Chapter 80

Regulation of Gene Transcription by Extracellular Stimuli

ALEXANDER J. NINFA

INTRODUCTION

Elsewhere in this book, examples of the regulation of intensively studied genes and operons are presented, and several examples of the global regulation of gene expression are discussed. As a prelude to that discussion, it may prove useful to present a brief overview of the mechanisms by which various extracellular stimuli bring about transcriptional regulation. This discussion will constitute the first part of this chapter. In the second part of this chapter, I will discuss in more detail the regulation of transcription by two-component regulatory systems, as this constitutes a common mechanism for the coupling of gene transcription to environmental stimuli. I will conclude the chapter with a few remarks on how the cumulative effects of specific and global controls result in physiological regulation.

OVERVIEW OF THE STIMULI AND MECHANISMS OF SENSATION

Bacteria regulate gene transcription in response to an amazing array of environmental stimuli. I like to think of these as falling into four general classes: stimuli that exert their effect nonspecifically on various cellular components, nutritional aspects of the environment, the presence of other cells of the same type, and the presence of (or location in) a suitable alternative niche, such as a host. Environmental conditions such as temperature, osmolarity, pH, and the presence of noxious conditions that evoke a specific or general protective response, such as DNA-damaging agents, constitute the first class, while responses to the presence or absence of minerals, energy sources, electron acceptors, and specific metabolites constitute the second class. Cellular responses to these environmental conditions are fairly easy to study, and thus they have received the most attention. By the third class I mean those instances where cells sense or signal to similar cells, and the fourth class is intended to include host-parasite interactions and those stimuli resulting in the formation of a different or "developed" cell type. Not surprising, the sensation of this array of stimuli occurs by many different mechanisms.

Most of these stimuli are sensed within the cell by soluble cytoplasmic components, either because the stimulus is able to penetrate the cell or by sensation not of the environmental stimulus itself but rather of the effects of the stimulus on cellular components or metabolism. For example, the presence of noxious conditions that cause DNA damage is sensed by monitoring the status of the DNA (105, 111; chapter 121). The condition of elevated temperature is sensed in several ways, with both the direct sensation of elevated temperature (94, 129, 144, 217) and the sensation of an effect of elevated temperature, probably the presence of misfolded proteins (129, 205, 206, 207), playing a role. Similarly, many nutritional stimuli are sensed indirectly by virtue of their effect on cellular metabolism. An example is the mechanism for the sensation of the presence of the preferred nitrogen source, ammonia, which is sensed indirectly by the measurement of intracellular concentrations of metabolites, such as glutamine and 2-ketoglutarate, that are significantly affected by the presence or absence of ammonia (1, 48). Additional examples are provided by other amino acid anabolic systems, where either the intracellular concentration of the amino acid itself or the presence of the appropriately charged tRNA is measured (chapters 55 and 81). Finally, for almost all catabolic systems, the substrate, or a compound derived from the substrate, is sensed within the cytoplasm. For certain of these, sensation of the inducer within the cell allows for additional global control by inducer exclusion (chapter 75). In catabolic systems, conversion of the stimulating substrate to the inducer allows for the additional evaluation of the cell’s capacity to successfully utilize the substrate (chapter 85). In all of these cases, the cell is not monitoring the environment directly; rather, it is monitoring itself: the integrity of essential components, metabolism, and its capacity for altering this metabolism.

The mechanisms of sensation of these intracellular stimuli and communication of this information to the transcriptional apparatus may be simple or complex, depending on the requirements of the individual system and on whether the stimulus elicits a global or specific response. In the simplest case, the stimulus directly alters the activity of a specific transcription factor, such as a repressor or activator, to regulate the appearance of specific gene transcripts. A prototypical example would be the ara system, where the activity of a specific transcription factor, AraC, is regulated by the inducer. In other cases, as noted above, a stimulating compound is converted directly to an internal inducer which controls the activity of a transcription factor. Examples of the latter phenomenon are found in the systems responsible for the utilization of lactose in Escherichia coli (chapter 85) and the utilization of histidine in Klebsiella aerogenes (189). The stimulus may also result in the synthesis of a distinct compound that serves as a second messenger, such as ppGpp, cyclic AMP, or homoserine lactone (24, 27, 78, 109, 131; chapters 85 and 92) which regulates the transcriptional apparatus directly (83) or through accessory transcription factors (34, 37, 53, 100). Finally, the stimulus may influence the transcriptional apparatus through a signal transduction cascade in which one or more proteins participate in the regulation of a transcription factor (see below). The common theme in all of these cases is that a transcription factor subject to control exists in two alternative states, corresponding to activity or the lack thereof, and the equilibrium between these states is regulated by the stimulus.

In contrast to those stimuli discussed above, certain stimuli are sensed at the surface of the cell; that is, the environment is directly monitored. The prototypical example of this phenomenon is found in the elegant diauxie experiments of Monod, first demonstrating catabolite repression. The wisdom of those experiments, as recently noted by Meadow and colleagues (133), was that a diauxic growth pattern was observed in cultures containing glucose and an additional carbon source that gives rise to glucose within the cell. Thus, extracellular glucose is the stimulus, as opposed to glucose per se. The mechanism by which extracellular glucose brings about inducer exclusion and catabolite repression and the related mechanism by which external β-glucosides regulate the transcription of genes necessary for their utilization (12, 13) are discussed in chapter 75. Other examples of stimuli that are sensed at the surface of the cell include pH (226), osmolarity (167), availability of P_i (chapter 87), glucose-6-P (86, 87, 92), availability of electron acceptors (197), the presence of or in host cells (59, 65, 178, 214), the presence of mating partners (227), and the presence of various small molecules that influence the cellular chemotaxis machinery (202, 215; chapter 73); of these, all except the last influence gene transcription. For most of these, the reason why sensation must occur at the cell surface can be rationalized. A common theme of these cases is the presence of specific transmembrane receptors that communicate the relevant information across the cell membrane. These receptors must exist in (at least) two states, corresponding to the presence or absence of the stimulus, and this serves as the switch by which a component of the transcriptional apparatus within the cell is controlled. In some cases, the cytosolic portion of a transmembrane receptor is known or implicated to be directly involved in transcriptional regulation (134, 226); in other cases, the cytosolic portions of transmembrane receptors regulate transcription indirectly by controlling the activity of transcription factors. Numerous examples of the latter phenomenon, in which transcription factors are controlled by reversible phosphorylation, will be discussed in the section on two-component regulatory systems.

HOW DOES THE STIMULUS BRING ABOUT THE REGULATION OF TRANSCRIPTION?

As noted above, in the simplest case the stimulus exerts direct allosteric control of the activity of a specific transcription factor. For example, binding of a ligand may favor a conformation of the protein that is better able to bind to DNA or to interact with RNA polymerase (35, 56, 100, 187). Alternatively, an allosteric mechanism may control the oligomeric state of a transcription factor, with only one of these species representing the active species. Other common mechanisms of control of transcription factors involve covalent modifications, such as phosphorylation (12, 13, 147, 200), methylation (105, 111), oxidation/reduction (102, 241), and proteolysis (76, 112), the consequences of which may similarly control the ability to bind to DNA or to interact with RNA polymerase. Noncovalent mechanisms such as activation (114, 203) and inhibition (108) are also seen, including titration phenomena where a transcription factor is sequestered in the absence of the stimulus (20, 23, 156, 157).

Global regulation of gene expression may be achieved by alteration of the components of the general transcriptional apparatus, that is, RNA polymerase and the chromosomal DNA. The specificity of RNA polymerase for promoter sequences is altered by highly regulated changes in the population of sigma subunits available for holoenzyme formation. E. coli and Salmonella typhimurium (official designation, Salmonella enterica serovar Typhimurium) contain a major sigma subunit (σ ⁷⁰) and at least five minor sigma factors (σ ²⁴, σ ²⁸, σ ³², σ ³⁸, and σ ⁵⁴ [14, 26, 49, 67, 72, 79, 106, 142, 155, 211; reviewed in references 117 and 118]); certain of these holoenzyme forms have absolute specificity for a given family of promoters (67, 72, 79); other holoenzyme forms have overlapping promoter specificity (211). Considerable flexibility in promoter use is therefore possible by the regulation of sigma factor synthesis (50, 52, 115, 128, 143, 144, 206, 222, 223) and activity (9, 23, 45, 135, 155, 156, 205, 207), with the latter process often involving sequestration of the sigma factor or the control of sigma factor stability. Additional mechanisms for regulation of the general transcriptional apparatus include regulatory covalent modification of the core subunits (36, 60, 158, 159), production of second messengers that are allosteric effectors of the enzyme (83), or factors that affect the suitability of the chromosome to serve as a template, such as by alteration of the superhelical density (chapter 51) or flexibility (39, 55, 169, 183; J. Feng, T. J. Goss, R. A. Bender, and A. J. Ninfa, in press) of the DNA.

Global control may also result from the recruitment of various genes into regulons subject to control by accessory transcription factors that are not components of the general transcription apparatus (i.e., not subunits of the polymerase). A given promoter may be subject to control by multiple transcription factors, in which case these controls may be concerted, antagonistic, or independent. An extreme example of this process is found in the regulation of transcription of the gene for manganese superoxide dismutase (sodA) in E. coli; in this case, four global regulators (ArcA, SoxQ, SoxS, and Fur) act to control transcription from a single promoter (42). Alternatively, a gene may be recruited into different regulons by the acquisition of separate promoters that are subject to distinct controls (reviewed in reference 40).

In many cases, transcription factors can be recognized as belonging to families of related proteins. The most important of these are the LysR family (69, 187), the AraC family (56, 171), the σ family (117), the Y-box family (238, 239), the response regulators (RR) of the two-component systems (which form several subfamilies), the CRP family (100; chapter 85), and the LacI-GalR family (chapter 85). The members of the first five of these families that are found in E. coli and S. typhimurium are listed in Table 1, and schematic illustrations of the "body plan" for these transcription factors are presented in Fig. 1 and 2.

Fig. 1Families of transcription factors. H-T-H, helix-turn-helix DNA recognition motif. For further details, see text.

Source:

Fig. 2Conserved domains found in two-component system regulators and arrangement of these domains in signal transduction systems. (A) Conserved domains found in HPK (top) and RR (bottom). The wide lines indicate the conserved domains, attached thin lines signify the usual location of nonconserved domains. Conserved sequence elements are indicated using the standard single-letter amino acid code, with very highly conserved residues appearing in boldface type. For further details, see text. (B) Arrangement of signalling domains in signal transduction systems. The HPK domain is indicated with crosshatched boxes, the RR domain is indicated with unfilled boxes, the nonconserved domains associated with the RR domain are indicated as vertically hatched boxes, and the nonconserved domains associated with the HPK domain are indicated by a thin line. (1) Single transmembrane HPK and single RR, as in the Omp system; (2) single transmembrane HPK with attached RR domain and single soluble RR, as in the Rcs system; (3) two HPK, with associated RR domain and a single soluble RR, as in the Arc/Cpx system; (4) two HPK and two RR, as in the Pho and Nar systems; (5) soluble HPK and single RR, as in the Ntr system; (6) one soluble HPK, one transmembrane HPK, and two RR, as in the Spo system; (7) unorthodox soluble HPK and two RR, as in the Che system; (8) unorthodox HPK, RR bearing two tandem RR domains, and another RR, as in the Frz system. For further details, see text and reference 161.

Source:

Table 1Families of homologous transcriptional regulators found in E. coli and S. typhimurium

The LysR family consists of more than 50 different transcriptional activators and repressors that share a highly conserved N-terminal DNA-binding domain consisting of a helix-turnhelix motif and flanking sequences (Fig. 1; reviewed in reference 187). For most members of the LysR family, the less-conserved C-terminal portion of the protein has a sensory function (e.g., see references 101 and 102); in the remaining cases a coinducer has not been identified and the protein is apparently always in the active conformation (e.g., see references 192 and 193; reviewed in reference 22). Transcription regulation by the LysR family members is likely to be due to a change in the DNA-binding properties associated with the binding of the coinducer. In several cases it has been shown that in the absence of the coinducer the LysR type of regulator binds to DNA sequences located at position –65 relative to the transcriptional start site of the regulated promoter; in the presence of coinducer, additional contacts with DNA are observed near position –35 and/or bending of the DNA is observed (reviewed in reference 187).

The AraC family of transcription factors consists of 27 proteins sharing a characteristic helix-turn-helix DNA-binding motif near the C terminus (chapter 85; also reviewed in references 56 and 171). The poorly conserved N-terminal portion of these proteins is involved in allosteric regulation by the coinducer and dimerization (35). In several cases, including the AraC protein, mutations in the N-terminal portion of the protein that alter the specificity for the coinducer have been obtained. In the case of the SoxS transcriptional activator, only the C-terminal DNA-binding domain is present, and in this case the protein is thought to be always in the active state, with regulation exerted solely by control of SoxS synthesis.

The σ family consists of the major and minor sigma factors of bacteria (reviewed in reference 117) with the exception of σ ⁵⁴ (185). Four highly conserved regions designated regions 1 to 4 have been noted (Fig. 1); secondary structure predictions and protease digestion studies suggest that regions 1 and 2 may constitute an N-terminal domain consisting primarily of β-sheet and that regions 3 and 4 may constitute a C-terminal domain consisting mostly of α-helix. Region 1 is found only in the primary sigma factors of various organisms. Several functions have been assigned to region 2, including the binding of the core polymerase, contact with DNA sequences near position –10 relative to the site of transcription initiation, and melting of the DNA strands surrounding the site of initiation. Region 4 contains a helix-turn-helix DNA-binding motif and is involved in recognition of DNA sequences at position –35. Similarity to region 4 of the σ family is observed in a large number of transcription factors in both eukaryotes and prokaryotes; these include the homeodomain proteins, the Cro protein of bacteriophage λ, and the transcriptional activator comprising the FixJ subfamily of two-component system RR proteins (93).

The Y-box family consists of 10 eukaryotic proteins involved in transcriptional activation and repression and translational regulation. These proteins share a highly conserved N-terminal domain involved in the recognition of nucleic acids; depending on the protein, RNA, double-stranded DNA, or single-stranded DNA may be bound. There are two examples in prokaryotes, the E. coli transcriptional activator of the cold shock regulon, designated CS7.4 (61), and a similar protein from Bacillus subtilis, designated CspB (234). Both of these prokaryotic homologs consist of just the highly conserved nucleic acid-binding domain; presumably, they are controlled solely by the regulation of their synthesis. The structure of the Csp protein has been solved; it consists of a five-stranded β-barrel that resembles staphylococcal nuclease and the bacteriophage fd gene 5 product (188).

Although much remains to be learned, several principles as to the mechanisms by which accessory transcription factors bring about the activation or repression of transcription initiation are beginning to emerge (34, 40, 83, 84, 85, 100, 233; chapters 54, 79, 82, and 83). Activators may work by facilitating the formation of a complex between polymerase and promoters (closed complex), by facilitating the isomerization reactions by which this complex is converted to the active transcription complex in which the DNA strands are melted surrounding the site of initiation (open complex), by stimulating promoter clearance and entry into the elongation phase of transcription, or by stimulating several of these processes (132). Virtually all activators and repressors that regulate genes transcribed by the major form of RNA polymerase, Eσ ⁷⁰, contact the promoter DNA at a site that would permit direct contact with bound RNA polymerase (40). Often additional binding sites for the transcription factor are present as well, and these serve to increase the occupancy of the site mapping adjacent to the site of polymerase binding (referred to as the proximal site [40]) by cooperative interactions (63, 140). The proximal site for transcriptional activators almost always overlaps position –40; proximal sites for repressors are more variable and may overlap any segment of the promoter contacted by RNA polymerase (40). From these positions, repressors may exclude polymerase from the promoter or engage polymerase in a complex that prevents the progression of the initiation process (204).

The activation of transcription initiation from promoters recognized by the Eσ ⁵⁴ form of RNA polymerase seems to be fundamentally different (reviewed in references 18, 103, 104, 124, 125, 165, and 200). Transcription by this form of RNA polymerase always requires an accessory transcriptional activator; these bind to sites distant from the region bound by RNA polymerase (e.g., see references 31, 72, 107, 148, 174, 175, 184, 240, and Feng et al., in press). These distant sites are functionally equivalent to the transcriptional enhancers of eukaryotic cells (148, 175). The enhancer-bound activators always stimulate the formation of the open transcription complex from the closed complex (163). In one case, the distantly bound activator also stimulates the formation of the closed complex (136). The activators of σ ⁵⁴-dependent promoters have an ATPase activity that is required for transcriptional activation (19, 53, 229); thus, they have been characterized as "molecular engines" (165). Interaction of the enhancer-bound activator and the closed complex requires the looping out of the intervening DNA (208). This DNA looping is often facilitated by the binding of the integration host factor (IHF) protein to a site located between the enhancer and the promoter (39, 183). Indirect repression of such promoters may be exerted by factors that bind between the enhancer and the promoter and prevent the formation of the required DNA loop (Feng et al., in press).

TWO-COMPONENT SYSTEMS AS A PARADIGM FOR THE REGULATION OF GENE TRANSCRIPTION BY ENVIRONMENTAL STIMULI

Signal transduction in response to environmental and internal stimuli in bacteria is often performed by pairs of proteins that form two superfamilies based on the presence of homologous protein domains, the so-called two-component systems. Indeed, these systems constitute the most frequently encountered signal transduction mechanism in bacteria. Based on the number of such signal transduction systems known and the fraction of the E. coli and S. typhimurium genomes that has been sequenced, it has been estimated that there are likely to be >50 different two-component systems in these organisms (200). The homologies reflect the use of protein-histidine and -aspartate phosphorylation and dephosphorylation reactions as mechanisms of regulation. In the sections that follow, I will discuss briefly the current state of our knowledge as to how these signal transduction components and the accessory components that interact with them bring about the regulation of cellular processes in response to the stimulus. I will focus on those systems found in the subject organisms of this treatise; however, at points it will be necessary to consider results from other organisms when these provide insights that are likely to be generally applicable. Numerous reviews in this area (7, 8, 28, 66, 145, 152, 161, 198, 200, 201, 202, 218, 219), as well as a recent treatise (75), are available, and subjects adequately covered in those reviews will only be briefly noted here.

SEQUENCE, STRUCTURE, AND CELLULAR LOCATION OF TWO-COMPONENT SYSTEM PROTEINS

The family of signal transduction systems known as the two-component systems was discovered on the basis of the conceptual translation of DNA sequences, which revealed that these regulatory systems contain two types of protein domains (157, 198; Fig. 2A). In keeping with an earlier nomenclature (200), I will refer to these domains as the histidine protein kinase (HPK) domain and the RR domain and use the same terms in discussion of the proteins containing these domains. More than one version of each type of domain may be present in a given signal transduction system (reviewed in references 161 and 200), and several different arrangements are depicted schematically in Fig. 2B. The sequence relationships shared by these two types of domains have been adequately reviewed recently (161, 201, 218, 219); therefore, I will here confine my remarks to the most important structural features of these domains.

The HPK domain, usually about 225 amino acids in all, contains five conserved motifs designated the H, N, D, F, and G boxes (201; Fig. 2A), although in many cases one or more of these elements are not recognizable (reviewed in references 161 and 200). The HPK domain is always found linked to associated domains; usually, the HPK domain comprises the C-terminal domain of the protein (Fig. 2B). It is likely that these associated domains, which are unrelated to one another, are involved in the control of the activities of the HPK domain (discussed below), although this has been only shown genetically in a few cases (3, 16, 41, 82, 86, 123, 137, 167, 180) and biochemically in two cases (21, 57, 58, 116, 215). Frequently, the N-terminal domain associated with an HPK domain contains two or more transmembrane segments, resulting in the localization of the HPK protein to the cytoplasmic membrane (e.g., see reference 54). In other cases, HPK is a soluble cytoplasmic protein (73, 74, 141, 149). In some cases, the HPK domain is associated with an RR domain in a single protein (e.g., see references 62, 88, and 214; reviewed in references 161 and 200).

The RR domain, containing about 125 amino acids, may exist as a separate protein, may be associated with an HPK domain in a single protein, or in the usual case, is associated with additional domains (Fig. 2B). In the latter case, it seems clear that the RR domain is used to control the activity of the associated domains, which are often involved in transcriptional regulation. Several very highly conserved sequences are found within the RR domain (218, 219): near the N terminus of the domain, several acidic amino acids are found near position 12, an aspartic acid is typically found near position 55, a threonine residue is found near position 87, and a lysine residue is found near position 105. The three-dimensional structure of the RR domain that constitutes the CheY protein has been solved by X-ray crystallography and nuclear magnetic resonance analysis (30, 199, 220). These studies indicate that the domain consists of a five-stranded β-sheet surrounded by α-helices, with the highly conserved aspartate and lysine residues clustered at one end of the domain. Except for those cases in which an RR domain is fused to an HPK domain, all of the proteins containing an RR domain are soluble, cytoplasmic proteins.

Examination of the domains associated with RR domains indicates that these fall into several classes on the basis of conceptual translation of the DNA sequence (reviewed in reference 161). One class consists of domains that activate transcription by RNA polymerase containing the major sigma factor, σ ⁷⁰ (OmpR subfamily). This domain probably activates transcription by contacting the α subunit of RNA polymerase (196). Another class consists of domains that activate transcription by RNA polymerase containing σ ⁵⁴ (NtrC or NR_I subfamily). Yet another class is also involved in transcriptional activation by an unknown mechanism (FixJ subfamily). It is interesting that in each of these cases, other proteins exist that lack the RR domain but contain the domain involved in transcriptional activation (reviewed in reference 200). In those cases in which an RR domain is absent, another mechanism is used to control the activity of the associated domain. For certain members of the NR_I subfamily, it has been directly demonstrated that the central domain alone is sufficient for transcriptional activation in vivo (68, 77). In addition to the subfamilies noted above, there are numerous cases in which domains associated with a RR domain are unrelated to any other known protein domains.

ACTIVITIES OF THE HPK AND RR PROTEINS

The HPK are so named because these proteins catalyze the ATP-dependent phosphorylation of a conserved histidyl residue within the H box, which is usually located at the N-terminal end of the conserved HPK domain. That is, they catalyze their own phosphorylation or "autophosphorylation" (70, 71, 82, 91, 99, 231). Upon interaction with the response regulator, phosphoryl groups are transferred from this phosphoryl-histidine to a conserved aspartyl moiety in the RR domain (e.g., see references 70, 71, 91, 126, 231, and 242). The phosphorylation of the RR domain always results in the stimulation, as opposed to the inhibition, of the activity of associated domains. In addition to this role, certain HPK proteins have been shown to stimulate the dephosphorylation of the phosphorylated RR (e.g., see references 6, 38, 74, 81, 126, and 147), which results in the inhibition of the activity of domains associated with the RR domain. These "kinase" and "phosphatase" activities are discussed below.

The site of autophosphorylation of HPK proteins has been directly demonstrated in three cases by digestion of the phosphorylated protein and purification and sequencing of the phosphorylated peptide (70, 146, 177). These studies showed that the highly conserved histidine moiety within the H box is the site of autophosphorylation. In addition, several studies have shown that alteration of this conserved histidine to a nonphosphorylatable amino acid results in proteins lacking the autophosphorylation activity (e.g., see references 17, 70, 90, 92, 150, and 197).

Less is known concerning the site of ATP binding by the HPK proteins, but circumstantial evidence suggests that residues in the G box may be involved. This glycine-rich sequence bears a resemblance to other ATP-binding sites, where the glycine residues have been shown to interact with the phosphoryl groups of ATP. A mutation (G313A) within this portion of HPK of the Ntr system, NR_II, was observed to diminish greatly the formation of a covalent adduct with ATP upon UV irradiation (150).

Even less is known concerning the binding of RR to HPK. Obviously, the RR domain must interact with the H box. For the CheA kinase, an additional domain adjacent to the H box has been shown to be involved in binding the CheY RR (139, 210). However, the structure of CheA is considerably different from most of the other HPK proteins, and it is therefore hard to generalize this finding to the other HPK proteins, which do not have a recognizable version of this sequence adjacent to the site of autophosphorylation. The CheA HPK forms a fairly stable complex with the CheY RR protein, and this complex is dissociated by ATP, which results in phosphorylation of CheY (210). However, similar experiments performed with the Ntr system HPK, NR_II (NtrB), and its cognate RR, NR_I (NtrC), suggest that the interaction of these proteins is transient and stimulated by ATP; thus, the interaction of HPK and RR of various two-component systems may be quite different.

The HPK proteins that have been studied in vitro have been found to exist as dimers under physiological conditions (e.g., see reference 149). For three of the HPK proteins, EnvZ, CheA, and NR_II, it has been observed that trans-intramolecular phosphorylation of the subunits within a dimer occurs (150, 209, 237, 244). That is, one subunit binds ATP and phosphorylates the other subunit within the dimer. This was shown by demonstrating that mutants containing alterations of the histidinyl site of autophosphorylation can be complemented for the autophosphorylation function by inactive proteins containing mutations that alter the highly conserved residues within either the D or G box at the C-terminal end of the HPK domain. Presumably, these two types of mutants define the "site" and kinase functions, with the kinase function requiring the residues in the D, F, and G boxes. In the case of the NR_II protein, autophosphorylation was shown to proceed exclusively by the trans-intramolecular mechanism (150), and it is likely that this requirement is shared by the other HPK proteins. Recent work with the CheA kinase indicates that inactive proteins with mutations in the D and G boxes cannot complement within hybrid dimers; both of these conserved regions must be intact within a given subunit for that subunit to have kinase activity (A. J. Wolfe, personal communication). The observation that the D and G boxes constitute the kinase function is in agreement with the observation that a related protein from mammalian mitochondria, branched-chain keto acid dehydrogenase kinase, which lacks an H box but contains D and G boxes, is a serine protein kinase (164). Similarly, the SpoIIAB protein of B. subtilis lacks an H box but contains D and G boxes; this protein is also a serine kinase (9, 45).

The finding that the autophosphorylation reaction proceeds by the trans-intramolecular mechanism raises the possibility that the autophosphorylation rate may be controlled by the regulation of dimerization or by effectors that affect the relative orientation of the subunits within the dimer. Such a mechanism may regulate the autophosphorylation of the CheA kinase controlling bacterial chemotaxis, and this is discussed further in chapter 73. It should be noted, however, that this mechanism does not occur for the NR_II protein, which seems to exist as a stable dimer under all conditions examined. The difference between these two enzymes is highlighted by the results of kinetic analysis using the method of isotope exchange: the rate of isotope exchange catalyzed by NR_II is first order with respect to the enzyme concentration, while that catalyzed by CheA is second order with respect to the enzyme concentration (E. G. Ninfa, Ph.D. thesis, Princeton University, Princeton, N.J., 1992).

As alluded to above, the HPK proteins catalyze the exchange of phosphoryl groups between ATP and ADP (151; Ninfa, Ph.D. thesis); this reaction has been used to study the kinetics of the autophosphorylation reaction for CheA and NR_II. The kinetics of CheA autophosphorylation and dephosphorylation have also been studied by pre-steady-state methods, and the binding of nucleotides by CheA has been studied by the column chromatography method of Hummel and Dreyer (212). These studies revealed that the dephosphorylated forms of the enzymes have considerable affinity for both ATP and ADP. Furthermore, the phosphorylated forms of CheA and NR_II have considerably higher affinity for ADP than for ATP, which when bound results in the reversal of the autophosphorylation reaction (212, 231). Thus, it has been noted that fairly small changes in the intracellular ATP/ADP ratio may have a significant effect on the rate of autophosphorylation (201). The phosphorylated form of the FixL HPK protein of Rhizobium meliloti shows a similar property, and it has been proposed that this may have physiological significance since the process regulated by FixL, nitrogen fixation, is associated with a condition of high energy charge (3).

Several lines of evidence suggest that the phosphotransfer reaction by which phosphoryl groups are transferred from the histidinyl moiety of HPK to the aspartate moiety of RR is catalyzed by the RR proteins. The primary observation was that a small segment of CheA containing the phosphorylated histidine was competent for the transfer of phosphoryl groups to the CheY RR protein (70). Further studies indicated that small-molecule phosphoryl group donors could substitute for HPK in bringing about the phosphorylation of the RR proteins (121). For example, phosphoramidate has proven to be very effective as a phosphodonor for all RR tested to date, and other molecules such as acetyl phosphate and carbamyl phosphate may serve as phosphoryl group donors in various cases (19, 44, 53, 130). For the NR_I protein, RR of the Ntr system, it was shown that phosphorylation of the protein with the small-molecule phosphodonors acetyl phosphate, carbamyl phosphate, and phosphoramidate resulted in the same properties as did phosphorylation by phosphotransfer from NR_II ∼P, including the acquisition of the capacity to activate transcription from the glnAp2 promoter in vitro (53). Thus, Sanders et al. have suggested that NR_I should be considered a protein phosphatase whose transient covalent intermediate activates transcription (181).

The RR are phosphorylated at the conserved aspartate moiety found near position 55; this was shown by reduction of the acyl phosphate with borohydride followed by cleavage of the proteins and mapping of the site (91, 181, 182). In other cases, mutation of this conserved aspartate has been shown to eliminate the activity of the associated domain (e.g., see references 29, 88, 120, and 138), although this is not universally true (87, 141). The acyl phosphate moieties in different RR have characteristic stabilities ranging from several seconds to several hours (reviewed in reference 145). Since these acyl phosphates are stabilized upon chelation of Mg²⁺ or upon denaturation of the phosphoprotein, this "self-catalyzed" dephosphorylation has been referred to as the autophosphatase activity (99). Little is known about the mechanism of the autophosphatase activity. Mutations in the highly conserved lysine near position 105, adjacent to the phosphorylated aspartate in the three-dimensional structure, have been observed to increase the stability of phosphorylated RR (29, 120, 138).

In addition to the autophosphatase activity, in several cases it has been shown that HPK stimulates the dephosphorylation of the phospho-RR (e.g., see references 6, 38, 74, 81, 116, 126, 147, 190, and 221). This activity has been referred to as the regulated phosphatase activity (99). It is not clear whether this activity reflects a stimulation of the autophosphatase activity of RR or a separate activity of HPK, although the former possibility seems more likely. In either case, it is clear from the genetic analysis of several systems that this activity is an important aspect of the regulation (5, 6, 16, 17, 32, 81, 90, 92, 95, 123, 126, 180, 197, 228, 245). Those HPK proteins having both kinase and phosphatase activities are both positive and negative regulators of their target systems. For these, mutations that result in a defect in the negative regulation have in some cases been shown to result in a defect in the phosphatase activity (e.g., see references 6, 95, and 147). Indeed, in many cases deletion of HPK results in a low-level constitutive expression of the target system at a level higher than seen with the wild type under noninducing (repressing) conditions, reflecting the absence of this phosphatase activity. A possible basis for the low-level constitutive expression of the target systems in mutants lacking the HPK protein will be discussed in a later section.

The mechanism of the regulated phosphatase activity has been examined for the case of the nitrogen regulatory proteins NR_II and NR_I and the osmoregulatory proteins EnvZ and OmpR. The wild-type NR_II HPK protein is only a potent NR_I ∼P phosphatase in the presence of another signal transduction protein known as protein II or P_II, the product of glnB, and in the presence of ATP (95, 96, 99, 147). P_II forms a complex with NR_II (K_d, 6 × 10^–9 M [E. S. Kamberov and A. J. Ninfa, submitted for publication]), and this complex brings about the rapid dephosphorylation of NR_I∼ P (97). This dephosphorylation does not seem to occur by a reversal of the phosphotransfer reaction; that is, no NR_II ∼P intermediate is observed (96). Further support for this point comes from the study of a mutant form of NRII in which the active-site histidine has been converted to the nonphosphorylatable residue asparagine (H139N). The NRII-H139N protein is not autophosphorylated owing to the absence of the site of autophosphorylation, but it is a potent NRI~P phosphatase even in the absence of P_II (95). This phosphatase activity is further stimulated by P_II and by ATP. If this activity is the same as the regulated phosphatase activity seen with wild-type NR_II, as seems likely, then the data would indicate that the conserved H-box histidine of HPK is not required for the phosphatase activity. In a number of other cases, it has been observed that conservative substitution of the H-box histidine results in proteins that are defective in positive regulation but retain the capacity for negative regulation of their target systems in vivo (e.g., see references 90, 92, and 197). By analogy, these mutant proteins are likely to retain the regulated phosphatase activity.

It is of interest that the NR_II-H139N protein noted above has phosphatase activity in the absence of the P_II protein. This may indicate that the wild-type protein has similar activity but that this activity is masked by the kinase activity of NR_II. P_II seems to have neither an effect on the rate or extent of autophosphorylation of wild-type NR_II (96) nor an effect on the rate of transfer of phosphoryl groups from NR_II ∼P to NR_I. Nor is P_II itself a phosphatase (96, 147). Thus, although much remains to be learned, it seems that the autophosphorylation activity of NR_II and the transfer of phosphoryl groups to NRI are not affected by P_II and that the phosphatase activity is a distinct activity that is stimulated upon the binding of P_II. Note that if the role of P_II were to inhibit the kinase activity of NR_II, then P_II would not be expected to increase the rate of NR_I ∼P dephosphorylation by the NR_II-H139N protein.

The phosphatase activity of the EnvZ HPK protein is also stimulated by ATP, and it has been shown that hydrolysis of ATP was not required, since nonhydrolyzable analogs could satisfy the nucleotide requirement (81). The stimulation of the phosphatase activities of NR_II and EnvZ by ATP raises the possibility that in vivo this activity may also be regulated by the energy charge.

Mutations that eliminate the regulated phosphatase activity have been found to cluster in the vicinity of the H-box histidine and to occur in the nonconserved N-terminal domains (e.g., see references 16, 123, 137, and 180). It has been noted that the former class of mutations may signify that regulation of the phosphatase activity occurs by control of the affinity of the phosphorylated form of RR for HPK and that this class of mutations may specifically alter this interaction (201). One such mutant form of NR_II binds to P_II as well as wild-type NR_II but completely lacks the regulated phosphatase activity (Kamberov and Ninfa, submitted).

HOW DO THE STIMULI CONTROL THE PHOSPHORYLATION OF THE RR DOMAIN?

Starting from first principles, there are several different ways that the stimuli might control the intracellular concentration of the phosphorylated form of RR. For example, either kinase or phosphatase activity could be essentially constant and the stimuli could simply elicit the other activity (single-regulation model). There are several ways that this could be accomplished. If the conformational requirements for the two activities of the bifunctional enzyme are not mutually exclusive, one of the two activities could be independently regulated by the stimulus. Alternatively, the bifunctional enzyme may exist in two conformations, corresponding to the kinase and phosphatase activities. Since populations of bifunctional enzymes are involved, the elicitation of a very potent activity from a small fraction of the enzyme molecules would then look like the single-regulation model. That is, if the kinase and phosphatase activities are exclusively catalyzed by different conformations of the enzyme in equilibrium, a slight shift in this equilibrium would look like single regulation if the stimulus-induced activity is the more potent. Mathematical modeling has been used to demonstrate that such a model could explain the reciprocal regulation of ompF and ompC transcription by the EnvZ-OmpR two-component system (179). Furthermore, the NR_I/NR_II system seems to work in this way, as noted above. If this line of reasoning is followed to the extreme, the stimulus may shift the equilibrium between the two enzyme conformations completely, in which case the bifunctional enzyme will have two stable states, each with one of the two activities (reciprocal-regulation model). Finally, more complex models are possible. For example, if the conformational requirements for the kinase and phosphatase activities are not mutually exclusive, these may be independently regulated by the same or separate stimuli (dual-regulation model). Since the various two-component systems have been optimized for diverse physiological functions, there is no reason to expect that a single mechanism will account for how the stimulus regulates the intracellular concentration of the phosphorylated RR, and the existing data bear this out.

The FixL HPK protein of R. meliloti regulates nif and fix gene transcription by virtue of its control of the phosphorylation state of the FixJ RR protein (reviewed in reference 3). This protein represents an extreme consolidation of functions in a single protein, since it is both sensor and HPK, with both kinase and regulated phosphatase activity. The sensory function is provided by an N-terminal domain that is associated noncovalently with a heme moiety. Although the native protein is membrane associated, owing to four transmembrane segments at the extreme N terminus, membrane association is not required for sensation or for proper regulation of FixJ phosphorylation in response to oxygen in vivo or in vitro (2, 116). The results of experiments in which target genes and various regulatory genes were transplanted into E. coli suggest that all necessary sensory functions are contained within FixL itself and that FixL is both a positive and a negative regulator of FixJ activity (43). This conclusion is fully supported by the results of experiments with purified components: both the autophosphorylation of FixL and the phosphatase activity of FixL are regulated by oxygen (116). Thus, FixL exhibits either the reciprocal-regulation model or the dual-regulation model as defined above. The phosphotransfer from FixL∼P to FixJ is not regulated by oxygen.

The case of bacterial chemotaxis represents the other extreme in terms of compartmentalization of function, since the kinase, phosphatase, and stimulus-sensing activities are contained on specialized, apparently monofunctional, enzymes. The CheA protein lacks the RR∼P phosphatase activity, and the chemotaxis system includes a separate protein, CheZ, that catalyzes the dephosphorylation of the CheY RR. Since this system is considered in depth in chapter 73, only two comments will be offered here. It is clear from the biochemical characterization of the system reconstituted from purified components that the autophosphorylation of the CheA kinase is controlled by the stimulus, which acts indirectly through transmembrane receptor proteins and the CheW protein (151, 191). Whether the phosphatase activity of CheZ is regulated by stimuli is not known.

In the case of nitrogen regulation (125), the HPK protein NR_II is a soluble cytoplasmic protein (149). As noted above, the autophosphorylation of this protein seems to be unregulated and constitutive, and the phosphatase activity is activated upon interaction with the P_II signal transduction protein (single-regulation model). The stimuli in this case apparently are small molecules indicative of the intracellular nitrogen status. These stimuli regulate the activity of P_II in two ways. First, the P_II protein directly binds to 2-ketoglutarate and ATP with high affinity and to glutamate with low affinity (97, 113), and the binding of these effectors to P_II regulates the interaction of P_II with NR_II and with other receptors of P_II (97; Kamberov and Ninfa, submitted). In addition, the P_II protein is subject to reversible uridylylation in response to these and other signals of nitrogen status (1), and only the nonuridylylated form of P_II is able to bind to NR_II (15, 97; Kamberov and Ninfa, submitted). Thus, in the case of NR_II, the stimuli act indirectly to control the phosphatase activity of NR_II by virtue of their effects on the P_II protein.

The two examples cited above, the Che and Ntr systems, highlight the role of accessory proteins in the detection of the environmental stimulus. Accessory proteins have been shown to play a similar role in a large number of other two-component systems, and only a few examples will be provided here. Transcription of the Pho regulon (chapter 87) is activated by the RR PhoB, whose phosphorylation state is primarily regulated by the PhoR HPK protein. PhoR is both a positive and a negative regulator of the Pho regulon and has both kinase and phosphatase activities (126). The stimulus in this case seems to be neither extracellular nor intracellular P_i but rather some feature or state of the major transport system for P_i known as the Pst system that is altered by external P_i. For example, the stimulus may be a conformation of one of the components of the Pst transport system that is only obtained when the transport system is bound to or actively transporting P_i. A component associated with this transport system, the PhoU protein, plays a role that is formally analogous to that of P_II in the Ntr system, in that it is required for the negative regulation of the Pho regulon under conditions of excess environmental P_i (chapter 87). Several models have been proposed to account for this role. For example, PhoU, possibly in association with other Pst components, may directly regulate the phosphatase activity of PhoR via a protein::protein interaction in a manner analogous to the Ntr P_II protein. Alternatively, PhoU, possibly in association with other Pst components, may synthesize a diffusible small-molecule effector that regulates PhoR activity (173). The latter model is based on the observation that the constellation of phosphorylated small molecules in the cell is altered in Pst and PhoU mutants (173).

The UhpB-UhpA two-component system regulates the biosynthesis of the UhpT permease required for the uptake of phosphorylated sugars by E. coli (92). Since deletion of the UhpB HPK in this system results in a low-level constitutive expression of the target promoter, it seems likely that this HPK has both kinase and phosphatase activity. Both the stimulus and sensor are known. The stimulus is extracellular glucose-6-P; it is detected by an accessory protein, UhpC, which has considerable similarity to the permease but lacks permease function (86, 87). One possible model for the regulation by external glucose-6-P is that it binds to and alters the conformation of UhpC and that this information is communicated to HPK by direct protein::protein contact between these two integral membrane proteins (86).

WHAT IS THE ROLE OF SMALL-MOLECULE PHOSPHODONORS?

As noted above, small-molecule phosphodonors can serve as substrates for the autophosphorylation of RR in vitro (19, 44, 53, 121, 130). Certain data suggest that this reaction occurs in vivo and may be of physiological significance, particularly for the case of the phosphorylated metabolic intermediate acetyl phosphate (53, 168, 225). Since the primary observations in this area have been reviewed recently (131, 224), I will confine my comments to the general outlines of the phenomenon. More information on the role of acetyl∼P in controlling two-component systems may be obtained from chapter 73; indeed, a role for an activated acetate compound in chemotaxis had been suspected for some time (236).

In many cases, deletion of the HPK protein of a two-component system does not result in the absence of expression of the target genes and operons but instead results in a low-level constitutive expression of these genes. Since deletion of the RR protein in these same cases results in the absence of expression of the target genes, it is clear either that an alternative mechanism for phosphorylation of RR exists or that the unphosphorylated RR has some capacity to activate transcription or both. Data from the Pho two-component systems suggest that for this system phosphorylation of the PhoB RR is required and there are several routes for the formation of the RR∼P. The Pho regulon (chapter 87) is activated by the phosphorylated form of the PhoB protein, which is primarily controlled by the PhoR HPK. In cells deleted for PhoR, a low-level constitutive expression of the regulon is observed, and this is dependent on a second HPK designated CreC (formerly PhoM). The CreC protein is apparently a part of a separate two-component system containing another RR, CreB, but retains the capacity to serve as a source of phosphoryl groups for PhoB (11). In the absence of both PhoR and CreC, there is very low expression of the Pho regulon, however, in cells containing additional mutations, the effect of which is the accumulation of intracellular acetyl phosphate, elevated expression of the Pho regulon is observed (110, 225). These results indicate that acetyl∼P is a positive activator of Pho regulon expression that acts through PhoB.

In the Ntr system the deletion of the NR_II HPK does not eliminate expression of the primary target, the glnA gene encoding glutamine synthetase. Genetic analysis indicates that in cells lacking NR_II, the activation of glnA expression is driven by acetyl~P: mutations resulting in the absence of acetyl~P greatly decrease activation of glnA, while mutations resulting in the accumulation of acetyl∼P result in elevated glnA expression (53). As in the case of the Pho regulon, these effects require the presence of the RR NR_I; thus, acetyl∼P is a positive regulator of glnA that acts through NR_I. The mechanism of this regulation was studied in vitro, where it was observed that acetyl∼P and certain other phosphorylated compounds directly phosphorylate NR_I and that the NR_I ∼P so formed is functionally indistinguishable from the NR_I ∼P formed by phosphotransfer from NR_II∼ P (53).

So far, all of the effects of acetyl∼P that I have discussed have been based on the phenotypes of mutant cells lacking an HPK protein. Indeed, the lack or accumulation of acetyl∼P seems to have no effect at all on the regulation of Pho or Ntr gene expression in cells containing PhoR and NR_II, respectively. The logical explanation for this is that the regulated phosphatase activities of PhoR and NRII prevent the inappropriate activation of the target regulons under conditions of phosphate and nitrogen sufficiency. However, recent results suggest that acetyl~P may have a role in the expression of the flagellar and chemotaxis genes in wild-type cells (194a). It has been known for some time that the production of flagella, and thus motility, is temperature sensitive, with cells producing few flagella at elevated temperatures. Cells grown at elevated temperatures contain elevated acetyl∼P; this is due to the temperature-sensitive nature of the acetate kinase activity responsible for the conversion of acetyl∼P to acetate (168). Mutations that prevent the synthesis of acetyl phosphate have the effect of increasing the motility and the number of flagella per cell at elevated temperatures, while mutations that result in the intracellular accumulation of acetyl∼P result in decreased motility and fewer flagella per cell even at the permissive temperature (194a). Thus, acetyl∼P is apparently a regulator of flagellar biosynthesis. Genetic evidence suggests that this regulation is exerted through the RR OmpR (194a), and this is supported by the observation that DNA sequences similar to OmpR-binding sites are found in the control region of the "master control operon" of the fla regulon (chapter 10). Furthermore, an effect of temperature on the regulation of the ompF and ompC genes by OmpR has long been known; high temperature results in the increased expression of ompC and decreased expression of ompF, a condition indicative of a higher intracellular concentration of OmpR∼P (reviewed in reference 167).

If acetyl∼P is indeed a global regulatory factor, it is reasonable to ask what this factor is signalling and how it might work in wild-type cells containing all of the HPK proteins. The intracellular concentration of acetyl∼P has been noted to be affected by the growth phase, carbon source, and temperature (130, 168). The growth phase and carbon source regulation of acetyl P may reflect the level of the intracellular acetyl coenzyme A pool, since acetyl∼P is formed from acetyl coenzyme A (reviewed in references 131 and 224). Acetyl coenzyme A is a central metabolite directly tied to virtually all anabolic processes, and an elevation in this pool may reflect a defect in one or more of these processes and thus a condition of stress. It should be noted that this is only one possible explanation and that much remains to be learned. The answer to the second of the questions raised above is also unknown, and several scenarios are possible. Regulation of two-component systems by acetyl∼P may reflect the evolutionary path by which two-component systems were developed and thus may be considered a "fossil," although the regulation of flagellar biosynthesis by acetyl∼P in wild-type bacteria seems to argue against this. Alternatively, acetyl∼P may provide a basal intracellular concentration of the phosphorylated RR that permits a more rapid transition to the activated state in response to environmental stimuli. Since many HPK and RR proteins are present at very low intracellular concentrations in unstimulated cells and the environmental stimuli frequently result, via positive autoregulatory loops, in an increase in the concentration of these proteins, acetyl∼P may provide a mechanism for priming these systems. Additional guesses as to possible roles of acetyl∼P are provided in a recent review (131).

The finding that acetyl∼P serves as a positive regulatory factor for several different two-component systems raises the possibility that the activity of these two-component systems may be coordinated by virtue of their common susceptibility to activation by acetyl∼P. For example, if a given RR is phosphorylated by acetyl∼P and subsequently dephosphorylated by the regulated phosphatase activity of the cognate HPK protein, then this should serve as a sink for acetyl∼P and thereby indirectly negatively regulate the expression of other two-component systems. Indeed, it has been recently shown that the concentration of acetyl∼P is increased 10-fold in cells deleted for NR_II (168). If we assume that in all cases the absence of stress conditions will be associated with HPK phosphatase activity and the stress state will be correlated with the absence of HPK phosphatase activity (as in the Ntr system), then the absence of stress should be correlated with a low intracellular acetyl∼P level and cellular distress should be correlated with an elevated acetyl∼P pool, due to the concerted action of the HPK proteins. This hypothesis has not been tested.

MECHANISMS FOR THE INTEGRATION OF EFFECTS OF DISTINCT STIMULI BY TWO-COMPONENT REGULATORY SYSTEMS

In addition to the possible roles of acetyl∼P and the cellular energy charge in coordinating the activity of distinct two-component systems, three other mechanisms for the regulation of two-component systems by distinct stimuli should be noted. The first of these, namely, the use of multiple HPK proteins to control the extent of phosphorylation of a single RR, has already been discussed. The primary example is again the Pho system, where both PhoR and CreC bring about the phosphorylation of PhoB. The system that regulates sporulation in B. subtilis provides another example of two HPK proteins (KinA and KinB) that act on a single RR protein (Spo0F [73]). These two HPK proteins seem to act at different points in the growth phase and with different effectiveness, with KinB serving to provide the low level of sporulation seen during active growth and KinA responsible for the much higher levels of sporulation seen in post-exponential-phase cultures (74).

A further level of complexity is seen with the system that regulates the final electron acceptor (nitrate- and nitrite-responsive gene expression) in E. coli (reviewed in reference 197). Signal transduction in this system involves two two-component systems with both distinct and overlapping functions and partially overlapping phosphotransfer capability (38, 170, 221). One of these two-component systems consists of the NarX HPK and the NarL RR, which are encoded by linked genes (153). The other consists of the NarQ HPK and the NarP RR, encoded by genes unlinked to each other and to narXL. Both HPK proteins are integral membrane proteins with a periplasmic domain of ∼115 amino acids, and membrane localization is apparently required for response to the stimuli, nitrate and nitrite. The periplasmic domain of these proteins contains a highly conserved 17-amino-acid segment that may reflect a common feature of the sensory mechanism. Both NarX and NarQ act as a NarL kinase, but only NarX is a NarL∼P phosphatase (38, 190). Genetic results suggest that only NarQ is a negative regulator of NarP; that is, NarQ may be a NarP∼P phosphatase (197). Various interactions of NarL and NarP with the target promoters are found (89); some promoters are repressed and some are activated by either protein, some are activated only by NarP, and some are activated only by NarL. Furthermore, these RR proteins may act either in concert or antagonistically depending on the promoter. The use of two two-component systems apparently provides for greater precision in the regulation of the large number of target genes in this stimulon.

The second mechanism for the integration of effects of distinct stimuli may be provided by the existence of additional phosphatase activities that act to dephosphorylate the RR∼P. The best examples of this phenomenon come from the Che system, as noted above, and from the sporulation system of B. subtilis. In the Spo system, distinct phosphatases have been shown to act on the various components of the phosphorelay system (154, 162).

Finally, the effects of different stimuli may be exerted independently of the phosphorylation/ dephosphorylation reactions of a two-component system by agents that regulate the expression or stability of the components or the activity of the associated domains of the RR. Examples of the latter scenario are provided by the Nar system described above, where the phosphorylated RR act in concert with Fnr to control transcription, and by the Rsc two-component system regulating colonic acid capsule synthesis in E. coli, where the activation of transcription by the RcsB RR is facilitated by an additional factor, RcsA, that is subject to independent control (203).

HOW DOES PHOSPHORYLATION OF THE RR DOMAIN CONTROL THE ACTIVITY OF ASSOCIATED DOMAINS?

There are several cases where it has been conclusively shown that phosphorylation of the RR domain controls the activity of the associated domains. For example, the transcription factors PhoB, OmpR, FixJ, VirG, Spo0A, and NR_I are transcriptional activators of the Pho, Omp, Fix/Nif, Vir, Spo, and Ntr regulons, respectively. In each of these cases, the regulation of transcription has been studied in vitro with purified components, and it has been observed that phosphorylation of the N-terminal RR domain is either required or greatly stimulates transcriptional activation of the target genes (2, 4, 25, 47, 80, 127, 147, 176). In the case of the CheB protein, an RR domain is associated with a C-terminal domain that has methylesterase enzymatic activity. Phosphorylation of the RR domain of CheB activates this methylesterase activity (122).

In several cases, it seems that the RR domain inhibits the activity of the associated domain and that the role of phosphorylation is to alleviate this inhibition (3, 46, 64, 68, 88, 122). In the case of the CheB protein, proteolytic cleavage removing the N-terminal domain activates the methylesterase activity of the C-terminal domain (122). Similarly, genetic deletion of the N-terminal RR domain of CheB results in a truncated protein with elevated methylesterase activity (122). The N-terminal RR domain of the DctD protein also seems to negatively regulate the activity of the associated domain (68). DctD is a transcriptional activator of the σ ⁵⁴-dependent dctA promoter in R. meliloti; constitutive transcription activation in intact cells is brought about by genetic deletion of the N-terminal RR domain of DctD (68). A similar phenomenon has been noted for the FixJ and Spo0A RR, which are transcriptional activators of the major form of RNA polymerase in their various organisms (3, 64).

The ArcB HPK protein contains an RR domain at its C terminus (Fig. 1), and this domain appears to act as a negative regulator of the phosphotransfer reaction between the H-box histidine of ArcB and the ArcA RR (88). Deletion of this ArcB RR domain does not prevent ArcA phosphorylation, but point mutations eliminating phosphorylation of the ArcB RR domain essentially eliminate ArcA phosphorylation. The simplest explanation for these and other data is that the ArcB RR domain sterically interferes with ArcA access to the H-box histidine and that phosphorylation of this domain results in a conformation that no longer interferes. A similar role may be envisioned for the RR domain at the extreme C terminus of the RcsC HPK protein, since in this case also deletion of the RR domain does not eliminate function of HPK (62).

While the data in the examples cited above are appealing, there is no reason to suspect that negative regulation of associated domains by the RR domain will prove to be the universal situation. In several cases, truncation of the N-terminal RR domains of transcription factors has not resulted in the activation of the associated "transcriptional activation domain" in vivo (noted in reference 160). As with all negative results, little can be said at this point about these experiments except that perhaps the results reflect destabilization of the remaining fraction of the protein as an unforeseen consequence of the deletion of the N-terminal domain.

There are two cases in which the RR domain has been definitely shown to have a positive regulatory role; however, both of these cases involve RR domains that constitute the entire protein (i.e., there are no associated domains). There is compelling evidence, both biochemical and genetic, that the CheY protein interacts with a component of the flagellar motor switch, FliM, to bring about reversal of the motor (chapters 10 and 73). This interaction requires the phosphorylated form of CheY (119, 232). The other example comes from the regulatory system controlling sporulation in B. subtilis (74). The Spo0F protein of this system, which consists of just the RR domain, is part of a phosphorelay system in which phosphoryl groups are transferred from either of two HPK proteins to Spo0F and subsequently transferred to a histidine moiety on the Spo0B protein, followed by transfer from Spo0B to the RR domain of the transcription factor Spo0A (33). Thus, Spo0F indirectly activates the transcriptional activation domain of Spo0A by supplying phosphoryl groups through the phosphorelay system.

Returning to the issue of those RR that contain an RR domain and an associated transcriptional activation domain, studies with the activator of the Ntr regulon, NR_I, suggest that phosphorylation and dephosphorylation of the RR domain control the oligomeric state of the transcription factor, which in turn controls its activity (166). The NR_I protein is a dimer consisting of an RR domain at the N terminus, a central domain characteristic of proteins that activate transcription in concert with σ ⁵⁴ RNA polymerase, and a C-terminal segment that contains a DNA-binding helix-turn-helix motif. The central or transcriptional activation domain contains an ATP-binding site (229), and phosphorylation of the N-terminal RR domain results in the activation of an ATPase activity of the central domain (19, 53, 229). This ATPase activity is required for the activation of transcription by NR_I ∼P. Various lines of evidence indicate that the active form of NRI∼P is a tetramer (or, less likely, a higher oligomer) of NR_I subunits, i.e., a dimer of the dimer. These data are covered extensively in chapter 86 and will only be noted briefly here. The ATPase activity of NR_I ∼P shows a second-order dependence on the NR_I ∼P concentration (53, 229). Furthermore, there are tandem NR_I-binding sites upstream from promoters activated by NR_I ∼P, and the binding of NR_I to these sites becomes highly cooperative upon phosphorylation of NR_I (230). Finally, an NR_I protein containing a C-terminal mutation that results in a defect in DNA binding has been obtained. This protein cannot bind to NR_I-binding sites but can stimulate the activation of transcription by a very low concentration of NR_I that is able to bind to the binding sites (229). The model inferred from this experiment is that an active dimer-dimer can activate transcription even if tethered to DNA at only a single NR_I-binding site.

Finally, it should be noted that in some cases the regulation of associated domains by the RR domain appears to be leaky. For example, OmpR could activate transcription in vitro in the absence of a phosphodonor if present in sufficiently high concentration (81). Also, a mutant form of the UhpA RR that cannot be phosphorylated due to alteration of the site (D54N) is active in vivo when overproduced from a multicopy plasmid (92). Such data suggest that the unphosphorylated form of these proteins is able to adopt the active conformation at a detectable level.

AMPLITUDE MODULATION OF THE BINARY SWITCH PROVIDES THE POSSIBILITY FOR A GRADUATED RESPONSE

Several different two-component systems provide excellent examples of how amplitude modulation of a binary switch can provide continuous variation and reciprocal regulation, and I will note here the single example of the Omp system. In this system, the ompF and ompC porin genes are reciprocally regulated in response to changes in osmolarity (actual stimuli unknown), with ompF transcription activated at low osmolarity and repressed at high osmolarity and ompC transcription only activated at high osmolarity. This reciprocal regulation is in part due to the regulation of the intracellular concentration of OmpR∼P, in that a low concentration optimally activates ompF transcription and a higher concentration represses ompF and activates ompC (reviewed in reference 167). The location of OmpR∼P-binding sites of various affinity at the two promoters thus governs when the promoters are activated and repressed as the concentration of OmpR∼P is varied (172, 213). Similar phenomena are seen in other systems (194; Feng et al., in press), and the promoters should be considered to be "hardwired" such that they are expressed only when the RR∼P concentration is within a defined range.

OVERLAPPING SPECIFIC AND GLOBAL CONTROLS PROVIDE FOR PHYSIOLOGICAL REGULATION

In closing, I would like to emphasize that it is evident that the "regulation of gene expression" occurs at many levels and that in many, if not all, cases separate regulatory mechanisms contribute to the overall physiological regulation. The examples of the former point are the many regulatory mechanisms discussed throughout this treatise, including regulation of transcription initiation, elongation, and termination, the regulation of mRNA stability and translation, and the regulation of protein stability, folding, secretion, and activity. There are also numerous examples for the latter point, many of which are found in this treatise. Separate regulatory control mechanisms may act in a concerted fashion on a single process, such as the regulation of gene transcription, or on distinct processes. Thus, the regulation of gene transcription by various environmental conditions is often only a part of the physiological response to that condition (e.g., see reference 216).

As an example of this phenomenon at the level of a single gene product, I like to consider the minor σ ³² factor, necessary for the transcription of genes comprising the heat shock regulon (chapter 88). The presence of the active σ ³² factor is regulated by the control of the initiation of transcription of its structural gene, rpoH (50), by the control of the stability of the mRNA, by the control of the translation of the mRNA (144), by the control of σ ³² protein stability (205), and by the control of protein activity (207). These separate mechanisms, probably acquired independently, contribute to the overall control of σ ³² activity and thus to the expression of heat shock genes.

In a different sense, layers of regulation result from the recruitment of individual genes into regulons, as noted earlier. These regulons are likely to have resulted from the extension of existing gene regulatory mechanisms to unrelated genes. Since a gene can be recruited into several different regulons (e.g., see references 10 and 186), various relationships between the different control mechanisms can exist; these may be cooperative, antagonistic, or independent. Moreover, separate promoters may render the transcription of a gene sensitive to distinct stimuli.

Finally, the activity of a gene product may be regulated by different stimuli by virtue of the duplication of the gene and the evolution of separate control mechanisms for each version of the gene. There are numerous examples of redundant genes in bacteria; for example, E. coli contains two molecular chaperone proteins, DnaJ and CbpA, of quite similar structure. Of these, the DnaJ protein is regulated as a member of the heat shock regulon, while the CpbA protein is induced by the transition into the stationary phase and by phosphate starvation (243).