coordinate carbon and nitrogen metabolism during nitrogen-limited growth when P_II is uridylylated. It will be suggested elsewhere in this review that such coordination might be a function of GlnK, a P_II-like protein.

NR_I regulon.

NR_I directly or indirectly controls the vast majority of Ntr genes. It is known to activate the expression of glnALG (GS and Ntr regulators), astCADBE (arginine catabolism), glnK-amtB (an alternate P_II and an ammonia transporter), nac (a ς⁷⁰-dependent transcriptional activator), and glnHPQ (glutamine transport) in E. coli. Several lines of evidence also suggest that it controls the expression of argT-hisJMPQ (arginine and histidine transport) and gltIJKL (glutamate-aspartate transport). In addition to these genes, microarray analysis suggests that NR_I might also activate b1012–b1006 (possibly for pyrimidine catabolism), chaC (calcium transport), ddpXABCDE (d-alanine–d-alanine metabolism), potFGHI (putrescine transport), yeaGH (unknown function), ygjG (a transaminase), and yhdWXYZ (amino acid transport) (176). In addition to activation, NR_I represses the two minor promoters, glnAp₁ and glnLp, of the glnALG operon.

Nac regulon.

There are two majors regulators of the Ntr response: NR_I and Nac. NR_I activates ς⁵⁴-dependent promoters, while Nac activates ς⁷⁰-dependent promoters. Nac is homologous to LysR (114). Unlike LysR, Nac apparently does not bind a ligand, which implies that it is constitutively active (16, 63).

Nac has been most intensively studied in Klebsiella aerogenes, where it activates genes for histidine, proline, urea, and d-alanine catabolism and represses glutamate dehydrogenase (16, 80, 106). It does not regulate the same genes in E. coli. E. coli lacks hut and ure operons, and Nac does not regulate the E. coli dad operon (16, 109). Nac deficiency in E. coli results in a slight derepression of glutamate dehydrogenase synthesis, slightly slower growth with cytosine as the nitrogen source, and slightly faster growth with arginine (114). The effect on arginine utilization is undoubtedly indirect, since synthesis of arginine catabolic enzymes does not require Nac (114) (see “astCADBE operon and catabolism of arginine and ornithine” below). Microarray analysis suggests Nac-dependent induction of b1440-1444 (probably for putrescine transport), codBA (cytosine metabolism), dppABCDF (dipeptide transport), fklB-cycA (d-alanine, d-serine, and glycine transport), gabDTP (γ-aminobutyrate [GABA] metabolism), nupC (nucleoside transport), ompF (outer membrane protein F), oppABCD (oligopeptide transport), yedL (unknown function), and yhiE (unknown function) (176). Nac-dependent control has been directly verified for the gab operon (S. Ruback and L. Reitzer, unpublished observation) but not for the other genes.

Why are there two general regulators of the Ntr response?

The main question concerning Nac is why there is a second Ntr regulator. Clearly, it is not necessary, since S. enterica serovar Typhimurium lacks it (114). We suggest that Nac-dependent control is important physiologically and serves a different function from NR_I-dependent control. NR_I-dependent genes respond to general nitrogen limitation, i.e., to intracellular glutamine, and not to specific induction mechanisms. The only known exception is the ast operon, which requires arginine-specific induction. In contrast, many Nac-dependent genes require both general and specific regulation: pyrimidines control the codBA operon by a complex process called reiterative transcription (126); GABA controls gab operon expression via the GabC repressor (Ruback and Reitzer, unpublished); and histidine and HutC control the hut operons in K. aerogenes (16). Nac may permit specific regulation, which may be difficult for ς⁵⁴-dependent promoters. This is illustrated in Fig. 6, which shows the regulatory sites for the single hutUH promoter of K. aerogenes. Other rationales for Nac have been proposed (16).

FIG. 6.

Binding sites for regulatory proteins at the hutUH promoter.

Function of the Ntr response.

It is likely that most of the ς⁵⁴-dependent Ntr genes have been identified using results from a microarray analysis and the complementary computer analysis of potential ς⁵⁴-dependent promoters. The Ntr genes can be divided into a few categories: GS, regulators, transport proteins, and catabolic enzymes.

Most Ntr genes specify transport proteins: amtB (ammonia), argT-hisJMPQ (arginine, lysine, ornithine, and histidine), b1006 (uracil?), b1440-b1444 (putrescine?), codB (cytosine), cycA (d-alanine, d-serine, and glycine), ddpXABCDE (d-alanyl–d-alanine), dppABCDE (dipeptides), gabP (GABA), glnHPQ (glutamine), gltIJKL (glutamate-aspartate), nupC (nucleosides), oppABCD (oligopeptides), potFGHI (putrescine), and yhdWXYZ (amino acids?). It has been suggested that a major function of the Ntr response is scavenging (176). However, Ntr proteins do not scavenge all amino acids. There are no Ntr-dependent transport systems for the aromatic amino acids, the branched-chain amino acids, threonine, methionine, or cysteine. An explanation for this pattern of expression may be that E. coli does not readily use these amino acids as nitrogen sources, which implies that their nitrogens are not readily available. In other words, E. coli generally has Ntr-dependent transport systems only for amino acids that can readily provide nitrogen for glutamate and glutamine synthesis.

Another class of Ntr genes specify enzymes for catabolic pathways. There are very few Ntr catabolic pathways, and unlike the transport genes, optimal synthesis usually requires specific induction. It should be noted that these catabolic pathways are not the major amino acid catabolic pathways, i.e., those that degrade amino acids that can be converted in one or two steps to intermediates of central metabolism, such as aspartate, glutamate, glutamine, serine, alanine, and glycine. An explanation for this observation is not apparent.

In summary, the function of the Ntr response is nitrogen assimilation when the intracellular glutamine level is low. This explains all the major aspects of the Ntr response: the regulation of GS activity and synthesis by glutamine, the regulation of the Ntr response by glutamine, and the reason why there are so many Ntr transport systems that scavenge nitrogenous compounds that have readily utilizable nitrogen.

glnALG (glnA-ntrBC) Operon

The glnALG operon codes for GS, NR_II, and NR_I, respectively (127, 129). All three products of this operon are required for nitrogen assimilation and the Ntr response (discussed above). ς⁵⁴ and NR_I∼P are required for transcription from the major promoter, glnAp₂, which has a score of 83.3. Nitrogen limitation activates glnA expression 11- to 24-fold, but appears to have little effect on glnL or glnG transcription (176). However, direct measurements of NR_I indicate that nitrogen limitation induces NR_I synthesis 14-fold (128). Minor promoters, glnAp₁ and glnLp, ensure basal synthesis of the important products of this operon (130, 157).

glnK-amtB Operon

Nitrogen limitation and NR_I are required for expression of the glnK-amtB operon (10, 160). Although the transcription start site has not been directly determined, the promoter has a score of 90.2 and potential binding sites for both Eς⁵⁴ and NR_I (Fig. 2) (160). In addition, nitrogen limitation increases the glnK transcript at least 10-fold (176). Therefore, this operon undoubtedly contains an authentic ς⁵⁴-dependent promoter. Both products of the operon contribute to the response to nitrogen limitation.

GlnK: a P_II paralog.

The existence of a P_II paralog was first suspected because of the rapid deadenylylation of GS in an E. coli glnB (P_II-encoding) mutant (160). The gene coding for this protein was cloned and called glnK (160). A glnK mutant has only a subtle phenotype (10). It has higher basal expression of an Ntr gene (glnK itself) in an ammonia-containing medium (even though the GlnK concentration should be low) and lower expression of an Ntr gene (again glnK) in a nitrogen-limiting medium. The mutant also has less of a lag during the transition to growth with arginine as a nitrogen source. This might result from higher basal expression of the ast operon, whose products degrade arginine. The phenotype of a glnB glnK double mutant, which lacks both P_II and GlnK, is more dramatic. It fails to grow in a nitrogen-rich minimal medium. The reason for this lethality is not known with certainty, but it may be related to uncontrolled phosphorylation of NR_I, which has been suggested to cause inappropriate overexpression of an Ntr gene (10). An alternate explanation is NR_I-dependent overexpression of a ς⁵⁴-dependent gene that is not normally regulated by NR_I.

Purified GlnK and P_II have similar activities, but the regulation of these activities is different (9, 58, 160, 161). However, the physiological relevance of many differences has not been established and is sometimes refuted by mutant phenotypes. The only safe basis for discussing the relevant properties of GlnK is when they account for the phenotype of mutants. One aspect of the mutant phenotype is the higher basal expression of an Ntr gene in a nitrogen-rich environment, which implies that GlnK suppresses this expression. One property of purified GlnK that accounts for this suppression is the relatively slow uridylylation of GlnK compared to that of P_II (9). This property is accentuated by the formation of GlnK-P_II heterotrimers (58, 161) and the inactivation of P_II-UMP by GlnK in such heterotrimers (161). The net effect is enhanced dephosphorylation of NR_I∼P and lower expression of Ntr genes. The second aspect of the phenotype of a glnK mutant is lower induced expression of an Ntr gene in a nitrogen-limited environment, which implies that GlnK stimulates Ntr expression. This is consistent with one property of purified GlnK. Although GlnK can efficiently stimulate the dephosphorylation of NR_I∼P via NR_II, α-ketoglutarate is more efficient in inhibiting the activity of GlnK than of P_II (9).

Why have GlnK?

The lower induction of an Ntr gene and higher basal expression in a glnK mutant suggest that GlnK sharpens the response to nitrogen availability. Perhaps the responsiveness of GlnK to α-ketoglutarate partially explains this effect. In this case, GlnK essentially restores the coordination of carbon and nitrogen metabolism (the responsiveness to the ratio of α-ketoglutarate to glutamine) that is lost when P_II is completely uridylylated, i.e., during nitrogen-limited growth, and is no longer responsive to α-ketoglutarate. A second function for GlnK has been found, but not in E. coli. GlnK is required for control of NifL, which inhibits the activity of NifA in Klebsiella species (65). NifA is the transcriptional activator required for expression of the nitrogenase gene cluster in Klebsiella species. Uridylylated and nonuridylylated GlnK can relieve repression. It is not known how GlnK mediates shutoff of the nif genes when ammonia is added, but it may interact with other proteins.

Product of amtB and ammonia transport.

The second gene of the operon, amtB, codes for an ammonia transporter. None of the phenotypes of the glnK mutant could be attributed to polar effects on amtB expression (10). An amtB mutant of S. enterica serovar Typhimurium has only a subtle phenotype (149). It is unable to utilize a low concentration of ammonia if the pH is less than 7. This phenotype implied that uncharged NH₃, not NH₄⁺, is transported. It was proposed that AmtB did not concentrate ammonia but only facilitated equilibrium across the membrane. This mechanism of transport might have significant physiological implications, which are discussed in “Physiological Function of ς⁵⁴” (below).

nac

Although E. coli, S. enterica serovar Typhimurium, and K. aerogenes are closely related, they differ in their ability to utilize certain nitrogen sources and in the regulation of some genes of nitrogen metabolism. For example, nitrogen limitation strongly represses glutamate dehydrogenase in K. aerogenes but not in E. coli (16, 127). The transcriptional regulator Nac accounts for many of these differences in regulation. Nac has been most extensively studied from K. aerogenes but has also been studied from E. coli. In contrast, S. enterica serovar Typhimurium lacks Nac (114).

The nac operon is monocistronic (114, 145). Transcription initiated from the K. aerogenes nac promoter requires ς⁵⁴ and NR_I∼P (16, 54, 106). Nitrogen limitation induces E. coli nac (114, 176), and computer analysis indicates a likely binding site for Eς⁵⁴ with a score of 85.2, which is very high. These results suggest that NR_I and Eς⁵⁴ are also required for E. coli nac expression. Nac negatively modulates its own synthesis in K. aerogenes by interfering with the interaction between NR_I and RNA polymerase (55, 114), and results from the DNA microarray analysis are consistent with such regulation in E. coli (176).

Catabolism of Arginine, Agmatine, Ornithine, Putrescine, and γ-Aminobutyrate

Arginine (via agmatine) and ornithine are both precursors for putrescine, which can be metabolized to GABA, and then to succinate (Fig. 7). Nitrogen limitation induces enzymes of GABA catabolism (175). Therefore, it was reasonable to propose that Ntr regulators affect the catabolism of all of these compounds, and some evidence is consistent with this regulation (147). However, recent studies with mutants containing targeted gene disruptions have indicated unsuspected pathways and a surprising complexity and redundancy of pathways and regulators. Only the enzymes of arginine and GABA catabolism require ς⁵⁴, while the enzymes of putrescine catabolism may not.

FIG. 7.

Metabolic relationships between ornithine, arginine, putrescine, and GABA. A thick black arrow indicates that nitrogen limitation induces the enzyme indicated. A reaction catalyzed by two (or more) enzymes is indicated by two arrows. The genes that specify the enzymes are shown when they are known. A dashed arrow indicates that the gene has yet to be identified.

astCADBE operon and catabolism of arginine and ornithine.

The five-step arginine succinyltransferase pathway catabolizes arginine (144). The pathway is named after the first reaction, which is the succinylation of the α-amino group of arginine. The astCADBE operon codes for the proteins of the pathway (144). Disruption of the operon in E. coli prevents growth with arginine as a nitrogen source and impairs but does not eliminate growth with ornithine (60, 144). It has been proposed that AstC (which catalyzes the deamination of succinylornithine) is one of at least two transaminases that can deaminate ornithine, which generates an intermediate of proline catabolism (144). The identity of the second transaminase is unknown.

Expression of the astCADBE operon in E. coli and S. enterica serovar Typhimurium requires either nitrogen limitation or entry into stationary phase (12, 60, 103, 144). There are two transcription start sites, which are separated by five bases (A. Kiupakis and L. Reitzer, unpublished results). Expression from the Ntr promoter requires NR_I∼P and ς⁵⁴, while expression from the other promoter requires Eς^S. SeqScan gives the ς⁵⁴-dependent promoter a score of 85.5. Some evidence suggests that transcription from one promoter prevents transcription from the other (60). One unusual feature of the ast operon in E. coli is that ArgR binds to the region between the Eς⁵⁴ and NR_I binding sites and has been proposed to stimulate the interaction between the two proteins (103). ArgR is required for optimal transcription of the E. coli ast operon but is not absolutely necessary (Kiupakis and Reitzer, unpublished). In contrast, ArgR appears to be required for the S. enterica serovar Typhimurium ast operon (103). The ast operon contains the only known E. coli NR_I-dependent promoter that also requires specific induction. Microarray analysis indicates that general nitrogen limitation (i.e., without arginine induction) increases ast transcription 7- to 11-fold (176), which is consistent with the results of a direct assay of the gene products (144). Arginine induces the enzymes three- to fourfold further (144), which is consistent with in vitro results (Kiupakis and Reitzer, unpublished). Another unusual aspect of ast expression is that a strain with a glnL (ntrB) deletion cannot utilize arginine as a nitrogen source (L. Reitzer, unpublished observation) but can still activate glnA expression, albeit not as rapidly (130). Expression of the ast operon requires phosphorylation of NR_I by both NR_II and small phosphodonors (B. L. Schneider, D. Fewell, and L. J. Reitzer, unpublished observation).

GABA and putrescine catabolism and the gabDTPC operon.

E. coli can utilize GABA as a nitrogen source (H. Kasbarian, S. Ruback, and L. Reitzer, unpublished results), although earlier studies indicated otherwise (50). A mutant with a gabDT deletion cannot utilize GABA as a nitrogen source (Ruback and Reitzer, unpublished). The distance between genes suggests the existence of a gabDTPC operon. Furthermore, transcript analysis indicates the presence of only one promoter for these genes, a promoter just upstream from gabD. Each gene of the putative operon has been implicated in GABA catabolism. GabT is a transaminase that deaminates GABA to succinic semialdehyde. GabD is an NADP-specific succinic semialdehyde dehydrogenase that oxidizes succinic semialdehyde to succinate. (An NAD-dependent succinic semialdehyde dehydrogenase is specified by the sad gene, and GABA or a product of GABA metabolism induces its synthesis [49].) GabP specifies a GABA permease (92). GabC appears to be a specific repressor since a deletion of gabC stimulates growth with GABA as a nitrogen source (Kasbarian et al., unpublished). Either nitrogen limitation or entry into stationary phase activates gab operon expression (12, 175; Kasbarian et al., unpublished). Expression of the gab operon requires Nac during nitrogen-limited growth, and Nac is required to reconstitute transcription in vitro (H. Kasbarian, A. Kiupakis, S. Ruback, and L. Reitzer, unpublished results). ς^S is required for expression during stationary phase.

GABA (via γ-aminobutyraldehyde) is a presumed intermediate in putrescine catabolism. Unexpectedly, a strain with a deletion of gabDT grew normally with putrescine or agmatine as a nitrogen source (Kasbarian et al., unpublished). Even more surprising is the observation that an rpoN mutant (ς⁵⁴ deficient) grew normally with putrescine as a nitrogen source (Kiupakis and Reitzer, unpublished). These results would suggest that nitrogen limitation is not required for induction of putrescine catabolic genes, but this is not the case. Microarray analysis suggests that nitrogen limitation activates the expression of two different putrescine transport operons: potFGHI (five- to sixfold) and b1440-1444 (five- to sevenfold) (176). In addition to these transport systems, E. coli possesses two ς⁵⁴-independent transport systems, the products of potABCD (which preferentially transport spermidine but also transport putrescine) and of potE (78). The genes of putrescine catabolism and the physiological function of the four transport systems have yet to be established.

ygjG.

Nitrogen limitation results in a three- to fivefold increase in the levels of steady-state ygjG transcripts (176). Expression of this gene requires ς⁵⁴, and the transcription start site has been determined (A. Kiupakis, R. Ye, and L. Reitzer, unpublished result). The score for this promoter is 70.7, which is low for an authentic ς⁵⁴-dependent promoter. The gene specifies a putative ω-transaminase which either removes the amino group from compounds with terminal primary amines (e.g., putrescine or ornithine), or adds amino groups to compounds with an aldehyde group (e.g., N-acetylglutamic semialdehyde, an intermediate in ornithine formation). Putrescine and compounds metabolized to putrescine activate ygiG expression, which suggests a possible role in putrescine catabolism. However, a mutant with a disruption of ygjG grows normally with putrescine as a nitrogen source, which suggests that YgjG is a redundant transaminase (C. Pybus and L. Reitzer, unpublished observation).

ς⁵⁴-dependent Amino Acid Transport Systems

More than half of the genes activated by nitrogen limitation in E. coli code for transport systems. Such activation usually does not require specific induction. Such regulation combined with the observation that mutational inactivation of the genes for ς⁵⁴-dependent amino acid transport system does not prevent growth on the respective amino acid suggests a scavenging function. In this section, the ς⁵⁴-dependent amino acid transport systems are considered within the context of the multiple transport systems for these amino acids.

(i) The three transport systems.

E. coli has three characterized arginine transport systems (described below), while S. enterica serovar Typhimurium has at least two (98). An early study suggested the presence of three periplasmic arginine binding proteins in E. coli (136). No strain has been constructed with mutations in all the characterized systems. Therefore, it is conceivable that there are other transport systems.

Only one of the three E. coli arginine transport systems requires ς⁵⁴ for its synthesis. It contains the periplasmic ArgT protein, also called the LAO protein, which binds lysine, ornithine, and arginine with high affinity (136). The argT gene and its product have been extensively studied in S. enterica serovar Typhimurium but not in E. coli. ArgT interacts with HisP of S. enterica serovar Typhimurium (98). HisJ, the periplasmic histidine binding protein, also interacts with HisP (5). An S. enterica serovar Typhimurium hisP mutant grows much more slowly with arginine as the nitrogen source than an argT mutant does, which suggests that HisP also interacts with an arginine binding periplasmic protein other than ArgT (98).

ς⁵⁴-independent arginine transport systems have been studied only in E. coli. One system contains AbpS, also called the arginine-ornithine protein, which binds arginine and ornithine in the periplasm with lower affinity than the LAO protein does (34). Early reports refer to this protein as a low-affinity, arginine-specific protein (41, 136). The abpS gene has been approximately mapped to min 63.5 of the most recent E. coli map (35). AbpS has been purified, and its size and amino acid composition have been determined (36). However, no gene near min 63 specifies a protein with the published amino acid composition. The nearest matches to this amino acid composition in the E. coli genome, in descending order, are the products of artI, artJ, and hisJ, which are located at min 19.4, 19.4, and 52.3, respectively. It is possible that the sequenced MG1655 does not contain abpS. The second ς⁵⁴-independent system consists of the artPIQM-artJ operons, which are at min 19 of the E. coli chromosome (169). ArtJ is a periplasmic protein that binds arginine but not ornithine (169). ArtI is another putative periplasmic binding protein, but it does not detectably bind any amino acid. Mutants with mutations in the ArtJ system do not exist, but overexpression increases arginine transport, which is consistent with a proposed function in arginine transport (169). Promoters precede artP and artJ, and neither appears to require ς⁵⁴ (169).

(ii) Repression by arginine.

Arginine represses all three E. coli transport systems (41, 137, 169). A possible mechanism of repression would involve ArgR, which mediates arginine repression for the enzymes of arginine synthesis. A computer analysis of ArgR sites in E. coli identified two sites in the art operon: one preceding artP and one preceding artJ (111). However, a missense mutation in argR had no effect on the kinetically detectable arginine transport systems (39). Instead of ArgR, ArgP and ArgK have been proposed to mediate arginine repression. ArgP is a transcriptional regulator required for synthesis of ArgK, which is required for arginine transport. Mutations in argP and argK affect both the ArgT and AbpS systems (40, 41, 137). Only one gene separates argP and argK, but they appear to be independently expressed. (argK is currently not listed in either GenBank or the latest E. coli genetic map. However, argK is ygfD, also called b2918 in GenBank.) ArgP is a LysR-type regulator that activates argK expression in the absence of arginine (37). ArgP complexed with arginine fails to activate argK expression and represses its own synthesis. ArgK has an ATPase activity that is apparently required for transport activity (158). ArgK also phosphorylates the periplasmic ArgT and AbpS (36), although this phosphorylation is not required for transport (38). ArgP is the previously characterized IciA, an inhibitor of the initiation of DNA replication (37, 154). An iciA mutant has no obvious phenotype, except for difficulty during dilution into fresh growth medium (154). It is conceivable that ArgP/IciA is a sensor of amino acid sufficiency that couples DNA synthesis with metabolism in some environments.

(iii) Transport and activation during nitrogen limitation.

Nitrogen limitation induces ArgT in E. coli and S. enterica serovar Typhimurium, and this induction does not require arginine (98, 176). The genetics and regulation of arginine transport have been studied in S. enterica serovar Typhimurium, and it is assumed that they will be similar in E. coli. In S. enterica serovar Typhimurium, loss of ArgT reduced but did not eliminate the binding of arginine to periplasmic proteins and an argT mutant grew normally with arginine as a nitrogen source, which implies a second transport system in S. enterica serovar Typhimurium during nitrogen-limited growth (98).

argT is adjacent to the hisJQMP operon, which codes for components of a histidine transport system. Transcript mapping in wild-type S. enterica serovar Typhimurium, not in an extensively studied dhuA1 mutant which appears to have a mutationally created promoter, shows a ς⁵⁴-dependent promoter immediately preceding argT but not immediately preceding hisJ (6). This is consistent with expression studies with reporter gene fusions, which failed to identify an Ntr promoter preceding hisJ (142, 143). The potential ς⁵⁴-dependent promoter preceding argT in E. coli has a score of 72.3. There is only one binding site for NR_I, and it is in the argT-hisJ intercistronic region and not upstream from argT (6). The function of this site is not clear. It does not appear to be necessary for expression, which would imply that NR_I activates the argT promoter without binding to DNA (143). There is precedent for NR_I-dependent transcription that does not require a DNA binding site (131, 171).

Histidine.

The genes and regulation of the HisJ transport system were discussed in the preceding section because of their relation to arginine transport. It is not known whether there are other histidine transport systems.

Glutamine.

Glutamine transport has been studied in both E. coli and S. enterica serovar Typhimurium. The kinetically dominant system requires GlnH, a high-affinity glutamine-specific binding protein in the periplasm (162). The glnHPQ operon specifies GlnH and two membrane proteins, which presumably interact with GlnH (119). Loss of GlnH unmasks a low-affinity glutamate-inhibitable glutamine transport system in E. coli (162) but has no effect on growth in S. enterica serovar Typhimurium (98). These results suggest a second glutamine transport system. Kinetic assays of glutamine transport also suggest two transport systems (11, 168). It is possible that the glutamate-inhibitable system requires the periplasmic glutamate-aspartate binding protein (see the next section), which also binds glutamine (167).

Expression of glnHPQ requires nitrogen limitation but not glutamine (18, 98, 168). Nitrogen limitation increases the production of glnHPQ transcripts five- to ninefold (176). The operon contains two promoters (118). Transcription from the downstream promoter, glnHp₂, requires ς⁵⁴ and NR₁ and is enhanced by IHF (44). The promoter has a score of 80.7. The factors that control the upstream promoter, glnHp₁, have not been examined.

Glutamate-aspartate.

Schellenberg and Furlong defined five transport systems for glutamate and aspartate in E. coli by a combination of genetic and biochemical experiments (140). There are no studies with gene fusions; therefore, the regulation and functions of the individual systems are poorly understood.

Nitrogen limitation induces a periplasmic protein that binds both glutamate and aspartate in S. enterica serovar Typhimurium (98). Such a protein has been purified and characterized from E. coli (167). The closest match in the entire SWISS-PROT database with the published amino acid composition of the glutamate-aspartate binding protein is gltI (also called ybeJ) from E. coli. The gltI product was the only protein in E. coli with the correct pI, size, and number of cysteines. gltI might be part of a gltIJKL operon. Despite the annotation of gltIJKL as part of a glutamate-aspartate transport system, no published evidence supports this possibility (17).

There are no potential ς⁵⁴-dependent promoters preceding gltI with a score greater than 60. However, the gene preceding gltI specifies an IS5 transposase, and the promoter region for the transposase gene contains a possible ς⁵⁴-dependent promoter with a score of 88, which is very high. Microarray analysis indicates a three- to sixfold increase of gltI transcription during nitrogen limitation, and NR_I-dependent activation (176). These results suggest that the ς⁵⁴-dependent promoter preceding the transposase gene can initiate gltIJKL transcription.

ddpXABCDE operon.

Microarray analysis shows that general nitrogen limitation (i.e., no specific induction) induces the ddpXABCDE operon 52- to 60-fold, which is more than any other operon. Expression appears to require NR_I (176). A potential ς⁵⁴-dependent promoter with a score of 76.6 precedes the operon. DdpX is a zinc-containing d-alanyl–d-alanine dipeptidase, while the other products of the operon appear to code for components of a dipeptide permease (100). There are two sources of d-alanyl–d-alanine: it is an intermediate in peptidoglycan synthesis, and it may be released during cross-linking of two diaminopimelic acids. Peptidoglycan cross-linking occurs in the stationary phase (156). Entry into the stationary phase induces the ddpXABCDE operon, and this induction requires ς^S (100). It is not known whether peptidoglycan remodeling also occurs in nitrogen-limited cultures. It has been proposed that the function of the ddpXABCDE products is to scavenge d-alanyl–d-alanine (176).

Peptide transport and ompF.

E. coli, S. enterica serovar Typhimurium, and other gram-negative bacteria digest peptides intracellularly after their passage through the outer membrane and transport via periplasmic binding protein-dependent transport systems. Nitrogen limitation activates the expression of dppABCDF and oppABCDF (176). The products of these operons are the major peptide transporters in E. coli and S. enterica serovar Typhimurium (120), and the periplasmic components of these systems, DppA and OppA, are among the most abundant proteins in the periplasm (1, 70, 121). In addition to its function as a transport protein, DppA is required for chemotaxis to peptides (108). Neither peptide transport operon contains a potential ς⁵⁴-dependent promoter. The apparent 10-fold dppABCDF induction appears to require Nac (176), which is consistent with the absence of a ς⁵⁴-dependent promoter. In contrast, microarray analysis provides no evidence for Nac-dependent regulation for the four- to sixfold oppABCDF induction and equivocal evidence for NR₁-dependent regulation (176). Induction by nitrogen limitation is most evident in medium with glutamine as the nitrogen source but not in medium with ammonia as the nitrogen source for strains with constitutively active Ntr regulatory proteins. These results suggest that induction may require specific nitrogen-containing compounds. This is consistent with the known regulators of oppABCDF expression, Lrp and the gcvB transcript (a regulatory RNA), which respond to leucine (or alanine) and glycine, respectively (31, 117, 159). (The gcvB transcript also controls dppABCDF expression.) Based on these considerations and the absence of a potential ς⁵⁴-dependent promoter, the effect of nitrogen limitation on oppABCDF may be indirect, i.e., independent of NR_I or Nac.

The first step in peptide transport is passage through the outer membrane, and nitrogen limitation results in 25-fold higher transcription of ompF, which codes for an outer membrane channel (176). The microarray analysis suggests very modest control by NR_I (two- to threefold), and stronger control by Nac (perhaps ninefold) (176). NR_I-dependent control cannot be ruled out since there is a weak potential ς⁵⁴-dependent promoter with a score of 64.4. However, like oppABCDF expression, the level of induction in mutants with constitutively active NR_II is dramatically stronger with glutamine (25-fold) than with ammonia (4-fold) (176), and Lrp has been implicated in ompF regulation (57). Perhaps nitrogen limitation controls ompF indirectly, as was suggested for oppABCDF.

Potential ς⁵⁴-Dependent Genes That Are Induced by Nitrogen Limitation

Nitrogen limitation results in a 20-fold increase in the expression of the putative b1012-1006 operon (176). The potential Eς⁵⁴ binding site preceding b1012 has a score of 76.8. The last gene of the putative operon codes for a possible uracil permease, which may suggest that this operon codes for enzymes of pyrimidine catabolism. This is consistent with the observation that [¹⁴C]uracil or [¹⁴C]thymine catabolism yields ¹⁴CO₂, even though E. coli cannot degrade these compounds as sole nitrogen sources (13).

Nitrogen limitation induces b2875-76 and b2882-85. Some of the genes in this region have been studied, and homology searches have suggested that they might participate in purine catabolism (172). The xdhA gene (b2866), which codes for one subunit of a recently discovered xanthine dehydrogenase, appears to have two promoters, and one of them might be ς⁵⁴ dependent (172). E. coli can utilize intermediates of purine catabolism, such as allantoin, as nitrogen sources anaerobically but not aerobically (46). The potential ς⁵⁴-dependent promoters preceding xdhA (b2866), ygeW (b2870), and b2878 have low scores (66.0, 75.2, and 71.5, respectively), but they might contribute to purine catabolism during anaerobic growth.

Nitrogen limitation activates the putative yhdWXYZ operon two- to ninefold, and appears to be NR_I dependent (176). The potential promoter for this operon has a score of 88.6. The products of this operon have homology to transport proteins for polar amino acids.

Nitrogen limitation activates the yeaGH operon two- to fourfold, but it is not clear whether regulation requires NR_I or Nac (176). The potential Eς⁵⁴-binding site has a score of 83.8. However, homology searches provide no clue to the function of the products.

ς⁵⁴-DEPENDENT GENES THAT ARE NOT INVOLVED IN NITROGEN METABOLISM

Formate Catabolic Genes and the FhlA Regulon

Formate metabolism.

The products of several ς⁵⁴-dependent operons contribute to formate metabolism during glucose fermentation. The products of glucose fermentation in E. coli in terms of total carbon from glucose are CO₂ (14.7%), ethanol (16.6%), acetic acid (12.2%), lactic acid (40%), formic acid (0.4%), succinic acid (7.2%), and cell constituents (∼10%) (reviewed in reference 24). Formate is a major intermediate even though it does not accumulate. Formate formation is linked to pyruvate metabolism because pyruvate formate lyase cleaves pyruvate to formate and acetyl coenzyme A (acetyl-CoA) during anaerobic growth. The production of fermentation products that require acetyl-CoA as a precursor, acetate and ethanol, necessarily generates a stoichiometric amount of formate. Therefore, about 14% of glucose carbon (8.3% concomitant with ethanol formation plus 6.1% with acetic acid formation) is converted to formate during fermentation.

The formate hydrogenlyase (FHL) complex catalyzes the disproportionation of formate to CO₂ and H₂, which accounts for the CO₂ (0.88 mol per mol of glucose) and H₂ (0.75 mol per mol of glucose) produced during fermentation. (Hydrogenases 1 and 2 recycle some of the H₂ as an electron acceptor. However, the final ratio of H₂ to CO₂ suggests that only about 15% of the H₂ is recycled.) The primary function of the FHL complex is pH homeostasis during fermentation (24). During the initial stages of fermentation, formate is excreted from the cell. As the acidic fermentation products accumulate and lower the pH, formate is imported into the cell, which induces FHL synthesis. FHL consumes all the formate produced and increases the pH. The FHL complex contains a formate dehydrogenase, hydrogenase 3, and intermediate electron carriers (24). The formate dehydrogenase component is termed FDH_H. (E. coli possesses two other formate dehydrogenases, which use electron acceptors other than protons [139].) FDH_H contains selenocysteine and binds iron, molybdenum, and cobalt. FDH_H is associated with the [Ni-Fe]-containing hydrogenase 3, which catalyzes electron transfer to protons. This electron transfer does not result in energy conservation.

The four confirmed ς⁵⁴-dependent operons of formate metabolism.

The requirement for ς⁵⁴ has been established by examination of mutant phenotypes, lacZ fusions, transcript analysis from mutant and wild-type strains, and transcription with purified components (20, 21, 73, 104, 107). The four confirmed ς⁵⁴-dependent operons code for components required for the FHL complex. The monocistronic fdhF specifies one subunit of FDH_H. The divergently transcribed polycistronic hyc and hyp operons code for components of FDH_H, hydrogenase 3, proteins required for processing of hydrogenase 3 and other hydrogenases, and two transcriptional regulators (24). The hydN-hypF operon specifies a protein required for FDH_H activity (possibly a component of electron transport) and a second protein required for processing of several hydrogenases (107). The promoters for the hyp, hyc, fdhF, and hydN-hypF operons have scores of 88.0, 79.1, 72.6, and 64.5, respectively. The last two are among the lowest scores for confirmed ς⁵⁴-dependent promoters.

The primary activator of these operons is FhlA (24). FhlA activates its own expression from the major promoter of the hyp operon, but secondary promoters ensure basal synthesis (24, 105). FhlA is homologous to other ς⁵⁴-dependent activators (reference 141 and references therein). Footprinting experiments have established the binding sites for FhlA (73). Like other ς⁵⁴-dependent activators, it can activate when bound to distant sites (104).

Several factors regulate FHL complex synthesis (22, 125). Low pH and formate induce its synthesis, while oxygen, nitrate, and glucose repress it (138). The formate and oxygen control are regulated via FhlA. Formate stimulates the ATPase activity of FhlA (74) and is required for in vitro transcription, which implies that formate binds FhlA (73). Oxygen control may be mediated by OxyS, which is induced by oxidative stress (3). OxyS is an abundant stable untranslated RNA that binds to the FhlA mRNA and blocks its translation (4). Glucose, pH, nitrate, and additional oxygen control is probably indirect, i.e., via regulation of the synthesis of the major formate transport system, which is part of the pfl (pyruvate formate-lyase) operon (93, 138). HycA is a regulator that antagonizes the activation of FhlA (unpublished results cited in reference 24). The mechanism of this regulation has not been characterized. Mo availability also regulates expression of the hyc operon (146). ModE is a sensor of intracellular Mo, and a ModE-Mo complex represses the modABCD operon, which specifies components of a Mo transport system. ModE-Mo also stimulates transcription of the hyc operon by binding to a site centered 190 bases from the start site of transcription (146). (The FhlA binding site is centered 100 bases from the transcription start site [73]). ModE-Mo is not required for expression, but its presence accounts for a two- to threefold stimulation. In addition, the MoeA protein, a component of Mo metabolism, stimulates hyc expression two- to threefold by an unknown mechanism (146). The multiplicity of regulators might strengthen the binding of Eς⁵⁴ to FhlA-dependent promoters and may account for their low scores.

Hydrogenase 4.

A 12-gene hyf operon at min 56 contains a possible ς⁵⁴-dependent promoter with a score of 68.7, and upstream sequences suggest the presence of FhlA binding sites. The expression of this operon has not been characterized or established. Homology analysis suggests that the products of the operon code for a putative hydrogenase 4, which catalyzes the same reactions as the FHL complex, and for proteins of respiration-linked proton translocation. Therefore, it was proposed that the products of the hyf operon specify an energy-conserving hydrogenase 4. The putative product of one gene of the operon, hyfR, is homologous to FhlA and may bind the same sites as FhlA because of conservation of the DNA binding residues (7).

ato Operon and Acetoacetate Catabolism

Loss of the ato operon results in failure to utilize acetoacetate as a carbon and energy source (reviewed in reference 43). The transcription start site has not been examined. Nonetheless, it is likely that this operon requires ς⁵⁴ for several reasons. First, an rpoN mutant (ς⁵⁴ deficient) cannot utilize acetoacetate as a carbon source (C. Pybus and L. Reitzer, unpublished). Second, expression of the ato operon requires AtoC, which is homologous to other activators of ς⁵⁴-dependent promoters (42, 43). Proteins homologous to AtoC usually activate ς⁵⁴-dependent promoters, although two activate ς⁷⁰-dependent promoters (59, 113, 173). The latter activators lack an essential region of the domain that interacts with Eς⁵⁴. The domain of AtoC that interacts with RNA polymerase is homologous throughout its length to the corresponding region of other ς⁵⁴-dependent activators (113). Therefore, it is likely that AtoC activates from a ς⁵⁴-dependent promoter. Finally, the score of the putative ς⁵⁴-dependent ato promoter is 85.9, which is very high.

The genes of the atoDAEB operon code for proteins of acetoacetate catabolism (42, 81). It has been proposed that atoE codes for an acetoacetate-specific transport system (42). The atoD and atoA genes specify the subunits of acetyl-CoA:acetoacetyl-CoA transferase, which catalyzes the transfer of CoA from acetyl-CoA to acetoacetate. AtoB specifies thiolase II, which catalyzes the formation of two molecules of acetyl-CoA from CoA and acetoacetyl-CoA (43). (Thiolase I is an enzyme in fatty acid β-oxidation.)

E. coli can utilize short-chain fatty acids (C₄ to C₆) such as butyrate (C₄) and valerate (C₅) as a carbon source and such catabolism requires the ato operon (43, 124). Butyrate catabolism requires the formation of butyryl-CoA by acetyl-CoA:acetoacetyl-CoA transferase, followed by dehydrogenation of the saturated fatty acid, hydration, and oxidation, which results in the formation of acetoacetyl-CoA. These reactions require enzymes of fatty acid degradation, products of the fadR regulon. Acetoacetyl-CoA is then degraded as described above. Butyrate does not induce either the ato or the fad genes. Therefore, growth with butyrate as the sole carbon source requires constitutive expression of both sets of genes (43).

AtoC is required for expression of the atoDAEB operon (82, 124). The nucleotide sequence of this region suggests that atoC is the second of two genes in an atoSC operon that is just upstream from the atoDAEB operon. AtoC is homologous to response regulators such as NR_I (NtrC), and the putative AtoS is homologous to sensor kinases, such as NR_II (NtrB). Acetoacetate or a product of acetoacetate metabolism probably binds AtoS and stimulates AtoC phosphorylation. Other aspects of regulation have not been examined. The possibility of IHF sites upstream from the atoD promoter has been suggested (42). Glucose blocks expression of the ato operon, and possible cyclic AMP sites that have been identified upstream from the atoD promoter may interfere with the AtoC-Eς⁵⁴ interaction (42).

prpBCDE Operon and Propionate Catabolism

Environmental propionate is the end product of several different fermentation pathways and can also result from β-oxidation of odd-chain fatty acids. Propionate is a membrane-permeable anion that can alter the internal pH of bacteria. The products of the prp operon degrade propionate. Most of the genetics of propionate catabolism and analysis of gene expression has been studied with S. enterica serovar Typhimurium. Expression of this operon requires ς⁵⁴ in S. enterica serovar Typhimurium, although the transcription start site has not been identified (122). In addition to ς⁵⁴, expression requires IHF and PrpR, which is homologous to NR_I (122). 2-Methylcitrate or a product of its metabolism has been proposed to bind PrpR and induce the operon (155). It is assumed that regulation in E. coli is similar. The putative E. coli promoter has the third highest score, 89.6, of known ς⁵⁴-dependent promoters.

The propionate catabolic pathway is called the methylcitric acid cycle (76, 153). The first reaction is the addition of CoA to propionate by PrpE, propionyl-CoA synthetase (75). In addition to PrpE, acetyl-CoA synthetase and possibly an enzyme of acetoacetate catabolism can catalyze this reaction (75, 133). The second reaction is catalyzed by methylcitrate synthase, the product of prpC, which generates the presumed inducer methylcitrate. Methylcitrate synthase also reacts with acetyl-CoA, although propionyl-CoA is the preferred substrate (76). The next reactions are a dehydration and hydration to form methylisocitrate with methylaconitate as an intermediate. An aconitase-like activity could conceivably catalyze these reactions, and PrpD might catalyze one or both of these reactions, but PrpD is not homologous to known aconitases (76). The last reaction is cleavage of methylisocitrate to succinate and pyruvate, which is catalyzed by PrpB, methylisocitrate lyase (76). The methylcitric acid cycle requires regeneration of oxaloacetate. The most obvious source of oxaloacetate is succinate. However, some strains of E. coli require the glyoxylate shunt for this oxaloacetate formation (153), and it has been proposed that S. enterica serovar Typhimurium generates oxaloacetate from pyruvate by the combined actions of phosphoenolpyruvate synthetase and phosphoenolpyruvate carboxylase (56). Several aspects of propionate catabolism are unusual. Strains lacking glutathione cannot utilize propionate as a carbon source (135). Strains lacking DNA polymerase I also fail to utilize propionate, which suggests that propionate or a product of propionate catabolism damages DNA (134). Finally, propionate is toxic in the absence of enzymes of the methylcitric acid cycle (64).

psp Operon and Phage Shock Response

The pspABCDE operon is unusual, and its regulation is perhaps the most complicated of the ς⁵⁴-dependent operons (see reference 112 for a review). This system was first discovered and studied by Peter Model and colleagues, who noticed that overexpression of a filamentous phage protein resulted in massive PspA synthesis in E. coli (25). It was subsequently shown that several different stresses also induce PspA synthesis: filamentous-phage infection, overexpression of some filamentous-phage proteins, overexpression of some outer membrane proteins (especially mutant forms), heat shock, ethanol, hyperosmotic shock, nutritional downshifts (passage into the late stationary phase of the growth cycle), proton ionophores and other uncouplers of oxidative phosphorylation (free fatty acids), and hydrophobic organic solvents (95, 112, 165). Various proteins are required to sense these stresses, and it is unlikely that there is a single inducing effector.

Mutants lacking PspA, PspB, or PspC have no dramatic phenotype during exponential growth. However, these mutants survive poorly in stationary phase in an alkaline environment (165). These mutants also have greater motility and slower protein translocation (112). PspA appears to maintain the proton motive force in stressed cells, and it has been proposed that this is the major function of PspA (94).

All the inducing stresses result in transcription from a single promoter that requires ς⁵⁴ (163). The promoter has a score of 80.0 (Table 3). Activation from the psp promoter requires PspF, which is specified by a gene adjacent to, but divergently transcribed from, the pspABCDE operon (91). PspF and IHF bind cooperatively to the psp promoter (89). One function of IHF is to increase the specificity of activation by preventing an interaction with RNA polymerase by other activators (52, 91). However, PspF (and possibly other ς⁵⁴-dependent activators) can activate the psp operon without the PspF binding site (the enhancer) during hyperosmotic shock (90). PspF lacks an amino-terminal regulatory domain, and it activates transcription without phosphorylation or an activating ligand (91). Instead, as described below, PspA controls PspF activity.

An unusual aspect of the psp operon is that four of the five genes specify regulators that control psp expression. PspA binds to PspF, which blocks its ability to activate transcription (51). PspA also inhibits at least one other ς⁵⁴-dependent activator, NR_I, and perhaps others (51). It is not known whether this inhibition of NR_I is important in vivo. A region of PspA has homology to the RNA polymerase binding region of ς⁵⁴-dependent activators, which suggests that the mechanism of inhibition may involve a nonproductive interaction with RNA polymerase (88). PspA is peripherally associated with the inner membrane. PspB and PspC are also components of the inner membrane, and they cooperate in activating transcription, probably by antagonizing the effects of PspA (112, 163). PspD may similarly antagonize PspA (unpublished results cited in reference 112). No function is apparent for PspE (112). The induction mechanism is stimulus specific. Induction by the gene IV product of the filamentous phage f1 requires PspB, PspC, and PspD, whereas induction by heat shock requires none of these proteins. Other stimuli may require one or more of these regulatory proteins (112). The IHF dependence also varies with the stimulus (164).

PspF negatively autoregulates its own synthesis by binding to the same sites that activate psp operon expression (88). PspA, PspB, or PspC does not affect this autoregulation, which implies that these proteins do not affect the binding of PspF to DNA (88).

Proteins of the heat shock response also affect psp induction. Many but not all of the stimuli that induce the psp operon also induce the heat shock response. However, loss of ς³², the heat shock sigma factor, results in higher and longer expression of the psp operon (25, 26, 112). The mechanism of this regulation is not known.

rtcBA Operon

The promoter for the rtcBA operon has the highest score (95.4) of known ς⁵⁴-dependent promoters. The existence of the promoter was shown by primer extension in wild-type E. coli and by failure to observe the transcript in an rpoN mutant (62). The ς⁵⁴-dependent transcript was the only detectable transcript in primer extension experiments. Possible IHF binding sites were identified between 46 and 68 bases upstream from the start site of transcription. Expression of the rtcBA operon appears to require the divergently transcribed rtcR, which has a deduced product that is homologous to other ς⁵⁴-dependent activators. Detectable expression of the operon requires artificial overproduction of RtcR. Deletion of its amino-terminal domain also increases expression, which suggests that this domain inhibits the activity of RtcR. The stimulus that controls the activity of RtcR is not known.

RtcA catalyzes the ATP-dependent formation of 2′,3′-cyclic phosphodiester from an RNA with a 3′ phosphate at its 3′ end. The function of this activity is unknown, although such cyclic intermediates may be required for RNA ligation reactions (19). This enzyme is found from E. coli to HeLa cell extracts. An E. coli strain with 90% of rtcA deleted had no phenotype when grown in Luria-Bertani or minimal M9 medium (62). The activity of RtcB is not known.

zraSR (hydHG), zraP, and the Response to Zn²⁺ and Pb²⁺

The products of zraSR (previously called hydHG) are a membrane-associated sensor kinase and a response regulator, respectively. They were initially implicated in the control of hydrogenase 3 synthesis (150), but this control was observed only in an fhlA mutant and was subsequently shown to be nonspecific (99). The gene divergently transcribed from zraSR, zraP, had been implicated in tolerance to high Zn²⁺; therefore, the effect of Zn²⁺ on gene expression was examined (99). In response to high Zn²⁺ or Pb²⁺ concentrations, ZraR and ZraS specifically activated zraP, which is divergently transcribed from zraSR, and also autogenously activated zraSR expression. Purified ZraR bound in the zraP-zraSR intergenic region. Metal-induced expression required ς⁵⁴ in vivo, and potential binding sites for Eς⁵⁴ were readily identifiable for zraP and zraSR, with scores of 88.6 and 76.2, respectively. In addition to these promoters, zraSR appeared to have a weak constitutive promoter, which ensures basal synthesis of the sensor and response regulator (99).

The most important system for Zn²⁺ tolerance is the zntA-zntR system, which codes for a Zn²⁺ efflux protein and a Zn²⁺ binding MerR-like transcriptional activator, respectively (15, 27, 132). Its loss results in Zn²⁺ hypersensitivity (132). In contrast, loss of the zraP-zraSR system is observable only in a longer lag during the transition to medium with Zn²⁺ (99). The precise physiological function of the zraP-zraSR system has yet to be determined. However, it is possible that it acts as a sensor of extracellular Zn²⁺ while the ZntR system responds to intracellular Zn²⁺ (99). It is not known whether the regulatory circuits of the ZntR and ZraR systems overlap.

OTHER GENES WITH HIGHLY RANKED POTENTIAL Eς⁵⁴ BINDING SITES

Computer analysis identified 25 properly oriented intergenic sites with a score greater than 76, which is a good indicator of an Eς⁵⁴-dependent promoter (discussed above). This section briefly describes the seven sites and the operons they potentially control, if they have not been discussed already. Nitrogen limitation does not affect the expression of any of these operons (176).

The site preceding rpoH, which codes for the heat shock sigma factor, has a score of 92.8, which is the second highest computer score for a Eς⁵⁴ binding site. It is normally spaced relative to the rpoH structural gene and the adjacent upstream gene. Furthermore, it is conserved among bacteria (123). There is an obvious rationale for having a ς⁵⁴-dependent promoter for rpoH. ς^H activates the synthesis of several proteases, which could transiently supply amino acids. However, we have been unable to demonstrate the existence of a transcript from the putative ς⁵⁴-dependent promoter from nitrogen-limited cells, which assumes, perhaps erroneously, that NR_I is the activator (A. Kiupakis and L. Reitzer, unpublished).

A possible b2710-ygbD operon has a potential ς⁵⁴-dependent promoter with a score of 86. BLAST analysis suggests that b2710 codes for a flavodoxin or a rubredoxin, a redox protein, and that YgbD has homology to oxidoreductases, such as rubredoxin reductase. The gene divergently transcribed from the b2710-ygbD operon is ygaA, and it specifies a potential ς⁵⁴-dependent activator that might regulate b2710-ygbD expression. Nothing else is known about this operon and its expression.

The site preceding the potential yfhKGA operon has a score of 83.9. The yfhKGA operon codes for a potential sensor kinase, a protein of unknown function, and a potential ς⁵⁴-dependent activator, respectively. The putative yfhKGA operon is upstream from and transcribed in the same direction as glnB, which specifies an important regulator of the Ntr response, P_II (102). glnB is not part of an operon containing yfhKGA, since the major glnB promoter precedes the glnB structural gene (102). Furthermore, glnB on a plasmid complemented the altered glnALG regulation in a glnB mutant, which implies that defects in the putative yfhKGA operon do not contribute to the altered regulation of the glnB mutant (102). Nothing else is known about the putative yfhKGA operon.

The potential Eς⁵⁴ binding sites preceding kch, topA, and yaiS have scores of 85.4,76.5, and 77.3, respectively. kch codes for a potassium channel, topA codes for topoisomerase I, and yaiS specifies a protein of unknown function. The distance between the putative Eς⁵⁴ binding site and the translational start site for kch (264 bases) or topA (201 bases) is larger than that for any known ς⁵⁴-dependent promoter (the range is from 34 to 136 bases). Furthermore, only 20 bases separate the 5′ end of putative Eς⁵⁴ binding site for kch from the adjacent upstream gene, which probably excludes the possibility of an activator binding site. The yaiS intergenic region (706 bases) is larger than that for any intergenic region containing an authentic ς⁵⁴-dependent promoter (the range is from 148 to 507 bases). The large intergenic regions for these three genes suggest that these sites are false positives.

The site preceding ybhK has a score of 78.8. Its product is homologous to RocR, a regulator of arginine catabolism in Bacillus subtilis, which is homologous to NR_I (30, 61). However, the homology does not extend to the ς⁵⁴ binding domain, which implies that YbhK is not an activator of ς⁵⁴-dependent genes.

ς⁵⁴-DEPENDENT ACTIVATORS

A discussion of the physiological function of ς⁵⁴ will account for all of the functional ς⁵⁴-dependent activators. The signature of these activators is a highly conserved activation domain that binds and hydrolyzes ATP and interacts with Eς⁵⁴. Almost all ς⁵⁴-dependent activators also have an amino-terminal regulatory domain and a small carboxy-terminal DNA binding domain. To identify ς⁵⁴-dependent activators, we used PspF (which lacks an amino-terminal regulatory domain) and the central activation domain of NR_I as probes for BLAST searches. Both probes identified the same proteins: AtoC, the product of b1201, FhlA, HyfR, NR_I, PrpR, PspF, RtcR, YfhA, YgaA, YgeV, and ZraR (HydH). The search also identified TyrR, an activator of ς⁷⁰-dependent genes of aromatic amino acid synthesis (45). It lacks a region of the activation domain that has been implicated in the interaction with Eς⁵⁴ (113). All the other proteins are homologous throughout the activation domain, which suggests that they activate ς⁵⁴-dependent promoters and not ς⁷⁰-dependent promoters. The functions of 10 of these proteins have already been discussed. The genetic contexts for the remaining two, the product of b1201 and YgeV, provide clues to their possible function and are discussed in this section. Table 4 summarizes the functions of these proteins and the genes that they activate. Curiously, if autogenous regulation is excluded (for ZraR), then only two of these regulators, NR_I and FhlA, are known to activate more than one operon.

TABLE 4.

Activators of ς⁵⁴-dependent promoters in E. coli

Activator	Function	Genes activated
AtoC	Acetoacetate catabolism	atoDAEB
b1201	Dihydroxyacetone catabolism?	b1200-b1199-ycgC?
FhlA	FHL	hyp and hyc, operons, fdhF, and hydN-hypF
HyfR	Hydrogenase 4 control?	hyf operon?
NRI	Nitrogen assimilation	argT-hisJQMP, astCADBE, b1012-b1006, ddpXABCDE, glnALG, glnHPQ, glnK-amtB, gltlJKL, nac, potFGHI, yeaGH, ygjG, and yhdWXYZ
PrpR	Propionate catabolism	prpBCDE
PspF	Phage shock	pspABCDE
RtcR	RNA terminal cyclase	rtcBA
YfhA	?	?
YgaA	?	?
YgeV	Purine catabolism control?	?
ZraR	Zinc tolerance	zraP, zraSR

The b1201 gene specifies an apparent ς⁵⁴-dependent activator. Genes flanking b1201 are transcribed in the opposite direction, which implies that b1201 is monocistronic. The b1201 gene is divergently transcribed from three genes, b1200-b1199-ycgC, which might form an operon. BLAST analysis suggests that b1200 and b1199 are homologous to a dihydroxyacetone kinase whereas YcgC is homologous to components of the phosphoenolpyrurate-dependent phosphotransferase system. Dihydroxyacetone kinase is an enzyme in the oxidative branch of glycerol fermentation (24). Klebsiella pneumoniae can ferment glycerol, but E. coli cannot. Nonetheless, a triply mutated E. coli can convert glycerol to dihydroxyacetone, which is initially excreted and subsequently metabolized (152). Furthermore, wild-type E. coli can use dihydroxyacetone as the sole carbon source (as long as the phosphate concentration is kept low) (86). Dihydroxyacetone kinase has yet to be assayed from E. coli. In Streptococcus faecalis, phosphotransferase system-dependent phosphorylation stimulates dihydroxyacetone kinase activity (48). Similarly, YcgC may be required for kinase activity in E. coli. In Citrobacter freundii, dihydroxyacetone kinase synthesis requires ς⁵⁴ (47). Therefore, it would not be surprising if similar control was found for the E. coli genes. Unfortunately, computer analysis does not identify a potential ς⁵⁴ promoter in the vicinity, perhaps because the ς⁵⁴-dependent promoter has been lost.

YgeV is a potential ς⁵⁴-dependent activator. Several genes in the vicinity of ygeV appear to code for enzymes of purine metabolism (172). Strains with a disruption of YgeV grow faster with aspartate as the sole nitrogen source without exogenous purines (H. Xi and L. Reitzer, unpublished observation). A strain with a disruption of the xdhABC operon, which codes for subunits of a xanthine dehydrogenase, has a similar phenotype (172). xdhABC appears to have a ς⁵⁴-dependent promoter and a ς⁵⁴-independent promoter (172). Since strains with a disruption of ygeV or xdhA have similar phenotypes, it is possible that YgeV is required for transcription from the ς⁵⁴-dependent xdhA promoter.

PHYSIOLOGICAL FUNCTION OF ς⁵⁴

Possible Relationship between the ς⁵⁴-Dependent Genes

With a nearly complete set of ς⁵⁴-dependent genes (verified or potential), it is appropriate to ask whether the physiological themes of these genes are related. Many ς⁵⁴-dependent genes are involved in nitrogen assimilation. These genes specify GS, the regulators NR_I and Nac, several transport systems, and a few catabolic operons. Is there a relationship between the ς⁵⁴-dependent genes of nitrogen metabolism and the other ς⁵⁴-dependent genes? One possibility is that there is no relation between them. This possibility makes one interesting prediction: ς⁵⁴ is present in excess, and expression of one ς⁵⁴-dependent gene will not affect any other. Considering the low level of ς⁵⁴ found in some strains (87), it is reasonable to question whether this is the case.

Another explanation for the apparent diversity of the ς⁵⁴-dependent genes is that certain conditions might make nitrogen assimilation very difficult, and several products of ς⁵⁴-dependent genes might remedy the problem. It has been suggested that genes of the FhlA regulon may have coevolved with the genes of nitrogen assimilation (101). pH homeostasis provides a rationale for such coevolution. The function of the FhlA regulon is to increase the pH in an acidic environment. This could help nitrogen assimilation, because of the mechanism of ammonia transport. The AmtB protein catalyzes facilitated diffusion of NH₃ but not NH₄⁺ (149). pH determines the extent of ionization. If the external environment is acidic relative to the cytoplasm, NH₃ will leak out of the more basic cytoplasm (either with the AmtB carrier or without the carrier, because NH₃ is membrane permeable). Subsequent protonation of the external NH₃ will make it difficult to bring NH₄⁺ back into the cell. The FhlA regulon, which increases the external pH, perhaps just locally, might alleviate this problem and facilitate nitrogen assimilation.

The psp operon may also alleviate a situation that impairs nitrogen assimilation. Since the psp operon responds to energy or nutrient limitation, it has been proposed that the ATP concentration controls pspABCDE expression (112, 165). The conditions that induce the psp operon may also modulate the expression of genes that require NR_I, which itself hydrolyzes ATP (166) and is severely inhibited by ADP (D. Fewell and L. Reitzer, unpublished observation). Another mechanism by which psp expression can affect nitrogen assimilation is based on the proposal that PspA maintains the proton gradient, whose collapse would impair energy generation. Under such conditions, energy-consuming nitrogen assimilation might be inappropriate. (It has been estimated that 1 g of E. coli requires about 57,000 μmol of ATP and that if all nitrogen is assimilated via GS, then the GS reaction itself consumes about 10,500 μmol of ATP. This is obviously a major strain on cellular resources.) In this context, it should be noted that the expression of the psp, ato, and prp operons are probably linked. Fatty acids can collapse the proton gradient and presumably induce the psp operon, and the subsequent products of fatty acid catabolism will induce the ato and prp operons.

If certain conditions make nitrogen assimilation difficult, then not only might the more active ς⁵⁴-dependent genes that are not directly involved in nitrogen assimilation alleviate these conditions, but also their expression might downwardly modulate the expression of the ς⁵⁴-dependent genes of nitrogen assimilation by competing for ς⁵⁴. This is especially plausible for strains with a low level of ς⁵⁴, such as strain MC4100, which may contain as few as 20 molecules of ς⁵⁴ per cell (87). Furthermore, PspA may have the ability to inactivate all ς⁵⁴-dependent transcription. If this is the case, expression of the psp operon might lower the availability of ς⁵⁴ and also inactivate the activators of ς⁵⁴-dependent genes. The inhibition of the activators may be important even if ς⁵⁴ is present in excess.

In summary, if the genes of the FhlA regulon and the psp, ato, and prp operons alleviate conditions that are detrimental to nitrogen assimilation, the vast majority of ς⁵⁴-dependent genes in E. coli have a function that is related to nitrogen assimilation.

Evolutionary Persistence of ς⁵⁴

The mechanism of ς⁵⁴-dependent transcription is complex and requires a large regulatory region. This raises the question why such cumbersome transcriptional control has been evolutionarily maintained. One advantage of ς⁵⁴-dependent control is the wide range of activity. PspA and GS (both products of ς⁵⁴-dependent operons) can become a few percent of the proteins of E. coli. Furthermore, expression of these operons can be completely suppressed. This could be important for enzymes of nitrogen assimilation, which consume energy and withdraw intermediates from central metabolic pathways, especially the citric acid cycle. The advantage of such absolute control may be to prevent the rapid and catastrophic depletion of resources. The potential for such a loss has been demonstrated by removal of just one layer of nitrogen assimilation control, the adenylylation system for GS, which can result in glutamate depletion (96).

The size constraints of ς⁵⁴-dependent promoters (the need for binding sites for Eς⁵⁴, distant activators, and perhaps a DNA bending protein) may counterbalance the potential advantages of ς⁵⁴-dependent promoters and also minimize the number of such promoters in a single organism. Such reasoning could account for the seemingly limited number of ς⁵⁴-dependent promoters in E. coli, which we estimate to be about 30 (discussed above). ς⁵⁴ is widespread among bacteria, and the ς⁵⁴-dependent operons code for proteins with a variety of functions (151). The sheer diversity of these functions suggests that these genes are not always associated with nitrogen assimilation. Nonetheless, the possible evolutionary pressure to maintain few ς⁵⁴-dependent promoters within a single organism may limit the function of ς⁵⁴-dependent proteins to a few physiologically related themes.