This Article
Right arrow Abstract Freely available
Right arrow Full Text (PDF)
Right arrow An author's correction has been published
Right arrow Alert me when this article is cited
Right arrow Alert me if a correction is posted
Right arrow Citation Map
Services
Right arrow E-mail this article to a friend
Right arrow Similar articles in this journal
Right arrow Similar articles in PubMed
Right arrow Alert me to new issues of the journal
Right arrow Download to citation manager
Right arrowReprints and Permissions
Right arrow Copyright Information
Right arrow Books from ASM Press
Right arrow MicrobeWorld
Citing Articles
Right arrow Citing Articles via HighWire
Right arrow Citing Articles via Google Scholar
Google Scholar
Right arrow Articles by Deutscher, J.
Right arrow Articles by Postma, P. W.
Right arrow Search for Related Content
PubMed
Right arrow PubMed Citation
Right arrow Articles by Deutscher, J.
Right arrow Articles by Postma, P. W.

 Previous Article  |  Next Article 

Microbiology and Molecular Biology Reviews, December 2006, p. 939-1031, Vol. 70, No. 4
1092-2172/06/$08.00+0     doi:10.1128/MMBR.00024-06
Copyright © 2006, American Society for Microbiology. All Rights Reserved.

How Phosphotransferase System-Related Protein Phosphorylation Regulates Carbohydrate Metabolism in Bacteria{dagger}

Josef Deutscher,1* Christof Francke,2 and Pieter W. Postma3

Microbiologie et Génétique Moléculaire, INRA-CNRS-INA PG UMR 2585, Thiverval-Grignon, France,1 Department of Molecular Cell Physiology, Biocentrum, Vrije Universiteit, Amsterdam, and Wageningen Centre for Food Sciences at CMBI, Radboud University, Nijmegen, The Netherlands,2 Swammerdam Institute for Life Sciences, BioCentrum, University of Amsterdam, Amsterdam, The Netherlands3

SUMMARY
INTRODUCTION
    Carbon Catabolite Repression
    The Phosphoenolpyruvate:Carbohydrate Phosphotransferase System
        The general PTS proteins EI and HPr.
        Enzyme II complexes.
        Organization of the PTS.
REGULATION OF CARBON METABOLISM IN GRAM-NEGATIVE ENTERIC BACTERIA
    EIIAGlc Is the Central Processing Unit of Carbon Metabolism in Enteric Bacteria
    Transcription Regulation by Crp/cAMP and Role of PEIIAGlc
        Regulation of cyaA and crp transcription.
        Regulation of adenylate cyclase activity.
        Regulation by Crp/cAMP.
        Growth rate effects on cAMP concentration.
        Secretion and breakdown of cAMP.
    Inducer Exclusion and Role of EIIAGlc
        Inducer exclusion is mediated by unphosphorylated EIIAGlc.
    Interactions of EIIAGlc
        Structure of EIIAGlc.
        Interaction with HPr and EIICBGlc.
        Interaction with GlpK.
        Interaction with LacY.
        Interaction with MelB.
        Interaction with MalK.
        Other interactions.
    Diauxic Growth
    Transcription Regulation by Mlc
        Repression by Mlc and its interaction with unphosphorylated EIICBGlc.
        The Mlc regulon.
    Role of Phosphoenolpyruvate
    CCR Mediated by Non-PTS Sugars and Catabolic Intermediates
    Reverse Inducer Exclusion or Retroregulation
    Regulation by EIIAGlc-Like Proteins
        Is EIIAGlc-mediated regulation limited to enteric bacteria?
    Mathematical Modeling of the PTS and Its Role in CCR
    Other Regulatory Mechanisms Involving the PTS in Enteric Bacteria
        Phosphorylation of EI by ATP.
        Chemotactic response to carbohydrates.
        Glycogen storage.
REGULATION OF CARBON METABOLISM IN LOW-G+C GRAM-POSITIVE BACTERIA: REGULATORY FUNCTIONS OF P-Ser-HPr
    HPr, the Central Processing Unit of Carbon Metabolism in Gram-Positive Bacteria
    Characteristics of ATP-Dependent HPr Phosphorylation
        Phosphorylation of HPr at Ser-46.
        HPr kinase also dephosphorylates P-Ser-HPr by producing PPi.
        Structure determination of HprK/P.
        Organization of the hprK operon.
        Metabolites regulate the antagonistic activities of HprK/P.
    P-Ser-HPr Regulates PTS Transport Activity by a Feedback Mechanism
    Role of P-Ser-HPr in CCR
        Loss of ATP-dependent HPr phosphorylation prevents CCR.
        The ptsH1 mutation has a pleiotropic effect on CCR and carbon catabolite activation.
    The Catabolite Response Element cre, an Operator Site for CCR and CCA
        A cis-acting element regulating CCR and CCA.
        Distribution of cre's.
    Catabolite Control Protein A Functions as a Catabolite Repressor or Activator
        A LacI/GalR-type trans-acting factor mediates CCR and CCA.
        CcpA-specific sequences.
        CcpA needs a corepressor.
    P-Ser-HPr Functions as a Catabolite Corepressor
        Interaction of P-Ser-HPr with CcpA.
        Binding of the P-Ser-HPr:CcpA complex to cre's.
        Deviations from the general CCR mechanism in gram-positive bacteria.
    P-Ser-Crh, a Second Catabolite Corepressor in Bacilli, Geobacilli, and Oceanobacilli
        A B. subtilis HPr-like protein without His-15.
        Binding of the P-Ser-Crh:CcpA complex to cre's.
        Distribution of Crh in gram-positive organisms.
    What Are the Functions of CcpB and CcpC?
    Role of P-Ser-HPr in the Virulence of Certain Pathogens
        P-Ser-HPr and the regulation of virulence genes in gram-positive pathogens.
        P-Ser-HPr and the regulation of virulence genes in gram-negative pathogens.
    Is P-Ser-HPr Involved in Inducer Expulsion?
        Simultaneous occurrence of inducer expulsion and P-Ser-HPr formation.
        Inducer expulsion in L. casei and L. lactis does not require P-Ser-HPr.
    Is P-Ser-HPr Involved in Inducer Exclusion?
        In vitro interaction of Ser46Asp mutant HPr with non-PTS permeases.
        Mutations preventing P-Ser-HPr formation abolish inducer exclusion.
REGULATION OF CARBON METABOLISM IN LOW-G+C GRAM-POSITIVE BACTERIA: REGULATORY FUNCTIONS MEDIATED BY PHis-HPr AND PEIIBs
    PEP-Dependent Phosphorylation of Non-PTS Proteins
    Regulation of Transcription Antiterminators by PTS-Mediated Phosphorylation
        Regulation of gene expression by transcription attenuation.
        rho-Independent terminators.
        RAT sequences are the binding sites for PTS-regulated antiterminators.
        Phosphorylation of BglG/SacY-type antiterminators by PTS proteins.
        Antiterminator-dependent induction is regulated via PEIIBs.
        PEIIBs are necessary for the phosphorylation of a histidine in PRD1.
        Some antiterminators need to be activated by phosphorylation in PRD2.
        CCR mediated by dephosphorylation of PRD2 of antiterminators.
        Distribution of BglG/SacY-type antiterminators in bacteria.
    Regulation of Transcription Activators by PTS-Mediated Phosphorylation
        PRDs in transcription activators.
        Domain organization in NifA/NtrC-type PRD-containing transcription activators.
        Antagonistic effects of PTS-mediated phosphorylation reactions on LevR activity.
        Distribution of LevR-like transcription activators.
        DeoR-type PTS-controlled transcription activators.
        Domain organization in DeoR-type PRD-containing transcription activators.
        PTS-mediated control of DeoR-type PRD-containing transcription activators.
        Distribution of DeoR-type PRD-containing transcription activators.
    Regulation of EIIA-Containing Non-PTS Transporters by PHis-HPr-Mediated Phosphorylation
        Phosphorylation of an EIIAGlc-like domain in certain non-PTS transporters.
        Phosphorylation of LacS stimulates the lactose/galactose exchange reaction.
    Regulation of Glycerol Kinase by PHis-HPr-Mediated Phosphorylation
        Glycerol metabolism in gram-positive bacteria requires functional EI and HPr.
        EI- and HPr-catalyzed phosphorylation regulates GlpK activity in gram-positive bacteria.
        PGlpK dephosphorylation leads to CCR via inducer exclusion.
        The phosphorylation loop of enterococcal GlpK binds FBP in the E. coli enzyme.
        GlpK phosphorylation in bacteria of the Thermus/Deinococcus group.
        Is GlpK the only carbohydrate kinase regulated by PHis-HPr?
SOME UNUSUAL PTS PATHWAYS AND PROTEINS
    PEP-Dependent Dihydroxyacetone Phosphorylation
    The PTSNtr
        Connection between carbon and nitrogen metabolism in E. coli.
        Transcription regulation of the TOL plasmid of P. putida.
        Other connections of the EIIANtr.
        Potential roles of EINtr.
        What is the function of the PTSNtr?
CONCLUSIONS AND PERSPECTIVES
ACKNOWLEDGMENTS
REFERENCES

arrow
SUMMARY
 
The phosphoenolpyruvate(PEP):carbohydrate phosphotransferase system (PTS) is found only in bacteria, where it catalyzes the transport and phosphorylation of numerous monosaccharides, disaccharides, amino sugars, polyols, and other sugar derivatives. To carry out its catalytic function in sugar transport and phosphorylation, the PTS uses PEP as an energy source and phosphoryl donor. The phosphoryl group of PEP is usually transferred via four distinct proteins (domains) to the transported sugar bound to the respective membrane component(s) (EIIC and EIID) of the PTS. The organization of the PTS as a four-step phosphoryl transfer system, in which all P derivatives exhibit similar energy (phosphorylation occurs at histidyl or cysteyl residues), is surprising, as a single protein (or domain) coupling energy transfer and sugar phosphorylation would be sufficient for PTS function. A possible explanation for the complexity of the PTS was provided by the discovery that the PTS also carries out numerous regulatory functions. Depending on their phosphorylation state, the four proteins (domains) forming the PTS phosphorylation cascade (EI, HPr, EIIA, and EIIB) can phosphorylate or interact with numerous non-PTS proteins and thereby regulate their activity. In addition, in certain bacteria, one of the PTS components (HPr) is phosphorylated by ATP at a seryl residue, which increases the complexity of PTS-mediated regulation. In this review, we try to summarize the known protein phosphorylation-related regulatory functions of the PTS. As we shall see, the PTS regulation network not only controls carbohydrate uptake and metabolism but also interferes with the utilization of nitrogen and phosphorus and the virulence of certain pathogens.


arrow
INTRODUCTION
 
From a few peaks rising above the fog we try to imagine what the hidden landscape underneath might look like.

—Pieter W. Postma, during a hiking trip in the Beaujolais region in October 1997.


arrow
INTRODUCTION
 
Carbon Catabolite Repression

Given a certain extracellular environment, a bacterium needs only a subset of the enzymes encoded by the genome to propagate, and therefore, it regulates gene expression differentially. For instance, in case a particular substrate is absent, the genes encoding the enzymes required for uptake and subsequent metabolism of the substrate are often repressed. Availability of the substrate then leads to the relief of repression. The classical example of this mode of transcription regulation is the lac operon of Escherichia coli (367). More complex responses of cells can occur when they are exposed to multiple nutrients at the same time. More than a hundred years ago, it was observed that growth on glucose can lower the activity of certain enzymes in bacteria and yeast. This phenomenon became known as the glucose effect (see also reference 744). One of the first descriptions of the glucose effect stems from the work of Dienert and was published in 1900 in Paris, France. He observed that Saccharomyces cerevisiae cells, which had been adapted to galactose (i.e., they were able to utilize galactose), rapidly lost the adaptation when the cells were exposed to glucose or fructose (188). This observation was later confirmed, for example, by Söhngen and Coolhaas, in Wageningen, The Netherlands, who measured the influence of temperature on the glucose effect in yeast (820). One of the first quantitative analyses of the glucose effect was carried out by Stephenson and Gale in Cambridge, England, who measured the activity of E. coli galactozymase. Galactozymase was understood as being the entity of enzymes necessary for the degradation of galactose by a specific organism. Those authors found that glucose exerted a strong repressive effect on galactozymase activity (more than sevenfold), while the utilization of glucose was not affected by galactose (836). In the early 1940s, Jacques Monod observed that when cultivated in a synthetic medium containing sucrose and dextran, the gram-positive bacterium Bacillus subtilis first utilized sucrose and then stopped growing for a certain period (lag phase or phase of adaptation) before the cells resumed growth by utilizing dextran (572). Owing to the biphasic growth he had observed, Monod called this phenomenon diauxie. He extended his studies with B. subtilis and found that diauxie was a more general phenomenon. When this organism was provided with sucrose or glucose, for example, and a second carbohydrate such as maltose, mannitol, inositol, or sorbitol, it also showed diauxic growth; i.e., sucrose or glucose was utilized first before the bacterium started to transport and metabolize the "less favorable" carbon source. He finally extended his studies to the gram-negative bacterium E. coli, which exhibited diauxie when grown in the presence of glucose or fructose, for example, as the preferred carbon source and xylose, arabinose, rhamnose, maltose, lactose, or glucitol as the less favorable sugar (572). It appeared that in each organism, a specific hierarchy existed for the utilization of carbon sources, with glucose usually being on top of it. In subsequent years, it was found that preferred sugars such as glucose, fructose, or sucrose, as long as they are present in sufficient amounts in the growth medium, repress the synthesis of the enzymes necessary for the transport and metabolism of less favorable carbon sources. The phenomenon therefore became known as carbon catabolite repression (CCR) (137). When the preferred carbon source is exhausted, bacteria first need to synthesize the enzymes necessary for the transport and metabolism of the less favorable carbon source (lag phase) before they can resume growth. The glucose effect reported at the beginning of the last century can therefore be considered CCR.

The regulatory mechanisms underlying CCR have since been intensively studied, and it was found that the preferential use of one carbon source over the other involves either the prevention of transcription activation or the repression of transcription. Here, we define CCR, somewhat more broadly, as the inhibitory effect of a certain carbon source in the growth medium on gene expression and/or the activity of enzymes involved in the catabolism of other carbon sources. Gene expression and enzyme activity are regulated via protein-DNA, protein-protein, and protein-metabolite interactions, and these interactions are in turn modulated via protein modification. Interestingly, the bacterial phosphoenolpyruvate (PEP):carbohydrate phosphotransferase system (PTS), which catalyzes the uptake and concomitant phosphorylation of numerous carbohydrates, plays a major role in bacterial CCR. However, quite different mechanisms have evolved in gram-negative and gram-positive organisms, and different PTS proteins are implicated in the two types of bacteria (for earlier reviews on the subject, see references 90, 756, and 845). In some bacteria, CCR is governed by other preferences and is possibly mediated by additional mechanisms. For instance, {alpha}-proteobacteria of the genera Sinorhizobium, Rhizobium, and Bradyrhizobium (83, 889) and {gamma}-proteobacteria of the Pseudomonas family (135) prefer acetate or tricarboxylic acid (TCA) cycle intermediates such as succinate over carbon sources such as glucose, fructose, or lactose. The underlying mechanisms have not yet been unraveled, but in pseudomonads, they involve the catabolite repression control (Crc) protein (22, 331, 332, 573, 747, 982), and rhizobacteria, they involve inducer accumulation (83).

In this review, we will focus on mechanisms that control carbohydrate transport and metabolism in PTS-containing bacteria, and we will concentrate on the regulatory roles played by the components of the system. We start by discussing the individual components of the PTS and their properties. Subsequently, we will describe their role in catabolite repression and in other regulatory mechanisms governing carbon metabolism. We will pay particular attention to the mechanisms that have developed in the enteric bacteria E. coli and Salmonella enterica serovar Typhimurium ({gamma}-Proteobacteria) and in low-G+C gram-positive bacteria (also known as Firmicutes) such as B. subtilis and Lactobacillus casei, but we will also refer to other proteobacteria and to gram-positive bacteria with high G+C content (also known as Actinobacteria). The two groups of gram-positive bacteria can be distinguished not only on the basis of the different G+C content of their DNA but also based on many other different characteristics. For example, actinobacteria have been shown to contain numerous proteins that have been detected only in members of this phylum so far (corynebacteria, streptomyces, mycobacteria, nocardiae, propionibacteria, bifidobacteria, leifsoniae, and rhodococci, to name just a few) (253). Important for this review is that most high-G+C gram-positive bacteria possess PTS components and transport sugars via this system but seem to be devoid of adenylate cyclase and HPr kinase/phosphorylase (41). Nevertheless, other PTS-related control mechanisms of carbon metabolism are operative.

The Phosphoenolpyruvate:Carbohydrate Phosphotransferase System

The PTS was discovered in E. coli by Kundig, Ghosh, and Roseman as a system that uses PEP to phosphorylate a number of hexoses, including N-acetylmannosamine, glucose, mannose, glucosamine, and N-acetylglucosamine (434). Subsequently, it was recognized that the PTS is in fact a transport system that catalyzes the uptake of numerous carbohydrates and their conversion into their respective phosphoesters during transport. After its discovery in E. coli, the PTS was found in many other bacterial species. The coupled transport and phosphorylation of carbohydrates were originally described as a two-step reaction catalyzed by two enzymes, enzyme I (EI) and enzyme II (EII), with the protein HPr as an intermediate phosphoryl donor:

Formula

Formula
However, this representation is a bit misleading in the sense that EI, HPr, and EII behave in a functionally identical manner. They accept a phosphoryl group from a donor and donate it to an acceptor (i.e., they transfer a group), thus cycling between the phosphorylated and unphosphorylated states (depicted in Fig. 1). It is precisely this aspect of the mechanism that is exploited by the cell in PTS-mediated regulation.


Figure 1
View larger version (36K):
[in this window]
[in a new window]

  		FIG. 1. Carbohydrate transport and phosphorylation by the PTS and their coupling to glycolysis. Carbohydrates are transported and concomitantly phosphorylated by the PTS. The phosphorylated carbohydrate feeds into glycolysis, normally at the glucose-6-P or fructose-6-P level. Two phosphoenolpyruvate molecules are usually formed in glycolysis, one of which is used to drive the transport and initial phosphorylation of the carbohydrate. As a result, the phosphorylation state of the PTS proteins depends on both the concentration of extracellular carbohydrates and the ratio of internal phosphoenolpyruvate and pyruvate. Abbreviations for enzymes (in boldface type) are as follows: Pgi, phosphoglucose isomerase; Pfk, phosphofructokinase; Fba, fructose-1,6-bisphosphate aldolase; Tpi, triose-phosphate isomerase; Gap, glyceraldehyde-3-phosphate dehydrogenase; Pgk, phosphoglycerate kinase; Pgm, phosphoglycerate mutase; Eno, enolase; Pyk, pyruvate kinase.

The basic composition of the PTS is in fact similar in all species studied so far (for reviews, see references 678 and 730). It is comprised of two "general" cytoplasmic components, EI and HPr, which are common to all PTS carbohydrates. Carbohydrate specificity resides in EII, and hence, bacteria usually contain many different EIIs. Each EII complex consists of one or two hydrophobic integral membrane domains (domains C and D) and two hydrophilic domains (domains A and B), which together are responsible for the transport of the carbohydrate across the bacterial membrane as well as its phosphorylation. In a sense, the EII complexes constitute parallel transport pathways connected to a common PEP/EI/HPr phosphoryl transfer pathway. E. coli contains at least 15 different EII complexes, and the existence and properties of these enzymes have been established by genetic, biochemical, and physiological studies. A similar number of PTSs is present in B. subtilis (178, 706). EII complexes are formed either by distinct proteins or by a single multidomain protein. Likewise, fusion proteins that contain EI and/or HPr domains exist. A prominent example of the latter is FPr, which consists of HPr and an EIIA domain and mediates the phosphotransfer in the uptake of fructose by E. coli and Salmonella enterica serovar Typhimurium (266). In Rhodobacter capsulatus, a similar fusion protein also contains an EI domain (958).

Genome sequencing projects have uncovered many proteins that are similar to either one of the two general PTS proteins or to parts of carbohydrate-specific EII complexes (345, 867, 935). For example, in E. coli, the protein DhaM is composed of an EIIAMan-like regulatory domain followed by an HPr and a truncated EI domain (304). In other cases, the newly discovered proteins are formed by fusions between domains that originate from both PTS and non-PTS proteins, e.g., a Clostridium acetobutylicum response regulator (built from HPr and an NtrC-like protein) (721) and the Streptococcus thermophilus LacS protein (built from a melibiose carrier and EIIAGlc) (673). The functions of most of these chimeric proteins remain unknown, but for some proteins, like LacS and DhaM, the function has been elucidated. Indeed, one of the present challenges is finding the function of the newly discovered PTS-like proteins, most of which have not been detected by any biochemical or mutant study. A very intriguing group consists of the paralogs of EI and HPr. In E. coli, at least five paralogs of each general PTS protein were uncovered (345, 720). We will discuss some of these proteins below (see "SOME UNUSUAL PTS PATHWAYS AND PROTEINS").

The general PTS proteins EI and HPr. EI is encoded by the ptsI gene. Sequence comparisons reveal that the EIs (about 570 residues; 63 kDa) from various gram-positive and gram-negative bacteria exhibit significant identity (345, 678). EI is autophosphorylated in the presence of Mg2+ (940) at the N-3 position of the imidazole ring of a conserved histidine (His-189 in E. coli EI) (16, 940), which is located on the N terminus of the protein. The C terminus contains the PEP binding site and is necessary for dimerization (114, 254, 798). The two-domain structure of the protein was discovered by high-sensitivity differential scanning calorimetry (478). Limited proteolysis of E. coli EI in vitro results in a specific split and provides a stable peptide representing the N-terminal domain (455). The N-terminal domain of EI (EI-N) (115) as well as the C-terminal domain (EI-C) (234) were cloned and characterized. EI-C was shown to complement EI-N both in vivo and in vitro; i.e., EI-N was phosphorylated in the presence of EI-C, PEP, and Mg2+ (234). A truncated protein composed of the first 259 amino acids has been synthesized and purified, and both the crystal (475) and solution (258) structures were resolved. It appears that phosphorylation does not drastically change the conformation of the N-terminal domain per se. Recently, the crystal structure of the C-terminal domain of EI of the thermophile Thermoanaerobacter tengcongensis was also elucidated (619). EI exhibits about 30% sequence similarity with pyruvate phosphate dikinase (PPDK) (328, 605, 670), and the various domains in both proteins have similar structures (619). PPDK, which converts ATP, Pi, and pyruvate into AMP, pyrophosphate (PPi), and PEP; PEP synthase, an enzyme that catalyzes the conversion of pyruvate and ATP into PEP, AMP, and Pi and that is also phosphorylated on a histidyl residue (590); and EI have related functions. For PPDK, a structural model with three domains has been proposed: a C-terminal PEP/pyruvate binding domain, an N-terminal nucleotide binding domain, and a P~His domain linked to the other two domains via two flexible polypeptide segments (328). Although the nucleotide and PEP/pyruvate binding sites are approximately 45 Å apart, the P~His domain is thought to bridge the distance by swiveling between the two other domains to enable the phosphoryl transfer. Recent data obtained from a comparison of several PPDK crystal structures are consistent with the proposed model (587). Similar to PPDK, a large distance (>20 Å) between the PEP/pyruvate binding site and the phosphorylatable histidine was observed in a model of EI in which the N-terminal and C-terminal EI structures were fused (619). Swiveling of the N-terminal domain by changing the position of the domains with respect to each other but retaining the structure of the individual domains reduces the distance significantly (to 7.4 Å). In the case of EI, the N-terminal domain contains the HPr binding site rather than the nucleotide binding site of PPDK and PEP synthase. This model was confirmed by recent studies in which the structure of full-length EI of Staphylococcus carnosus was determined (516). The phosphohistidine and HPr-binding domain are clearly separated from the C-terminal dimerization and PEP binding domain. The structure also revealed the extensive interaction surface related to the dimer. Thermodynamic analyses of the influence of several effectors on the conformational stability of EI of Streptomyces coelicolor showed that at a low pH, EI is partly unfolded and that the presence of PEP causes structural changes (356). Two other recent studies on the structure of EI (639) and its C-terminal domain (638) suggest a relatively large structural variability for monomeric EI, which progressively diminishes when EI dimerizes and binds Mg2+ and PEP. The observed effects are explained by a swiveling mechanism, as described for PPDK. The structure thus provides additional evidence for the proposed conformational change necessary to facilitate phosphotransfer.

HPr (about 90 residues; 9 to 10 kDa) is encoded by the ptsH gene and has been purified from a variety of organisms (678). Similarity is most pronounced around the active-site histidyl residue, His-15 in the HPr of most enteric bacteria and firmicutes (183, 681, 941). HPr is phosphorylated at the N-1 position of the imidazole ring of His-15. The crystal and solution structures of HPrs from E. coli and B. subtilis as well as a few other species are known (for reviews, see references 33, 123, 329, and 936). The molecule forms an open-faced ß-sandwich consisting of a four-stranded antiparallel ß-sheet that is covered at one side by one short and two long {alpha}-helices. The active-site His-15 is located in the N-terminal part of the first long {alpha}-helix and is exposed to the solvent.

The solution structures of the complex between HPr and the N-terminal domain of EI (260) and between HPr and EIIAGlc of E. coli (929) have been determined. The absence of large changes of the chemical shift values for the backbone atoms upon complexation indicates that the interactions of HPr with EI and EIIAGlc do not require considerable conformational changes in any of the domains of the PTS binding partners. The binding interface of HPr and glycogen phosphorylase (see "REGULATION OF CARBON METABOLISM IN GRAM-NEGATIVE ENTERIC BACTERIA") was also mapped, and the interaction surface of the three structures was compared (930). EI and EIIAGlc interact with the same narrow region of HPr, whereas the region of interaction with glycogen phosphorylase is somewhat wider (929, 930). This is consistent with the higher binding affinity reported for HPr with glycogen phosphorylase than for HPr with EI or EIIAGlc (259, 799). A similar interaction surface was reported for E. coli HPr binding either EIIAMtl (139, 908) or EIIAMan (949) and B. subtilis HPr binding EIIAGlc (123). The key interacting residues are located in helices 1 and 2 and the loops preceding helix 1 and following helix 2. Studies with a large number of ptsH mutants, which show lowered affinity for binding to EI, suggest that the same residues are important for the interaction between EI and HPr (85).

In most low-G+C gram-positive bacteria and a few gram-negative organisms, HPr can also be phosphorylated by an ATP-dependent protein kinase on a seryl residue, e.g., Ser-46 in B. subtilis. We will discuss the importance of the second regulatory phosphorylation in detail below (see REGULATORY FUNCTIONS OF P-Ser-HPr). The regulatory phosphorylation is not part of the phosphoryl transfer to carbohydrates, but phosphorylation of the seryl residue slows the phosphoryl transfer from P~EI to HPr at least 100-fold. E. coli HPr contains a Ser-46 residue but lacks an HPr kinase. Nevertheless, replacement of Ser-46 in E. coli HPr by an aspartyl residue lowers the affinity of P~EI for HPr almost 1,000-fold (589). This effect is most likely caused by electrostatic repulsion, as the mutation caused by the replacement of Ser-46 with an aspartyl residue (Ser46Asp) leads to only weak structural changes (952).

Enzyme II complexes. The carbohydrate specificity of the PTS resides in the EIIs, which consist of an integral membrane domain facing both the periplasmic space and the cytoplasm and cytoplasmic domains or proteins. All EIIs are built basically in a similar modular manner (764) and can consist of up to four separate proteins (for reviews, see references 464, 678, 730, and 813). The PTSs were classified into four (super)families with distinct evolutionary origins on the basis of the phylogenies of the EIIs (759) as follows: (i) the glucose-fructose-lactose superfamily, comprised of the glucose family (e.g., E. coli EIIAGlc/EIICBGlc or B. subtilis EIICBAGlc [CB, CBA, etc., indicate the domain order in multidomain EIIs]), the fructose-mannitol family (e.g., E. coli EIICBAMtl), and the lactose family (e.g., L. casei EIIALac/EIICBLac); (ii) the ascorbate-galactitol superfamily, comprised of the ascorbate family (e.g., E. coli SgaA/SgaB/SgaT) (358, 989) and the galactitol family (e.g., E. coli EIIAGat/EIIBGat/EIICGat) (610); (iii) the mannose family (e.g., E. coli EIIABMan/EIICMan/EIIDMan or B. subtilis EIIALev/EIIBLev/EIICLev/EIIDLev [a mannose- and fructose-specific PTS]); and (iv) the dihydroxyacetone family (e.g., E. coli EIIA-HPr-EIDha [DhaM] [304] or EIIADha in firmicutes).

The presence of a common preserved domain in the latter two families (411) might indicate a similar evolutionary origin of the two families and thus the existence of a third superfamily. The sequence-based classification of the various PTSs is supported by X-ray crystallography and nuclear magnetic resonance (NMR) studies, which clearly show that the structures of the EIIA (474, 617, 818, 858, 906, 955) and EIIB (1, 2, 98, 459, 623, 778, 906) domains/proteins belonging to the various classes are quite different. Information on the structure of the integral membrane domain EIIC (and EIID) is limited. The membrane topologies of these proteins have been studied using Cys replacement mutagenesis and chemical modification by thiol reagents for EIIBCABgl (960) and EIICBAMtl (918).

Extensive work on the intricate properties of the EII complexes and the detailed mechanisms by which they transport and phosphorylate carbohydrates has been performed in the laboratories of G. T. Robillard, B. Erni, and G. R. Jacobsen. The results that focus predominantly on EIICBAMtl, EIICMan/EIIDMan, and EIICBGlc have been reviewed extensively (369, 730, 813), and we will therefore discuss only a few important aspects. Some more recent papers included work on the properties of a cyclized protein formed by the soluble EIIAGlc and the membrane-bound EIICBGlc (812), the dimerization of EIICBAMtl (903, 904), sugar recognition by the EIICMan/EIIDMan and EIICBGlc of E. coli (256), the membrane topologies of EIIBCABgl (960) and EIICBAMtl (918), and transient-state kinetics of enzyme EIICBGlc (544).

The glucose-specific EII complex of enteric bacteria consists of two distinct proteins, the cytoplasmic protein EIIAGlc (gene crr) and the membrane-associated protein EIICBGlc (gene ptsG), in which the EIIB domain is hydrophilic and in contact with the cytoplasm and the EIIC domain is buried within the membrane. The EIIA and EIIB domains are defined as the domains that receive the phosphoryl group from P~HPr and P~EIIA, respectively. Whereas the phosphorylated residue in the EIIA domain/protein is a histidyl residue, that in the EIIB domain/protein can be either a histidyl residue (mannose family) or a cysteyl residue (all other families). The latter was first demonstrated in E. coli EIIMtl (634), and the specific phospho-cysteyl residue was thereafter identified in a number of cases by analytical methods (547, 633). The phosphoryl group of the EIIB domain is transferred to the carbohydrate after translocation of the carbohydrate by the EIIC domain across the membrane and delivery at the inside face of the cytoplasmic membrane. In the case of the mannitol-specific E. coli EIIMtl or the glucose-specific B. subtilis EIIGlc complex, the EII is a single polypeptide that contains the two hydrophilic domains EIIA and EIIB as well as the transmembrane domain EIIC. Again, in other EII complexes, such as the mannose-specific E. coli EIIMan, phosphoryl groups are carried by two histidyl residues present in the A and B domains of the cytoplasmic EIIABMan, whereas translocation of the carbohydrate requires the two transmembrane subunits EIIC and EIID.

Organization of the PTS. Phosphoryl transfer between the PTS proteins proceeds by an in-line associative mechanism (48). The transfer is accommodated by the formation of heterodimeric transition complexes, and considering the kinetics of the transfer, the heterodimeric association should be transient. The solution structures of different transition complexes, including EI and HPr (260), HPr and EIIAGlc (124, 929), HPr and EIIAMtl (139), HPr and EIIAMan (949), EIIAGlc and EIIBGlc (98, 270), and, finally, P~EIIAChb and Cys10Ser mutant EIIBChb (398), could be determined. The formation of the heterodimeric complexes has a significant effect on the flux control properties of the individual components of the PTS (738, 896, 901).

Based on the observation that E. coli EI and HPr are attached to membrane fragments of E. coli (757), it was speculated that the PTS components associate into a metabolon (611), which would improve PTS function. However, it is not clear how phosphotransfer from one component to the next could take place in a fully associated system without linkers to provide room for the domains to move, and furthermore, as it was calculated that diffusion should not limit glucose influx (239, 240), there should be no functional driving force to fully associate into an EI-HPr-EII complex. Formation of larger complexes by EI linked to yellow fluorescent protein has been observed in cells grown on PTS as well as non-PTS carbohydrates when they reached stationary phase and high cell densities (637). The observed distribution of the fluorescence over the cells was punctate and sometimes even bipolar (after induction of yellow fluorescent protein-EI with IPTG [isopropyl-ß-D-thiogalactopyranoside]). The aggregation appeared to be reversible, as it disappeared after the addition of a carbohydrate, which led to renewed growth of the cells and dispersion of the fluorescence. Thus, phosphotransfer activity within the PTS, as invoked by the addition of a carbohydrate, in fact seems to promote the overall dissociation of the PTS components rather than their association into a metabolon.

Notwithstanding, several PTS proteins were experimentally proved to associate to functional homodimers. The first PTS protein that was found to functionally dimerize was EI (432, 560; also reviewed in reference 114). Later, it was shown that the N-terminal domain does not dimerize (115) but that the C-terminal domain is responsible for dimerization (86). In fact, the association properties of the isolated C-terminal domain are similar to that of entire EI, although the binding constants are orders of magnitude larger (638). It was proposed that the monomer/dimer transition of EI could potentially regulate uptake flux (940), and as a result, the transition has been extensively studied. The idea was supported by the fact that PEP phosphorylates only dimeric EI, and the rate of association/dissociation is very slow, especially compared to sugar uptake (114, 539). The monomer/dimer transition has been studied by various analytical techniques, and many physiological conditions have been tested. It was observed that P~EI has a 10-fold increased dimerization constant with respect to unphosphorylated EI and that the dimerization of EI is promoted by PEP and Mg2+ and inhibited by pyruvate (190-192). It was concluded that the stimulatory effect of Mg2+ and PEP on EI association results from a change in conformation. Association of both normal EI and the nonphosphorylatable mutants H189E, H189A, and H189Q was studied by carrying out sedimentation equilibrium experiments (639). The association constant appears to be very sensitive to pH, temperature, and ionic strength and may vary by as much as 700-fold. It was concluded that the ligands Mg2+ and PEP are the major determinants for dimerization and that their binding leads to conformational changes. A model that accommodates the known facts about the dimerization was presented. A potential regulatory function of EI dimerization with regard to the role of PEP is discussed below.

Several membrane-spanning EIIs have also been shown to dimerize, such as EIICBAMtl (730) and EIICBGlc (213, 214, 548). Proteome analysis of stable oligomeric protein complexes confirmed the existence of EIICBAMtl and EIICBGlc oligomers in the membrane of E. coli (826). In addition, oligomers of EIINag and EIITre were detected, which complies with a generalized picture in which the PTS permeases function as dimers (or oligomers). In contrast, it is generally assumed that HPr and EIIA do not form functional homooligomers. However, several experimental observations suggest the opposite. Biochemical evidence strongly suggested that EIIALac forms homotrimers (176, 319). This suggestion was supported by the crystal structure of EIIALac from Lactococcus lactis, which also revealed a homotrimer (818). In addition, it was reported that four molecules of EIIAGlc associate with one EIICBGlc homodimer (213). Moreover, multimeric EIIAGlc was observed during gel filtration (786) and in crystals (224). Dimeric EIIAChb and P~EIIAChb were observed when ultracentrifugation was carried out (398). EIIABMan was found in the dimeric form both in the cytoplasm and linked to the membrane (216, 826). Also, EIIANtr, a protein homologous to EIIAs that are functional in sugar transport, crystallizes as a dimer (73), although it seems to be a monomer in solution (927). Crh, an HPr-like protein of B. subtilis, was shown to form homodimers in crystals (644), and F29W HPr from Geobacillus stearothermophilus (previously Bacillus stearothermophilus) forms similar domain-swapped dimers in crystallization experiments (827). On the other hand, several measurements performed with the protein in solution suggested that F29W HPr forms monomers. Recently, in vitro experiments provided kinetic evidence that all components of the E. coli glucose PTS form functional dimers (R. Bader, J. G. Blom, K. J. Hellingwerf, M. A. Pelletier, H. V. Westerhoff, and C. Francke, unpublished results).

Although the supposed primary function of the PTS is the transfer of the phosphoryl group from PEP to carbohydrates, all reactions up to and including the EIIs are reversible. The equilibrium constant, Keq, for the reaction PEP{iff}P~EI{iff}P~HPr is about 80 for the E. coli enzymes (539), whereas the Keq for the phosphoryl group transfer between HPr and EIIAGlc from E. coli has a value of between 1 and 1.5 (540). Thus, the overall Keq for the phosphoryl group transfer from PEP to EIIAGlc is in the order of 80 to 120. Furthermore, the Keq for the reaction P~EIIAGlc + EIICBGlc{iff}EIIAGlc + P~EIICBGlc was determined to be 12 (539), very close to the estimated value of 7 (737). The number signifies that the phosphotransfer potential of the thiophospho bond in P~Cys-EIICBGlc is very close to that of the phosphoamidate bond in phosphohistidine of the other PTS proteins. Only the final step, P~EIIBGlc + glucose{iff}EIIBGlc + glucose-6-phosphate (glucose-6-P), is virtually irreversible, with a measured forward rate of 3.2 x 106 M–1 s–1 (539). These values indicate that under physiological conditions, the phosphoryl transfer reactions catalyzed by the PTS between PEP and P~EIIBGlc can run in both directions, from one high-energy P~enzyme intermediate to the next or back to the previous reactions. This differs from most eukaryotic signal transduction systems in which the seryl, tyrosyl, and threonyl residues that are often phosphorylated represent low-energy phosphoenzyme intermediates (837). The reversibility of the phosphoryl group transfer in the PTS has regulatory consequences because it allows the metabolic network to control the phosphorylation state of the PTS proteins in a number of ways, and we will discuss how numerous cellular processes are regulated by the phosphorylation states of certain PTS proteins below (see Table 1 for a summary).


View this table:
[in this window]
[in a new window]

  		TABLE 1. The PTS components and their various non-PTS interaction and/or phosphorylation partners


arrow
REGULATION OF CARBON METABOLISM IN GRAM-NEGATIVE ENTERIC BACTERIA
 
In the following sections, we will discuss the numerous regulatory functions carried out by EIIAGlc in enteric bacteria. This protein not only is involved in the regulation of adenylate cyclase, and therefore in CCR, but also interacts with several non-PTS permeases and glycerol kinase to inhibit their activity. The latter phenomenon is described by the term inducer exclusion. We will present new insights into the complex regulation of the intracellular cyclic AMP (cAMP) level and the binding of EIIAGlc to adenylate cyclase and to non-PTS permeases.

EIIAGlc Is the Central Processing Unit of Carbon Metabolism in Enteric Bacteria

The integration of hundreds of enzymatic reactions into a metabolic network and the complex response of such a network to environmental changes require global regulation at the level of enzyme activity and gene transcription (89, 592, 917). A bacterium that is confronted with changes in the supply of nutrients can adapt its metabolic potential through the induction of specific catabolic operons (most often induced by the substrate) (90, 94). More drastic changes may require global adaptation and therefore the induction of complete regulons (89, 592, 917). Transcriptome data indicate that when E. coli cells that have been growing on a rich carbon source such as glucose are exposed to a carbon source allowing only slow growth, the number of induced genes increases progressively (485). Low growth rates result in the expression of many transport systems and catabolic genes for nutrients that might be absent but that would stimulate growth if present. Results obtained by high-throughput nutrient screening of glucose-grown cells and cells grown under carbon limitation confirmed the effects observed by transcriptome analysis: cells grown under carbon limitation do utilize a large variety of carbon sources, whereas cells growing on glucose do not (the result of CCR) (360). The large-scale transcriptional responses in E. coli are mediated via a limited number of global regulators (527). In enteric bacteria, global regulation of carbon metabolism involves several {sigma} and transcription factors such as Crp (cyclic AMP receptor protein) (reviewed in references 76, 95, 96, 416, and 636), Mlc (666), FruR (145, 267, 268, 762), CsrA (carbon storage regulator) (451, 742, 743, 752, 861), and {sigma}54 (25, 150, 843). The activities of some of these regulators are directly affected by the PTS.

The importance of the PTS in the regulation of carbon metabolism in enteric bacteria became apparent when ptsHI mutants containing defective EI and/or HPr were isolated. These mutants failed to grow not only on PTS carbohydrates but also on a large number of non-PTS carbon sources like lactose, melibiose, glycerol, and maltose (for reviews, see references 677, 678, and 763). Analysis of suppressor mutations, which restored growth of ptsHI mutants on the non-PTS carbon sources (765, 767), revealed that these mutations were in the crr gene (carbohydrate repression resistant), which was later shown to be the structural gene for EIIAGlc (542, 543, 786). While ptsHI-crr triple mutants grew on the non-PTS carbon sources, they did not resume growth on PTS carbohydrates. These results indicated that not only is EIIAGlc involved in glucose transport and phosphorylation via the glucose PTS, it also regulates the transport and/or metabolism of several non-PTS carbon sources.

The changes of structure or charge induced by phosphorylation or dephosphorylation determine the regulatory role of EIIAGlc. P~EIIAGlc is required for the activation of adenylate cyclase. S. enterica serovar Typhimurium and E. coli crr mutants exhibit a low residual adenylate cyclase activity, generally 5 to 20% of the activity found in the parental strain (227, 469, 595). Unphosphorylated EIIAGlc can bind to and inhibit numerous non-PTS proteins, including the permeases for lactose (LacY) and melibiose (MelB), the ATP-hydrolyzing component of the maltose transport system (MalK), and glycerol kinase (GlpK). As a result, induction of the genes involved in the uptake and metabolism of these substrates is prevented, and hence, the inhibitory process was called inducer exclusion.

Both control mechanisms, inducer exclusion and modulation of adenylate cyclase activity, have a physiological function and allow the cell to choose between different carbon sources as long as the uptake of glucose and other PTS carbohydrates leads to the dephosphorylation of EIIAGlc. The connections between carbohydrate uptake, the phosphorylation state of EIIAGlc, adenylate cyclase activity, and the intracellular concentration of metabolites such as PEP and pyruvate relate metabolic flux directly to the regulation of solute uptake and gene expression. In Fig. 2, a summary of the mechanisms underlying CCR in enteric bacteria is given. Although EIIAGlc-mediated regulation has been found only in enteric bacteria so far, this signal transduction system has served and serves as a model for other regulatory processes that involve reversible phosphorylation, such as PTS-regulated transcription activation or antitermination.


Figure 2
View larger version (38K):
[in this window]
[in a new window]

  		FIG. 2. Mechanisms underlying CCR and inducer exclusion in enteric bacteria. The import of glucose and other PTS carbohydrates leads to net dephosphorylation of the PTS proteins (including EIIAGlc and the B domain of EIIBCGlc) and thereby to inducer exclusion and recruitment of the transcription regulator Mlc to the membrane. The upper left part of the figure shows that unphosphorylated EIIAGlc blocks the import of lactose, maltose, and melibiose and the phosphorylation of glycerol by binding to the respective transporter or kinase. The upper right part of the figure shows recruitment of Mlc by unphosphorylated EIIBCGlc, which prevents the regulator from binding to its target sites on the DNA. In the absence of glucose and in the presence of phosphoenolpyruvate, the PTS proteins are found mainly in the phosphorylated state. The central part of the figure shows that phosphorylated EIIAGlc activates adenylate cyclase (AC) but probably only in the presence of an unknown adenylate cyclase activation factor (ACAF). Adenylate cyclase binds phosphorylated as well as unphosphorylated EIIAGlc (see reference 632). The bottom part of the figure shows the effect of the activated transcription factors (free Mlc and Crp:cAMP) on the transcription of the genes encoding Crp, adenylate cyclase, Mlc, and EIIBCGlc, respectively. The inset shows the Crp:cAMP concentration dependence of crp transcription (deduced from data reported in references 310 and 311). The arrow indicates the physiological concentration of activated Crp in exponentially growing cells in the absence of PTS carbohydrates.

Transcription Regulation by Crp/cAMP and Role of P~EIIAGlc

In enteric bacteria, Crp is one of the few truly global regulators, as it controls the expression of a vast number of genes/operons (34, 291, 527, 990). Crp is activated by binding cAMP, which is synthesized from ATP by adenylate cyclase (gene cyaA) (for recent papers on the structure of the Crp/cAMP regulator, see references 316, 471, 671, and 888). Different genes and operons require different levels of Crp/cAMP for full expression because the Crp/cAMP complex can adopt several conformationally active states and because the affinity for the recognition sites on the DNA varies (149, 316, 417, 484, 853, 888).

The concentration of Crp/cAMP is tightly regulated by the PTS. Soon after the discovery of cAMP in E. coli, it was observed, for example, that the addition of glucose to cells growing on lactate lowers the concentration of cAMP (508). Later results confirmed this finding (212, 384), and experiments with toluenized cells established that adenylate cyclase activity is inhibited by glucose (318, 648-652). Other PTS carbohydrates were also shown to inhibit cAMP synthesis, provided that the cognate EII complex was intact and induced (317, 767). Studies with ptsHI and crr mutants demonstrated that adenylate cyclase activity is low in both types of mutants (227, 469, 595). Based on these and other studies (see references 678 and 763), a regulatory mechanism for the activation of adenylate cyclase by P~EIIAGlc was proposed. In this scheme, transport and phosphorylation of glucose or other PTS carbohydrates decrease the extent of phosphorylation of the PTS proteins, including EIIAGlc, and thus lower the activity of adenylate cyclase, whereas growth on non-PTS carbohydrates, particularly poor carbon sources like lactate and succinate, results in fully phosphorylated PTS proteins and the activation of adenylate cyclase. However, this model does not explain how growth on non-PTS carbon sources such as glucose-6-phosphate or gluconate can lead to intracellular cAMP levels similar to, or sometimes lower than, those observed in glucose-grown cells (212, 337). Moreover, the inhibitory effect of glucose-6-phosphate is found only in intact but not in toluenized cells (318), and we will discuss this important point later.

Regulation of cyaA and crp transcription. Most of our current knowledge regarding Crp/cAMP-mediated regulation is derived from the study of cyaA, crp, and/or ptsHI-crr mutants (76, 416, 636, 677, 678, 763, 890). The effect of different carbon sources on growth, gene expression levels, cAMP concentrations, and Crp levels has been tested with these mutants. The addition of exogenous cAMP was used to distinguish between the effects caused by concentration changes of either cAMP or Crp. One class of mutants, designated crp* or crp(In), produced Crp, which is active without cAMP. Unfortunately, an interpretation of the mutant data is not easy, as the factors involved in cAMP-mediated regulation are interdependent, as depicted in Fig. 2.

Inactivation of the cyaA gene disables bacteria like E. coli and S. enterica serovar Typhimurium from growing on a large number of carbon sources (76), while cyaA overexpression appears to be detrimental (699). Inactivation of crp also affects the cAMP concentration by leading to a drastic increase in the rate of cAMP synthesis (75, 170, 241, 363, 679). The increase in the cAMP concentration depends on an intact crr gene. In crp crr double mutants, cAMP overproduction is markedly reduced (from about 60-fold of the wild-type level in the crp strain to only 4-fold of the wild-type level in the double mutant) (143, 170).

The phenomena described above are related to feedback regulation on the level of transcription. Expression of cyaA is negatively regulated by Crp/cAMP in both E. coli (10, 387, 395, 576) and S. enterica serovar Typhimurium (221), whereas expression of crp is both positively (310) and negatively (9, 140, 311) affected by Crp/cAMP. As a result, strong crp expression occurs at low and high Crp/cAMP concentrations (due to reduced inhibition and increased stimulation, respectively), whereas low crp expression is observed at intermediate Crp/cAMP concentrations (310). The concentration of Crp/cAMP in exponentially growing cells is an order of magnitude higher than that allowing minimal crp transcription (310, 365). Consequently, a decrease in the cAMP concentration caused by the addition of glucose to these cells should lower the transcription of crp and hence the concentration of Crp. Glucose-grown E. coli cells indeed contain less Crp, and this was shown not to be due to faster degradation of the crp mRNA (364, 365). The glucose effect on crp expression is absent in mutants lacking either adenylate cyclase or Crp, consistent with the postulated role of Crp/cAMP. The growth stage of E. coli cultures also has an influence on the amount of Crp (365). These variations might explain why some authors reported complete relief from glucose repression after the addition of cAMP (647), whereas others observe only partial relief (932).

Expression of the EIIAGlc-encoding crr gene is almost completely independent of Crp/cAMP (857). In contrast, expression of the EIICBGlc-encoding ptsG gene is almost fully repressed in cyaA or crp mutants (727). Owing to the low level of the glucose transporter EIICBGlc in these mutants, the addition of glucose to cyaA or crp mutants does not lead to the same extent of dephosphorylation of P~EIIAGlc as in the wild-type strain. The regulation of crp transcription by factors other than Crp/cAMP adds further complexity to catabolite repression data. crp transcription is negatively regulated by both the DNA binding protein Fis (280), the transcription of which is in turn regulated by Crp/cAMP (591), and SpoT (378), the protein that mediates the stringent response and that catalyzes the synthesis of ppGpp (guanosine tetraphosphate).

Regulation of adenylate cyclase activity. Whether P~EIIAGlc exerts its regulatory effect on adenylate cyclase directly or indirectly has been an open question for a long time. Early experiments with permeabilized cells containing an intact PTS showed that adenylate cyclase activity is stimulated by potassium and phosphate and inhibited by {alpha}-methylglucoside ({alpha}-MG) and pyruvate (476, 477, 701), which is in line with a stimulatory effect by P~EIIAGlc. However, in vitro experiments with purified adenylate cyclase were not conclusive (963). The effect of EIIAGlc phosphorylation on the activation of adenylate cyclase was studied in crr mutants producing EIIAGlc in which the phosphorylatable His-90 (197) and the catalytically important His-75 had been changed. In comparison to wild-type EIIAGlc, His75Gln EIIAGlc synthesized in a crr mutant caused an approximately fourfold activation of adenylate cyclase. In contrast, neither His90Gln nor His90Glu EIIAGlc activated adenylate cyclase (700, 852). The latter two mutant proteins are not phosphorylated by P~His-HPr, while His75Gln EIIAGlc is. Takahashi and coworkers also found that His75Glu EIIAGlc, which is barely phosphorylated in vitro (700), can slightly activate adenylate cyclase (852). In wild-type EIIAGlc, His-75 and His-90 are in very close proximity, and the replacement of His-75 with a negatively charged Glu possibly mimics phosphorylated His-90. Although these results indicate that phosphorylation of His-90 of EIIAGlc is important for the activation of adenylate cyclase, they do not prove a direct interaction. Recent results with adenylate cyclase tethered to the membrane showed that although both P~EIIAGlc and unphosphorylated EIIAGlc bind with similar affinity to the protein, it is activated by P~EIIAGlc only when an additional, not-yet-identified factor from cell extracts is present (632) (Fig. 2).

From studies with truncated adenylate cyclase, it can be concluded that the catalytic activity resides in the N-terminal domain, with the C-terminal domain required for EIIAGlc-mediated regulation (632, 699, 746). Starting from an E. coli ptsHI-crr deletion strain containing, in addition, an Asp414Asn mutation in cyaA, which prevents the large increase in adenylate cyclase activity in a crp mutant (143), suppressor mutants affected in the cyaA gene could be isolated. Most suppressor mutations were located in the region around Asp-414 and resulted in a truncated adenylate cyclase that was about half the size of the wild-type enzyme (144). The truncated molecules produce about 10 times more cAMP in a crr strain than the wild-type enzyme. These results suggest that the N-terminal domain of adenylate cyclase is active in the absence of P~EIIAGlc and that the C-terminal domain acts as an inhibitor that may be removed/inactivated by P~EIIAGlc.

Mutation of crp leads to a drastic increase of extracellular cAMP. This increase is 50-fold diminished when the C-terminal 48 amino acids of adenylate cyclase are cut off. In contrast, the increase is diminished by only fourfold when the CRP/cAMP-mediated transcriptional regulation of the cya gene is prevented by replacing the cya promoter with the constitutive bla promoter (363). Therefore, it was concluded that the regulation of cAMP production occurs mainly posttranslationally (at the level of enzyme activity) and not at the level of transcription. However, because the regulation of cyaA transcription is affected by the Crp/cAMP concentration, which in turn is affected by CyaA activity, that conclusion does not seem to be justified. This can be illustrated by the changes in adenylate cyclase activity caused by a crr null mutation. Adenylate cyclase activity measured in crr mutants amounts to only 5% to 20% of the activity measured in wild-type strains (227, 469, 595), and the reintroduction of plasmid-borne crr increases it four- to fivefold, thus approaching levels of wild-type activity (477, 700). If EIIAGlc was the only factor interacting with and stimulating adenylate cyclase, crr mutants and strains producing adenylate cyclase missing the last 48 amino acids should exhibit the same phenotype. However, it is evident that the positive effect on adenylate cyclase caused by the expression of plasmid-borne crr in a crr mutant is much smaller than the negative effect (50-fold) caused by the deletion of the 48 C-terminal amino acids of adenylate cyclase in a crp mutant. This difference probably owes to the absence of the Crp-dependent effect of cAMP on transcription in a crp mutant.

Regulation by Crp/cAMP. Despite the enormous amount of experimental data, the control of cAMP and Crp levels in enteric bacteria is still not completely understood. Regulation occurs not only at the level of transcription but also at the level of enzyme activity. These two processes affect each other, making it difficult to quantify their individual contributions. Nevertheless, the general properties of the regulatory mechanism are clear. The immediate response of cells to the addition of a rapidly metabolizable carbohydrate involves regulation only at the level of enzyme activity. The addition of glucose or another PTS sugar causes the dephosphorylation of P~EIIAGlc. This deactivates adenylate cyclase and hence lowers the concentration of cAMP and Crp/cAMP inside the cell. On a longer time scale, regulation on the transcriptional level becomes important. At normal physiological levels of cAMP and Crp, the crp gene is negatively regulated by Crp/cAMP. Thus, a reduced activity of adenylate cyclase leads to lower expression of crp. As a result, the concentration of Crp/cAMP drops even further. On the other hand, a lower Crp/cAMP concentration leads to an increased expression of cyaA, and more adenylate cyclase will accumulate, thereby counteracting the decrease in activity and raising the Crp/cAMP concentration. In principle, the elevated pool of adenylate cyclase provides a higher potential for cAMP production once the glucose is depleted, although the Crp concentration will be lower.

Some results cannot be explained by the proposed mechanism. For instance, in S. enterica serovar Typhimurium crp* (595) and E. coli crp* cya (851) mutant strains, the addition of glucose lowers the concentration of Crp* and hence repression. However, because Crp* is constitutively active, i.e., it does not need cAMP for activity, its concentration should not be affected by the presence or absence of glucose. The occurrence of a component that is able to interact with Crp in the presence of glucose could explain the observed decrease in the Crp concentration. In fact, such a mechanism regulates the activity of the repressor Mlc (588): when glucose is present, EIICBGlc is dephosphorylated and binds the transcription factor Mlc, thus relieving repression. It is tempting to speculate that unphosphorylated EIICBGlc might also bind Crp, but as yet, there is no experimental proof to support this assumption.

Factors other than those mentioned above have been reported to influence the activity of adenylate cyclase, but their physiological relevance remains uncertain. A decrease in adenylate cyclase activity, for instance, was observed upon lowering the membrane electrochemical potential (649). Mutation of fruB (formerly fruF) can have a negative (272) or positive (225) effect on adenylate cyclase activity. Also, when GTP or the elongation factor Tu, a GTP-binding protein (702), was added to purified adenylate cyclase, it stimulated its activity almost twofold (829).

Growth rate effects on cAMP concentration. Although early studies reported hardly any effect of the growth rate on the cAMP concentration under glucose limitation (529, 957), more recent studies report a clear relationship, with elevated cAMP concentrations occurring at low growth rates (613, 614). During growth on glucose in continuous culture, expression of lacZ appears to be independent of ppGpp but directly related to the cAMP concentration and inversely related to the growth rate (438); i.e., growth rate and cAMP concentration are also inversely related. The effect of the growth rate on the cAMP concentration can be explained in terms of the stimulation of adenylate cyclase activity by P~EIIAGlc. The phosphorylation state of the PTS components in these experiments should depend on the growth rate because the cells were grown in a chemostat under carbohydrate-limiting conditions, and the growth rate was varied by changing the rate of carbon source supply. As a result, at high growth rates, less P~EIIAGlc and hence less cAMP should be formed than that at slow growth rates.

Secretion and breakdown of cAMP. Makman and Sutherland (508) observed that following the addition of glucose, starved E. coli cells halt cAMP synthesis and excrete cAMP into the medium. In a later study, the presence of other metabolizable sugars was shown to elicit the same effect (758). Interestingly, crp mutants excrete much higher amounts of cAMP than wild-type cells (75, 241, 679). A putative low-affinity cAMP exporter (Km of approximately 10 mM) was identified (275). However, as the internal cAMP concentration is generally in the micromolar range, excretion is not thought to be relevant to cAMP-mediated regulation.

The breakdown of cAMP by the endogenous cAMP phosphodiesterase (80) provides another possible way to affect the cAMP concentration. However, cAMP phosphodiesterase has a low affinity for cAMP (Km of between 0.5 and 0.8 mM) (80, 604), making it unlikely that this enzyme exerts much, if any, control over cAMP levels. The findings that, in mutants lacking cAMP phosphodiesterase, glucose still lowers the cAMP level (508) and that CCR exerted by various carbon sources is as strong as that in wild-type cells (337) also argue against a physiological role of the cAMP phosphodiesterase.

Inducer Exclusion and Role of EIIAGlc

Soon after the discovery of the PTS, it became clear that ptsHI mutants have a pleiotropic phenotype; i.e., they fail to grow not only on PTS carbohydrates but also on non-PTS carbon sources such as lactose, melibiose, maltose, and glycerol. Based on a number of growth and transport studies with several ptsHI-crr mutants, a model in which unphosphorylated EIIAGlc inhibits systems catalyzing the transport and subsequent metabolism of these non-PTS sugars in enteric bacteria was proposed (765-767, 770; also see reference 678 for more references). An important observation strengthening the model was made in so-called "leaky" ptsH or ptsI mutants of S. enterica serovar Typhimurium (773), where "leaky" means that these mutants retain low EI or HPr activity. The "leaky" mutants do not grow on PTS sugars, but they do grow on the non-PTS sugars melibiose, maltose, and glycerol. Moreover, mutants growing on the latter sugars are hypersensitive to repression by the glucose analog {alpha}-MG. This phenotype can be explained by the fact that a low amount of EI or HPr strongly reduces the phosphorylation flux via the general PTS proteins EI and HPr, causing rapid and nearly complete dephosphorylation of EIIAGlc in the presence of {alpha}-MG. The phenomenon was called inducer exclusion because the presence of unphosphorylated EIIAGlc prevents either the import of these sugars or their subsequent metabolism, and as a consequence, bacterial cells are devoid of the inducer for the corresponding operons (see reference 502).

Inducer exclusion is mediated by unphosphorylated EIIAGlc. During inducer exclusion, unphosphorylated EIIAGlc binds directly to either the respective permease (lactose or melibiose), the ATP-binding subunit of the ATP binding cassette (ABC) transporter (maltose), or, in the case of glycerol, GlpK, the cytoplasmic enzyme catalyzing the formation of the inducer (inducer exclusion is depicted in Fig. 2). Initial evidence for a direct interaction between EIIAGlc and the inhibited proteins came from studies showing that the lactose transport activity of LacY in E. coli vesicles is strongly reduced when the vesicles contain partially purified unphosphorylated EIIAGlc (189, 559). Similar results were obtained with whole cells synthesizing either LacY or the melibiose transporter (MelB) (561). Direct binding of unphosphorylated EIIAGlc to membrane preparations of an E. coli strain overproducing LacY was demonstrated by Osumi and Saier (624), who also found that the binding of EIIAGlc correlates with the amount of transporter and is enhanced by the presence of lactose, while the phosphorylation of EIIAGlc abolishes it. These results were confirmed and quantified by Nelson et al. (593, 597). Direct binding and inhibition by unphosphorylated EIIAGlc were also established for the MalK component of the maltose transport system (163, 675) and for GlpK (166, 676).

The uptake systems for raffinose (876), galactose (559), and arabinose (338) are also subject to inducer exclusion. Raffinose uptake via the plasmid-encoded raffinose permease RafB is inhibited by {alpha}-MG, and the effect is enhanced in "leaky" ptsI mutants (876). It was suggested that galactose transport in E. coli would also be under the control of inducer exclusion (7, 385). Indeed, the uptake of galactose by membrane vesicles containing galactose permease was reported to be diminished by the addition of unphosphorylated EIIAGlc (559). Nevertheless, other experiments from the same group (562, 767) did not reveal an effect of the PTS on galactose fermentation by intact cells. The authors of those reports argued that the absence of an effect probably results from the full induction of the galactose permease in intact cells, thereby reducing the effectiveness of inhibition by EIIAGlc, as inducer exclusion is mediated by a 1:1 stoichiometric interaction (559). Nevertheless, E. coli gal operon expression is submitted to CCR, as was demonstrated by the early experiments of Stephenson and Gale (836).

Efficient binding of unphosphorylated EIIAGlc to the target protein occurs only when the substrate of the target is present. For example, EIIAGlc binds strongly to the lactose transporter LacY only in the presence of a ß-galactoside (597, 624, 800, 825). This mechanism ensures that EIIAGlc, the concentration of which is fairly constant in E. coli and S. enterica serovar Typhimurium cells, will not be wasted on useless binding, i.e., under conditions in which the substrate is not present in the environment. The necessity of substrate binding indicates that the target proteins probably have to undergo a conformational change before unphosphorylated EIIAGlc can bind. Evidence for conformational changes associated with substrate binding has been presented for LacY (245, 390), MelB (586), MalK (71, 581), and GlpK (355, 654, 874).

Interactions of EIIAGlc

Structure of EIIAGlc. EIIAGlc directly interacts with many different proteins as it serves in PTS phosphoryl transfer, inducer exclusion, and probably activation of adenylate cyclase (Table 1). Although most interaction partners are structurally unrelated, it appears that the binding surface of EIIAGlc is always similar. The structure of EIIAGlc is hardly affected by binding to its target proteins (98, 124, 270, 355, 929) or by phosphorylation (642). E. coli EIIAGlc forms an antiparallel ß-sandwich in which His-90 and His-75 of the active site are located off center on one face of the sandwich (224, 643, 955). Besides His-90, the residue that carries the phosphoryl group in E. coli P~EIIAGlc (197), His-75, appears to be very important for the phosphocarrier activity. Replacement of His-75 with Gln results in strongly diminished phosphoryl group transfer (686). A structural study of His75Gln mutant EIIAGlc indicated no significant structural changes around the active site compared to the wild-type protein (643). Two possible explanations were proposed: (i) His-75 stabilizes the negative charge of P~His-90, or (ii) the active site contains a proton relay network also involving Thr-73. The latter hypothesis was recently falsified by replacing Thr-73 with Ser, Ala, or Val and measuring the effect on the phosphotransfer rates to and from HPr (545). These rates were barely affected by the mutations.

The active site of E. coli EIIAGlc is surrounded by a ring of solvent-exposed hydrophobic residues flanked by some polar and charged residues. A similar arrangement was found for the EIIAGlc of Mycobacterium capricolum (349) and the EIIA domain of B. subtilis EIICBAGlc (122, 474). The NMR structure of E. coli P~EIIAGlc shows that the phosphoryl group bound to His-90 (197) protrudes from the ring of hydrophobic residues that surround the active site (642). Supposedly, the disturbance of the hydrophobic binding surface of EIIAGlc prevents the interaction between P~EIIAGlc and the target proteins. Furthermore, the introduction of a negative charge on the protein surface contributes to the diminished interaction.

Interaction with HPr and EIICBGlc. The structures of B. subtilis HPr (330) and the EIIA domain of B. subtilis EIICBAGlc (474) were used to construct a model of the HPr:EIIAGlc transition complex (327). Herzberg concluded that (i) upon the formation of the transition complex, no major conformational change occurs and (ii) the central part of the binding surface consists mainly of hydrophobic residues. These assessments are sustained by the NMR solution structures of the HPr:EIIAGlc transition complexes of B. subtilis (124) and E. coli (929). The transition complex perfectly accommodates the binding of a phosphoryl group by both the active-site histidine of EIIAGlc and that of HPr in either an associative or dissociative mechanism. The active-site histidine of EIIAGlc (His-90 in E. coli) is located in a shallow depression of a slightly concave surface (224, 643, 955), whereas the active-site histidine of HPr is located at the end of the first {alpha}-helix on a convex protrusion of the protein (330). The binding of the phosphoryl group in B. subtilis EIIAGlc is stabilized by His-68 (His-75 in E. coli) (474) and by Arg-17 in HPr (330). In addition, two aspartyl residues are perfectly positioned to form alternate ion pairs with Arg-17 of HPr. The hydrophobic binding surface containing the active site of EIIAGlc includes several valine and phenylalanine residues (Phe-41, -71, and -88 and Val-40, -46, and -96 in E. coli) (929).

It appears that the binding surfaces of EIIAGlc for HPr and EIIBGlc are almost identical (Table 2). The soluble EIIB domain of E. coli EIICBGlc was isolated, and the solution structure of the EIIAGlc:EIIBGlc transition complex was determined (98, 270). The interaction surface consists of a central hydrophobic patch composed of the phenylalanine and valine residues mentioned above, and the same two aspartyl residues form alternate ion pairs with arginine residues. P~EIIAGlc transfers its phosphoryl group to Cys-421 of the glucose permease EIICBGlc (92, 547, 618). This residue is located at the C terminus of ß1 of the EIIB domain on a hydrophobic protrusion (203). Two arginine residues neutralize the negative charge added to Cys-421 upon phosphorylation of EIIBGlc (98). Like Cys-421 (92), the presence of these residues was reported to be compulsory for effective glucose transport and phosphorylation (449). Although the central interaction parts are very similar, the residues forming the edges of the binding surfaces are different between the transition complexes HPr:EIIAGlc and EIIAGlc:EIIBGlc (98, 929). The edges of the binding surface of HPr contain only positively charged residues, whereas those of EIICBGlc include both positively and negatively charged amino acids. Consequently, the edges of the binding surface of EIIAGlc for HPr include many negatively charged and some polar residues, while those of EIIAGlc for EIIBGlc include a polar residue and both negatively and positively charged amino acids.


View this table:
[in this window]
[in a new window]

  		TABLE 2. Residues involved in the interaction of EIIAGlc with partner proteinsa

Sequence comparisons of EIIAGlc and EIIBGlc from various bacteria reveal that the interfacial residues are strongly conserved (98). The few changes are mostly conservative, thus preserving the central network of intermolecular hydrophobic interactions. A higher variability is observed for the peripheral polar and charged residues. There are, for instance, several positively charged residues in E. coli EIIAGlc replaced with polar or negatively charged residues in the B. subtilis protein. Although these substitutions abolish some salt bridges, their contribution to the stability of the complex appears to be moderate, as illustrated by the similar apparent Km values for heterologous (PTS proteins from different organisms) and homologous phosphoryl transfer (98, 723). Point mutations in several well-conserved residues of EIIAGlc also lower the phosphoryl transfer activity but only by between 20 and 70% (822). It was concluded from these observations (98) that multiple interactions contribute to the affinity between PTS protein partners.

The N terminus of EIIAGlc is disordered in the structures of both the HPr:EIIAGlc and EIIAGlc:EIIBGlc transition complexes and is therefore probably not involved in the protein/protein interactions. Nevertheless, Meadow et al. (538, 541) established that although the naturally occurring N-terminal truncation (the first seven residues are absent, and the truncated EIIAGlc therefore migrates faster on nondenaturing gels and was called EIIAfastGlc) has no effect on phosphoryl transfer between HPr and EIIAGlc, it reduces the phosphoryl transfer to EIICBGlc by about 97%. Recent measurements by the same group confirm these observations with genetically truncated EIIAGlc (deletion of the first 7 and 16 residues) (545). Simultaneously, they showed that phosphoryl group transfer to the soluble EIIBGlc domain remains unaffected. Misko et al. (559) found that compared to intact EIIAGlc, EIIAfastGlc has a lower affinity for LacY-containing membranes. These findings suggest that the N terminus plays a role in the recruitment of EIIAGlc to the E. coli inner membrane. Indeed, Wang et al. (926, 928) showed that the N terminus interacts with anionic lipids isolated from E. coli, suggesting that the N terminus serves as a kind of membrane anchor.

Interaction with GlpK. The crystal structure of the E. coli EIIAGlc:GlpK complex has been determined (355). GlpK with bound glycerol and ADP forms tetramers in which each GlpK subunit interacts with one EIIAGlc molecule. Indeed, complete inhibition of glycerol kinase activity occurs only when at least four molecules of EIIAGlc are present per GlpK tetramer (900). The structure of EIIAGlc in the complex is very similar to that of EIIAfastGlc (955), except that residues 1 to 11 are in contact with the surface of GlpK. Like EIIAfastGlc, the EIIA domain of B. subtilis EIICBAGlc lacks the seven N-terminal residues. Nevertheless, the protein complements an E. coli ptsHI-crr mutant defective in GlpK regulation (723). Therefore, the N-terminal residues cannot be very important for the inhibition of glycerol kinase (355). The inhibitory contact between the two proteins is mediated by a 310 helix of GlpK (residues 472 to 481), which protrudes into the concave surface of the active site of EIIAGlc. The contact surface of EIIAGlc is very similar to that in the EIIAGlc:EIIBGlc complex (Table 2). The hydrophobic residues surrounding the active-site histidine are important, and binding involves negatively and positively charged residues.

The organization of the residues around the active site of EIIAGlc in the EIIAGlc:GlpK complex suggests a possible role for Zn2+ in the association. Indeed, crystals soaked in a 20 mM ZnCl2 solution adopted a Zn2+ ion at the expected position (223). Subsequent in vitro studies revealed that the presence of the metal ion lowers the Ki of EIIAGlc for GlpK by about 60-fold (223, 616). The Zn2+ ion is liganded by His-75, His-90, and a water molecule in EIIAGlc and by Glu-478 in GlpK. The stimulatory effect of Zn2+ on EIIAGlc-mediated GlpK inhibition disappeared when Glu-478 was replaced with Cys, Asp, His, or Gln (657).

It is not known how the interaction with unphosphorylated EIIAGlc inhibits glycerol phosphorylation. The more than 30-Å distance between the sites of EIIAGlc binding and glycerol phosphorylation rules out a direct effect. A conformational change of glycerol kinase is part of the catalytic cycle, and so it was proposed that the binding of EIIAGlc could lock the protein in a closed state (355). The lower inhibition by EIIAGlc observed for some mutant GlpKs affected in the domain associated with "opening" and "closing" GlpK is consistent with this concept (655, 658).

Interaction with LacY. LacY is an H+ symporter that uses electrochemical potential to drive the uptake of several galactosides such as lactose and melibiose (390, 391, 942). The recently resolved LacY structure confirms previous predictions that the protein contains 12 hydrophobic transmembrane domains connected by relatively hydrophilic loops (4). The major determinants of substrate binding are located at the interfaces of helices IV, V, and VIII (300, 444, 754). In contrast, the cytoplasmic loop between helices IV and V together with the flanking helical domains and some residues in the large cytoplasmic loop connecting helices VI and VII are important for the interaction of LacY with EIIAGlc (338, 800, 823). Based on the analysis of various lacY mutants, a model for LacY function in galactoside transport has been proposed (3, 391, 942). LacY is assumed to have a conformation in which the sugar binding site is facing the periplasm ("outward"). The binding of the sugar induces a structural change, which exposes the substrate binding site to the cytoplasm ("inward"). To explain the observation that EIIAGlc binds to LacY with higher affinity when its substrate is present (597, 624, 800, 825), the "outward"/"inward" transition is assumed to make the EIIAGlc interaction site more easily accessible (800, 823). Evidence supporting this concept comes from studies of a Cys154Gly mutant LacY. The mutation abolishes both lactose transport and binding of EIIAGlc and was originally thought to lock the protein in the "outward" conformation (819, 825, 902). However, the crystal structure of Cys154Gly LacY indicates that the protein is locked in an "inward" conformation (4). It was then proposed that EIIAGlc binds to the "outward" conformation of LacY, thus preventing the structural change to the "inward" conformation and, consequently, the entry of the bound sugar. Other observations support this interpretation. When galactosides are present at saturating amounts, 125I-labeled unphosphorylated EIIAGlc binds in a 1:6 stoichiometry to wild-type LacY (in inside-out vesicles) (824, 825). The replacement of residues Ile-129 and Lys-131 (these residues are not located in the EIIAGlc interaction site but are located on central helix IV, which is involved in substrate binding) with Cys allows more EIIAGlc molecules to bind to LacY but does not affect the apparent equilibrium dissociation constant, KD, of the EIIAGlc/LacY complex. In addition, different sugars lead to different binding stoichiometries, with the preferred substrates melibiose and lactose causing maximum binding of EIIAGlc. The data were explained by assuming that LacY exists in two (or more) populations: one population that binds EIIAGlc ("outward") and one that doesn't ("inward"). The observed EIIAGlc binding stoichiometry was thought to reflect the ratio of these populations, which should be affected by the transported sugar (and in mutant LacYs by the conformational flexibility). According to the model, the binding stoichiometry should not affect the dissociation constant of the EIIAGlc/LacY complex, which is consistent with the observations. It is noteworthy that in contrast to the data reported by Sondej et al. (824, 825), Nelson et al. (597) reported an EIIAGlc/LacY binding stoichiometry of 1:1. The reason for the discrepancy is not clear.

The interaction surface of LacY and EIIAGlc has been probed by analyzing the effect of various mutations. Residues of the central loop of LacY were found to be involved in the interaction (338, 951). Changing Thr-7 or Met-11 to Ile prevents EIIAGlc from binding (597, 625). However, a deletion of residues 2 to 8 hardly affects the regulation of LacY by EIIAGlc (338). Other residues of LacY involved in binding EIIAGlc were identified by Cys-scanning mutagenesis (823). Although some of the LacY residues important for EIIAGlc binding are conserved in MalK, MelB, and RafB, no consensus EIIAGlc binding motif could be proposed (823). In contrast to the binding surfaces of other EIIAGlc target proteins (HPr, EIIBGlc, and GlpK), the LacY binding surface seems to contain only a small number of hydrophobic residues. Several E. coli EIIAGlc (crr) mutants defective in inducer exclusion of LacY substrates have been identified (822, 987). Some of these residues are also important for the interaction of EIIAGlc with HPr, EIIBGlc, and GlpK (Table 2).

Interaction with MelB. Similar to LacY, the melibiose permease MelB is an ion (Na+/H+) sugar symporter with 12 membrane-spanning segments (74, 586, 966). S. enterica serovar Typhimurium melB mutants synthesizing an inducer exclusion-resistant melibiose permease have been isolated (439) (Table 2). The mutated residues all lie on one side of an {alpha}-helical stretch situated in the cytoplasmic C-terminal region of MelB. Other substitutions located on the same {alpha}-helical stretch diminished inhibition of the permease by EIIAGlc. Cutting off the last 22 amino acids of MelB reduces inducer exclusion by fourfold, and mutants lacking more residues exhibit low melibiose and methyl-{alpha}-thiogalactoside transport even in the absence of inducer exclusion.

Interaction with MalK. Maltose uptake in E. coli requires the malE-encoded periplasmic maltose binding protein and the multisubunit ABC transporter MalFGK2 (72, 226, 447). The soluble MalK subunits are tightly associated with the two permease subunits MalF and MalG (354, 580). Binding of maltose-carrying maltose binding protein to MalFGK2 stimulates ATP hydrolysis by MalK and therefore maltose transport (118, 161, 187). Stimulation of ATP hydrolysis is cooperative, suggesting that the MalK subunits interact with each other (159, 160, 226). MalK possesses an N-terminal ATPase domain (present in all ATP-binding proteins of ABC transporters) and, in certain bacteria including E. coli and S. enterica serovar Typhimurium, a specific C-terminal regulatory domain (66, 187, 782, 828). In the presence of maltose and glucose, EIIAGlc is assumed to bind to the regulatory domain, thus preventing maltose import. In addition, if no maltose is present, the C-terminal MalK domain is proposed to bind and inhibit the transcription activator MalT (380, 627, 828), which controls the expression of the mal regulon (65, 70).

Several malK mutants exhibiting weak inducer exclusion of maltose but normal inhibition of MalT by MalK and vice versa have been isolated (163, 431) (Table 2). In addition, the binding of monoclonal antibodies directed against linear epitopes extending from either amino acids 113 to 123 or 352 to 361 in the MalK sequence is diminished by the presence of EIIAGlc (828). The scattered location of the mutations (Table 2) and the EIIAGlc/antibody competition sites made it difficult to propose a binding site for EIIAGlc (66, 828). In contrast to Thermococcus litoralis MalK (187), biochemical studies suggest that in the functional MalFGK2 complex of E. coli, the C termini of the MalK subunits are in close contact (774). This assumption was confirmed by resolving the structures of three different forms of the MalK dimer (117). Several amino acids, the replacement of which leads to weak maltose exclusion, lie on the same face of the dimer, thus identifying a potential binding site for EIIAGlc. The binding of EIIAGlc to the nucleotide binding domain of one subunit and the regulatory domain of the other could well prevent the "tweezers-like motion" necessary to accommodate the cycle of hydrolysis and subsequent transport (117).

Other interactions. Using ligand fishing, yet another interaction partner of unphosphorylated E. coli EIIAGlc was recovered (419). The protein was called fermentation/respiration switch protein (FrsA; formerly YafA) because frsA disruption causes increased respiration on several sugars including glucose, and overexpression results in increased fermentation. EIIAGlc forms a 1:1 complex with FrsA. Thus, in E. coli, the balance between fermentation and respiration seems to be regulated by the PTS. Furthermore, the E. coli genome has been screened for proteins containing a potential EIIAGlc binding site using sequence patterns deduced from mutagenesis studies of LacY (823). The search has yielded over 35 potential binding partners including HPr and MalK but excluding MelB and FsrA. Therefore, whether the recovered list of proteins is significant or not remains an open question.

Diauxic Growth

As mentioned in the introduction, diauxic growth was first described by Monod (572), who observed that when two carbon sources are supplied simultaneously, one (e.g., glucose or fructose) is often preferred over the other (e.g., glucan, maltose, lactose, or galactose). When the preferred carbon source is exhausted, growth is arrested. After a certain time (lag time), the genes involved in the catabolism of the second carbon source are expressed, and growth resumes. Inducer exclusion and/or poor expression of catabolic genes due to low Crp/cAMP levels is responsible for this phenomenon (Fig. 2).

As outlined above, EIIAGlc-mediated inducer exclusion prevents the uptake of carbohydrates such as lactose, melibiose, glycerol, and maltose and, consequently, the induction of the related catabolic genes. Expression of these genes is also regulated by Crp/cAMP, and growth on glucose reduces the cAMP levels by lowering the concentration of P~EIIAGlc. Even when induced, non-PTS sugar transporters are present at relatively low concentrations compared to those of EIIAGlc. The preferential uptake of glucose during the first part of diauxie is probably ensured by the inducer exclusion mechanism. In contrast, the induction of gene transcription during the lag phase should be caused by the entry/formation of the inducer as well as by rising concentrations of Crp/cAMP. This concept complies well with the available experimental data on glucose-lactose, glucose-maltose, and glucose-melibiose diauxie. An example is glucose-melibiose diauxie in S. enterica serovar Typhimurium, where inducer exclusion of melibiose is responsible for the preferential use of glucose (621). The addition of extracellular cAMP eliminates the lag period, an observation also reported for glucose-lactose diauxie (365, 502). However, in no case could the addition of extracellular cAMP prevent the preferred utilization of glucose over the second carbon source (242).

The concentrations of PEP and P~EIIAGlc, which are low during growth on glucose, rise strongly when glucose is exhausted (335), which switches off the inducer exclusion mechanism and allows the second sugar to enter and eventually leads to the induction of the respective catabolic genes. In the case of glucose-lactose diauxie, the lac operon is initially repressed by LacI, as the presence of glucose does not allow the entry of lactose. When glucose is depleted, lactose is taken up via the small amount of LacY formed despite repression, the disaccharide is converted by ß-galactosidase into allolactose, which binds to and inhibits LacI, and transcription commences (6). The effect of the entry of the inducer on transcription can be mimicked by the inactivation of lacI or by the addition of the nonmetabolizable inducer IPTG. Indeed, both abolish repression of the lac operon and the lag period in glucose-lactose diauxie but not the preferential uptake of glucose (362, 409).

While these results suggest that the repression of the catabolic lac genes is mainly a consequence of inducer exclusion, other data indicate that Crp/cAMP is also involved in the regulation of the lac operon. Four observations have lent support to the latter concept. First, the complete repression of the lac operon in cyaA and crp mutants shows that its expression requires a basal level of Crp/cAMP (6, 438). Second, the addition of extracellular cAMP abolishes the lag phase when switching from glucose to lactose utilization (362, 365, 621). Third, the addition of lactose to a lacI mutant or to IPTG-induced cells leads to a 1.5-fold increase in LacZ production, although inducer exclusion plays no role in this case (362, 409). An up to sixfold increase of LacZ is observed when the E. coli lacI strain KB53 switches from glucose to lactose utilization (425). Fourth, Crp/cAMP and LacI both bind in a cooperative manner to the lac promoter (40, 351, 470).

It is commonly assumed that the lag phase in diauxie is related to the time period required by the cells to synthesize the proteins necessary for the catabolism of the second sugar once glucose is exhausted. The addition of external cAMP abolishes the lag phase. It was therefore not surprising that a strong (sevenfold) transient increase in cAMP levels was observed during the switch from glucose to lactose metabolism (362). In addition, during the lag phase, Crp levels increased slightly before dropping again. We mentioned above that the low concentration of P~EIIAGlc present in cells growing on glucose stimulates the synthesis (but not activity) of adenylate cyclase due to a relief from repression by Crp/cAMP (763). At the onset of the lag phase, the rise in the P~EIIAGlc concentration together with the relatively high amount of adenylate cyclase should cause a transient surge in cAMP before it should drop again as a result of the down-regulation of the cyaA gene due to the increased concentration of Crp/cAMP and possibly by a decrease in EIIAGlc phosphorylation.

As described above, external cAMP is expected to increase the synthesis of the proteins necessary for lactose catabolism despite the presence of glucose, which can still prevent lactose catabolism via inducer exclusion. However, the addition of cAMP to E. coli cells growing on glucose leads to only a slight increase in ß-galactosidase activity (362). This makes it difficult to explain why the addition of extracellular cAMP abolishes the lag phase but not the repression of ß-galactosidase. It is possible that the slight increase of LacY and LacZ synthesis in the presence of high (extracellular) concentrations of cAMP might be sufficient to shorten the lag phase by allowing a more rapid induction of the relevant genes as soon as glucose is exhausted. However, this assumption does not explain why a lag phase is still observed in a crp* mutant in which Crp activity is independent of cAMP (362). In this case, the activity of Crp* is possibly too low in the presence of glucose due to inhibition by an as-yet-unidentified mechanism (365, 851). This would be in line with the observed lowered Crp concentration in the presence of glucose. If glucose regulates the interaction of Crp with an effector and if the regulation would also depend on the concentration of cAMP, the addition of cAMP would, at least partly, relieve repression.

The experimental data discussed in this section concern mostly glucose-lactose diauxie. Although it is likely that the same concept holds for diauxie observed with other sugars that are sensitive to inducer exclusion, i.e., maltose and melibiose, there is little experimental evidence supporting this assumption. Indeed, experiments by Holtman et al. (340) indicate that inducer exclusion plays only a minor role in glucose-glycerol diauxie. The proteins involved in glycerol metabolism are encoded by the genes of the glp regulon. Similar to the lac operon, it is positively controlled by Crp/cAMP and negatively controlled by a repressor (GlpR) (479, 480, 944). The inducer sn-glycerol-3-phosphate is produced from glycerol by glycerol kinase (GlpK), the activity of which is regulated by the allosteric inhibitors fructose-1,6-bisphosphate (FBP) and EIIAGlc (487, 616, 676, 994). Reverse genetics has been used to distinguish between FBP- and EIIAGlc-mediated allosteric regulation of GlpK (340). Surprisingly, glpK mutants that are insensitive to regulation by FBP exhibit no lag phase, consume glucose and glycerol simultaneously, and have only a transient induction of GlpK activity after the cells have depleted glucose. By contrast, glpK mutants that are insensitive to inducer exclusion exhibit diauxie similar to wild-type strains. Therefore, it was concluded that allosteric regulation by metabolic intermediates and poor expression due to low Crp/cAMP levels are probably the dominant mechanisms in glucose-glycerol diauxie. It was proposed that this unexpected behavior might be related to the fact that the inducer of the glp operon is also an intermediate of lipid metabolism.

The supposed direct link between the phosphorylation state of EIIAGlc, glucose uptake, and adenylate cyclase activity has been questioned (614). In a glucose-limited chemostat, it was observed that the cAMP concentration rises sharply (8- to 10-fold) when the external glucose concentration drops below 300 µM, a concentration more than 20-fold higher than the Km of glucose transport (5 to 15 µM). Notley-McRobb and Ferenci claimed that this result ruled out a direct role of the PTS in cAMP production and therefore that another signaling molecule would be required. However, Kremling et al. (425) reported that the lac operon is induced when the glucose concentration drops below 100 µM, which is much closer to the Km. In addition, we found that in glucose-consuming E. coli cells, the cAMP concentration rises steeply between 120 and 30 µM glucose (M. Hoorneman, R. Bader, and P. W. Postma, unpublished results). At the same time, EIIAGlc, which is predominantly unphosphorylated at higher glucose concentrations, becomes mostly phosphorylated. The latter observation complies with the concept that P~EIIAGlc stimulates adenylate cyclase activity.

Transcription Regulation by Mlc

When E. coli grows on PTS carbohydrates, the levels of both the general PTS proteins and the respective EIIs are often increased. The synthesis of EII complexes is usually controlled by sugar-related mechanisms. Operons encoding sugar-specific PTS components are often flanked by or contain a gene encoding either a specific repressor (operons encoding hexitol-specific PTS) (461, 465, 466), an antiterminator (the E. coli bgl operon [287] or the L. casei lac operon [288]), or a transcription activator (the B. subtilis lev operon) (518). These are examples of single operons controlled by their respective inducers. In enteric bacteria, the expression of several operons encoding either sugar-specific PTS components or the general PTS proteins is controlled by an additional mechanism responding to the phosphorylation state of a PTS protein (Fig. 2).

Repression by Mlc and its interaction with unphosphorylated EIICBGlc. The utilization of a PTS substrate, such as glucose, by enteric bacteria increases the amount of not only EIICBGlc and the mannose-specific EIIs (215, 420, 769, 839) but also EI and HPr (172, 727, 771, 839). Unfortunately, a detailed analysis of transcription regulation of glucose-activated genes is complicated by the fact that most of them are also under the control of Crp/cAMP, i.e., partly repressed by the presence of glucose. For instance, cyaA or crp mutants have lower levels of EI, HPr, and several EIIs (727). On the other hand, E. coli wild-type cells grown on glucose exhibit increased levels of these proteins compared to cells grown on lactate. The relatively elevated synthesis in the presence of glucose appeared to require the activity of the repressor Mlc. Overexpression of the E. coli mlc gene leads to an increased colony size during growth on glucose-containing solid medium (hence the name mlc, for making large colonies), which is due to retarded glucose uptake and the reduced formation of acetate (343). This growth behavior led to the discovery of mlc in the first place.

Earlier studies of E. coli mutants that were affected in growth had already suggested that ptsG and ptsM expression might be controlled via a repressor. Suppressor mutants that were able to grow anaerobically on glucose, mannose, and glucosamine were obtained from E. coli ptsG strains, which cannot utilize the above-mentioned carbon sources in the absence of oxygen (732). The suppressor mutants were also 2-deoxy-D-glucose (2DG) sensitive under anaerobic conditions, and hence, the locus/gene was called dgsA. The toxic 2DG is transported mainly via EIICMan/EIIDMan, and indeed, after anaerobic growth, dgsA mutants exhibited elevated mannose and 2DG transport activities compared to the parental strain. DgsA was therefore assumed to be a repressor for the man operon. In fact, mlc was later found to be allelic with dgsA (662).

Strains overproducing Mlc (DgsA) were constructed, and the protein was purified (407, 857). It was found that the presence of a Zn2+ ion is essential for binding to the DNA (780). The amino acid sequence of the 44-kDa Mlc is about 40% identical to that of NagC (343, 579). NagC is a repressor/activator involved in N-acetylglucosamine uptake and metabolism (646, 667, 669, 922), and its activity is induced by N-acetylglucosamine-6-phosphate (668). Both Mlc and NagC belong to the ROK family, which contains repressors, open reading frames (ORFs), and kinases (314, 877). The DNA binding sequences of Mlc and NagC (663) and the putative sugar binding sites of Mlc and the structurally similar ROK family members glucokinase (GlcK) of E. coli and a putative fructokinase (FrcK) of B. subtilis (780) appear to be very similar. Nevertheless, in contrast to NagC, GlcK, and FrcK, Mlc does not seem to bind glucose or metabolites that are derived from it. Gel mobility shift assays showed that these molecules do not affect the interaction of Mlc with its various DNA target sites (167, 407, 408, 664). In addition, activation/deactivation of Mlc by phosphorylation seems unlikely, as no phosphorylation of Mlc has been detected (407).

While studying the expression of ptsHI-crr and ptsG, which are both regulated by Mlc, it became clear that the PTS controls Mlc activity (171, 172, 664, 665, 857). de Reuse and Danchin (171) concluded that transcriptional regulation of the pts operon is a consequence of an increase in the level of unphosphorylated EIICBGlc. Plumbridge (665) then suggested that unphosphorylated EIICBGlc might interact with Mlc (Table 1). Affinity chromatography demonstrated that Mlc indeed binds to unphosphorylated, but not to phosphorylated, EIICBGlc (856). Mlc was found to be sequestered to the membrane when P~EIICBGlc was dephosphorylated during the uptake of glucose (458, 588). In addition, the overproduction of EIICBGlc resulted in the derepression of the Mlc-controlled operons ptsG and ptsHI-crr (588). Finally, surface plasmon resonance experiments and mobility shift assays confirmed that Mlc binds to EIICBGlc but not to EI, HPr, or EIIAGlc (588). The affinity of Mlc for unphosphorylated EIICBGlc (KD = 10–7 M) and for its various DNA target sites (KD = 10–8 to 10–7 M) is similar (588).

Initial studies with mutant EIICBGlc carrying various deletions suggested that Mlc binds to the EIIB domain only when the EIIC-EIIB linker (residues 320 to 390) is present (458). Certain mutations in the EIIC domain interfere with Mlc-mediated regulation (615, 661, 988). In addition to dgs mutants, another class of mutants that also exhibited a phenotype resembling that of mlc strains was isolated. E. coli manXYZ strains grow only poorly on glucosamine, which is also slowly transported via EIICBGlc. Starting from a manXYZ strain, several mutants that were able to grow efficiently on glucosamine were obtained. The mutants, which synthesized EIICBGlc constitutively, were called umgC (uptake of {alpha}-methylglucoside control), and the umgC mutations were mapped close to the ptsG gene (383). UmgC was hence originally postulated to be a repressor for ptsG. Investigation of the original mutants and some new umgC mutants revealed that these mutations actually map within ptsG (615, 661, 988). Interestingly, some of these mutations affect the putative cytoplasmic loops in the EIICGlc domain and cause an up to fivefold increase in {alpha}-MG uptake rates compared to wild-type cells as well as elevated ptsG mRNA levels under noninducing conditions. The effects of umgC mutations resemble those found in mlc null mutants (615), and it was therefore assumed that they increase the affinity of EIIBGlc for Mlc. The binding of Mlc to EIIB is very sensitive to changes in the surface-exposed residues in the vicinity of the phosphorylatable Cys (795), and specific mutations in the EIIC domain possibly affect the arrangement of these residues in the EIIB domain, thereby affecting the binding of the transcription factor. However, the purified EIIB domain alone was found to be sufficient to bind Mlc, but anchoring to the membrane is necessary to prevent Mlc-mediated repression (795, 855). Nevertheless, other effects resulting from the membrane sequestration, like the induction of conformational changes or the masking of binding domains, might contribute to the regulation of Mlc (855).

Mlc has been crystallized, and its structure has been solved (780). The protein appears in two distinct homodimeric forms in the crystallographic unit, and the difference in the distances of the helix-turn-helix (HTH) DNA binding domains between the two forms suggests that only one should be able to bind to the DNA. The helix-turn-helix DNA binding domain of Mlc is stabilized by the C-terminal helix of the protein, and this structural feature seems to be crucial for Mlc function (780). Although deletion of the first 9 C-terminal amino acids does not affect Mlc activity, deletion of the first 18 C-terminal amino acids prevents repression and EII binding (795). The loss of function is accompanied by a transition from the tetrameric form, which is found in dilute solutions (588), to a dimeric form (795).

The Mlc regulon. DNase I footprinting experiments uncovered Mlc binding sites in the regulatory regions of five operons/genes in E. coli. They include ptsHI (407, 665, 857); ptsG (408, 664); manXYZ (662); malT, the positive regulator of the maltose regulon (167); and mlc itself (167). Transcription of these operons/genes is repressed by the binding of Mlc to a site overlapping the downstream promoter (except for the pts operon, where Mlc binds close to P0) and is stimulated by Crp/cAMP, which binds upstream from Mlc. To find out to what extent Mlc and Crp/cAMP contribute to the regulation of the operons, expression of lacZ fusions in wild-type strains was compared to that in strains devoid of adenylate cyclase, Crp, Mlc, or EIICBGlc.

A null mutation of the mlc gene caused a threefold increase in manX expression in glycerol-grown cells, and a similar increase was observed for malT (167). Synthesis of a ManX-LacZ fusion protein is very low in a cyaA mutant and is increased 14-fold by the addition of cAMP. The control exerted by Crp/cAMP appears to be stronger than that of Mlc, as a fivefold-higher amount of ManX-LacZ is present in cells grown on glycerol than in cells grown on glucose (662).

Glucose-grown cells also exhibit expression of mlc that is about twofold lower than that of cells grown on glycerol (167, 808). Whereas transcription regulators are often associated with one of the transcription units that they control, mlc is monocistronic. The coding sequence of the gene is preceded by a Crp/cAMP binding site, two promoters (579), and an Mlc binding site overlapping the second promoter (167). Although both promoters are recognized by RNA polymerase containing the housekeeping sigma factor {sigma}70 (E{sigma}70), P2 is also recognized by the heat shock factor {sigma}32 complex (E{sigma}32) (808). In vitro transcription assays corroborated the finding that transcription is initiated from both promoters and that Crp/cAMP has a positive effect on P2 in the presence of E{sigma}70 but a negligible effect in the presence of E{sigma}32. Crp/cAMP inhibits transcription from P1, and Mlc inhibits transcription from P2. The synthesis of {sigma}32 in response to a heat shock leads to the overexpression of mlc. Nevertheless, the Mlc-to-EIICBGlc ratio remains fairly constant because ptsG transcription is also induced by E{sigma}32.

The EIICBGlc-encoding ptsG gene is expressed from two promoters, which are under the control of Mlc (408, 664). One Mlc binding site overlaps P1 and precedes a Crp/cAMP target site, and the other is upstream from P2. Crp/cAMP-dependent expression from P1 prevails, as P2 transcripts amount to only 10% of the total ptsG mRNA (664). crp or cyaA null mutants do not express ptsG (408, 409, 727), while ptsG is up to 18-fold overexpressed in mlc null mutants (depending on the growth conditions) (373, 664). In wild-type cells, growth on glucose causes an eightfold induction of ptsG expression compared to growth on non-PTS carbon sources (215, 664). It follows that although affected by Crp/cAMP, the expression of ptsG is regulated mainly by Mlc. In addition, other factors have been implicated. It appears that the repression of ptsG transcription from promoters P1 and P2 by Mlc and the activation of transcription from promoter P1 by Crp-cAMP are enhanced by Fis (807), one of the major histone-like proteins of E. coli (279). In addition, under conditions of nutrient limitation, ptsG expression is repressed by E{sigma}s (RpoS), and this repression seems to act in synergy with that of Mlc (792). Furthermore, using ligand fishing with the promoter region of E. coli ptsG, it was found that the two-component response regulator ArcA, when phosphorylated, binds the promoter region at three positions: two between P2 and P1 overlapping the Crp/cAMP binding sites and one next to P1 overlapping the Mlc binding site (373). P~ArcA also binds to the promoter region of the ptsHI-crr operon but with lower affinity. The binding of P~ArcA leads to a twofold reduction in ptsG transcription. The following link between the Arc two-component system and the glucose PTS might exist. An elevated glycolytic flux might cause a redox imbalance and hence lead to the activation of the response regulator ArcA, which is phosphorylated when reducing equivalents accumulate under aerobic conditions (271). P~ArcA then binds to the ptsG promoter region and represses the transcription of ptsG (373), thus reducing the uptake of glucose and the glycolytic flux.

A different mode of regulation of the EIICBGlc concentration is related to the stability of the RNA message. It was established that ptsG mRNA stability is reduced in an RNase E-dependent manner by metabolic blocks early in glycolysis (accumulation of glucose-6-P or fructose-6-P) (209, 410, 577, 578). Destabilization of the mRNA depends on enolase (578), a major component of the RNase E-containing degradosome, and the RNA chaperone Hfq (394). The chaperonin Hfq promotes the annealing of specific small RNA (sRNA) to target mRNA, thereby affecting the translation of the message (269, 842, 895). The ptsG mRNA is destabilized by an Hfq-binding sRNA called SgrS (898). Furthermore, expression of sgrS (formerly ryaA) is controlled by the transcription activator SgrR (formerly YabN) and becomes activated under conditions of {alpha}-MG-6-P (glucose-6-P) accumulation. However, Hfq-catalyzed annealing of the SgrS sRNA to ptsG mRNA alone is not enough to explain the observed instability of the ptsG mRNA. Kawamoto et al. (394) showed that the presence of the first two transmembrane helices of translated PtsG (i.e., localization near the membrane) is essential for an effective ptsG mRNA breakdown under conditions of hexose-6-P accumulation.

Transcription of the ptsHI-crr operon is initiated from three promoters (172, 173, 237, 753). P0 and P1 each yield two different transcripts: a short transcript coding for HPr only and a long transcript encompassing all three genes (172). The third promoter is located at the 3' end of ptsI and allows crr transcription. Transcription from P2 is independent of Crp/cAMP and Mlc (857) and accounts for about 80% of the total crr-containing mRNA (172). As a result, growth on glucose hardly affects the concentration of EIIAGlc, whereas it increases the number of ptsH and ptsI transcripts three- to fourfold (171, 172, 727, 857). This induction is due to an increase in transcription from P0. In the absence of glucose, Mlc binds to the P0 promoter region and prevents the binding of the RNA polymerase (407, 665, 857). Although the presence of Crp/cAMP is essential for expression from P0 (173, 665, 750, 857), this promoter is controlled mainly by Mlc (407, 857). The P1 promoter is preceded by a second Crp/cAMP binding site, and expression from this promoter is weakly controlled by Crp/cAMP (173) but is not affected by Mlc (407, 857). Next to the Crp/cAMP is a FruR binding site (751), the inactivation of which has no effect on ptsHI-crr expression (665, 857). Recently, it was found that crr expression from P2 is affected by the histone-like nucleoid-structuring protein (427), but the physiological relevance of this finding remains to be established.

Repression by Mlc and activation by Crp/cAMP are directly related to the phosphorylation state of EIICBGlc and EIIAGlc, respectively. As explained above, during growth on glucose, unphosphorylated EIIAGlc and EIICBGlc prevail, with the latter sequestering the repressor Mlc to the membrane. As a result, EIICBGlc will be synthesized, whereas the amount of Mlc will decrease slightly due to the low amount of Crp/cAMP, which will further increase EIICBGlc synthesis. Repression by Mlc is relieved not only by growth on glucose but also by growth on other rapidly metabolized PTS substrates such as N-acetylglucosamine and mannitol and to a lesser extent by growth on fructose and mannose (407, 661). Lactose, melibiose, and sucrose have no effect. Mlc repression is also affected by maltose, although maltose is not a PTS sugar. However, maltose degradation yields intracellular glucose and glucose-1-P. The import of maltose can therefore relieve Mlc repression inasmuch as intracellular glucose causes P~EIICBGlc dephosphorylation (678).

Role of Phosphoenolpyruvate

The effects of PTS carbohydrates on the transport of non-PTS carbon sources and on cAMP synthesis seem to result mainly from changes in the phosphorylation state of EIIAGlc. Early data on the phosphorylation state of PTS proteins showed that in galactose-grown S. enterica serovar Typhimurium cells, EIIAGlc was mainly phosphorylated, whereas after the addition of {alpha}-MG, P~EIIAGlc rapidly became dephosphorylated (596). In the meantime, the phosphorylation state of EIIAGlc in E. coli cells growing on a number of PTS carbohydrates has been determined. When grown on glucose, more than 95% of the EIIAGlc is unphosphorylated. A lower percentage of unphosphorylated EIIAGlc was detected in cells grown on other PTS carbohydrates such as mannose, fructose, and mannitol (335). Interestingly, EIIAGlc is also mainly dephosphorylated in wild-type cells grown only in rich medium, whereas in cya or crp mutants, EIIAGlc is almost completely phosphorylated (852).

An unexpected finding was that EIIAGlc is also partly unphosphorylated in cells grown on several non-PTS carbon sources, including lactose, melibiose, maltose, arabinose, and glucose-6-phosphate (334, 335). For instance, in cells growing on lactose or glucose-6-phosphate, approximately 70% and 60% of the EIIAGlc is unphosphorylated, respectively. A correlation was found between the phosphorylation state of EIIAGlc and the ratio of the intracellular concentrations of PEP and pyruvate, the substrate and product of EI phosphorylation, respectively. A decrease in this ratio leads to a decrease in EIIAGlc phosphorylation. In harvested and washed E. coli cells, the intracellular PEP concentration ranges from 2 to 4 mM, and that of pyruvate ranges from 0.2 to 1.2 mM. The addition of glucose changes the concentration to about 0.3 mM PEP and 2.5 mM pyruvate within 15 s (335). For other carbohydrates, the changes are weaker and take slightly longer (30 to 60 s). The low PEP-to-pyruvate ratio observed during the metabolism of several carbon sources is expected to lead to poor phosphorylation of the PTS proteins, as the first phosphoryl transfer steps are reversible (540, 737, 940). An important corollary of these results is that carbon sources, whose transport/metabolism lowers the PEP-to-pyruvate ratio and is catalyzed by enzymes that are sensitive to inhibition by EIIAGlc, should slow their own uptake. This has been confirmed for lactose. Cells lacking EIIAGlc or producing a mutated LacY permease that is insensitive to inducer exclusion (inhibition by EIIAGlc) exhibit elevated lactose influx (336).

It has been observed that when nonmetabolizable PTS substrates were added to starved cells of S. enterica serovar Typhimurium (839) or L. lactis (871), their uptake is initially fast and slopes off quickly. The phenomenon has been used as an argument for a possible regulatory role of the monomer/dimer transition of EI (639, 940). However, the observation can be explained by the rapid drop of the PEP concentration upon the addition of sugar (335, 527, 871), which will cause a continuous decrease of the uptake rate until the PEP concentration reaches steady-state levels. In addition, model calculations (C. Francke, unpublished results) suggest that the initial uptake (first seconds) by starved cells might be extremely fast owing to a surplus of phosphoryl groups that are attached to PTS proteins in the nonfluxing state (equilibrium) compared to the state where phosphoryl group transfer towards the incoming sugar has commenced. The strongly diminished uptake rate that is sometimes observed can be explained by the accumulation of phosphorylated (nonmetabolizable) products and equilibrium, which will finally be reached under these conditions. In conclusion, although the extremely slow monomer/dimer transition rate of EI certainly shows regulatory potential, so far, uptake experiments can be perfectly explained by mechanistic models (239, 737) that do not take into account the dimerization of EI. In contrast, these models indicate that the PEP-to-pyruvate ratio is an important factor in regulation.

CCR Mediated by Non-PTS Sugars and Catabolic Intermediates

To determine the effect of Crp/cAMP on the expression of catabolic genes without the interference of inducer exclusion, lacZ expression was studied in the presence of the nonmetabolizable inducer IPTG (337), and mal operon expression was studied in the absence of its inducing sugar (210, 211). Lactose, gluconate, and glucose-6-P exert strong CCR on lacZ expression, even in a crr mutant, and a clear correlation between cAMP/Crp levels and lacZ expression was observed (337). Glycerol-3-P, glycerol, and other non-PTS sugars (i.e., arabinose, rhamnose, and especially xylose) cause strong repression of malT and malK, although phosphorylation of EIIAGlc is only slightly lowered (210, 211). Repression by glycerol is absent in a glpK mutant and enhanced in a glpD mutant (GlpD transforms glycerol-3-P into dihydroxyacetone-P), suggesting that repression is caused by elevated glycerol-3-P levels. Similar to glycerol-3-P, glucose-6-P and gluconate also lower malT expression (211). While growing a pgi mutant (Pgi, phosphoglucose isomerase, converts glucose-6-P into fructose-6-P) in the presence of glucose-6-P has only a weak effect on the phosphorylation state of EIIAGlc (334, 335), malT expression is still repressed under these conditions (211). If glucose-6-P is normally metabolized, mainly unphosphorylated EIIAGlc is present (334), and the repression of catabolic genes can be achieved via both inducer exclusion and low adenylate cyclase activity. However, if gluconate-6-P or glycerol-3-P accumulates, these metabolites seem to exert CCR even at high levels of P~EIIAGlc. Although the phosphorylation state of EIIAGlc is barely altered, the cAMP concentration drops severalfold upon the addition of glucose-6-P or glycerol-3-P, whereas the concentration of Crp remains fairly constant (210, 337). A low cAMP level during growth on glucose-6-P has also been reported (201). It was therefore concluded that these metabolites interfere with the activation of adenylate cyclase by P~EIIAGlc (211).

2-Ketobutyrate, a precursor of leucine, is another metabolite involved in CCR. The addition of 2-ketobutyrate causes a strong transient decrease of the cAMP concentration in E. coli wild-type cells but not in ptsI or crr null mutants (152). The link between the PTS and the metabolism of 2-ketobutyrate is not clear. However, it has previously been reported that 2-ketobutyrate (which resembles pyruvate) accepts the phosphoryl group from P~EI (768). This could reduce the phosphorylation state of EIIAGlc and thereby reduce the activity of adenylate cyclase, especially when 2-ketobutyrate accumulates inside the cell. A transient accumulation of 2-ketobutyrate and a concomitant drop in cAMP concentration were indeed observed when cells growing on glucose were shifted from anaerobic to aerobic conditions (152).

Reverse Inducer Exclusion or Retroregulation

It is generally believed that PTS sugars, particularly glucose, are the preferred substrates in enteric bacteria and that they are on top of the hierarchy of carbohydrate utilization. However, one would expect that the overproduction of non-PTS target proteins of EIIAGlc, such as GlpK or LacY, would lower the amount of free EIIAGlc. Because the EIIAGlc phosphorylation site is usually part of the interface in the EIIAGlc/non-PTS protein complexes, the phosphoryl group transfer through the glucose PTS will be diminished. This should lower the uptake of glucose via the PTS and at the same time diminish or even prevent inducer exclusion. The latter phenomenon has indeed been observed and was called escape from inducer exclusion (595) or desensitization (760). About 25% of the amount of EIIAGlc present in S. enterica serovar Typhimurium wild-type cells is sufficient for maximal uptake of {alpha}-MG (901). Lowering the amount of EIIAGlc below this level causes a decrease in the rate of {alpha}-MG transport. Interestingly, in cells producing high levels of GlpK, the rate of {alpha}-MG transport decreases by adding glycerol (734, 735), which reinforces the interaction between GlpK and EIIAGlc. The competition of various target proteins for EIIAGlc and the resulting consequence for regulation in intact cells have also been studied. If different EIIAGlc target proteins are simultaneously present in the cell, they could compete for EIIAGlc, especially when they outnumber the EIIAGlc molecules. Competition between the lactose carrier and GlpK for EIIAGlc was indeed observed in vitro (166). Similarly, inducer exclusion of glycerol or maltose in intact cells is relieved after the induction of the lac operon and the addition of thio-ß-galactoside, which strengthens the interaction between LacY and EIIAGlc (761). In the same vein, inducer exclusion of maltose is relieved after the induction of GlpK synthesis and the addition of glycerol. Increasing the amount of EIIAGlc by introducing a crr-containing plasmid restores inducer exclusion (594).

Regulation by EIIAGlc-Like Proteins

The crr mutation was isolated as a suppressor mutation in ptsHI strains, which restored growth on many non-PTS carbohydrates. No other mutations giving rise to this general suppressor phenotype have been reported, suggesting that EIIAGlc is unique in being able to regulate PTS-mediated inducer exclusion and to activate adenylate cyclase in enteric bacteria. Therefore, it came as a surprise that a residual level of inducer exclusion was observed in S. enterica serovar Typhimurium crr deletion strains (595). This effect is probably mediated via EIIs containing an EIIAGlc domain such as EIICBANag and EIIBCABgl. In bacteria grown on a carbohydrate that induces one of the EIIs, antibodies against EIIAGlc cross-react with membrane-associated proteins (786). In addition, E. coli EIIBCABgl and Klebsiella pneumoniae EIICBANag completely restore {alpha}-methyl-glucoside transport in a crr mutant (921), and EIIAGlc complements a C-terminal deletion in EIICBANag (923). The EIIA domain of EIICBANag restores inducer exclusion in a strain lacking both EIIAGlc and EIICBANag (899). In addition, several EIIAGlc-like domains of proteins from gram-positive organisms, such as EIIAGlc of B. subtilis EIICBAGlc or Streptococcus thermophilus LacS, restore inducer exclusion in E. coli (302, 723).

Is EIIAGlc-mediated regulation limited to enteric bacteria? Many bacteria contain an EIIAGlc-like protein or domain, but whether EIIAGlc plays a regulatory role has been investigated for only a few organisms, such as B. subtilis, M. capricolum, Haemophilus influenzae, S. thermophilus, and S. coelicolor.

Similar to E. coli ptsI mutants, B. subtilis ptsI mutants are not able to grow on the non-PTS carbon source glycerol (265). However, in contrast to E. coli, inactivation of EIIAGlc does not restore growth of the B. subtilis ptsI strain on glycerol (282). It will be discussed below that in gram-positive bacteria, EIIAGlc does not interact with GlpK but that this enzyme is regulated via P~His-HPr-mediated phosphorylation (158). In M. capricolum, glycerol kinase is also not inhibited by EIIAGlc (992).

Sequencing of the H. influenzae genome revealed genes encoding EI, HPr, and EIIAGlc (ptsHI-crr operon) as well as genes encoding an EIIBCFru-like protein (fruA) and a protein in which two FPr domains, each containing a phosphorylatable histidine (His-300 and His-424), are fused to EIIAFru (EIIAFru-FPr-FPr; fruB) (500, 717). In H. influenzae, only fructose, but not glucose, is taken up via a PTS, and a ptsI mutant can therefore ferment glucose (499). Interestingly, the development of competence, which in H. influenzae requires cAMP and Crp (105, 196), was lowered about 70-fold in a crr mutant (305) and 250- to 500-fold in a ptsI mutant (305, 499). The addition of extracellular cAMP restored the development of competence. It is likely that H. influenzae P~EIIAGlc stimulates adenylate cyclase activity in a manner similar to that observed in enteric bacteria. No EIICBGlc exists in H. influenzae, and thus, the presence of glucose does not lead to direct dephosphorylation of P~EIIAGlc. Nevertheless, the uptake of fructose will lower the amount of P~EI and P~HPr and consequently will also lower the amount of P~EIIAGlc.

Although the main uptake system for glucose in S. coelicolor is GlcP, a non-PTS permease of the major facilitator superfamily (911), CCR by glucose was observed (21, 217, 530, 912). Glucose kinase (443, 911) and the accumulation of glycolytic intermediates (694) are implicated in mediating the effect. Although S. coelicolor contains EI, HPr, EIIAGlc (designated EIIAcrr), and EIIBCGlc homologs as well as other sugar-specific EII complexes, PTS proteins do not seem to be involved in CRR in this high-G+C gram-positive organism (58, 629). In contrast to E. coli, the crr gene of S. coelicolor precedes ptsI, and ptsH is located somewhere else on the chromosome. EIIAcrr restores growth of an E. coli crr strain on glucose (392). HPr from S. coelicolor is efficiently phosphorylated by B. subtilis EI and PEP but is barely modified by B. subtilis HprK/P and ATP (97, 630). In fact, S. coelicolor is missing HPrK/P, and the deletion of ptsH has no effect on the glucose repression of galactokinase (97) or glycerol kinase (612). The absence of fluctuations in the cAMP concentration (112) restricts the potential role of the PTS to inducer exclusion-mediated mechanisms. However, although expression of the S. coelicolor crr gene in an E. coli ptsHI-crr deletion strain significantly reduces maltose uptake and although a specific interaction between E. coli MalK and S. coelicolor EIIAcrr was established by using surface plasmon resonance spectroscopy (392), no evidence showing that the PTS protein exerts a similar function in S. coelicolor has been obtained.

Mathematical Modeling of the PTS and Its Role in CCR

The study of complex metabolic and regulatory networks is facilitated by the availability of mathematical models (463). As a consequence of the vast amount of reliable experimental data on individual reaction rates, it has become possible to make such models for the PTS. A kinetic model for the glucose PTSs of E. coli and S. enterica serovar Typhimurium has been constructed (737) and converted spatiotemporally to unravel the possible effects of diffusion on PTS function (239, 240). The flux data obtained earlier in vivo (901) and in vitro (738) are described reasonably accurately by these models. When using the measured cellular protein concentrations, the computations indicate that most PTS components should be present as heterodimeric phosphoryl transfer complexes. Lowering the total EIIAGlc concentration in the calculation, for example, by binding to non-PTS components (735), has no significant effect on either the flux of phosphoryl groups or the concentration of unphosphorylated EIIAGlc. Consequently, inducer exclusion will hardly have an effect on glucose influx, which was also observed experimentally (734). During the uptake of a rapidly metabolizable PTS sugar, the concentration of unphosphorylated EIIAGlc is predicted to be low. The low concentration of unphosphorylated EIIAGlc implies that the affinities between non-PTS transporters or GlpK and unphosphorylated EIIAGlc should be stronger than originally predicted. In fact, this prediction is in line with the interaction data for GlpK and EIIAGlc. Initially, the Ki was reported to be 16.6 µM, but in the presence of 0.1 mM Zn2+, it was found to be as low as 0.28 µM (223). When assuming a simple one-to-one association mechanism between GlpK and unphosphorylated EIIAGlc and a predicted concentration range for unphosphorylated EIIAGlc of 2.5 to 0.25 µM (239), the ratio of active/inactive GlpK can vary from 1.1 to 0.11. The prediction is also supported by the reported KDs for EIIAGlc from LacY (1.0 µM) (825) and FrsA (0.2 µM) (419).

A remarkable outcome of the spatiotemporal modeling is that although diffusion is not limiting for glucose uptake, it prevents an equal distribution of unphosphorylated EIIAGlc throughout the cell, with a significantly (30%) higher concentration near the cytoplasmic membrane (239, 240). LacY, MelB, MalK, and probably also GlpK (920) are associated with the cytoplasmic membrane, which implies that diffusion limitation enhances inducer exclusion. The calculations also clearly indicate that the PTS is a dual sensing system. In accordance with the experimental data, both the glucose and the PEP concentrations are predicted to affect the phosphorylation state of EIIAGlc (335, 736). The same conclusion can be drawn from the response towards carbohydrate pulses of two other detailed kinetic models of the PTS (including glycolysis) (426, 777). The first model (426) was analyzed further with respect to the time hierarchy of the responses, and the second model (777) was analyzed further with respect to the regulatory roles of the various PTS components.

A mathematical model providing a different view of the PTS was presented by Thattai and Shraiman (868). Those authors attempted to describe the competition of different sugar-specific PTSs for the phosphoryl groups provided by the common PTS components EI and HPr and derived phase diagrams for the uptake rates of the "competing" PTS sugars. Their procedure revealed that there is only one "nontrivial switching" phenotype generated by the PTS. This phenotype corresponds to diauxic growth (winner-takes-it-all behavior). However, the calculations did not include the regulation of enzyme activity by catabolic intermediates and transcription regulation via Crp/cAMP and Mlc. In addition, the model is based on the assumption that the concentrations of EIIAs and EIIBCs are correlated and that the sugar uptake rate is maximized, while the phosphoryl group flux is limiting. In the case of glucose, these assumptions are problematic, as the expression of EIIAGlc and that of EIICBGlc are not correlated (see the section on the mlc regulon) and transport controls the flux of phosphoryl groups (240, 737, 901). Nevertheless, hierarchical utilization of PTS carbohydrates is a common feature of both gram-negative and gram-positive bacteria (see also reference 678).

Several mathematical models that describe the expression of the lac operon exist (626, 776, 802, 953, 978, 979). The model described by Wong et al. (953) contains regulation by inducer exclusion, cAMP, and LacI but lacks regulation of cyaA and crp expression. The expression of lacI is assumed to be constitutive, and the cAMP level is linearly proportional to the glucose transport rate. Intracellular phosphorylation of glucose by EIICBGlc is not considered. Nevertheless, for cells growing on glucose and lactose, the experimentally observed preferential utilization of glucose, the short lag phase when shifting to lactose utilization, the accompanying sharp rise in the cAMP concentration, and the induction of lacZYA transcription can be reproduced. However, in contrast to the experimental data (362), the concentration of cAMP remains high after the lag phase, and the time course predictions are incorrect. The models by the group of Mackey focus on the proposed bistable nature of the lac operon. A bistable system can convert a graded signal into a switch-like response; i.e., when the inducer concentration passes a certain threshold, the system is switched on. At the same time, to switch off and return to the initial state, the inducer concentration has to drop below a smaller threshold value (hysteresis). The initial models, which did not include the mechanisms responsible for CCR (978, 979), described the induction of the lac operon during growth on lactose well. By including experimental data on ß-galactosidase expression, the models predicted bistability for the lac operon depending on the concentration of extracellular lactose and the growth rate. When activation/repression by Crp/cAMP and inducer exclusion were included (776), bistability was maintained. The potential for bistable behavior of the lac operon was further proved by the observed history dependence (i.e., a hysteretic dependence) of the induction of fluorescent reporter proteins synthesized from genes expressed from the lac or gat promoter (626). Based on these induction experiments, a mathematical model was constructed, which separates the effects of inducer exclusion (acting on LacI activity) and Crp/cAMP, as E. coli MG1655, which was used in these studies, lacks GatR. Its gat promoter is regulated solely by Crp/cAMP and can thus be used as a specific reporter of Crp/cAMP effects. It was found that repression by glucose is mediated by both inducer exclusion and Crp/cAMP and that the effect of the latter is stronger. Similarly, the calculations reported by Santillán and Mackey (776) indicate that the effect of activation/repression by Crp/cAMP and inducer exclusion on the induction of the lac operon are cumulative so that both the sensitivity of the system towards glucose and the concentration of lactose required to induce the system are elevated. However, in agreement with the conclusions drawn from "wet" experiments, the calculated effective ß-galactosidase concentration appears to be more sensitive to inducer exclusion than to Crp/cAMP (see the section on diauxic growth).

A comprehensive model of diauxic growth was presented by Kremling et al. (425). Repression of ptsG by Mlc, inducer exclusion, and non-PTS transport of glucose were included, but the formation of heterodimeric phosphoryl group transfer complexes was not taken into account. Parameters were taken from the literature or otherwise estimated and optimized on the basis of growth experiments with isogenic mutants (cyaA, lacI, and ptsG) obtained from E. coli strain LJ110, a well-characterized derivative of the E. coli K-12 reference strain W3110. The model perfectly reproduces biomass yield, glucose and lactose consumption, the preferential use of glucose, and the induction of lacZ expression. In agreement with the experimental data, the model predicts high concentrations of unphosphorylated EIIAGlc during growth on glucose, followed by a rise in the P~EIIAGlc and cAMP concentrations when glucose is exhausted. When intracellular glucose is produced from lactose and phosphorylated by the PTS, P~EIIAGlc is predicted to drop again. A similar model describes the transport and metabolism of the non-PTS sugar glycerol and the PTS substrate sucrose as well as glp and scr gene expression (931). Again, the model reproduces the experimental data obtained with E. coli K-12 derivatives well. The models described by Wang et al. and Kremling et al. (425, 931) were used to construct a large-scale stochastic model of glucose, lactose, and glycerol metabolism (up to pyruvate), including transcriptional regulation (688). The model reveals potential effects of history and stochasticity, the latter especially in transcription, on reaction network behavior towards the changes in available nutrients.

Other Regulatory Mechanisms Involving the PTS in Enteric Bacteria

Phosphorylation of EI by ATP. PEP-dependent phosphorylation of EI was originally thought to be the only route to generate P~His-HPr (435, 436). However, two decades after the discovery of the PTS, PEP-independent EI phosphorylation was discovered (236). E. coli acetate kinase (AckA) and [{gamma}-32P]ATP phosphorylate EI in a reversible reaction (236). When catalyzing the formation of acetyl~P from acetate and ATP, AckA is transitorily phosphorylated on a glutamyl residue (238, 881), probably at Glu-387 in E. coli AckA (817). PEP and P~AckA most likely phosphorylate the same histidyl residue in EI (16), as the phosphoryl group of P~AckA can be passed to EIIAGlc via EI and HPr. A second protein kinase that also phosphorylates EI at the expense of ATP was found in E. coli (154, 155). It was partially purified, but its activity vanished during purification due to the loss of a cofactor, which was shown to be NAD(P)+. The reaction catalyzed by this EI kinase is reversible; i.e., it can transfer the phosphoryl group of P~EI back to ADP, and the backward reaction requires the presence of NAD+. In its absence, EI kinase acts as a phosphatase and converts P~EI to EI and Pi. It is not yet clear whether phosphorylated EI kinase is transitorily formed during P~EI dephosphorylation. Although AckA- and EI kinase-mediated phosphorylation of EI have been demonstrated in vitro, whether they have physiological importance remains to be shown.

Chemotactic response to carbohydrates. Motile bacteria such as E. coli and B. subtilis can be attracted to or repelled from various chemical and physical environmental signals. These signals can be processed via different pathways (see reference 11), but ultimately, they affect the frequency with which the rotation direction of the flagella changes. An increase in counterclockwise rotation results in smooth, straight movement of the bacterium (e.g., towards the attractants), whereas an increase in clockwise rotation leads to "tumbling" and a change in direction (e.g., away from the attractant) (see reference 501). The most important chemotaxis signal processing pathway comprises chemoreception in the cytoplasmic membrane by methyl-accepting chemotaxis proteins (MCPs) (for reviews on chemotaxis, see references 77, 82, 838, and 850). Although the number of proteins involved in the chemotactic signaling pathway varies in different bacteria, the main mechanism follows similar principles (850). The chemotaxis pathway of enteric bacteria is displayed in Fig. 3. An incoming chemotactic signal is relayed to the flagellar motor via the CheA/CheY two-component system. The autophosphorylation activity of the ATP-dependent histidine kinase CheA is stimulated by its CheW-mediated recruitment to the MCPs. P~CheA transfers its phosphoryl group to an aspartyl residue of CheY, the response regulator. P~CheY binds to the flagellar motor protein FliM, and as a consequence, the tumbling frequency increases. To prevent the accumulation of P~CheY, the protein has autophosphatase activity, and in various bacteria (including E. coli), dephosphorylation is stimulated by a protein called CheZ. This signal transduction pathway is controlled by an adaptive feedback mechanism involving MCP methylation by the methyltransferase CheR, which stimulates CheA activity, and demethylation by the methylesterase CheB, which inhibits CheA activity. CheB is activated by phosphoryl transfer from P~CheA. Upon chemoreception, the MCP lowers the activity of CheA considerably, and as a result, the concentration of P~CheY decreases. The bacterium therefore tumbles less frequently and thus effectively moves up a concentration gradient of chemoattractant. Simultaneously, receptor demethylation diminishes due to lower levels of P~CheA/P~CheB, and CheA will be reactivated, leading finally to an adaptation of the system.


Figure 3
View larger version (43K):
[in this window]
[in a new window]

  		FIG. 3. Mechanism of chemotaxis in E. coli and role of the PTS in carbohydrate chemotaxis. In the absence of chemotactically active molecules (top left), CheA autophosphorylates at a histidyl residue. This reaction is stimulated by CheW, which recruits CheA to the MCP receptor protein. P~CheA transfers its phosphoryl group to CheY, and P~CheY binds the flagellar motor FliM, thereby evoking clockwise (CW) rotation of the flagella, which leads to tumbling of the bacterium. In the presence of chemotactically active molecules (top right), the autophosphorylation activity of CheA is inhibited, and as a result, the concentration of P~CheY drops. FliM molecules will no longer be complexed with CheY, which favors counterclockwise (CCW) rotation of the flagella. This results in smooth swimming towards increasing concentrations of the chemotactically active substance. The presence of an efficiently metabolizable PTS carbohydrate (or the absence of PEP) induces a similar response, as under these conditions, EI is present mainly in an unphosphorylated form. In fact, dephospho-EI inhibits the autophosphorylation of CheA (center) and therefore also favors smooth swimming. The sensitivity of the system towards the signal is regulated through methylation and demethylation of the MCPs (bottom), which are catalyzed by CheR and CheB, respectively. MCP methylation stimulates the autophosphorylation of CheA, and demethylation inhibits it.

E. coli shows chemotaxis towards many nutrients, including several PTS sugars (8, 462). By using a variety of null mutants, it was established that the chemotrophic response towards PTS sugars was elicited by their transport and not by their binding or metabolism (8, 297, 462, 641, 948). It was established that chemotaxis towards PTS sugars at relatively high concentrations (>1 mM) is also mediated by the chemoreceptors Aer and Tsr (296), which sense redox state and proton motive force, respectively (865). In E. coli, chemotaxis by PTS sugars can be abolished by knocking out CheA, CheY, CheW, EI, HPr, or the corresponding sugar-specific EII complexes, whereas the removal of the MCPs, CheR, or CheB has no significant effect (8, 297, 460, 462, 498, 550, 609, 641, 745, 863, 864; see also reference 678). Despite the fact that PTS-mediated sugar uptake and Che protein-mediated chemotaxis involve similar phosphoryl transfer reactions, no phosphoryl transfer between PTS components and Che proteins has been demonstrated (379). Rather, in vitro experiments established that unphosphorylated EI interacts with CheA and thereby inhibits the autokinase activity of CheA by up to 10-fold (497). In addition, CheA activity drops about threefold upon shifts of the PEP concentration in the physiologically relevant range (from 2 to 0.3 mM). Further evidence for an effect of the phosphorylation state of PTS proteins on CheA activity was provided by kinetic analyses of the chemotactic response in E. coli wild-type cells and several che null mutants (498). The response, which was induced by flash photolysis of caged D-glucose and {alpha}-MG, shows rapid kinetics and high signal sensitivity (Km for D-glucose is about 10 nM). Considering the rapidity and sensitivity of the response, those authors proposed that transport-induced dephosphorylation of P~EI affects CheA activity but has little influence on the PEP concentration (the intracellular concentration of PEP is about 200-fold higher than that of EI). This hypothesis holds only when phosphorylation of EI by PEP would be a slow process controlling glucose influx. However, because even at high glucose concentrations, nearly all control over the flux resides in the glucose transporter (901), it is highly unlikely that EI phosphorylation exerts control at low glucose concentrations. Rather, PTS model calculations predict that the extent of EI phosphorylation rapidly responds to changes in the PEP concentration (239, 240, 777). At high intracellular PEP concentrations, mainly P~EI and little EI have been calculated to be present. Thus, even a slight reduction in the PEP concentration (as anticipated for the chemotaxis experiments) (497) could cause a relatively significant increase of unphosphorylated EI, which would lower CheA activity. This could in turn slightly lower the amount of P~CheY and lead to a strong chemotactic response, as the binding of P~CheY to the flagellar motor is highly cooperative (133).

In contrast to enteric bacteria, chemotaxis towards PTS carbohydrates in B. subtilis is mediated by the methyl-accepting chemotaxis protein McpC, which is also a sensor for several amino acids (582). Deletion of mcpC abolishes chemotaxis towards the carbohydrates D-glucose, D-mannitol, D-fructose, trehalose, N-acetylglucosamine, and {alpha}-methyl-D-glucoside (261, 428). Nevertheless, chemotaxis in B. subtilis is affected by the PTS, and the cytoplasmic domain of McpC is thought to receive a transport-related signal from the PTS (428). As a consequence, EIIMtl, EIIFru, and EIITre null mutations resulted in a loss of chemotaxis towards the related carbohydrate. Deletion of ptsH also prevents B. subtilis from moving up a sugar gradient of mannitol and fructose (261). The response towards glucose is more complex and involves McpA (313). A ptsH or mcpA deletion strain shows chemotaxis towards glucose (which is taken up via a non-PTS transporter by the ptsH mutant), whereas a double mutant does not (261). Binding studies with purified CheA and EI indicate that P~EI, but not EI or PEP, binds CheA and thereby inhibits its autophosphorylation. In B. subtilis, P~CheY exerts an effect opposite that in E. coli (61, 62, 945) (i.e., P~CheY leads to smooth swimming), and therefore, the inhibition of CheA activity by phosphorylated EI in B. subtilis is consistent with CheA inhibition by unphosphorylated EI in E. coli. However, experimental findings with tethered cells raise doubts about the proposed mechanism. For instance, the addition and removal of mannitol had effects similar to those of the addition of glucose on the frequency of counterclockwise rotation in a ptsH mutant (261), although this mutant does not transport mannitol. It is not clear how the phosphorylation state of the PTS proteins could be affected by a sugar that is not transported and metabolized. In addition, conflicting results were obtained when the effect of glucose on the rotation of tethered cells was studied (261, 428).

Glycogen storage. Many bacteria, including E. coli and S. enterica serovar Typhimurium, store carbohydrates in the form of glycogen. They produce it from glucose-1-P by the sequential action of ADP-glucose pyrophosphorylase (GlgC) and glycogen synthase (GlcA) (see reference 683). When needed, glycogen is reconverted to glucose-1-phosphate by glycogen phosphorylase (GlgP) (962). The related E. coli genes are located in the glgCAP operon, which is positively regulated by Crp/cAMP (684) and negatively regulated by CsrA (38, 486, 962). B. subtilis possesses a glgBCDAP operon (glgB codes for a glycogen branching enzyme, and glgD codes for a second ADP glucose pyrophosphorylase), which is presumably negatively regulated by CcpA/P-Ser-HPr (178). Surface plasmon resonance-based ligand fishing revealed a strong interaction between E. coli glycogen phosphorylase (GlgP) and HPr; indeed, GlgP was the only protein "fished" out of a cell extract by immobilized HPr (799). The interaction between the two proteins was confirmed by mobility shift assays and sedimentation equilibrium centrifugation. Subsequent competition experiments established that the binding was highly specific. Neither B. subtilis and Mycobacterium capricolum HPrs nor E. coli NPr and diphosphoryl transfer proteins (FPr) bind to E. coli GlgP, and vice versa, E. coli maltodextrin phosphorylase, which is very similar to GlgP (352, 797), does not bind HPr (418, 799). The affinity of GlgP for P~His-HPr is four times higher than that for unphosphorylated HPr. Binding of HPr stimulates the formation of GlgP dimers and tetramers. Only the binding of HPr, but not of P~His-HPr, leads to a significant increase in GlgP activity (2.5-fold) (799). The concentration of HPr (531) is much higher than that of GlgP (116), and therefore, a major part of cellular GlgP will be complexed with HPr or P~His-HPr. As a result, GlgP activity will be determined by the phosphorylation state of HPr. Conversely, overproduction of GlgP should lead to the sequestration of most HPr and thereby to the inhibition of the PTS, as was indeed observed (418).

Based on these results, the following model for the regulation of glycogen breakdown was proposed for E. coli. Cells growing on glucose start to accumulate glycogen during the late exponential growth phase and the onset of stationary phase (418, 683, 684). Under these conditions, the concentration of phosphorylated PTS proteins is relatively high (797). cAMP levels will therefore rise (653), and expression of the glgCAP operon will be stimulated. The activity of GlgP is low, because mainly P~His-HPr is present, and glycogen will accumulate. Later, when growth of the cells has advanced far into stationary phase, the PEP concentration drops, PTS proteins will become dephosphorylated, and the cAMP concentration and expression of the glgCAP operon will decrease. However, GlgP is now activated by binding unphosphorylated HPr, and glycogen breakdown will therefore commence and dominate over glycogen synthesis.


arrow
REGULATION OF CARBON METABOLISM IN LOW-G+C GRAM-POSITIVE BACTERIA: REGULATORY FUNCTIONS OF P-Ser-HPr
 
HPr, the Central Processing Unit of Carbon Metabolism in Gram-Positive Bacteria

Most low-G+C gram-positive organisms do not possess adenylate cyclase, and the cAMP-dependent EIIAGlc-controlled CCR mechanism of gram-negative bacteria can therefore not be operative in these organisms. In fact, the general PTS protein HPr turned out to be the master regulator of carbon metabolism in gram-positive bacteria. Similar to EIIAGlc in gram-negative bacteria, HPr carries out its diverse regulatory functions in response to changes in its phosphorylation state. In low-G+C gram-positive bacteria, HPr becomes phosphorylated not only by PEP at His-15 but also by ATP at Ser-46. As a consequence, four different forms of HPr exist in these organisms: dephospho-HPr, HPr phosphorylated at either His-15 (P~His-HPr) or Ser-46 (P-Ser-HPr), and doubly phosphorylated HPr (571, 893). The rapid metabolism of various sugars affects the activities of a bifunctional protein kinase/P-protein phosphorylase, which responds to changes of the ATP, Pi, PPi, and FBP concentrations, and of EI, which responds to alterations of the PTS phosphotransfer activity and the PEP-to-pyruvate ratio (335). These two enzymes control the concentration of the various forms of HPr, which regulate carbon metabolism via protein-protein interactions (HPr and P-Ser-HPr) or the phosphorylation of non-PTS proteins (P~His-HPr) (Table 1). In gram-positive bacteria, HPr therefore functions as the "central processing unit" for carbon metabolism, as its phosphorylation state is determined by various signals (input), which in turn allow it to phosphorylate or to interact with numerous other proteins (output). In the following sections, we discuss the characteristics of the enzyme catalyzing the phosphorylation/dephosphorylation of HPr at Ser-46, the role of P-Ser-HPr in CCR and inducer exclusion, and the regulation of non-PTS proteins (antiterminators, transcription activators, carbohydrate transporters, and catabolic enzymes) via phosphorylation by P~His-HPr and/or P~EIIBs.

Characteristics of ATP-Dependent HPr Phosphorylation

In the following sections, we will describe the enzyme HPr kinase/phosphorylase (HprK/P), which catalyzes the ATP-dependent phosphorylation of HPr (and certain HPr paralogs) at Ser-46 as well as the dephosphorylation of P-Ser-HPr. We will include structural studies of HprK/P and the complexes with HPr and P-Ser-HPr and discuss the regulation of its two opposing activities by metabolites. The organization of hprK with specific genes, some of unknown function, in gram-positive bacteria will be discussed, and finally, results that suggest a function of P-Ser-HPr in the regulation of PTS activity by a feedback mechanism will be provided.

Phosphorylation of HPr at Ser-46. Isocitrate dehydrogenase from E. coli was the first bacterial protein shown to be phosphorylated by an ATP-dependent seryl protein kinase (257). In the early 1980s, when Deutscher, Reizer, and Saier investigated a possible implication of protein phosphorylation in inducer expulsion (185, 710, 712), a process where certain sugar phosphates that have accumulated in bacteria are dephosphorylated and expelled as soon as the cells are exposed to rapidly metabolizable carbon sources such as glucose or mannose, a second seryl-phosphorylated bacterial protein was detected in Streptococcus pyogenes and was identified as being HPr of the PTS (185). Unlike PEP-dependent EI-catalyzed phosphorylation at the catalytic His-15 of HPr, which occurs in both gram-negative and gram-positive organisms (Fig. 1) (262, 941), phosphorylation of HPr at a seryl residue, which requires a specific protein kinase and ATP (185) or GTP (709), was originally believed to be restricted to gram-positive organisms with low G+C content.

The site of ATP-dependent phosphorylation in HPr of Enterococcus faecalis was identified as being Ser-46 (183). An equivalent of Ser-46 is present in HPr of all low-G+C gram-positive bacteria, and the surrounding sequence is usually well conserved (Fig. 4). However, in the HPr of some organisms, the phosphorylatable serine is in a slightly different position (position 45 in HPr of Bacillus thuringiensis subsp. israelensis [400] and position 47 in HPr of Mycoplasma genitalium [713]). Replacement of Ser-46 in B. subtilis HPr with an alanine, tyrosine, or aspartate prevents the FBP-stimulated ATP-dependent phosphorylation of HPr (206, 722), whereas HPr carrying a threonine at position 46 is still slowly phosphorylated (722). Mutations at Ser-46 also affect the PEP-dependent, EI-catalyzed phosphorylation at His-15. In a mutant complementation assay carried out with rate-limiting amounts of B. subtilis HPr (20 µM), the phosphoryl carrier activity was lowered to 35, 30, and 20% when Ser46Ala, Ser46Tyr, or Ser46Thr HPr, respectively, was used in place of wild-type HPr (206) and was lowered to 10% when Ser46Asp HPr was used (722). Phosphorylation of HPr at Ser-46 lowered the rate of PEP-dependent phosphorylation by 2 orders of magnitude (177). In addition, the apparent Km value for the PEP-dependent phosphorylation of P-Ser-HPr was 15-fold higher than the apparent Km value determined with HPr (723). In E. coli HPr, Ser-46 is located within a hydrophobic patch implicated in the interaction with EI (259) and EIIAs (139, 691, 949). The fact that Ser-46 of HPr faces the well-conserved Glu-84 of EI in the HPr:EI complex (260) probably explains why HPrs of gram-positive bacteria carrying a negative charge at Ser-46, due to either phosphorylation or mutation, interact poorly with EI. Vice versa, phosphorylation of HPr by EI at His-15 slowed its phosphorylation at Ser-46 by the ATP-dependent HPr kinase (177). In the HPr:HPr kinase complex, helix {alpha}4 of HPr kinase interacts with the region around His-15 (228, 533). Phosphorylation at His-15 probably diminishes the affinity between the kinase and its protein substrate. Indeed, in two-hybrid experiments, an interaction of B. subtilis HPr kinase with Ser46Ala mutant HPr could be detected. This interaction was prevented when EI was additionally produced in the yeast cells, suggesting that EI efficiently competes with HprK/P for binding to HPr and/or that EI phosphorylates HPr at His-15, which lowers its affinity for HprK/P (P. Noirot, S. Poncet, and J. Deutscher, unpublished results).


Figure 4
View larger version (25K):
[in this window]
[in a new window]

  		FIG. 4. Alignment of the first 55 amino acids of HPr proteins from firmicutes (B.s., B. subtilis; E.f., E. faecalis; L.c., L. casei; S.s., S. salivarius; S.c., S. carnosus), from a spirochete (T.p., T. pallidum), from proteobacteria with a Ser-46 region strongly resembling the corresponding sequence in HPr of firmicutes (X.f., Xylella fastidiosa; N.m., N. meningitidis), and from other gram-negative bacteria (E.c., E. coli; H.i., H. influenzae; V.c., V. cholerae) (V. cholerae possesses a second HPr in which the region around Ser-46 more strongly resembles the corresponding region in the HPr of firmicutes). The arrows indicate the amino acids His-15 and Ser-46. In HPrs from organisms, in which these residues are phosphorylated, they are shown in boldface type. Conserved regions around the phosphorylation sites are boxed.

Gram-negative enteric bacteria do not possess HPr kinase, although their HPrs usually contain Ser-46 and exhibit about 45% overall sequence identity to HPrs from gram-positive organisms. Nevertheless, the region around Ser-46 is quite different. This difference is probably responsible for the failure of HPr kinase from gram-positive bacteria to phosphorylate E. coli HPr (177, 722). HPrs of Mycoplasma capricolum, in which amino acids next to Ser-46 were replaced with the corresponding amino acids of E. coli HPr, were poorly phosphorylated with ATP (991). Based on the solution and crystal structures of HPrs from E. coli, E. faecalis, and B. subtilis, the presence of a lysine in position 49 of E. coli HPr, which is a glycine in HPrs of most gram-positive bacteria (Fig. 4), was assumed to represent a major obstacle for the interaction of HPr kinase with HPr from E. coli and presumably with HPrs from other gram-negative bacteria (374).

Interestingly, several gram-negative nonenteric organisms possess an HPr kinase, and the sequence around Ser-46 of their HPrs resembles that of HPr from low-G+C gram-positive bacteria (64) (Fig. 4). In contrast, Streptomyces coelicolor, a gram-positive organism with high G+C content in its DNA, does not possess a protein exhibiting significant sequence similarity to HPr kinase (55), and the sequence around Ser-47 of its HPr barely resembles the sequence around Ser-46 in HPrs of low-G+C gram-positive bacteria. Nevertheless, HPr from S. coelicolor can be phosphorylated by the HPr kinase from B. subtilis (97, 630), suggesting that despite the sequence differences, the structure around Ser-47 sufficiently resembles the structure around Ser-46 in B. subtilis HPr to allow its phosphorylation.

HPr kinase also dephosphorylates P-Ser-HPr by producing PPi. A soluble enzyme that specifically dephosphorylated P-Ser-HPr was purified to near homogeneity from E. faecalis (181). It had a molecular mass of 30 kDa (on sodium dodecyl sulfate-polyacrylamide gels), but it escaped the authors' attention that the purified enzyme must have also exhibited HPr kinase activity. In fact, it was later discovered that similar to the first described bacterial serine protein kinase, isocitrate dehydrogenase kinase/phosphatase (450), HPr kinase is also bifunctional and possesses P-Ser-HPr dephosphorylation activity (423). HPr kinase of E. faecalis has a molecular mass of 33 kDa, which is close to the molecular mass detected for the major protein in the P-Ser-HPr phosphatase preparation. Inorganic phosphate (Pi) was reported to stimulate P-Ser-HPr dephosphorylation (181). This was surprising, as Pi was assumed to be one of the products of P-Ser-HPr dephosphorylation and was therefore rather expected to inhibit this reaction. The paradox was resolved by demonstrating that P-Ser-HPr dephosphorylation is not a hydrolysis reaction. Dephosphorylation of either 32P-Ser-HPr in the presence of Pi or P-Ser-HPr in the presence of 32Pi leads to the formation of radioactive 32PPi. Pi thus functions as a substrate, and lowering the Pi concentration below the µM range almost completely prevents the dephosphorylation of P-Ser-HPr (557). In analogy to the usual phosphohydrolysis reaction, this novel protein dephosphorylation mechanism was called phospho-phosphorolysis, and the responsible enzyme was termed HPr kinase/P-Ser-HPr phosphorylase (HprK/P). Because P-Ser-HPr dephosphorylation is reversible, PPi can replace ATP as the phosphoryl donor for HPr phosphorylation. The kinetic parameters indicate that PPi-dependent HPr phosphorylation is actually the energetically favored reaction (557). To allow the efficient dephosphorylation of P-Ser-HPr in the cell, the resulting PPi therefore needs to be hydrolyzed. In B. subtilis and other bacilli, PPi hydrolysis seems to be achieved by YvoE, which exhibits pyrophosphatase activity (557). The yvoE gene is located in the hprK operon of bacilli (Fig. 5), and since YvoE exhibits sequence similarity to P-glycolate phosphatases and stimulates P-Ser-HPr dephosphorylation, it was mistakenly assumed to be P-Ser-HPr phosphatase (251). It is likely that organisms missing a yvoE gene in the hprK operon use one of usually several YvoE homologs present in bacteria to hydrolyze PPi formed during P-Ser-HPr dephosphorylation.


Figure 5
View larger version (27K):
[in this window]
[in a new window]

  		FIG. 5. The gene context of hprK in bacteria of the phylum Firmicutes. The hprK gene is followed by lgt in all sequenced genomes of the firmicutes except in L. mesenteroides, Oenococcus oeni, and some clostridiae. In addition, there are other genes associated with hprK that appear to be conserved. They include two genes without a known function (in L. lactis, E. faecalis, and several streptococci [S. agalactiae, S. mutans, S. pneumoniae, S. pyogenes, S. suis, S. thermophilus, and S. uberis]) and a glycerol-3-phosphate dehydrogenase gene and a UTP-glucose-1-phosphate uridylyltransferase gene (in E. faecalis) followed by a thioredoxin reductase gene (in L. acidophilus, L. johnsonii, L. gasseri [these species lack the uridylyltransferase], L. brevis, L. plantarum, P. pentosaceus, L. mesenteroides, and O. oeni [the latter two species lack the lgt gene]). Staphylococci contain YvoF, a protein with a hexapeptide transferase motif (S. aureus, S. epidermidis, S. haemolyticus, and S. saprophyticus). YvoF, together with YvoE (a pyrophosphatase), is also present in B. anthracis, B. cereus, B. thuringiensis, Exiguobacterium species, L. innocua, and L. monocytogenes, while YvoD, YvoE, and YvoF are found in the bacilli B. clausii, B. halodurans, B. licheniformis, B. stearothermophilus, B. subtilis, G. kaustophilus, and O. iheyensis.

Cold-chase ATP labeling of HPr in Listeria monocytogenes crude extracts suggested that HprK/P of this organism cannot dephosphorylate P-Ser-HPr (129). However, later experiments with purified HprK/P established that L. monocytogenes HprK/P functions as P-Ser-HPr phosphorylase, similar to the B. subtilis enzyme (326).

Interestingly, although Mycoplasma pneumoniae hprK mutant cells cannot form P-Ser-HPr, crude extracts prepared from this mutant were still able to dephosphorylate P-Ser-HPr. The dephosphorylation activity could be attributed to a PP2C-type P-protein phosphatase, which, in accordance with its B. subtilis homolog, was called PrpC (306). M. pneumoniae HprK/P is unusual, as it can phosphorylate HPr at relatively low ATP concentrations (831). In addition, in vivo P-Ser-HPr formation is stimulated by the presence of glycerol in the growth medium (307). Although the M. pneumoniae enzyme was reported to dephosphorylate P-Ser-HPr in vitro in a Pi-dependent manner, similar to HprK/Ps from other organisms (831), the P-Ser-HPr dephosphorylation activity exhibited by PrpC seems to be of physiological relevance. While an M. pneumoniae wild-type strain grown in the presence of glucose and glycerol contained only low amounts of P-Ser-HPr and doubly phosphorylated HPr, more than half of the HPr was found to be present as P-Ser-HPr and doubly phosphorylated HPr in the prpC mutant (306). Because a B. subtilis strain synthesizing Val267Phe HprK/P, which still functions as a kinase but has lost its phosphorylase activity (see "P-Ser-HPr Regulates PTS Transport Activity by a Feedback Mechanism"), contains more than 95% of its HPr as P-Ser-HPr (571), it is unlikely that PrpC of this organism plays a major role in P-Ser-HPr dephosphorylation. In B. subtilis, PrpC and the corresponding protein kinase, PrkC, were reported to play important roles in stationary-phase cells and to control the phosphorylation state of the translation elongation factor EF-G (248). It is interesting that HprK/Ps of certain gram-negative bacteria such as Neisseria meningitidis (S. Poncet, M.-K. Taha, M. Larribe, and J. Deutscher, unpublished results) and Brucella melitensis (S. Poncet, M. Dozot, X. de Bolle, J. J. Letesson, and J. Deutscher, unpublished results) exhibit no or very low P-Ser-HPr dephosphorylation activity, and although these organisms do not contain a homolog of PrpC, it is possible that they possess another P-Ser-HPr phosphatase.

Structure determination of HprK/P. The crystal structures of truncated HprK/P from L. casei (missing the N-terminal 120 amino acids) (229) and of full-length HprK/Ps from Staphylococcus xylosus (515) and M. pneumoniae (13) have been determined. The crystallized proteins form hexamers composed of two layers of trimers. Size exclusion chromatography and equilibrium sedimentation showed that L. casei HprK/P also forms hexamers in solution (229). B. subtilis HprK/P was reported to form hexamers at a neutral pH but was reported to form monomers and dimers at pH 9.5 (697). HprK/Ps do not exhibit similarity to eukaryotic protein kinases (312, 866) or P-protein phosphatases (43, 805) but resemble PEP carboxykinase and nucleosidediphosphate kinases (229, 252, 748). A P loop (or Walker motif A [GXXGXGKS]) that is usually located between amino acids 150 and 170 serves as a nucleotide binding site. The crystal structures of L. casei and S. xylosus HprK/Ps revealed that Pi binds to the same position in Walker motif A as the ß-phosphate of the nucleotide. This explains the inhibitory effect of Pi on the kinase activity, as ATP and Pi compete for the same binding site. Vice versa, ATP inhibits the phosphorylase activity of HprK/P (181). The inhibitory effect of ATP was stronger at low Mg2+ concentrations. Mutant studies with hprK from various organisms suggested that Walker motif A is important not only for the kinase but also for the phosphorylase function (315, 571, 831).

The dual role of Walker motif A was confirmed by solving the crystal structures of HprK/Ps complexed with HPr or P-Ser-HPr, which also allowed the prediction of a detailed mechanism for HPr phosphorylation and P-Ser-HPr dephosphorylation (228). His-140 of L. casei HprK/P forms a hydrogen bond to Asp-179, which in turn acts as a base during HPr phosphorylation by forming a hydrogen bond to the hydroxyl group of Ser-46 in HPr. The structure of the HprK/P:P-Ser-HPr complex, which was obtained by cocrystallizing HPr, HprK/P, and PPi, allowed the identification of the amino acids cooperating in the nucleophilic attack of Pi on the P-Ser bond and revealed that the binding of HPr and P-Ser-HPr to HprK/P involves nearly identical interfaces. Compared to the HprK/P:HPr complex, an additional contact is made between the phosphoryl group in P-Ser-HPr and Arg-245 of a neighboring HprK/P subunit (228). The loop containing Arg-245 is too flexible to be traced in the HprK/P and HprK/P:HPr structures, but it is stabilized in the HprK/P:P-Ser-HPr complex due to the interaction of Arg-245 with the phosphoryl group (228, 229). During P-Ser-HPr dephosphorylation, Asp-179 (present in protonated form) and His-140 (donates a proton to Ser-46 of HPr) (228) play an inverse role compared to the kinase reaction. The interaction site of S. aureus HPr with S. xylosus HprK/P was also determined by NMR, and the HPr interface appears to be similar in solution and in the crystal (533).

The crystal structure of HprK/P from S. xylosus revealed that a Pi ion binds not only to Walker motif A but also to the N-terminal domain (515). It was fixed by three conserved arginines located in two different subunits, one in the upper and one in the lower HprK/P trimer. The second Pi therefore bridges the two trimers (515), but the Pi bridge is not essential for hexamer formation, as truncated L. casei HprK/P (229) and M. pneumoniae HprK/P without bound Pi (13) also form hexamers. The function of the N-terminal domain, which protrudes from the central part like a propeller, is unknown. Truncated L. casei HprK/P that is missing this domain is active as a kinase and phosphorylase and responds to all known effectors (229). As will be discussed below, the N-terminal domain is naturally truncated in HprK/Ps of {alpha}-proteobacteria. Surprisingly, the N-terminal domain of HprK/P exhibits a fold similar to that of the N-terminal part of MurE (13), an enzyme implicated in the formation of cytoplasmic precursors for cell wall synthesis (284).

The structure of HprK/P with bound ATP or GTP has not yet been solved. If the nucleotide would occupy the same position as that in the closely related adenylate kinase, it would clash with the central K3 loop of a neighboring HprK/P subunit. Structural changes of ATP and/or HprK/P must therefore accompany nucleotide binding (13, 229, 515). Likewise, structural changes of HprK/P might be responsible for the observed cooperative binding of ATP (372).

Organization of the hprK operon. About 15 years after the discovery of the ATP-dependent phosphorylation of HPr at Ser-46 (185), at almost the same time, three laboratories succeeded in identifying the HprK/P-encoding hprK gene (named ptsK in one publication) from three different organisms. In the first report, purification of E. faecalis HprK/P and sequencing of its N terminus and internal peptides allowed the amplification of the 5' part of the hprK gene by PCR. This information was subsequently used to identify hprK of B. subtilis within the sequenced genome (251) and to obtain the complete hprK gene of E. faecalis (423). In the second report, purification and N-terminal sequencing of B. subtilis HprK/P was used to directly identify hprK of B. subtilis (709). Finally, purification of HprK/P and probing of a DNA library with a degenerate oligonucleotide derived from the N-terminal amino acid sequence allowed the identification of hprK from Streptococcus salivarius (84). More recently, the hprK genes from other low-G+C gram-positive bacteria including S. xylosus (357), L. casei (198), M. pneumoniae (830), and G. stearothermophilus (127) have been cloned and sequenced, and the corresponding HprK/Ps were purified and characterized. Genome sequencing revealed the presence of hprK in most low-G+C gram-positive organisms.

In the firmicutes, hprK is usually the first gene within an operon (Fig. 5), which is composed of five genes in B. subtilis. In most gram-positive bacteria, the second gene is the prolipoprotein diacylglyceryl transferase-encoding lgt gene, which carries out the diacylglyceryl lipidation of lipoproteins at a cysteyl residue. The frequent association of hprK with lgt might indicate that the functions of the two enzymes are somehow related, although no experimental evidence for this assumption has been obtained so far. A possible link between the PTS and Lgt is further suggested by the observation that in Enterobacteriaceae (E. coli, S. enterica serovar Typhimurium, Yersinia pestis, etc.), lgt is preceded by ptsP (T. Doan, personal communication), which encodes an EI with an N-terminal extension resembling the GAF domain in NifA. The hprK operon of G. stearothermophilus, Oceanobacillus iheyensis, L. monocytogenes, Listeria innocua, and the seven bacilli, whose genomes have been sequenced, contains the pyrophosphatase-encoding yvoE gene followed by yvoF downstream of lgt. B. subtilis, Bacillus clausii, Bacillus halodurans, Bacillus licheniformis, and G. stearothermophilus contain an additional gene, yvoD, inserted between lgt and yvoE. In S. aureus and Staphylococcus epidermidis, yvoE is missing, and yvoF directly follows lgt. The yvoDEF genes are absent from most other low-G+C gram-positive organisms. O. iheyensis possesses two genes encoding HprK/Ps, with one followed by lgt. It is not known whether the HprK/P ortholog encoded by the other hprK gene can phosphorylate HPr. In the genus Clostridium, hprK and lgt are located in different regions of the genome. In Clostridium acetobutylicum, hprK is preceded by a gene encoding a protein with significant sequence similarity to the glycerol-inducible E. coli GlpX, which resembles fructose-1,6-bisphosphatases (859).

Quite unique is a presumed hprK gene in Fusobacterium nucleatum ATCC 25586 and F. nucleatum subsp. vincentii, which encodes two complete HprK/P proteins fused together in tandem. The two HprK/P domains exhibit 22% sequence identity to each other and about 30% identity to HprK/Ps from gram-positive bacteria. However, a conserved Walker motif A is present only in the C-terminal domain. The corresponding sequence in the N-terminal domain is strongly altered and does not resemble Walker motif A. A prerequisite for HPr phosphorylation is the deprotonation of the hydroxyl group of Ser-46 of HPr by the concerted action of conserved histidyl and aspartyl residues (His-140 and Asp-179 in L. casei HprK/P). In F. nucleatum HprK/P, these residues are present in the C-terminal domain but are replaced with Glu and Lys, respectively, in the N-terminal domain. In contrast, Arg-245, which interacts with the phosphoryl group of P-Ser-HPr and probably plays a major role in P-Ser-HPr dephosphorylation, is present only in the N-terminal domain. It is therefore tempting to assume that in this duplicated enzyme, the two activities have been separated and that the C-terminal domain carries the kinase activity, whereas the N-terminal domain might function as P-Ser-HPr phosphorylase. The two sequenced F. nucleatum strains also contain HPr (FN1782) with conserved His-15 and Ser-46 regions and EI (FN1793), but although the genes are located close to each other on the chromosome, they do not seem to be organized in an operon.

The organization of the hprK gene in certain gram-negative bacteria will be discussed below in more detail.

Metabolites regulate the antagonistic activities of HprK/P. In vitro experiments suggested that the two antagonistic activities associated with HprK/P are regulated in the cell in response to changes in the concentrations of FBP, ATP, and Pi (177). FBP clearly stimulates the kinase activity of B. subtilis HprK/P (372), whereas its effect is less pronounced with the L. casei enzyme (198). The effect of FBP is usually stronger when low, nonphysiological concentrations of ATP are used. In addition, the stimulatory effect of FBP on HPr phosphorylation was more evident when the experiments were carried out in the presence of a few mM Pi (177, 198, 423), which lowers the HPr kinase activity (84, 181, 251). The main physiological function of FBP might therefore be to prevent the inhibition of HprK/P by Pi. For the S. xylosus enzyme, activation by FBP was observed only when HprK/P was present at low concentrations (357). Surprisingly, the Streptococcus salivarius enzyme was not at all stimulated by FBP (84). Similarly, the M. pneumoniae enzyme is fully functional at relatively low ATP concentrations, and its activity is also not stimulated by FBP (831). The intracellular concentrations of FBP, ATP, and Pi vary largely depending on whether the cells utilize a favored carbon source or not. During the uptake of a rapidly metabolizable carbon source such as glucose, fructose, or mannose by L. lactis, the concentrations of Pi, ATP, and FBP are about 5, 8, and 30 mM, respectively (528, 598, 873). Under these conditions, HprK/P functions as an HPr kinase and not as P-Ser-HPr phosphorylase. In contrast, in the absence of a rapidly metabolizable carbon source, the concentration of Pi increases to about 50 mM, whereas the concentrations of FBP and ATP drop to 1 to 3 mM (528, 598, 873). The effect of growth on a preferred or less favorable carbon source on P-Ser-HPr formation was confirmed with S. salivarius cells. By carrying out crossed immunoelectrophoresis experiments, the fractions of the various forms of intracellular HPr in S. salivarius cells grown in glucose-containing medium were found to be 45% for P-Ser-HPr and 50% for doubly phosphorylated HPr, whereas only little P~His-HPr (5%) and almost no dephospho-HPr could be detected (263). Cells grown on galactose, a less efficiently metabolized carbon source, also exhibited a high amount of doubly phosphorylated HPr (44%) and little P~His-HPr (3%), whereas P-Ser-HPr was lowered to 18%, and dephospho-HPr increased to 35%. A similar observation was made for Streptococcus bovis, where an inverse correlation between P-Ser-HPr formation and the Pi concentration was observed. Cells growing on glucose contained more than half of the HPr as P-Ser-HPr, and the Pi concentration was in the low mM range. When glucose was exhausted, P-Ser-HPr disappeared, and the Pi concentration went up to 30 mM (26). In in vitro experiments, 20 mM Pi completely prevents the phosphorylation of HPr by 5 mM ATP, even when 25 mM FBP is present (177).

The observation that PPi is formed during P-Ser-HPr dephosphorylation and used as a phosphoryl donor by HprK/P suggested that the PPi concentration affects the intracellular amount of P-Ser-HPr. B. subtilis cells grown in the presence of glucose contain 10-fold more FBP than cells grown on succinate (from 1.4 to 14 mM). A similar increase was observed for PPi (from 1.2 to 6 mM) (557). This metabolite distribution is probably the reason why about 65% of the HPr in glucose-grown B. subtilis cells is present as P-Ser-HPr, 35% is present as HPr, and little is present as P~His-HPr and doubly phosphorylated HPr (571). In contrast, in succinate-grown cells, mainly dephospho-HPr was present, together with a few percent of P~His-HPr. In in vitro experiments, HPr phosphorylation with B. subtilis HprK/P was maximal above 5 mM PPi but negligible below 1 mM PPi. The approximately 6 mM PPi present in cells growing on glucose is therefore expected to be partly responsible for the increase of P-Ser-HPr, whereas in the absence of a rapidly metabolizable carbon source, the low PPi and high Pi concentrations probably allow efficient P-Ser-HPr dephosphorylation. ATP-dependent HPr phosphorylation by B. subtilis HprK/P was maximal in the presence of 10 mM FBP, whereas it almost disappeared below 2 mM FBP (372). On the contrary, FBP was not required for efficient PPi-dependent HPr phosphorylation (557).

An endogenous low-molecular-weight HprK/P inhibitor was detected when soluble extracts of Lactobacillus brevis were separated by size exclusion chromatography (715). However, no further details about the nature of the bacterial HPr kinase inhibitor were reported. Because the HPr kinase activity was detected by autoradiography after phosphorylation with [{gamma}-32P]ATP, it is possible that the reported HPr kinase inhibitor was PPi. If the PPi concentration in the assay mixtures exceeded the employed [{gamma}-32P]ATP concentration (10 µM) by severalfold, primarily nonradioactive (and therefore undetected) P-Ser-HPr would have been formed. Ironically, in the same publication, it was reported that the presence of PPi can completely inhibit [{gamma}-32P]ATP-dependent HPr phosphorylation (715). Inhibition of [{gamma}-32P]ATP-dependent HPr phosphorylation by PPi was also reported for HprK/P from Treponema denticola and was suggested to be probably due to the competition of the two phosphoryl donors for binding to HprK/P, as P-Ser-HPr was again detected by autoradiography (278). A synthetic inhibitor of HprK/P has also been identified. It is a heterocyclic bis-cationic compound, which was reported not to bind to the Walker motif (696).

In conclusion, the two opposing activities of HprK/P and consequently the ratio of the various HPr forms seem to be regulated mainly via metabolites. In agreement with this assumption, the amount of HprK/P present in B. subtilis cells was reported to have no influence on the ratio of the different HPr forms (59). The amount of HprK/P probably determines only how fast cells can adapt to changes in carbohydrate availability but apparently has no influence on the equilibrium reached under a certain condition.

Interestingly, in oral streptococci (glucose or lactose grown) (660, 893) and S. thermophilus (lactose grown) (134), a major part (up to 75%) of the HPr is converted into doubly phosphorylated HPr. A possible role of (P~His,P-Ser)-HPr in the regulation of the lactose transporter LacS of S. thermophilus will be discussed below (see "Regulation of EIIA-Containing Non-PTS Transporters by P~His-HPr-Mediated Phosphorylation").

P-Ser-HPr Regulates PTS Transport Activity by a Feedback Mechanism

P-Ser-HPr is a poor substrate for PEP-dependent phosphorylation by EI, although phosphorylation at Ser-46 causes only minor changes in the HPr structure (29, 952). One of these changes leads to an extension of the short ß-helix, which is capped by Ser-46 at the N-terminal site. This alteration was assumed to affect the interaction of Met-48 with a hydrophobic pocket in EI, thus partly explaining why phosphorylation at Ser-46 drastically slows the EI-catalyzed phosphorylation at His-15. However, the main reason for the poor phosphorylation of His-15 in P-Ser-HPr is probably an electrostatic repulsion between the phosphoryl group at Ser-46 and residue Glu-84 of EI (260). As discussed in the preceding section, in glucose-grown cells, a major fraction of HPr is present as P-Ser-HPr or doubly phosphorylated HPr, which, compared to P~His-HPr, has a much lower affinity for EIIAs (723), but there is little P~His-HPr.

If P~His-HPr is present at low concentrations, it is expected to control the PTS sugar uptake rate. The efficient phosphorylation of HPr at Ser-46 during growth on glucose might thus reduce PTS-catalyzed sugar uptake. A ptsH1 mutant (Ser46Ala replacement in HPr) was therefore expected to show deregulated PTS transport activity. However, a B. subtilis ptsH1 mutant did not exhibit elevated uptake of glucose or other rapidly metabolizable PTS sugars (184), indicating either that additional regulatory mechanisms are operative or that in B. subtilis, the small amount of P~His-HPr detected in cells growing on glucose is sufficient to fully catalyze sugar transport via the PTS. In contrast, an S. xylosus hprK mutant, also unable to phosphorylate HPr at Ser-46, exhibited reduced growth in the presence of glucose, although glucose-grown hprK mutant cells showed a two- to fourfold-elevated initial glucose transport rate compared to the wild-type strain. It was therefore concluded that the growth deficiency of the S. xylosus hprK mutant is probably due to uncontrolled glucose uptake leading to the accumulation of deleterious amounts of metabolites (357).

In various gram-positive bacteria, the presence of glucose exerts an inhibitory effect on the uptake of other PTS sugars. This inhibitory effect could be due partly to the low amounts of P~His-HPr detected in glucose-grown cells, which probably aggravates the competition between the EIIAs for their common phosphoryl donor. Competition was expected to be less severe in a ptsH1 mutant that is unable to form P-Ser-HPr. The effect of the ptsH1 mutation on the simultaneous uptake of the two PTS substrates mannitol and glucose was measured with B. subtilis cells. If phosphorylation at Ser-46 of HPr and competition for the remaining P~His-HPr were involved in glucose-mediated inhibition of mannitol uptake, elevated mannitol transport should occur in the ptsH1 mutant. Indeed, mannitol uptake in the presence of glucose was two- to fivefold higher in the ptsH1 mutant than in the wild-type strain (184, 974).

Evidence for a participation of P-Ser-HPr in the regulation of PTS transport activity was also provided by experiments with L. lactis vesicles. During the preparation of L. lactis vesicles, larger proteins such as HprK/P and EI remain entrapped in the vesicles, whereas almost all HPr is lost (970). An influence of the ATP-dependent phosphorylation of HPr at Ser-46 on PTS-catalyzed sugar transport could be demonstrated by 2DG uptake measurements using L. lactis vesicles into which B. subtilis HPr and various glycolytic intermediates had been electroporated (971). 2DG is transported via a PTS, and electroporation of HPr or coelectroporation of Ser46Ala mutant HPr with ATP and FBP into the vesicles stimulated 2DG uptake. In contrast, coelectroporation of wild-type HPr with ATP and FBP had no stimulatory effect. In addition, only slow 2DG uptake was detected with the P-Ser-HPr-resembling Ser46Asp mutant HPr (952), irrespective of whether it was electroporated alone or together with ATP and FBP. Similar results were obtained when the uptake of methyl-ß-D-thiogalactoside (TMG) by L. lactis vesicles was studied (970). In addition to FBP, several other metabolites including gluconate-6-P, glycerate-2-P, and fructose-6-P were reported to inhibit TMG uptake when electroporated into L. lactis vesicles together with B. subtilis HPr. Moreover, when wild-type HPr was electroporated into L. lactis vesicles, the uptake of fructose was inhibited by the presence of glucose, while only slight inhibition occurred when Ser46Ala HPr was electroporated in place of HPr (974).

Although providing no direct proof, these in vitro results suggested that the formation of P-Ser-HPr during the utilization of a rapidly metabolizable carbon source can lower the amount of P~His-HPr to such an extent that it is insufficient to sustain the simultaneous uptake of two PTS carbohydrates. This hypothesis was supported by in vivo experiments using the hprK(Val265Phe) allele. The Val265Phe replacement causes an almost complete loss of the P-Ser-HPr phosphorylase activity but has little effect on the FBP-activated HPr kinase function (571). When grown in the absence of a rapidly metabolizable carbohydrate, a B. subtilis wild-type strain contains little P-Ser-HPr, whereas in strains carrying the Val265Phe hprK allele, almost all HPr is accumulated as P-Ser-HPr. The Val265Phe hprK strain grows barely or not at all on media containing the PTS sugar glucose, fructose, maltose, or mannitol as the sole carbon source. Similar results were obtained with a B. subtilis strain in which the hprK gene had been replaced with the L. casei Val267Phe hprK allele (571). A strain carrying the Val267Phe hprK allele transports glucose and mannitol at rate that is more than 10 fold lower than that of an integrant expressing wild-type L. casei hprK (571). The inability of B. subtilis strains expressing the B. subtilis Val265Phe or the L. casei Val267Phe hprK allele to grow on PTS sugars is due to strongly reduced PTS transport activity and, as will be explained below, to a kind of "permanent CCR." Expression of the EIICBAGlc-encoding ptsG, which is not submitted to CCR, was not repressed by the presence of the L. casei Val267Phe hprK allele. The assumption that the accumulation of P-Ser-HPr in the Val267Phe hprK strain results in slow PTS phosphoryl group transfer, which in turn prevents PTS sugar uptake, was confirmed by introducing a ptsH1 mutation in the Val267Phe hprK strain (no P-Ser-HPr formation). It restored growth on PTS sugars, whereas introducing a ccpA mutation, which prevents CCR, had only a slight effect on PTS transport (571).

As mentioned above, glucose-grown S. salivarius (263, 893) and B. subtilis (571) cells contain very little P~His-HPr. Nevertheless, the small fraction present as P~His-HPr (about 5%) seems to be sufficient to fully catalyze PTS-mediated uptake of glucose and probably other rapidly metabolizable PTS sugars in gram-positive bacteria. Similarly, less than 20% of the HPr present in wild-type S. enterica serovar Typhimurium cells can fully sustain PTS-catalyzed glucose uptake (901). However, P~His-HPr probably starts exerting flux control when two or more PTS sugars are simultaneously transported or when the natural balance of the different forms of HPr is disturbed by, for example, the conversion of an additional fraction of HPr to P-Ser-HPr, as is the case in B. subtilis strains expressing the Val265Phe hprK allele (571).

Role of P-Ser-HPr in CCR

The following sections will deal with the role of P-Ser-HPr and P-Ser-Crh in CCR/carbon catabolite activation (CCA). We will discuss some early results that suggested a connection between CCR and P-Ser-HPr formation and the first confirmation of this connection by construction of a B. subtilis ptsH(Ser46Ala) mutant. The ptsH(Ser46Ala) mutation has a pleiotropic effect and leads to a relief from CCR for numerous genes and operons. We will mention the interaction of P-Ser-HPr and P-Ser-Crh with the catabolite repressor/activator CcpA (Table 1) and how these interactions allow CcpA, a LacI/GalR-type DNA binding protein, to interact with the catabolite response elements (cre's), operator sites which precede genes that are sensitive to CCR/CCA. The occurrence of Crh, an HPr paralog in which His-15 is replaced with a Gln, in bacilli, oceanobacilli, and geobacilli will also be discussed.

Loss of ATP-dependent HPr phosphorylation prevents CCR. Heterofermentative lactobacilli such as L. brevis and Lactobacillus buchneri were reported to contain HPr and HprK/P but to be devoid of EI and EIIs (715). Owing to the presumed absence of a functional PTS, there seemed to be no obvious regulatory role for the ATP-dependent phosphorylation of HPr at Ser-46 in these organisms. Although the report about the absence of EI and EIIs from heterofermentative lactobacilli later turned out to be wrong (an L. brevis ptsHI operon has been cloned and sequenced) (193) and the presence of a fructose-specific PTS induced under anaerobic conditions has been established (772), it initiated the search for additional functions of P-Ser-HPr, one of them being the regulation of CCR.

A connection between the phosphorylation of HPr at Ser-46 and CCR was suggested by the observation that CCR in B. subtilis requires the metabolism of the repressing sugar (490, 606). In gram-negative bacteria, the transport of a carbohydrate via the PTS is sufficient to elicit CCR. By studying CCR in mutants defective at various steps of glycolysis, the number of glycolytic intermediates potentially serving as a signal for CCR in B. subtilis could be narrowed to four candidates: dihydroxyacetone phosphate, glyceraldehyde-3-phosphate, 1,3-diphosphoglycerate, and FBP (606). As FBP, the main activator of HPr kinase, was among these metabolites, ATP-dependent phosphorylation of HPr was suspected to play a role in CCR (184).

The ptsH1 mutation has a pleiotropic effect on CCR and carbon catabolite activation. To test the proposed hypothesis that P-Ser-HPr plays a role in CCR, Deutscher and coworkers studied CCR in a B. subtilis ptsH1 mutant (Ser46Ala replacement). Indeed, several enzymes, including gluconate kinase, mannitol-1-P dehydrogenase, inositol dehydrogenase, and glucitol dehydrogenase, which are all sensitive to CCR in a wild-type strain, were not or only partly repressed by glucose and other rapidly metabolizable sugars in the ptsH1 mutant (184). Moreover, while the expression of a translational fusion (integrated at the amyE locus) of the promoter region and the 5' part of gntK (encodes gluconate kinase) to lacZ was sensitive to CCR in a wild-type strain, it was not repressed by glucose in the ptsH1 mutant (184). In addition, if a B. subtilis strain carrying a deletion of the ptsGHI genes, which encode EIICBAGlc, HPr, and EI, respectively, was transformed with a plasmid that allowed the production of the P-Ser-HPr-resembling Ser46Asp HPr (952), partial repression of the gnt operon was observed even in the absence of a repressing sugar (184). These results clearly established the crucial role of the phosphorylation of HPr at Ser-46 in CCR of B. subtilis.

The ptsH1 mutation turned out to be an important tool to study the implication of P-Ser-HPr in CCR in B. subtilis in more detail (184, 206). All B. subtilis ptsH1 mutants used in various laboratories are derivatives of ptsH1 strain SA003. They were obtained by either introducing additional mutations into SA003 or transforming specific mutants with chromosomal DNA of ptsH1 strains carrying an antibiotic resistance cassette inserted about 6 kb downstream from ptsH (184). The ptsH-linked antibiotic resistance cassette was recovered from strain MO481 (P. Stragier, unpublished results). The use of the ptsH1 mutant demonstrated that many other B. subtilis genes and operons are partly or completely relieved from CCR, such as the acetyl coenzyme A (acetyl-CoA) synthetase-encoding acsA gene (984) as well as the lev (523, 526), xyl (148), xyn (249, 250), and glp (158) operons (for a review, see reference 178). In a few cases, the effect of the ptsH1 mutation on CCR depends on the growth conditions. CCR of the B. subtilis hut operon, for example, is not altered in a ptsH1 mutant grown in minimal medium, whereas CCR in this strain completely disappears when it is grown in complex medium (985). A similar but less pronounced growth medium-dependent effect on CCR was observed with the iol operon (184). The mechanisms underlying the growth-dependent differences in CCR in ptsH1 mutants are not understood but might be related to the different nitrogen sources present in the media (220).

Synthesis of the plasmid-encoded B. thuringiensis larvicidal protoxin Cry4A by stationary-phase cells is also sensitive to CCR. The repressive effect of glucose on Cry4A production disappeared in a ptsH1 mutant (Ser45Ala replacement in B. thuringiensis HPr) (399). In L. casei, the ptsH1 mutation causes a relief from CCR of 6-P-ß-galactosidase and N-acetyl glucosaminidase and from diauxic growth in media containing glucose and either lactose, maltose, or ribose (919). An L. lactis ptsH1 mutant is relieved from CCR of the ptbA-bglH operon (569). These results confirm that P-Ser-HPr plays an important role in CCR of not only bacilli but also other gram-positive organisms.

A chromosomal B. subtilis ptsH2 mutant (Ser46Thr replacement) has also been constructed. Although the purified mutant HPr is slowly phosphorylated in ATP-dependent phosphorylation assays (722), the B. subtilis ptsH2 mutant exhibits a relief from CCR of the gnt operon identical to that of the ptsH1 mutant (M. Steinmetz and J. Deutscher, unpublished results). In contrast, an L. casei ptsH2 mutant shows only partial relief from CCR and diauxic growth (919). It is therefore likely that Ser46Thr HPr from L. casei is more efficiently phosphorylated than the B. subtilis mutant protein.

Some genes and operons are more strongly expressed in the presence of glucose or other rapidly metabolizable carbohydrates. This phenomenon is known as CCA. In B. subtilis, the expression of several genes such as alsS ({alpha}-acetolactate synthase) (726, 983), ackA (acetate kinase) (886), and pta (phosphotransacetylase) (685) is stimulated by the presence of glucose. CCA of these B. subtilis genes is absent from ptsH1 mutants. Expression of the L. lactis las operon containing the genes encoding the glycolytic enzymes phosphofructokinase, pyruvate kinase, and lactate dehydrogenase is also activated by the presence of glucose in the growth medium. The stimulatory effect of glucose on las operon expression disappeared in an L. lactis ptsH-disrupted strain transformed with a plasmid allowing the synthesis of L. lactis Ser46Ala mutant HPr (493), whereas transformants containing a plasmid carrying L. lactis wild-type ptsH exhibited normal glucose-mediated las activation.

The involvement of P-Ser-HPr in CCR and CCA was further supported by studying mutants that were unable to form P-Ser-HPr due to hprK inactivation. Such mutants have been obtained, for example, with B. subtilis (251, 315, 709), S. xylosus (357), and L. casei (198). In all three organisms, the inactivation of hprK causes a relief from CCR, but for several B. subtilis catabolic operons, the releasing effect in hprK mutants is markedly stronger than that in ptsH1 mutants (249, 523, 685). This phenomenon was observed for many other genes when transcriptome analysis was carried out with B. subtilis hprK and ptsH1 mutants (491). As will be explained below, the difference between ptsH1 and hprK mutants is due to the presence of Crh, a B. subtilis HPr paralog, which is also phosphorylated by HprK/P (250) and which can partly substitute for P-Ser-HPr in CCR and CCA.

The Catabolite Response Element cre, an Operator Site for CCR and CCA

A cis-acting element regulating CCR and CCA. Groundbreaking work in the laboratory of Glenn Chambliss led to the discovery of two other components involved in CCR in B. subtilis. Both were identified by isolating mutants in which the {alpha}-amylase-encoding amyE gene was relieved from CCR. Two independently isolated mutants were affected in a 14-bp imperfect palindromic sequence called amyR1, which is located just downstream from the amyE promoter and which exhibits similarity to the lac operator site of E. coli (601, 603). In both strains, replacement of the G:C base pair in position 8 of amyR1 with an A:T base pair was responsible for the observed CCR resistance (603, 939). Expression of a plasmid-borne cat gene under the control of the amyE promoter followed by mutated amyR1 was insensitive to CCR, while the simultaneous expression of the chromosomal amyE (promoter followed by wild-type amyR1) was repressed by glucose (602). These findings suggested that amyR1 is a cis-acting element. A series of mutants in which one or two positions of amyR1 were altered was constructed, and most of these mutants were partly or fully relieved from CCR of amyE. Based on mutational analysis and on the occurrence of similar imperfect palindromic sequences in the promoter regions of many other catabolite-repressed B. subtilis genes, the following consensus sequence was proposed for the catabolite response operator: TGWNANCGNTNWCA (939). Slightly different consensus sequences have been proposed by others (353, 563). The general term catabolite response element (cre) has been introduced for these operator sites (421).

Distribution of cre's. Imperfect palindromic sequences resembling the proposed consensus cre sequence are located in the promoter region or at the 5' ends of many catabolite-repressed transcription units. For some of these cre's, their implication in CCR or CCA has been experimentally confirmed (178). By searching the B. subtilis genome with a degenerate cre sequence as a query, Miwa and coworkers detected a total of 126 tentative cre's (565). Out of these 126 putative cre sites, 32 were inserted into a plasmid between the spac promoter and the lacZ gene, and the effect of these sequences on the glucose repression of ß-galactosidase synthesis was tested. In 22 constructs, expression of the lacZ gene was repressed by glucose (565). By using a slightly different search method, 61 additional potential B. subtilis cis elements for CCR could be detected, including several cre's preceding operons known to be submitted to CCR. A list of the about 160 identified or presumed B. subtilis cre's was presented by Deutscher et al. (178). As the transcription units regulated by these cre's contain, on average, two to three cistrons, about 400 genes, which represent 10% of the B. subtilis genome, are estimated to be regulated by CCR or CCA. A similar number was obtained when transcriptome analyses were carried out using a ccpA mutant (see below for ccpA) (575, 980). The number of genes presumed to be regulated by cAMP/Crp-mediated CCR in E. coli is in the same order (34).

A unique potential cre precedes the B. subtilis uxaC gene from the yjm operon, which codes for the enzymes necessary for glucuronide/galacturonate transport and metabolism (549, 729). While most cre's are composed of an imperfect palindromic sequence, the putative cre of the yjm operon is a perfect palindrome. Moreover, the 14-bp-long putative yjm cre is part of a longer perfect palindrome composed of a total of 26 bp (TCAAAATGTTAACGTTAACATTTTGA [the 14-bp consensus cre is in boldface type]). To date, it is not known whether this extended palindrome plays a special role in CCR. The ß-galactosidase-encoding mbgA gene of Bacillus megaterium contains two cre's overlapping by 1 bp. Mutations in either one affect CCR of the mbgA gene (803). Two potential cre's overlapping to a much larger extent (by 4 bp) begin at positions 278 and 287 within the B. subtilis gudB (former ypcA) gene, which encodes one of two glutamate dehydrogenases present in this organism and the expression of which is sensitive to the presence of glucose. The second B. subtilis glutamate dehydrogenase, RocG, is also submitted to CCR (53). The first 8 bp in the two presumed gudB cre's are identical (GTGAAGGCGgtgaaggcgCTTTCA [the first potential cre is in boldface type, the second is in italics, and the repeated sequences are underlined and in lowercase letters]). Interestingly, a rocG mutant is not able to utilize proline, ornithine, or arginine as the sole carbon source (growth on these carbon sources requires an active glutamate dehydrogenase), establishing that wild-type B. subtilis contains only low GudB activity. However, spontaneous revertants that were able to grow on these compounds appeared with high frequency. Surprisingly, in four independently isolated revertants, the GTGAAGGCG direct repeat was deleted, providing a gudB1 allele with only a single cre (TGAAGGCGCTTTCA) (54). It is not clear whether the elevated GudB activity in the spontaneous mutants is due to the change in the two overlapping cre's or to the deletion of the -V-K-A- repeat in the protein. The B. subtilis iol (563) and ara (361) operons also contain two functional cre's, which, however, are not located next to each other. One cre precedes the first gene (mmsA and araA, respectively); the other is located within the second gene (iolB and araB). Several other catabolite-repressed or -activated genes or operons contain more than one potential cre in the promoter region or within the 5' end (93, 290, 564, 886), but in several cases, only one of the cre's is operative in CCR or CCA.

Two divergently transcribed genes or operons can potentially be regulated by a single cre. This seems to be the case for the putative L. casei lev cre, which overlaps the transcription start site of the levABCDX operon and the –10 box of the levR gene, which encodes the transcription activator for the lev operon. Simultaneous repression of levABCDX and the transcription activator gene levR presumably leads to very efficient CCR (534).

The L. casei lev operon is only one of many examples establishing the presence of functional cre sites in bacteria other than B. subtilis (Table 3). In S. xylosus, deletion of the cre located in front of the malRA operon (for maltose/maltotriose utilization) causes a complete relief from CCR for {alpha}-glucosidase (204). Similarly, deletion of a cre located right behind the transcription initiation site of the fructan hydrolase-encoding fruA gene of Streptococcus mutans causes a relief from glucose repression (93, 946). CCR of the ß-amylase-encoding bamM gene from B. megaterium is mediated by a cre located in the DNA region encoding the signal peptide of this extracellular enzyme (457). A cre site is also present in front of the P-ß-glucosidase-encoding bglH gene of Lactobacillus plantarum (511, 512) and the pepQ gene of Lactobacillus delbrueckii (779). These examples confirm that the cre-dependent CCR/CCA mechanism is widely distributed within gram-positive bacteria with low G+C content. Interestingly, the HPr-encoding ptsH gene of several streptococci is also preceded by a potential cre site, and expression of the ptsHI operon in S. salivarius and Streptococcus bovis is up-regulated in the presence of glucose (CCA) (894).


View this table:
[in this window]
[in a new window]

  		TABLE 3. Genes and operons in gram-positive bacteriaa containing functional cre's or being regulated by the P-Ser-HPr:CcpA complex

cre's can either overlap the promoter, overlap the transcription start site, or be located between the transcription start site and the initiation codon or within the 5' part of catabolite-repressed genes (Table 3). Binding of a target protein to cre sites located within the promoter region is thought to interfere with the binding of the RNA polymerase holoenzyme, while cre's situated downstream from the promoter are assumed to be regulated via a road block. As will be discussed below, most transcription units submitted to CCA contain a cre located in front of the promoter region.

Catabolite Control Protein A Functions as a Catabolite Repressor or Activator

A LacI/GalR-type trans-acting factor mediates CCR and CCA. Since cre's exhibit similarity to the E. coli lac and gal operators, it was tempting to assume that a LacI/GalR-type repressor might be involved in CCR of B. subtilis and other gram-positive bacteria (603). By carrying out transposon mutagenesis, a B. subtilis strain in which amyE expression was resistant to catabolite repression could be isolated (322). Indeed, this strain carried the transposon inserted into a gene encoding a LacI/GalR-type repressor, which was called CcpA (catabolite control protein A) (938). Interestingly, genetic mapping revealed that the ccpA gene is located close to the previously described B. subtilis alsA1 mutation, which prevented the production of acetoin in glucose-grown cells (983). The failure of the alsA1 mutant to produce acetoin during growth on glucose is due to the almost complete absence of the alsS-encoded acetolactate synthase. The alsS gene forms an operon together with the acetolactate decarboxylase-encoding alsD (726) and is induced by a CCA mechanism. Similar to the alsA1 mutation, the ccpA mutation abolishes acetoin production. Introduction of an SPß prophage carrying intact ccpA restored acetoin production in the ccpA and alsA1 mutants, strongly suggesting that ccpA and alsA are identical loci (322). Indeed, the alsA1 mutation was identified as a G-to-A exchange within the Shine-Dalgarno box of the ccpA gene (at position –14 with respect to the ccpA start codon) (566). This mutation lowers the synthesis of CcpA about 10-fold. The concentration of CcpA in wild-type B. subtilis was calculated to be 0.6 µM (566).

Disruption of ccpA not only prevented the inhibitory effect of glucose on {alpha}-amylase activity and the induction of acetolactate synthase but also affected CCR of many other B. subtilis enzymes (178, 184, 299, 566), thus establishing the pleiotropic character of the ccpA mutation. Whole-genome analysis turned out to be a powerful tool for detecting CcpA-regulated transcription units (575, 980). Interestingly, three B. subtilis ccpA mutants that specifically affected CCA of the alsSD operon, while CCA or CCR of other transcription units was not or only slightly altered, were isolated (887). Two of the mutations affect conserved regions in the DNA binding domain of CcpA (A18V and R48C replacements), whereas the third mutation was in a less conserved region in the core domain (Pro286Leu replacement).

The effect of ccpA inactivation on the synthesis of catabolic enzymes could also be demonstrated by comparing the protein patterns of two-dimensional gels obtained with crude extracts from B. subtilis (880) or E. faecalis (453) wild-type and ccpA strains. E. faecalis proteins isolated from a few spots exhibiting an altered intensity in the ccpA mutant compared to the wild-type strain were identified by microsequencing. They include a protein with strong similarity to the {alpha} subunit of L-serine dehydratase, an enzyme catalyzing the conversion of serine to pyruvate, as well as glycerol dehydrogenase and the two subunits of dihydroxyacetone kinase (DhaKL). The genes encoding the three latter proteins are present in the dha operon (453), which also contains the gene for an EIIA of the mannose class PTS (see "SOME UNUSUAL PTS PATHWAYS AND PROTEINS"). In addition, galKETR operon expression as well as the synthesis of 13 proteins identified as glucose starvation proteins were no longer sensitive to CCR in the E. faecalis ccpA mutant. Two-dimensional gel electrophoresis with B. subtilis crude extracts showed that in a ccpA mutant, seven tricarboxylic acid cycle enzymes (CitC, CitG, CitH, CitZ, OdhA, SdhA, and SucC/SucD) were relieved from CCR (880). The citZ (403) and citG genes and the odhAB, sdhCAB, and sucCD operons are preceded by potential cre's (178, 565). Interestingly, E. faecalis and B. subtilis ccpA mutants contain a much larger amount of P-Ser-HPr than wild-type strains (453, 492). The increased percentage of P-Ser-HPr slows glucose uptake by the B. subtilis ccpA mutant when grown on synthetic medium (492), which in turn seems to prevent CCA of the gapA operon, which encodes the enzymes of the lower part of glycolysis (492, 880). When a ptsH1 mutation, which prevents the formation of P-Ser-HPr, was introduced into the ccpA strain, the activating effect of glucose on gapA operon expression (230) was restored (492). Expression of gapA is controlled by the repressor CggR, which is inactivated by elevated amounts of FBP (194). A similar repressive effect of the elevated amount of P-Ser-HPr in a ccpA mutant was observed for the B. subtilis glutamate synthase-encoding gltAB operon. Introduction of a ptsH1 mutation restored gltAB expression in a ccpA background (925). The mechanism leading to enhanced P-Ser-HPr formation in ccpA mutants is not known. In addition, when grown in complex medium, ccpA mutants seem to efficiently transport PTS sugars, because glucose utilization by a B. subtilis ccpA mutant represses the synthesis of several enzymes submitted to a second, CcpA-independent CCR mechanism. For example, CCR of the {alpha}-glucosidase- and glycerol kinase-encoding genes is not diminished in a B. subtilis ccpA mutant (184).

Genes regulated by CcpA include not only carbon catabolic genes but also genes related to carbon metabolism in a larger sense. The L. delbrueckii subsp. lactis CcpA ortholog PepR1 stimulates the expression of the prolidase-encoding pepQ gene and seems to affect the expression of pepX, pepI, and brnQ, which encode two other peptidases and a branched-chain amino acid transporter, respectively (779). A connection between CCR and amino acid metabolism also exists in Streptococcus gordonii and Streptococcus rattus, where the arginine deiminase operon is repressed by CcpA (195, 298), and in B. subtilis, where CcpA stimulates the expression of the ilv-leu operon, which encodes the enzymes for the synthesis of branched-chain amino acids (810, 882). A connection between CCR and the utilization of nitrogen sources was provided by the finding that the expression of rocG, which encodes the main glutamate dehydrogenase of B. subtilis, is subject to CcpA-mediated CCR (53). A connection between carbon and nitrogen metabolism in B. subtilis was further supported by the finding that the {sigma}54-encoding sigL gene is regulated by CcpA (128). Transcriptome analysis with B. subtilis ccpA, hprK, and ptsH1 mutants revealed that numerous nitrogen and phosphorus metabolic genes are submitted to CCR or CCA (491). CcpA of Lactobacillus pentosus represses the activity of xylulose 5-phosphate phosphoketolase, the central enzyme of the phosphoketolase pathway (674). In Clostridium perfringens, inactivation of ccpA diminishes the expression of the enterotoxin-encoding cpe gene during entry into the stationary growth phase (913). S. mutans ccpA mutants exhibit diminished fructosyltransferase and glucosyltransferase gene expression (88) and biofilm formation (947).

CcpA-specific sequences. CcpA belongs to the family of LacI/GalR-type repressors but contains characteristic sequences located at the end of the DNA binding domain and within the core protein and can therefore easily be distinguished from other members of this family. These regions are important for CcpA function, because mutations affecting amino acids in the CcpA-specific sequences prevent or lower CCR of a xylA-lacZ fusion in B. megaterium (422). Interestingly, polyclonal antibodies directed against CcpA from B. megaterium do not recognize E. coli LacI (441). Phylogenetic analysis shows that all low-G+C gram-positive bacteria for which the genomes have been sequenced contain a ccpA ortholog (J. Deutscher and C. Francke, unpublished observation). The gene is probably always expressed, as a protein exhibiting the approximate size of CcpA and cross-reacting with the B. megaterium CcpA antibody was present in the crude extracts of all gram-positive bacteria tested (441). For several of these bacteria, including E. faecalis, L. casei, Lactobacillus pentosus, L. lactis, L. monocytogenes, Clostridium saccharobutylicum (formerly C. acetobutylicum P262) (162), and S. xylosus, a role of their CcpAs in CCR was demonstrated experimentally. Either their ccpA genes complement a B. subtilis ccpA mutant or B. subtilis ccpA complements ccpA mutants of these organisms. Genes affected by ccpA mutations in bacteria other than B. subtilis are listed in Table 3.

CcpA needs a corepressor. Interestingly, overexpression of the C. saccharobutylicum ccpA ortholog regA in an E. coli strain harboring the C. saccharobutylicum staA gene inhibited the expression of staA, which encodes a starch-degrading enzyme. As the corepressor of CcpA is absent from E. coli (see the next section), these results imply that, if overproduced, CcpA can bind to cre's without a corepressor. Therefore, regulation of ccpA expression represents one way to control the binding of CcpA to cre's. Indeed, in a few organisms (S. xylosus, L. pentosus, and L. delbrueckii), ccpA is preceded by a cre, and its expression is submitted to autoregulation (204, 507, 574). In L. monocytogenes, the synthesis of CcpA was elevated when the cells were exposed to prolonged salt stress (200). By contrast, expression of B. subtilis ccpA is constitutive and is not influenced by the presence or absence of glucose (566). CCR in B. subtilis is therefore not regulated via ccpA expression. Instead, additional factors must be involved in the CCR signal transduction pathway. Because CcpA at physiological concentrations was not able to bind to the cre of the catabolite-repressed B. subtilis gnt operon (566), it was likely that these additional factors would function as corepressors of CcpA.

P-Ser-HPr Functions as a Catabolite Corepressor

Interaction of P-Ser-HPr with CcpA. When the effects of the B. subtilis ptsH1 mutation on CCR were compared to those of a ccpA mutation, it was found that all enzymes that are partly or completely relieved from CCR in the ptsH1 mutant are also relieved from CCR in the ccpA mutant (184). In addition, two of the enzymes that remained repressed by glucose in the ptsH1 mutant (GlpK and {alpha}-glucosidase) remained repressed in the ccpA mutant (184). The similar pleiotropic phenotypes observed for ptsH1 and ccpA mutants suggested that P-Ser-HPr and CcpA participate in the same CCR mechanism and might interact. Binding of P-Ser-HPr to CcpA was indeed demonstrated by elution retardation experiments (182). In the presence of FBP, elution of P-Ser-HPr from a Ni-nitrilotriacetic acid column saturated with His-tagged CcpA was clearly retarded compared to the elution of HPr or P~His-HPr, while doubly phosphorylated (P~His,P-Ser)-HPr was only slightly retarded.

The specific interaction between P-Ser-HPr and CcpA was confirmed by NMR measurements (381). The formation of the P-Ser-HPr:CcpA complex could be observed in the absence of FBP, probably owing to the high protein concentrations used in the NMR studies. The NMR measurements also allowed the determination of the residues of P-Ser-HPr interacting with CcpA. In addition to the regions around the active-center His-15 and the regulatory Ser-46 (amino acids 43, 44, and 46 to 56), the interface includes amino acids 21 to 27. The regions at amino acids 21 to 27 and 46 to 56 are well conserved within the HPrs of gram-positive bacteria but are quite different in the HPrs of most gram-negative organisms (Fig. 4).

The size of CcpA (37 kDa) is too large to allow a determination of its interface in the P-Ser-HPr:CcpA complex by NMR measurements. However, based on the crystal structure of PurR (788), a LacI/GalR-type repressor missing the CcpA-specific sequences, and the failure of various mutant CcpAs to form a complex with P-Ser-HPr, a presumed surface-exposed area including Tyr-89, Tyr-295, Ala-299, and Arg-303 was proposed to be part of the interface in the P-Ser-HPr:CcpA complex (422). The four amino acids presumed to participate in P-Ser-HPr binding are located in CcpA-specific regions that are absent from other members of the LacI/GalR repressor family. The crystal structure of the B. megaterium P-Ser-HPr:CcpA complex revealed that residues Tyr-295, Ala-299, and Arg-303 are located on helix IX of the CcpA N subdomain (which directly follows the DNA binding domain) and indeed interact with P-Ser HPr (787). Val-300 and Leu-304 of CcpA are also part of this tight interface, which includes Ile-47, Met-48, and Met-51 of P-Ser-HPr. Replacement of Ile-47 with Thr in S. salivarius HPr leads to the disappearance of diauxic growth and the CCR observed with the wild-type strain during lactose and maltose utilization (263). However, crossed immunoelectrophoresis with HPr antibodies revealed that the amount of P-Ser-HPr in the ptsH(Ile47Thr) mutant was similar to that observed in the wild-type strain. The ptsH(Ile47Thr) mutation was therefore assumed to diminish the affinity of the mutant P-Ser-HPr for CcpA (263). The surface-exposed hydrophobic patch formed by Ile-47, Met-48, and Met-51 is also involved in the interaction of HPr with EI (260) and EIIAs (139, 691, 949).

NMR studies revealed an affinity of CcpA for P-Ser-HPr that was about 50-fold higher than that for HPr (381). This effect must be due mainly to electrostatic repulsion, as no significant structural changes accompany the phosphorylation of HPr at Ser-46 (29, 381). The electrostatic effects were assumed to involve the negatively charged phosphorylated Ser-46 of HPr and a positively charged counterpart in CcpA. According to the crystal structure, Arg-303 and Lys-307 indeed form hydrogen bonds with the phosphoryl group in P-Ser-HPr. Tyr-89, the fourth CcpA residue predicted to be involved in the interaction with P-Ser-HPr, is located on helix I of CcpA. Its close contact to helix IX probably explains why a mutation affecting Tyr-89 prevents the interaction with P-Ser-HPr (422). The P-Ser-HPr:CcpA example therefore proves that based on genetic and biochemical studies, reliable predictions about protein-protein interfaces can be made.

Similar to gram-negative bacteria, where PEP-dependent phosphorylation of EIIAGlc is important for CCR (see "REGULATION OF CARBON METABOLISM IN GRAM-NEGATIVE ENTERIC BACTERIA"), phosphorylation of HPr at His-15 affects CCR in gram-positive bacteria, probably by weakening the P-Ser-HPr:CcpA interaction. Indeed, the region around His-15 of P-Ser-HPr makes contacts to helix IX of subunit 1 and helices I and II of subunit 2 in the regulatory domains of the CcpA dimer (787). One P-Ser-HPr molecule therefore interacts with both subunits of the CcpA dimer. Asp-69 and Asp-99 on helices I and II in the regulatory domain of subunit 2 of the CcpA dimer form hydrogen bonds with Arg-17 in P-Ser-HPr. The crystal structure provides an explanation as to why CcpA exhibits only a weak affinity for (P-Ser,P~His)-HPr (184) and why replacing His-15 with Glu or Ala caused an almost complete relief from CCR (707). When a ptsH1 mutant strain was transformed with a plasmid carrying the His15Ala ptsH allele, the Ser46Ala mutant HPr synthesized from the chromosomal ptsH1 allele allowed the uptake of glucose via the PTS, while the His15Ala mutant HPr synthesized from the plasmid-located ptsH allele could be phosphorylated at Ser-46 and was therefore expected to be active in CCR. However, only a weak repressive effect of glucose on gluconate kinase activity was observed in the ptsH1 transformant (707), suggesting that seryl-phosphorylated His15Ala mutant HPr has only a low affinity for CcpA (707). Indeed, in the P-Ser-HPr:CcpA complex, Asp-296 forms hydrogen bonds with the N{delta}1 of His-15 and the backbone amide nitrogen of Ala-16, which enhances the affinity between the two proteins. Phosphorylation at N{delta}1 of His-15 in P-Ser-HPr by PEP and EI or replacing His-15 with Glu or Ala probably prevents the interaction of Asp-296 with HPr and, in the case of (P-Ser,P~His)-HPr and His15Glu mutant P-Ser-HPr, probably even leads to electrostatic repulsion. The participation of His-15 in the interaction with CcpA provides a link between CCR and PEP-dependent phosphorylation of PTS proteins (182).

Several mutations leading to "permanent" CCR have been described. Replacing Ser-46 of HPr with a negatively charged Asp provides a mutant protein resembling P-Ser-HPr (952), which was therefore able to bind to CcpA immobilized on a Ni-nitrilotriacetic acid column (707). In agreement with this finding, overexpression of the ptsH(Ser46Asp) allele from a plasmid led to partial CCR of the B. subtilis gnt operon even in the absence of a repressing sugar (184). Two other classes of mutations that caused similar "permanent" CCR have been described. Several B. megaterium ccpA mutants that exhibit partial CCR of the xylose utilization genes in the absence of a repressing sugar have been isolated (442). In contrast to wild-type CcpA, the mutant CcpAs are thought to interact with the xyl cre even in the absence of a corepressor (442). The mutations affected the DNA binding domain (Glu77Leu replacement) or the core domain of CcpA. Permanent CCR was also observed with a B. subtilis mutant producing Val265Phe mutant HprK/P, which exhibits normal kinase but strongly reduced P-Ser-HPr phosphorylase activity, which leads to the accumulation of P-Ser-HPr (571). Similar to the HPr(Ser46Asp)-producing strain, CCR in the hprK(Val265Phe) mutant occurs even in the absence of a repressing sugar. Expression of a xynB'-lacZ fusion in the hprK(Val265Phe) mutant grown in a medium containing xylose and succinate as the sole carbon sources is 100 times lower than that in a wild-type strain (571). This permanent repressive effect is responsible for the failure of the hprK(Val265Phe) mutant to grow on CCR-sensitive sugars such as gluconate, glucitol, and ribose. Inactivation of ccpA or the insertion of a ptsH1 mutation in the hprK(Val265Phe) strain restored growth on the above-mentioned sugars (571). These results confirm that P-Ser-HPr formation represents the main signal for CCR in gram-positive bacteria. When present at elevated concentrations in the cell, P-Ser-HPr is sufficient to lead to strong CCR, while glycolytic intermediates do not seem to be essential.

Binding of the P-Ser-HPr:CcpA complex to cre's. The above-described results indicate that in gram-positive bacteria with low G+C content, P-Ser-HPr functions as a corepressor by allowing the repressor CcpA to bind to cre sites. This concept was first tested with the B. subtilis gnt operon, which possesses a cre located within the 5' end of gntR, the first gene of this operon (247). Footprint experiments revealed that CcpA or P-Ser-HPr alone does not bind to the gnt cre. By contrast, if P-Ser-HPr and CcpA were present together in the reaction mixture, a specific protection of a region corresponding to the gnt cre was observed, while no protection occurred with CcpA and dephospho-HPr (247). Mutations affecting the gnt cre and causing a relief from CCR prevented binding of the P-Ser-HPr:CcpA complex. Similar results were obtained with cre's from many other catabolite-repressed or -activated transcription units. In addition, by isolating the DNA:protein complex and separating its components on a denaturing gel, it was demonstrated that P-Ser-HPr and CcpA bind together to cre sites (30).

The binding affinity of the P-Ser-HPr:CcpA complex for a cre site was first studied with circular dichroism spectroscopy. Based on association/dissociation-related changes in the spectrum, an apparent KD of 4.5 µM was determined for the formation of the ternary complex P-Ser-HPr:CcpA:optimized B. megaterium xyl cre (A:T in position 10 replaced with T:A) (381). A 1:1:2 (cre:CcpA dimer:P-Ser-HPr) binding stoichiometry was proposed. Because no interaction with the cre could be observed when 0.5 mM HPr and CcpA was used, binding of P-Ser-HPr to CcpA was estimated to increase the affinity of the repressor for its DNA targets by more than 2 orders of magnitude (381). Besides, equilibrium constants for the binding of the P-Ser-HPr:CcpA complex to the B. megaterium xyl cre and the B. subtilis amyE cre were determined by electrophoretic mobility shift experiments and were in the orders of 5 x 106 and 3.5 x 108 M–1, respectively (30, 404). A KD (0.6 nM) that was much lower than the one determined by circular dichroism spectroscopy (4.5 µM) was reported for the formation of the ternary complex of B. subtilis CcpA, P-Ser-HPr, and the xylA cre (794).

The crystal structure of the triple complex composed of the CcpA dimer and two P-Ser-HPr molecules from B. megaterium and a cre has been solved (787). It revealed that the N-terminal HTH motifs and the hinge helices of the CcpA dimer make 32 phosphate contacts to the DNA. They interact with 10 bp out of the 16 bp forming the pseudopalindromic cre, an artificial operator site not present in B. megaterium. However, a reverse complementary sequence identical to bp 2 to 15 of the employed cre is present in Streptococcus iniae and probably controls expression of the lactose permease- and lactose oxidase-encoding lctPO operon (cre located 52 bp upstream from the start codon) (J. Deutscher, unpublished observation). In the ternary complex, important contacts are made with the central C:G and G:C base pairs via Leu-55, the "leucine lever." These contacts are responsible for DNA kinking and minor groove expansion. DNA bending by the P-Ser-HPr:CcpA complex (35°) is smaller than what was observed for LacI (40°) and PurR (50°). The HTH motif exhibits sufficient flexibility to allow CcpA binding to the variable half sites of pseudopalindromic cre's. The plasticity of the HTH motif also explains the capacity of CcpA to interact with the various cre's of an organism, which generally exhibit significant sequence differences.

A comparison of the structure of the triple complex with that of unliganded CcpA (apoCcpA) revealed that the interaction of P-Ser-HPr with CcpA causes significant structural changes, primarily in the regulatory N subdomain, which allow CcpA to interact with cre's (787). The N subdomain in both subunits of "activated" CcpA is rotated on average by 6° compared to apoCcpA. In addition, the position of helix IX (contains Arg-303) in "activated" CcpA is different from its location in apoCcpA. The interaction of Arg-303 of CcpA with the phosphoryl group in P-Ser-HPr requires the rotation of the side chain of Arg-303, which initiates a series of structural changes culminating at Thr-61, which is located at the core/DNA binding junction. The rotation of Tyr-91, which in apoCcpA interacts with Thr-306 via a hydrogen bond, towards the CcpA dimer interface plays an important role in this process, which ultimately allows high-affinity binding of CcpA to cre sites (787).

The P-Ser-HPr:CcpA interaction mode seems to be independent of the absence or presence of a cre site. The crystal structure of a complex formed between B. subtilis P-Ser-HPr and CcpA missing the first 57 amino acids was very similar to the corresponding structure in the B. megaterium P-Ser-HPr:CcpA complex (106). Interestingly, the B. subtilis protein complex contains two sulfate ions located close to each other. Both sulfate ions interact with Asp-276, and because an FBP molecule could easily be modeled into the structure with its two phosphoryl groups superimposed on the sulfate ions, it was predicted that the region around Asp-276 represents the FBP binding site.

Based on the above-described results, a general CCR and CCA mechanism, as outlined in Fig. 6 can be proposed for low-G+C gram-positive bacteria. The increase in FBP in cells growing on a rapidly metabolizable carbohydrate such as glucose stimulates the HprK/P-catalyzed formation of P-Ser-HPr, which forms a complex with CcpA and allows the LacI/GalR-type repressor to bind to the cre's. For all genes of gram-positive organisms known to be submitted to CCA, the cre is located upstream from the promoter, and binding of the P-Ser-HPr:CcpA complex stimulates transcription, probably by a mechanism similar to that reported for Crp- or Cra-mediated activation of gene expression (142, 645). cre's located upstream from the promoter are found in the catabolite-activated B. subtilis ackA (886), glg (178), pta (685), and ilv-leu (810, 882) transcription units and the pepQ and las genes of L. delbrueckii and L. lactis, respectively (494) (Table 3). CcpA-mediated CCA of the B. subtilis als operon occurs probably via an indirect mechanism, as no obvious cre is located in front of the als promoter (882). The cre of the CCR-sensitive S. xylosus malRA operon is also located in front of the promoter (205). However, the mal cre and the mal promoter are separated by only 8 bp. This might be close enough to enable the P-Ser-HPr:CcpA complex bound to the cre to prevent binding of the RNA polymerase holoenzyme. In carbon catabolite-activated genes (ackA, pta, ilv-leu, and pepQ), the cre and the corresponding promoter are separated by at least 12 bp.


Figure 6
View larger version (46K):
[in this window]
[in a new window]

  		FIG. 6. Mechanisms underlying CCR in firmicutes. The uptake of glucose and other rapidly metabolizable PTS sugars (top left) leads to a net dephosphorylation of the PTS proteins. The center and bottom of the figure show that the high concentration of FBP present in cells growing on a rapidly metabolizable carbohydrate stimulates the HPr kinase activity of the bifunctional HprK/P and the formation of P-Ser-HPr. P-Ser-HPr interacts with CcpA, and the protein complex binds to the cre operator sites on the DNA. The promoter regions are indicated as –10 and –35. CCA occurs when the cre is located upstream from the promoter, while CCR requires a cre located within or downstream from the promoter. High concentrations of Pi present in resting cells favor the pyrophosphate-producing dephosphorylation of P-Ser-HPr by HprK/P. The top right part of the figure shows that P-Ser-HPr probably interacts with certain non-PTS carbohydrate transport systems, such as the maltose transporter from L. casei, and thereby inhibits their transport activity. Phosphorylation of LacS by P~His-HPr stimulates the lactose/galactose exchange reaction (in S. thermophilus). In the presence of rapidly metabolizable PTS substrates, the low level of P~His-HPr does not allow sufficient phosphorylation of GlpK, which leads to the inactivation of GlpK and to inducer exclusion. Pyr., pyruvate.

Deviations from the general CCR mechanism in gram-positive bacteria. When Ramseier et al. studied the in vitro binding of B. megaterium CcpA to an "optimized" B. subtilis xyl cre, in which the G:C and A:T base pairs in positions 3 and 10 were replaced with T:A, an interaction was observed in the absence of P-Ser-HPr (695). Moreover, the presence of P-Ser-HPr or the structurally related Ser46Asp mutant HPr (952) inhibited the interaction of CcpA with the modified xyl cre. In addition to the two mutations introduced intentionally in the B. subtilis xyl cre, there is also an ambiguity for position 9, which was previously reported to be a T (421) but is a C in the published genome sequence (437). In the nearly identical B. megaterium xyl cre (only one base pair difference) (290), there is a C at position 9. As the experiments reported by Ramseier et al. (695) were carried out with the sequence carrying a T in position 9, the unexpected results obtained in the CcpA binding studies might be due to this sequence difference. Indeed, when footprint experiments were carried out with B. megaterium CcpA and an optimized B. megaterium xyl cre, which differs from the optimized B. subtilis xyl cre used previously (695) only at position 9 (C:G instead of T:A), results similar to those reported for the B. subtilis gnt cre were obtained; i.e., binding of B. megaterium CcpA to the xyl cre was observed only when P-Ser-HPr was present (290).

CcpA has been reported to bind in vitro to the B. subtilis ccpC (401) and ackA cre's (886) in the absence of a corepressor. Nevertheless, the stimulating effect of glucose on ackA expression disappeared in a ptsH1 crh double mutant (for crh, see the next section) (886). The discrepancy between in vivo and in vitro results could not be explained.

Glucose-6-P allowed P-Ser-HPr-independent binding of CcpA to the B. subtilis gnt cre (564) and the optimized B. megaterium xyl cre (290), albeit only under nonphysiological conditions. For the B. subtilis gnt cre, an effect could be detected only when glucose-6-P was used at more than 30 mM, which is far above the 4 mM glucose-6-P detected in glucose-grown B. subtilis cells (233). For the B. megaterium xyl cre, an effect of glucose-6-P on the binding of CcpA could be observed only at pH values below 5.4 (290), i.e., far below the cytoplasmic pH of 7 to 8 reported for B. megaterium cells (503). Glucose-6-P was therefore not considered to play a role in regulating binding of CcpA to the cre's but was suggested to mimic a possibly physiologically relevant unknown effector (290).

In footprint experiments, CcpA was found to bind weakly to the B. subtilis amyE cre in the absence of a corepressor (405). CcpA binding was stimulated by P-Ser-HPr and further enhanced by FBP (404). In in vitro transcription studies, CcpA in the 500 nM range was able to specifically inhibit amyE expression. In the presence of 2.7 µM P-Ser-HPr, inhibition of amyE expression occurred at slightly lower CcpA concentrations (around 100 nM) and was more efficient (90% inhibition). However, the P-Ser-HPr effect was considered to be nonspecific, as expression from control promoters was also inhibited. In addition, the ptsH1 mutation had no influence on CCR of the amyE gene (924). Surprisingly, when NADP or NADPH was present, specific inhibition of in vitro transcription from the amyE promoter was observed at very low CcpA concentrations (in the 5 nM range), but NADP(H) was less efficient (70% inhibition) than P-Ser-HPr. Moreover, in footprint experiments, the addition of NADP(H) had only a weak effect on the protection of the amyE cre by CcpA. It was therefore concluded that NADP(H) does not function as a corepressor by enhancing the binding of CcpA to the cre but, rather, that it allows CcpA to inhibit the transcription machinery, thus leading to reduced amyE expression (404). CcpA has been proposed to directly interact with and to inhibit RNA polymerase, and NADP/NADPH was thought to stimulate the interaction of CcpA with RNA polymerase (406). The suggested role of NADP and NADPH cannot be related to the redox state of the cells, as both molecules were similarly efficient in the stimulation of CcpA-mediated inhibition of amyE expression, while NAD and NADH had no effect. With the exception of dormant spores, the concentration of the NADP(H) pool was reported to remain fairly constant in B. megaterium cells under various growth conditions, including growth in the absence and presence of glucose (801). It is therefore not clear to which signal the interaction of NADP(H) with CcpA responds.

In several instances, CCR of catabolic genes is mediated by an indirect mechanism. For example, cre sites can be located in front of genes encoding transcription regulators of catabolic genes. This is the case for B. subtilis CcpC, which controls the expression of citZ and citB (401) and for LevR of L. casei, which regulates the expression of the levABCDX operon (534). Similarly, the acoABCL operon of B. subtilis is controlled by the transcription activator AcoR, the gene of which is preceded by a cre recognized by CcpA (12). While a cre site also precedes citZ and the L. casei levABCDX operon, a cre seems to be absent from acoABCL, which is therefore only indirectly regulated by CcpA. A cre site has also been identified within the sigL gene of B. subtilis, which codes for {sigma}54, which is necessary for the transcription from several "–12, –24" promoters (128).

The regM gene of S. mutans encodes a homolog of CcpA. Surprisingly, the disruption of regM increased the glucose repression of {alpha}-galactosidase, mannitol-1-P dehydrogenase, and P-ß-galactosidase, suggesting that RegM functions as a positive regulator for the expression of the corresponding genes. In addition, regM inactivation affected neither diauxic growth in media containing glucose and a less efficiently metabolizable sugar (816) nor expression of the fructanase-encoding fruA gene, although deletion or mutation of the cre preceding fruA led to a relief from CCR (946). Moreover, FBP, which stimulates HPr kinase activity in most bacteria, slightly inhibited HPr kinase from S. mutans (869). It therefore seems that S. mutans follows a CCR mechanism different from that operative in most other gram-positive organisms.

P-Ser-Crh, a Second Catabolite Corepressor in Bacilli, Geobacilli, and Oceanobacilli

A B. subtilis HPr-like protein without His-15. Although most catabolite-repressed genes and operons of B. subtilis are partly or completely relieved from CCR in both ccpA and ptsH1 mutants, the degree of relief can be quite different for the two mutants. For example, CCR of the amyE gene and the hut, cta, xyn, and iol operons is only slightly affected by the ptsH1 mutation (less than 20%), whereas these five transcription units are completely relieved from CCR in ccpA mutants (184, 250, 488, 924, 985). Similarly, CCA of the phosphotransacetylase-encoding pta gene is not affected in a ptsH1 mutant but is completely lost in a ccpA strain (685). Less pronounced differences in CCR or CCA between ptsH1 and ccpA mutants were observed for the glucitol dehydrogenase-encoding gutB (184), the acetyl-CoA synthetase-encoding acs (984), and the acetate kinase-encoding ackA (886) genes as well as for the lev operon (250, 526). These results suggested that in B. subtilis, the binding of CcpA to certain cre's is regulated not only by P-Ser-HPr but that a second corepressor might exist.

Evidence for a potential second corepressor came from the B. subtilis genome sequencing project, which led to the discovery of a gene encoding an HPr-like protein (219). Although this protein exhibits 45% sequence identity to HPr, there is a striking difference: His-15, the site of PEP-dependent phosphorylation in HPr, is replaced with a Gln in the HPr paralog. As a consequence, the HPr-like protein is not phosphorylated by PEP and EI (250). Nevertheless, the sequence around Ser-46 resembles that in HPr of gram-positive bacteria, and HprK/P-mediated, ATP-dependent, FBP-stimulated phosphorylation at Ser-46 in Crh could be demonstrated (250). These results suggested that the phosphorylated HPr paralog might play a role in CCR similar to that of P-Ser-HPr, and it was therefore called catabolite repression HPr (Crh).

However, when the crh gene is disrupted, no effect on CCR is observed. The participation of P-Ser-Crh in CCR can be detected only in a ptsH1 background. The residual CCR of the lev, iol, xyn (250), and hut (985) operons as well as the remaining CCA of the pta (685) and ackA (886) genes observed in ptsH1 mutants disappeared almost completely when crh was inactivated. Complete relief of the acsA (984), xyn (249), and lev (522) operons from CCR was also observed in a ptsH1 crh1 double mutant in which Ser-46 of both HPr and Crh was replaced with Ala.

In vitro, Crh forms a mixture of monomers and dimers (644), which, similar to EI, are in slow equilibrium. The structure of the monomer and the interface of the dimer were determined by NMR spectroscopy (222). The NMR data suggested that domain swapping of the ß1 strand is implicated in B. subtilis Crh dimer formation, and this was confirmed by the crystal structure (389). The Crh monomer strongly resembles HPr, except for a few specific differences. Their structural similarity explains why Gln15His mutant Crh can be phosphorylated by PEP and EI (524). When the Gln15His mutant Crh is overproduced in a B. subtilis ptsH deletion strain, it partly restores PTS transport activity and growth on PTS substrates or glycerol and allows the full activation of several transcription regulators possessing a PTS regulation domain (PRD) (156, 524). GlpK and PRD-containing transcription regulators (see "REGULATORY FUNCTIONS MEDIATED BY P~His-HPr AND P~EIIBs") are stimulated by PEP-dependent phosphorylation via EI and HPr (158, 844).

Binding of the P-Ser-Crh:CcpA complex to cre's. To determine the binding characteristics of the P-Ser-Crh:CcpA complex, footprint experiments were carried out with cre's from transcription units, which are not completely relieved from CCR in ptsH1 mutants. The xyn cre site was indeed specifically protected by the P-Ser-Crh:CcpA complex, but the effect was slightly weaker than that with P-Ser-HPr:CcpA (249). The use of various mixtures of P-Ser-HPr and P-Ser-Crh did not increase the protective effect compared to the protection obtained with only a single corepressor. For a few catabolic genes, the binding of P-Ser-Crh:CcpA to the cre's was stimulated by FBP, and in some cre's, certain positions became hypersensitive towards DNase I digestion after binding the repressor:corepressor complex. P-Ser-HPr:CcpA and P-Ser-Crh:CcpA protected identical DNA regions, and FBP always had comparable effects on the DNA binding affinity of the two protein complexes. For example, FBP increased the protective effect of the P-Ser-HPr:CcpA and P-Ser-Crh:CcpA complexes against DNase I digestion of the xyn cre (249) and the pta cre (685). In contrast, FBP did not enhance the protective effect of both CcpA:corepressor complexes in footprint experiments with the lev cre (523). Both protein complexes rendered several nucleotides located outside the lev cre hypersensitive towards DNase I digestion. Expression of the lev operon is regulated by the transcription activator LevR, which binds to an upstream activating sequence (UAS) centered at about 130 bp upstream from the lev promoter. LevR was therefore assumed to exert its stimulating effect on lev operon expression by DNA bending (523). The lev cre is located between the UAS and the {sigma}54-dependent "–12, –24" promoter. Binding of the P-Ser-HPr:CcpA or P-Ser-Crh:CcpA complexes to the lev cre supposedly prevents LevR-induced DNA bending by altering the DNA structure, which might be responsible for the hypersensitivity towards cleavage during footprint experiments. Although in the case of the pta cre, only weak protection by P-Ser-HPr:CcpA or P-Ser-Crh:CcpA was observed, several nucleotides located outside the cre site became hypersensitive to DNase I digestion. When FBP was included in the reaction mixtures, the hypersensitive bands disappeared, and the protective effect on the pta cre was enhanced (685). The physiological significance of these in vitro observations remains to be determined.

In addition to the fact that certain cre's are not recognized by the P-Ser-Crh:CcpA complex (when the ptsH1 mutation leads to a complete relief from CCR or CCA), P-Ser-Crh was reported to have a 10-fold-lower affinity for CcpA than P-Ser-HPr. These differences are surprising, as, with the exception of two amino acids (His-15 and Thr-20), all P-Ser-HPr residues involved in the interaction with CcpA are conserved in Crh, and the crystal structures of the two proteins revealed a very similar fold. A possible explanation was that the dimerization of Crh (389) might have an influence on the formation of the P-Ser-Crh:CcpA complex and its interaction with cre sites. However, the crystal structure of the triple complex P-Ser-Crh(B. subtilis):CcpA(B. megaterium):cre (same cre as that used for the P-Ser-HPr:CcpA structure) (787) revealed that the two P-Ser-Crh molecules bound to the CcpA dimer are present exclusively in monomeric form (789). The binding of P-Ser-Crh:CcpA to the cre induces a kink (DNA bend angle of 31°) in the central conserved C-G that is almost identical to that observed in the triple complex with P-Ser-HPr (DNA bend angle of 35°). Nevertheless, specific differences between P-Ser-HPr:CcpA and P-Ser-Crh:CcpA occur in the contact region including His-15/Gln-15 and Ala-20/Thr-20 of HPr/Crh, respectively. In the complex with P-Ser-HPr, His-15 hydrogen bonds with the side-chain carboxyl of Asp-296 of CcpA. In contrast, Gln-15 of Crh makes only a weak contact to Asp-296 (distance of 4 Å), but instead, hydrogen bonds with the side chain of Arg-324, which, compared to the P-Ser-HPr:CcpA structure, needs to be rotated to come sufficiently close to the interface to allow the contact with Gln-15. Interestingly, an equivalent of Arg-324 seems to be present in CcpAs of only those organisms possessing Crh (789). If this assumption holds, Bacillus pseudofirmus, Bacillus sphaericus, and Exiguobacterium sibiricum 255-15 should contain a Crh ortholog, as their CcpAs contain an equivalent of Arg-324, whereas G. stearothermophilus, in contrast to Geobacillus kaustophilus, should be devoid of Crh, as its CcpA is missing an equivalent of Arg-324 (replaced with a Glu). The presence of Crh in the above-mentioned four organisms is uncertain, as no or only incomplete genome sequences are presently available.

In conclusion, the structures of the complexes of P-Ser-HPr and P-Ser-Crh with CcpA and a cre exhibit extensive similarity, and the few differences that were observed do not explain why the B. subtilis P-Ser-HPr:CcpA complex binds to several cre sites, which, according to genetic data, are not or only barely targeted by the P-Ser-Crh:CcpA complex. The corresponding transcription units are partly or completely relieved from CCR in ptsH1 and hprK mutants (gnt, mtl, etc.) (184). To better understand the different affinities of the two corepressor:CcpA complexes for certain cre's, it might be useful to determine the structure of a triple complex of P-Ser-HPr, CcpA, and a cre that does not interact with the P-Ser-Crh:CcpA complex.

Distribution of Crh in gram-positive organisms. In addition to B. subtilis (250) and B. megaterium (794), Crh has been detected in most other bacilli (Bacillus anthracis, B. clausii, B. cereus, B. halodurans, B. licheniformis, B. thuringiensis, and B. weihenstephanensis) as well as in G. kaustophilus and O. iheyensis, which are both closely related to bacilli. The various Crhs exhibit a high degree of sequence identity (between 61 and 83%), whereas they show only 41 to 47% sequence identity to the HPrs of their respective organisms. Interestingly, B. licheniformis possesses two HPr-like proteins containing a Gln in place of His-15. One of them has a Ser-46 and therefore corresponds to Crh (83% sequence identity with Crh of B. subtilis). The other protein exhibits only 44% sequence identity to B. subtilis Crh, and its Ser-46 is replaced with an Asp, which probably mimics permanent phosphorylation at position 46 (Deutscher, unpublished). The function of the latter protein is unknown. crh is the last gene (or penultimate gene in B. subtilis) of the yvc operon. The last gene of the B. subtilis yvc operon (250), yvcN, is absent from the other organisms, and O. iheyensis also misses the first gene, yvcI.

The genome sequences of other gram-positive bacteria revealed that staphylococci, clostridia, lactobacilli, lactococci, enterococci, etc., also have a yvc operon, but none of them contains a crh gene. Although the in vivo and in vitro results unequivocally established that P-Ser-Crh can contribute to CCR in B. subtilis, the fact that crh is present only in bacilli and a few close relatives and that crh inactivation in wild-type B. subtilis has no effect on CCR raises serious doubts about whether CCR is the primary function of Crh. P-Ser-HPr seems to be sufficient to mediate CCR in gram-positive organisms, independently of whether they possess Crh or not. In addition, numerous B. subtilis operons are completely relieved from CCR in a ptsH1 mutant, suggesting that their cre is not recognized by the P-Ser-Crh:CcpA complex. So far, only one P-Ser-Crh-specific function could be established. Expression of the Mg2+-citrate transporter-encoding B. subtilis citM gene is strongly repressed during growth on glucose but is also slightly repressed when grown on glutamate/succinate as the sole carbon source. Whereas repression of citM by glucose is mediated via CcpA in complex with either P-Ser-HPr or P-Ser-Crh, repression by glutamate/succinate completely disappeared in a crh-disrupted strain (934). Interestingly, crh expression, which is much weaker than ptsH expression, is highest in cells grown on one of the two TCA cycle intermediates, citrate or succinate (286).

It is possible that additional functions of Crh are related to the genes cotranscribed with crh. It is noteworthy that the yvcJ gene organized in an operon with crh resembles yhbJ from the E. coli rpoN operon. YhbJ possesses Walker motifs A and B, but its function is not known. Because in many hprK-containing gram-negative bacteria, yhbJ and other genes of the E. coli rpoN region are organized in an operon with hprK, it is tempting to assume that Crh might somehow play a role in the regulation of SigL ({sigma}54), the RpoN ortholog of B. subtilis.

An HPr paralog containing a conserved Ser-46 but missing His-15 (His-15 is replaced with a Phe) is also present in the mollicute Ureaplasma urealyticum (also called Ureaplasma parvum serovar 1) (308). The sequence around Phe-15 (amino acids 12 to 22) completely differs from all known HPrs, while the other parts of the protein resemble HPr from gram-positive and gram-negative bacteria. In contrast to other mollicutes, U. urealyticum lacks EI as well as an HPr with His-15. Mollicutes generally lack CcpA but contain an hprK gene, as is also the case for U. urealyticum (308). The role of P-Ser-HPr formation in these organisms therefore remains unknown (307).

What Are the Functions of CcpB and CcpC?

A second trans-acting factor implicated in CCR in B. subtilis was identified. This protein, which was called catabolite control protein B (CcpB), exhibits 30% sequence identity to CcpA, and especially, the helix-turn-helix motifs of the two proteins are very similar. Indeed, ccpB disruption leads to a partial relief from CCR of the gnt and xyl operons in cells grown on solid medium (113). However, when the cells are grown in liquid media at a high agitation rate, no effect of the ccpB mutation is observed. In contrast, mutant cells grown at a low agitation rate exhibit partial relief from CCR. Based on these results, it was proposed that CcpB would affect CCR in response to changes in environmental conditions, such as oxygen supply, cell density, etc. However, the DNA targets and the factors controlling the binding of CcpB to these targets are not known. The structures of the P-Ser-HPr:CcpA and P-Ser-Crh:CcpA complexes (787, 789) revealed that those sequences that are specific for the CcpA repressor subfamily (422) are important for the interaction with P-Ser-HPr. The fact that these sequences are absent from CcpB argues against the assumption that it might use P-Ser-HPr as a corepressor (113).

A third trans-acting factor called CcpC is involved in CCR of the B. subtilis citB and citZ genes (which code for aconitase and citrate synthase, respectively) (386) and in CCR of the L. monocytogenes genes encoding a CitB ortholog and a presumed glutamine transporter (lmo0847) (402). However, CcpC is not a member of the LacI/GalR family but is a member of the LysR family of transcription regulators. In the absence of an effector, B. subtilis CcpC functions as a repressor by binding to both a dyad symmetry element centered at position –66 with respect to the transcription start site of citB and a repeat of only the downstream arm of the dyad symmetry element at position –27. In the presence of citrate, the putative inducer for citB and citZ, CcpC, binds only to the complete dyad symmetry element (at position –66), which leads to transcription activation. Mutations affecting the CcpC binding sites in front of citB cause derepressed expression of only citB, whereas ccpC inactivation stimulates the expression of both citB and citZ (386). The dyad symmetry element shows no similarity to cre's. Nevertheless, expression of the CcpC-regulated citB and citZ genes is elevated in a ccpA mutant (880), probably owing to two indirect effects. Firstly, a cre site is present in front of citZ (403). Inactivation of ccpA therefore leads to enhanced synthesis of the citrate synthase CitZ and probably to elevated amounts of citrate, the inducer of citB (403). Secondly, CcpA regulates ccpC expression by binding to a cre site located between the two promoters of the ccpC gene (401). Disruption of ccpA leads to elevated ccpC expression and, in the presence of high concentrations of citrate, to enhanced citB and citZ expression.

It seems contradictory that B. subtilis cells growing on rapidly metabolizable carbon sources such as glucose activate the expression of glycolytic enzymes but repress enzymes of the TCA cycle. However, during growth on glucose, this organism secretes large amounts of acetate into the medium (685) and also produces acetoin when approaching stationary phase (726). To divert the carbon flux towards acetate/acetoin production, the expression of the genes encoding the enzymes necessary for the synthesis of these metabolites (acetate kinase, phosphotransacetylase, acetolactate synthase, and acetolactate decarboxylase) is stimulated, whereas TCA cycle enzymes, cytochromes, and cytochrome oxidases are repressed. Once the rapidly metabolizable carbon source is exhausted, B. subtilis begins to utilize the previously secreted acetate and acetoin. Under these conditions, the synthesis of the enzymes for acetate and acetoin production is no longer activated, and TCA cycle enzymes, cytochromes, and cytochrome oxidases are derepressed. This coordinated regulation probably provides B. subtilis with a growth advantage, as not all bacteria are able to efficiently utilize acetate and acetoin.

CcpB and CcpC do not seem to be directly regulated by P-Ser-HPr. Nevertheless, the phosphoprotein has been shown to interact with B. subtilis RbsR (Table 1), a LacI/GalR-type repressor, which controls the expression of the ribose operon (583). When electrophoretic mobility shift assays were carried out, only P-Ser-HPr, but not HPr, allowed RbsR to interact with a DNA fragment containing the putative RbsR binding site, which also represents a perfect cre site. It was therefore proposed that this site might be the target for both RbsR and CcpA (731). The positively charged residues of RbsR that were predicted to interact with the phosphoryl group of P-Ser-HPr according to a structure model differ from the positively charged residues located in the interface of the P-Ser-HPr:CcpA complex (787). So far, no in vivo data that would attribute a physiological role to the observed in vitro P-Ser-HPr/RbsR interaction have been published.

Role of P-Ser-HPr in the Virulence of Certain Pathogens

P-Ser-HPr and the regulation of virulence genes in gram-positive pathogens. Surprisingly, in certain pathogens, P-Ser-HPr formation affects the activity of transcription regulators of virulence genes. The best-studied example is the Crp/Fnr-type transcription activator PrfA of L. monocytogenes (207). It regulates the expression of several virulence genes, including the hemolysin-encoding hly gene and the phospholipase-encoding plcA gene, which together with the prfA gene are clustered within a 10-kb region on the chromosome (129). Growth of L. monocytogenes on rapidly metabolizable carbon sources such as glucose and fructose leads not only to CCR of catabolic genes (50) but also to repression of the PrfA-controlled genes (558). While repression of catabolic genes requires CcpA, repression of the virulence genes is CcpA independent, since it persists in ccpA mutants (50). Nevertheless, P-Ser-HPr, the corepressor of CcpA, seems to play a role in the regulation of PrfA activity. A B. subtilis strain containing L. monocytogenes prfA under the control of the spac promoter and the lacZ reporter gene fused to the PrfA-activated hly promoter exhibits lacZ expression only in the presence of IPTG (804). Expression of lacZ and thus PrfA activity are inhibited by glucose and fructose, indicating that CCR of PrfA-controlled genes is also operative in B. subtilis (179, 326). Interestingly, introduction of the hprK(Val267Phe) mutation in the B. subtilis strain, which leads to the conversion of more than 95% of the HPr into P-Ser-HPr (571), strongly inhibits PrfA activity even in the absence of a repressing sugar (326). Replacement of Ser-46 in HPr with an alanine (ptsH1 mutation) prevented the inhibitory effect of the hprK(Val267Phe) mutation, suggesting that P-Ser-HPr is directly or indirectly involved in PrfA regulation. In contrast, inactivation of ccpA or introduction of the crh1 mutation in the hprK(Val267Phe) strain did not stimulate PrfA activity (326). The effect of P-Ser-HPr on PrfA activity seems to be indirect and based primarily on its poor phosphorylation by enzyme I, which probably leads to barely phosphorylated EIIAs and EIIBs. In agreement with this concept, deletion of ptsH or the enzyme I-encoding ptsI also led to strong inhibition of PrfA activity (326). PrfA therefore seems to require a fully functional PTS and might be either stimulated by P~His-HPr, P~EIIAs, or P~EIIBs or inhibited by dephosphorylated EIIAs or EIIBs. A connection between the PTS components and PrfA was also suggested by the observation that the overproduction of PrfA inhibits glucose utilization via the PTS (517). In contrast, the utilization of glucose-6-P via the hexose phosphate transport protein (UhpT) was not affected by PrfA overproduction. In addition, in contrast to glucose, fructose, cellobiose, etc., glucose-1-P did not inhibit PrfA activity, although it is probably converted to glucose-6-P and metabolized via glycolysis (728). The failure of glucose-1-P, which is not transported via the PTS, to inhibit PrfA activity might therefore be related to its different mode of transport. In conclusion, these results suggest a direct or indirect effect of PTS components necessary for the uptake of glucose, fructose, cellobiose, etc., on PrfA activity.

In contrast to PrfA, the S. pyogenes virulence gene regulator Mga seems to be controlled by the classical CcpA-dependent CCR mechanism. Mga functions as a transcription activator for the expression of numerous genes encoding mainly surface-associated proteins involved in pathogenesis. Expression of the Mga regulon is growth phase dependent, and the presence of rapidly metabolizable carbon sources such as glucose stimulates the synthesis of members of the M-protein family (emm, mrp, arp, and enn) (424, 535). In addition, Mga controls genes encoding fibronectin-binding proteins (sof and fbx) and a collagen-like protein (sclA) (14, 15). Mga also controls its own expression by interacting with two 59-bp binding motifs located within the mga promoter region (536). In addition, the –35 region of the first of the two mga promoters is preceded by a DNA sequence that almost perfectly matches the cre consensus sequence. Purified CcpA was found to bind to the presumed cre site (K. S. McIver, personal communication). Its location upstream from the –35 promoter region suggests that mga is submitted to CCA. This is in agreement with the previous observation that the presence of glucose stimulates the synthesis of the M protein.

The Streptococcus pneumoniae CcpA ortholog RegM represses {alpha}-glucosidase and ß-galactosidase activities, but the inactivation of ccpA (regM) also attenuates the virulence of this organism. The mutation has no impact on the expression of several established pneumococcal virulence genes but, rather, affects the expression of genes from the capsular polysaccharide synthesis (cps) locus (273). Capsular polysaccharides are known to be important factors for host cell adhesion. Indeed, the ccpA mutant is severely attenuated for nasopharyngeal colonization and lung infection in mice (366). This effect might partly be related to the differences in surface protein composition observed between the wild type and the ccpA mutant.

P-Ser-HPr and the regulation of virulence genes in gram-negative pathogens. Genes encoding proteins with significant similarity to HprK/P are present in numerous gram-negative bacteria ({alpha}-, ß-, {gamma}-, and {delta}-proteobacteria), spirochetes, green sulfur bacteria, Fibrobacteres, and fusobacteria (41, 64). The existence of HprK/P homologs in gram-negative bacteria came as a surprise. Due to the absence of Ser-46 phosphorylation in the HPrs of E. coli and other Enterobacteriaceae, HprK/P was originally thought to exist only in gram-positive organisms. Interestingly, HprK/P-containing gram-negative bacteria possess an HPr in which the sequence around Ser-46 differs from the corresponding sequence in HPrs of Enterobacteriaceae and instead resembles those in HPrs from gram-positive organisms (64) (Fig. 4). For the gram-negative bacteria Neisseria gonorrhoeae and N. meningitidis (64), the spirochete T. denticola (278), and the {alpha}-proteobacteria Agrobacterium tumefaciens (S. Poncet, A. Khemiri, I. Mijakovic, A. Bourand, and J. Deutscher, unpublished results) and B. melitensis (S. Poncet, M. Dozot, A. Bourand, J. Deutscher, J. J. Letesson, and X. de Bolle, unpublished results), ATP-dependent phosphorylation of HPr by their HprK/Ps has been demonstrated in vitro. In several of the above-mentioned bacteria, ptsH and sometimes also ptsI are located in the hprK operon. Nevertheless, most of these bacteria seem to contain neither CcpA nor an EIIB or EIIC, excluding the possibility that P-Ser-HPr plays a role in the regulation of PTS transport activity or CcpA-mediated CCR in these bacteria.

In addition to ptsH and ptsI, the hprK operons of gram-negative bacteria frequently contain a gene encoding an EIIA of either the fructose/mannitol or the mannose class PTS (64, 345). The organization of the hprK operon in certain gram-negative pathogens and symbionts suggests a role of HprK/P in cell adhesion and virulence (64). In these proteobacteria, the hprK operon usually contains between 1 and 10 genes of the E. coli rpoN region or genes encoding a two-component system. Genes coding for an EIIA-like protein (PtsN) and an HPr-like protein (PtsO or NPr) are also present in the rpoN operon of E. coli and other enterobacteria (553) (for the regulatory functions of RpoN, PtsN, and PtsO, see "SOME UNUSUAL PTS PATHWAYS AND PROTEINS"). In hprK-containing ß-, {gamma}-, and {delta}-proteobacteria, hprK is usually located between rpoN and yhbJ (64). In a few {alpha}-proteobacteria, the hprK and yhbJ genes are also located next to each other. Interestingly, the crh gene of B. subtilis is also organized in an operon with a yhbJ homolog (64), while in Ruminococcus albus, hprK is followed by murB and a yhbJ-like gene (Deutscher, unpublished). These observations strongly suggest a functional connection between the PTS and the nucleotide binding protein YhbJ.

For ß- and {gamma}-proteobacteria containing a combined hprK/rpoN operon, HprK/P was proposed to control the phosphorylation state of the EIIAFru-like PtsN (64, 724), the gene of which often precedes hprK. In these organisms, signals causing enhanced HPr kinase and diminished P-Ser-HPr phosphorylase activity were assumed to lead to elevated amounts of P-Ser-HPr, which would slow the phosphorylation of the EIIA (180, 723). In N. meningitidis, such a phosphorelay system seems to somehow control the interaction with host cells, as the inactivation of hprK strongly reduces cell adhesion. The inactivation of hprK possibly leads to the enhanced formation of P~PtsN. Dephospho-PtsN was therefore proposed to participate in a signal transduction pathway controlling cell adhesion (64).

Similarly, in several {alpha}-proteobacteria, the colocalization of hprK with genes encoding a two-component system suggests a role of HprK/P in the regulation of cell adhesion (64, 345). The A. tumefaciens EnvZ/OmpR-like two-component system ChvG/ChvI, which is encoded by the genes preceding hprK, controls the expression of the virulence genes virB and virE (472), which encode components of a type IV secretion system delivering oncogenic DNA to susceptible plant cells (100). Inactivation of chvG or chvI prevents tumor formation on the host plants (108). Similarly, inactivation of the chvG/chvI-like bvrR and bvrS genes of the pathogenic animal endosymbiont Brucella abortus affects cell invasion and virulence of this organism (821). Inactivation of the ChvG-like sensor kinase ExoS in the plant symbiont Sinorhizobium meliloti prevents the production of succinoglycan, the main exopolysaccharide necessary for the successful invasion of nodules on the host plant alfalfa (126). It has been proposed that HprK/P and the PTS proteins might somehow control the activity of the two-component systems (64). In {alpha}-proteobacteria, the hprK operon usually contains a gene encoding an EIIA of the mannose class PTS and sometimes also the genes for EI and HPr, whereas the gene for an EIIA of the fructose class PTS is located next to the rpoN gene (S. Poncet, unpublished observation). HprK/P was thought to control the phosphorylation state of the EIIAs. An EIIA or P~EIIA was in turn suggested to interact with the two-component system (64), or alternatively, a P~EIIA might phosphorylate one of the proteins of the two-component system. Interestingly, during a search for genes affecting the virulence of Brucella melitensis, inactivation of the EIIAFru-encoding gene was found to lower the pathogenicity of this organism (169).

{alpha}-Proteobacteria as well as Coxiella burnetii contain a truncated HprK/P missing the N-terminal domain, which resembles the N-terminal part of MurE (13), the enzyme catalyzing the formation of cytoplasmic precursors for cell wall synthesis (284). Similar to the artificially truncated L. casei HprK/P (229), the natural absence of the first 130 amino acids of A. tumefaciens HprK/P does not affect the known HprK/P functions. Nevertheless, the A. tumefaciens enzyme is not able to efficiently use PPi as a phosphoryl donor and to dephosphorylate P-Ser-HPr in the presence of Pi (I. Mijakovic, A. Khemiri, A. Bourand, S. Poncet, and J. Deutscher, unpublished results). However, the absence of these functions from A. tumefaciens HprK/P is likely due to sequence differences in the central loop compared to HprK/P from gram-positive organisms. Mutations affecting this loop in L. casei or B. subtilis HprK/P strongly diminish the phosphorylase activity (571).

Interestingly, all known bacteria with a truncated HprK/P possess an EI with an N-terminal extension resembling the N-terminal DNA binding domain of NifA, a transcription activator for certain {sigma}54-dependent promoters. EINtr (PtsP) of E. coli, which, together with the HPr paralog PtsO, is assumed to catalyze the phosphorylation of PtsN, also contains this NifA domain (see "SOME UNUSUAL PTS PATHWAYS AND PROTEINS").

Is P-Ser-HPr Involved in Inducer Expulsion?

In the next few sections, we will deal with the regulatory phenomena inducer expulsion and inducer exclusion in low-G+C gram-positive bacteria, which were both suggested, based mainly on in vitro data, to be related to P-Ser-HPr formation. However, while in vivo experiments confirmed a participation of P-Ser-HPr in inducer exclusion, inducer expulsion also occurs when P-Ser-HPr cannot be formed due to the inactivation of hprK or the replacement of the Ser-46 of HPr with an alanine.

Simultaneous occurrence of inducer expulsion and P-Ser-HPr formation. Inducer expulsion is a regulatory phenomenon that has been known for at least 40 years. Halpern and Lupo observed that the presence of certain carbon sources accelerates the exit of {alpha}-MG from E. coli cells that had accumulated this nonmetabolizable PTS substrate (309). Glucose had the strongest effect and accelerated {alpha}-MG expulsion about sixfold compared to cells to which no carbon source was added. When discussing the expulsion of nonmetabolizable carbohydrates, at least two different forms need to be distinguished, as they follow different mechanisms: the expulsion of non-PTS sugars and PTS sugars. The latter are accumulated as phospho derivatives inside bacterial cells and are first dephosphorylated before they are expelled. We will discuss the expulsion of PTS sugars, because ATP-dependent HPr phosphorylation was discovered in connection with TMG expulsion from S. pyogenes cells (185, 710).

Streptococci and lactococci transport nonmetabolizable sugar derivatives such as TMG or 2DG via the PTS and accumulate these carbohydrates as P derivatives. Interestingly, these nonmetabolizable P-sugars are rapidly expelled when the cells are exposed to a well-metabolizable carbon source such as glucose, sucrose, or mannose (712, 872). Expulsion turned out to be a two-step process: the accumulated P-sugars are first intracellularly dephosphorylated before the dephosphorylated sugars are expelled (710). The expulsion could be prevented when the ATP synthesis was inhibited by poisoning the cells with arsenate, fluoride (712), or iodoacetate (872). It has subsequently been established that in poisoned cells, the dephosphorylation of the accumulated P-sugars is inhibited (710). Arsenate- or fluoride-poisoned S. pyogenes cells synthesize little ATP, but they can metabolize arginine via the deiminase pathway and thereby generate ATP. The presence of arginine indeed restored sugar-P dephosphorylation and inducer expulsion in arsenate- or fluoride-poisoned S. pyogenes cells (710). It was therefore concluded that the first step of inducer expulsion depends on the presence of intracellular ATP and that the sugar-P phosphatase catalyzing this step might be regulated directly or indirectly via ATP-dependent protein phosphorylation. The search for this presumed P-protein led to the discovery of P-Ser-HPr: under conditions provoking inducer expulsion, a low-molecular-weight protein became phosphorylated by ATP (185, 710). This protein was identified as P-Ser-HPr (185), and it was therefore assumed that P-Ser-HPr might play a role in inducer expulsion, probably by controlling the activity of the sugar-P phosphatase (708, 710).

Experiments aimed at studying in vitro TMG-6-P expulsion by L. lactis vesicles supported this concept. When we discussed the effect of P-Ser-HPr formation on PTS-catalyzed carbohydrate uptake, we mentioned that vesicles prepared from L. lactis cells contain EI, HPr kinase, and a sugar-P phosphatase but had lost most of the HPr (970). As a result, L. lactis vesicles exhibit only about half the TMG uptake rate measured with intact cells (970). The full TMG uptake rate was restored when 100 µM B. subtilis HPr was electroporated into the L. lactis vesicles. Coelectroporation of wild-type or mutant HPrs with various metabolites was used to test whether phosphorylation of HPr at Ser-46 affects expulsion of TMG from the vesicles. L. lactis vesicles were able to accumulate TMG-6-P, which, similar to in intact cells, was expelled when glucose was added. Interestingly, vesicles into which HPr had been electroporated exhibited glucose-induced TMG expulsion that was about three times faster than that of vesicles that had been electroporated in the absence of HPr or in the presence of Ser46Ala mutant HPr (970). In addition, a cytoplasmic sugar-P phosphatase was activated about sevenfold when glucose was added to toluenized vesicles into which HPr had been electroporated prior to toluenization. Coelectroporation of FBP and HPr similarly enhanced the sugar-P phosphatase activity, and stimulation was even stronger when ATP was included. Because glucose and FBP stimulate the ATP-dependent phosphorylation of HPr at Ser-46 in L. lactis vesicles, it was assumed that the sugar-P phosphatase catalyzes the first step in inducer expulsion and that it is allosterically regulated by P-Ser-HPr. In agreement with this concept, electroporation of the P-Ser-HPr-resembling Ser46Asp HPr stimulated the intravesicular sugar-P phosphatase even in the absence of FBP (970), whereas only a weak increase of the phosphatase activity was observed with FBP and Ser46Ala mutant HPr. In addition to this cytoplasmic phosphatase, a membrane-associated L. lactis sugar-P phosphatase was also reported to be activated by P-Ser-HPr (973). Similarly, S. bovis contains a cytoplasmic and a membrane-associated sugar-P phosphatase (138). However, the activity of only the latter enzyme was stimulated by Ser46Asp mutant HPr. Membrane-associated, Ser46Asp HPr-activated sugar-P phosphatases were also detected in S. pyogenes and E. faecalis, two other organisms exhibiting inducer expulsion (712, 967). Inducer expulsion in all these bacteria was therefore assumed to be triggered by P-Ser-HPr-mediated activation of the membrane-associated sugar-P phosphatase. However, as heterologous systems were used for the in vitro electroporation experiments (HPr from B. subtilis and vesicles from streptococci, lactococci, or enterococci), these results needed to be confirmed by in vivo experiments.

Inducer expulsion in L. casei and L. lactis does not require P-Ser-HPr. Similar to L. lactis ML3, L. casei strains 64H and BL23 transport TMG via a lactose-specific PTS, accumulate it as TMG-6-P, and expel it from the cells as TMG when glucose or another rapidly metabolizable carbon source is added (111, 198). An L. casei mutant synthesizing Ser46Ala HPr (ptsH1) was constructed from strain BL23. Although P-Ser-HPr cannot be formed in this strain, it expelled accumulated TMG-6-P similarly to the wild-type strain when it was exposed to glucose or mannose. Identical results were obtained with ptsH mutants in which Ser-46 or Ile-47 had been replaced with a threonine (198). These results established that the expulsion of TMG in L. casei does not depend on P-Ser-HPr. To exclude the possibility that an HPr-like protein capable of substituting for P-Ser-HPr in inducer expulsion might exist in L. casei, TMG expulsion was also studied in an L. casei hprK mutant (198). Compared to inducer expulsion by a wild-type strain, no effect of the hprK mutation on glucose- or mannose-triggered expulsion of accumulated TMG-6-P was observed.

Because most experiments suggesting a role for P-Ser-HPr in inducer expulsion were carried out with L. lactis vesicles, a ptsH1 mutant of this organism was constructed and tested for the expulsion of preaccumulated TMG-6-P and 2DG-6-P. Expulsion of both nonmetabolizable sugar derivatives was observed in the ptsH1 mutant. However, whereas 2DG expulsion was nearly identical in the wild type and the mutant, TMG expulsion was slowed by about 2.5-fold in the ptsH1 strain (569). Because L. lactis possesses only HPr and no HPr-like proteins (68), these results unequivocally established that inducer expulsion in this organism does not depend on the presence of P-Ser-HPr. Nevertheless, HPr seems to play an indirect role in TMG expulsion in L. lactis, which might partly explain the vesicle results, which led to the conclusion that P-Ser-HPr participates in inducer expulsion. Differences in the phosphorylation state of the EIIBCLac, which was proposed to catalyze not only TMG uptake but also TMG expulsion (719), in the wild-type and the ptsH1 strains were suggested to be responsible for the indirect effect of the ptsH1 mutation on TMG expulsion (569).

Is P-Ser-HPr Involved in Inducer Exclusion?

In vitro interaction of Ser46Asp mutant HPr with non-PTS permeases. As explained above (see REGULATION OF CARBON METABOLISM IN GRAM-NEGATIVE ENTERIC BACTERIA), the inducer exclusion mechanism in enteric bacteria is mediated by dephospho-EIIAGlc, which binds to several target proteins (permeases and catabolic enzymes), thereby inhibiting their activity and preventing the entry or the intracellular formation of the corresponding inducer. Several lines of evidence indicated that P-Ser-HPr might play a similar role in inducer exclusion in gram-positive bacteria (Fig. 6). In L. brevis, the galactose/H+ symporter GalP also catalyzes the uptake of TMG (193, 739). The addition of glucose arrests TMG uptake and causes the rapid efflux of preaccumulated TMG from the cells. Both effects of glucose were ascribed to P-Ser-HPr, as its structural homolog, Ser46Asp HPr (from B. subtilis) (952), labeled with 125I was able to bind to the galactose/H+ symporter of L. brevis (972). Galactose as well as galactose derivatives stimulated the binding of 125I-labeled Ser46Asp HPr to the galactose/H+ symporter about 10-fold. Coelectroporated metabolites known to stimulate the kinase activity of HprK/P (FBP, gluconate-6-P, and glycerate-2-P) as well as extravesicular glucose inhibited TMG uptake by L. brevis vesicles preloaded with B. subtilis HPr by electroporation about 10-fold (969). No inhibitory effect of glucose or its metabolites on TMG uptake was observed when Ser46Ala mutant HPr was used in place of wild-type HPr. Moreover, Ser46Asp mutant HPr slowed TMG uptake even in the absence of extravesicular glucose or intravesicular metabolites. Studies of 2DG uptake by L. brevis vesicles electroporated with wild-type or mutant HPrs and with glycolytic intermediates provided results that were almost identical to those obtained for TMG uptake (968). In addition, by expressing the L. brevis galP gene (encodes the galactose/H+ symporter) in a B. subtilis wild-type strain and an hprK mutant, both B. subtilis strains gained the capacity to efficiently take up TMG. When the B. subtilis strains also synthesized L. brevis wild-type or Ser46Ala or Ser46Asp mutant HPr, a strong inhibition of TMG uptake was observed with the strain producing Ser46Asp HPr (193), suggesting that the P-Ser-HPr-resembling Ser46Asp HPr causes inducer exclusion. However, when the strain producing L. brevis wild-type HPr was grown in the presence of glucose, which stimulates the formation of P-Ser-HPr, no inducer exclusion occurred. In addition, since the experiments with L. brevis vesicles were carried out in vitro with B. subtilis wild-type or mutant HPrs instead of the physiologically relevant L. brevis P-Ser-HPr, the proposed implication of P-Ser-HPr in inducer exclusion needed to be confirmed in vivo.

Mutations preventing P-Ser-HPr formation abolish inducer exclusion. The importance of ATP-dependent HPr phosphorylation for inducer exclusion could be established by studying this regulatory process in L. casei and L. lactis mutants that are unable to form P-Ser-HPr. To detect a potential influence of the ptsH1 and hprK mutations on inducer exclusion in L. casei, transport studies with maltose and ribose were carried out (198, 919). In L. casei, maltose and ribose are taken up by ABC transport systems (570). When wild-type cells are grown in a medium containing glucose and either ribose or maltose, diauxic growth with a long-lasting lag phase is observed. In addition, maltose uptake is immediately arrested when glucose is added to maltose-transporting L. casei cells (198, 919), suggesting that glucose inhibits maltose uptake via an inducer exclusion mechanism. Interestingly, in ptsH1 (919) and hprK (198) mutants, glucose no longer exerts its inhibitory effect on maltose transport. These results imply that the phosphorylation of HPr at Ser-46 plays a role in inducer exclusion in L. casei. Inactivation of the ccpA gene had no effect on inducer exclusion (919), confirming that the P-Ser-HPr-dependent inhibitory effect of glucose on maltose uptake is due exclusively to an inhibition of the transport step and not to CCR of the maltose operon mediated via the P-Ser-HPr:CcpA complex. Additional evidence for P-Ser-HPr-mediated maltose exclusion from L. casei cells came from direct measurements of sugar consumption in the growth medium. In wild-type cells, maltose consumption is instantaneously arrested when glucose is added to the incubation mixture and does not restart before glucose is exhausted. In contrast, in the hprK and ptsH1 mutants, the addition of glucose causes only a short transient stop of maltose consumption, and the two sugars are subsequently simultaneously utilized (198, 919).

Interestingly, ribose transport was stimulated by the presence of glucose. Glucose stimulation of ribose uptake has also been reported for Lactobacillus sakei. Dephosphorylation of the PTS proteins in the presence of glucose was suggested to be responsible for the stimulating effect, as the inactivation of ptsI also led to enhanced ribose uptake (834, 835).

The repressive effect of glucose on maltose uptake and maltose consumption had also disappeared in an L. casei strain producing Ser46Thr mutant HPr, which is a poor substrate for HprK/P. Unexpectedly, inducer exclusion in L. casei was also prevented by the ptsH(Ile47Thr) mutation, although the amount of P-Ser-HPr detected in the S. salivarius ptsH(Ile47Thr) mutant was similar to that in the parental strain (263). The hydrophobic patch on the surface of HPr, which is formed by Ile-47, Met-48, and Met-51 in gram-positive bacteria (375), is involved in the interaction of P-Ser-HPr with CcpA (381, 787) and in the interaction of HPr with EI (260) and various EIIAs (139, 691, 949). The finding that the L. casei ptsH(Ile47Thr) mutant does not exhibit inducer exclusion suggests that this hydrophobic patch might also be important for the interaction of P-Ser-HPr with non-PTS permeases regulated by inducer exclusion.

Inducer exclusion studies were also carried out with an L. lactis ptsH1 mutant. Similar to L. casei, this organism takes up maltose and ribose via specific ABC transport systems (68). The inhibitory effect of glucose on the transport of the two non-PTS sugars in the wild-type strain had completely disappeared in the ptsH1 mutant (569).

An approximately twofold decrease in the P-Ser-HPr level was reported to lead to an almost complete loss of inducer exclusion in S. salivarius cells (870). The lower amount of P-Ser-HPr was due to specific point mutations in the ptsH gene or the ptsH leader sequence, which did not affect the uptake rate of PTS substrates, although the metabolism of PTS sugars was slowed. Wild-type S. salivarius cells immediately stop the uptake of the non-PTS sugar lactose or galactose when glucose or fructose is added to the growth medium and resume the metabolism of the non-PTS sugars only when glucose or fructose is exhausted (660). In contrast, the uptake of lactose or galactose by the ptsH mutants containing twofold-lower amounts of P-Ser-HPr was not affected by the presence of the PTS sugar glucose or fructose (870). To exert inducer exclusion, gram-positive bacteria therefore seem to require high levels of P-Ser-HPr.

In summary, the above-described results support the concept that inducer exclusion in gram-positive bacteria is regulated via the metabolite-activated formation of P-Ser-HPr. Considering the similarities between the maltose uptake systems from E. coli and those from the lactic acid bacteria L. casei and L. lactis, it is tempting to assume that the exclusion of maltose follows a mechanism analogous to that in gram-negative organisms, except that P-Ser-HPr instead of dephospho-EIIAGlc binds to MalK (Fig. 6) (Table 1). This assumption is supported by a comparison of the sequences of the nucleotide binding and regulatory domains of MalKs from E. coli (117) and L. casei BL23 (570). The two proteins are composed of very similar N-terminal nucleotide binding domains (amino acids 1 to 241) (60% identity), while the C-terminal regulatory domains (amino acids 242 to 371) exhibit only 19% sequence identity (Deutscher, unpublished). The sequence difference in the regulatory domain might reflect the binding of different effectors: EIIAGlc in gram-negative bacteria and P-Ser-HPr in gram-positive bacteria.


arrow
REGULATION OF CARBON METABOLISM IN LOW-G+C GRAM-POSITIVE BACTERIA: REGULATORY FUNCTIONS MEDIATED BY P~His-HPr AND P~EIIBs
 
PEP-Dependent Phosphorylation of Non-PTS Proteins

Besides their role as phosphocarriers during PEP-dependent phosphorylation of carbohydrates transported via the PTS, P~His-HPr and several P~EIIBs also phosphorylate non-PTS proteins and regulate their activities. Based on the characteristics of the phosphorylation site, three different types of non-PTS proteins phosphorylated by P~His-HPr and/or P~EIIB can be distinguished. In the simplest case, an EIIA domain, the usual phosphoryl acceptor of P~His-HPr within the PTS phosphorylation cascade, or an HPr domain is fused to the target protein. These PTS protein domains are not active in sugar transport but, rather, regulate the activity of the fusion proteins in response to their phosphorylation state. For instance, an HPr domain is fused to the response regulator HprR of a C. acetobutylicum two-component system (721). EIIA domains are fused to non-PTS sugar transporters (673) and to transcription activators such as MtlR of G. stearothermophilus (324) or LevR of B. subtilis (178), which control the expression of operons encoding sugar-specific PTS components (Table 1). The transcription regulators also contain an EIIB domain (295) for which neither phosphorylation nor a regulatory function could be demonstrated so far.

The second class of proteins phosphorylated by P~His-HPr and/or P~EIIBs contains a PRD (844). The PRDs seem to have specifically evolved to control the RNA binding activity of transcription antiterminators and the DNA binding function of transcription activators in response to phosphorylation by PTS proteins (Table 1). PRDs exhibit no obvious similarity to PTS proteins. Nevertheless, they usually contain two histidyl residues, which are phosphorylated by P~His-HPr or a P~EIIB. The sequences around the two phosphorylatable histidyl residues are conserved. Most PRD-containing proteins possess two PRDs either organized in tandem or sometimes separated by EIIA and EIIB domains (178, 295) (Fig. 7). Depending on which of the usually four conserved histidyl residues in the two PRDs is phosphorylated, the regulatory domains stimulate or inhibit the activity of their antiterminator or transcription activator (520, 884).


Figure 7
View larger version (32K):
[in this window]
[in a new window]

  		FIG. 7. Domain structure of the transcription antiterminator LicT and the transcription activators LicR and LevR of B. subtilis. The N-terminal RNA or DNA binding domain and the two regulatory domains PRD1 and PRD2 together with their conserved histidyl residues are indicated. Histidyl residues, which have been shown to become phosphorylated, are in boldface type. When the phosphorylation exerts a positive effect on the activity of the regulator, they are shown in red, whereas blue indicates the sites of negative regulation. For LicT, its four conserved histidyl residues become phosphorylated. Phosphorylation in PRD1 inhibits, while phosphorylation in PRD2 stimulates, LicT activity. Most other antiterminators of the BglG/SacY family also contain four conserved histidines, but sometimes, not all of them can be phosphorylated. LevR contains an NtrC-like central domain and EIIAMan- and EIIBGat-like domains inserted between the complete PRD1 and a truncated PRD2. The positive site of phosphorylation is His-585, and the negative site is His-869. The conserved histidyl residues His-506 and His-567 in PRD1 and the EIIBGat phosphorylation site (Cys-718) do not seem to become phosphorylated. An identical organization can be observed in most other NifA/NtrC-type PRD-containing transcription activators. However, a few LevR-like regulators contain a complete PRD2 with an additional potentially phosphorylatable histidyl residue. The DNA binding domain of LicR resembles that of DeoR, and PRD1 and PRD2 are followed by an EIIBGat-like and EIIAMtl-like domain. An identical domain organization can be observed in other DeoR-type PRD-containing transcription activators. All four conserved histidines in PRD1 and PRD2 are sites of positive regulation, while LicR activity is inhibited by phosphorylation at His-559 in the EIIAMtl-like domain.

The third class of non-PTS proteins phosphorylated by P~His-HPr contains GlpKs from low-G+C gram-positive bacteria (Table 1). These GlpKs contain neither a PTS protein domain nor a PRD but developed a novel P~His-HPr phosphorylation site, i.e., a histidine that is usually surrounded by three aromatic amino acids (Tyr and Phe) (109). The phosphorylation site is located on a surface-exposed loop.

Here, we will discuss the complex regulation of PRD-containing antiterminators and transcription activators in both gram-positive and gram-negative bacteria. The activity of these transcription regulators is controlled by multiple (up to fivefold) PEP-dependent PTS-catalyzed phosphorylations. The phosphorylation events can have antagonistic effects on the antitermination and transcription activation function (Table 1). The observations that these transcription regulators occur mainly in low-G+C gram-positive bacteria and that there are only a few representatives in proteobacteria and high-G+C gram-positive organisms will also be discussed. We will subsequently describe the P~His-HPr-mediated phosphorylation of lactose and raffinose transporters with an EIIAGlc domain and of a conserved histidyl residue in GlpK, regulatory processes so far observed only in gram-positive organisms (Table 1).

Regulation of Transcription Antiterminators by PTS-Mediated Phosphorylation

Regulation of gene expression by transcription attenuation. In addition to the most common mechanisms regulating gene expression, i.e., transcription activation or repression, bacteria also developed a mechanism based on transcription termination/antitermination, i.e., transcription attenuation, which was first described for the E. coli trp operon (60). Two principal mechanisms of transcription termination have been reported for prokaryotes: factor-dependent termination and intrinsic termination (965). In the first case, the RNA polymerase is halted at a terminator located at the end of a transcription unit. Interaction with a {rho} factor is necessary to dissociate the DNA:RNA polymerase complex and to terminate transcription. In contrast, transcription units submitted to intrinsic termination contain a {rho}-independent terminator. When the {rho}-independent terminator is located between the transcription start site and the start codon (Fig. 8), the expression of the corresponding gene or operon usually occurs from a constitutive promoter. However, under noninducing conditions, transcription stops when the terminator is formed on the nascent mRNA. Formation of the terminator leads to the disruption of the elongation complex and to the release of short, incomplete transcripts. Under inducing conditions, an antiterminator binds to the mRNA in front of the terminator or to a region overlapping the terminator, thereby preventing the formation of the disruptive stem-loop structure and allowing the RNA polymerase to complete transcription of the corresponding gene or operon (Fig. 8). Antiterminators can be either proteins or oligonucleotides. When tRNAs function as antiterminators, binding to their target site on the mRNA is usually prevented when they are charged with their corresponding amino acid (136). Polypeptide antiterminators frequently respond to changes in the concentration of intracellular metabolites, which serve as cofactors and allow the corresponding antiterminator to bind to its target on the mRNA. For example, glycerol-3-P binds to and activates the B. subtilis antiterminator GlpP (749). Another mode of regulation was reported for antiterminators controlling the expression of genes and operons encoding either sugar-specific PTS components or enzymes catalyzing the extracellular degradation of polysaccharides to mono- or oligosaccharides, which are subsequently taken up via the PTS. This group of polypeptide antiterminators is regulated by reversible PTS-catalyzed phosphorylation. They possess an N-terminal RNA binding domain (first 55 amino acids) usually followed by two regulatory domains, each containing two conserved histidyl residues (Fig. 7), which are phosphorylated or dephosphorylated depending on whether the corresponding carbohydrate is present in the growth medium.


Figure 8
View larger version (33K):
[in this window]
[in a new window]

  		FIG. 8. Transcription regulation by the PTS via PRD-containing transcription antiterminators. (A) In the absence of the corresponding inducer, full-length transcription of several PTS-encoding genes/operons is inhibited owing to the formation of a terminator structure (t, yellow) on the nascent mRNA upstream from the start codon. Under these conditions, the corresponding antiterminator cannot bind to its RNA target, RAT (blue), because the EIIB is mainly phosphorylated and transfers its phosphoryl group to PRD1 of the antiterminator. The absence of a repressing sugar is expected to also allow phosphorylation at the activating domain (PRD2) by P~His-HPr. However, the negative effect of phosphorylation at PRD1 is dominant. The RAT sequence can also form a stem-loop, which, however, was calculated to free less energy than the terminator t. Interestingly, in most antiterminator-controlled PTS operons, the two sites RAT and t overlap (green), and the formation of the terminator therefore prevents the formation of the RAT stem-loop and vice versa. (B) If an inducer is present, the EIIB as well as PRD1 of the corresponding antiterminator will be present mainly in an unphosphorylated form. Because antiterminator-controlled PTSs are usually low-capacity sugar transporters, there will be sufficient P~His-HPr to guarantee activating phosphorylation in PRD2. The activated antiterminator binds to its RAT and thus favors the formation of the RAT stem-loop, thereby preventing the formation of the terminator stem-loop, as part of it (in green) is already used for the RAT stem-loop. (C) If, in addition to the inducing sugar, a repressing carbohydrate is present, the amount of P~His-HPr will be low in the cells. In firmicutes, P-Ser-HPr will also be formed, which further lowers the amount of P~His-HPr. These conditions prevent activating phosphorylation at PRD2, and most antiterminators are therefore inactive, although the presence of the inducer probably prevents the phosphorylation in PRD1. Dephosphorylation of PRD2 in the presence of a rapidly metabolizable PTS sugar therefore represents a CcpA-independent, P-Ser-HPr-dependent CCR mechanism. ptsH1 as well as licT(Pia) but not ccpA mutants are relieved from this type of CCR.

{rho}-Independent terminators. The B. subtilis levansucrase-encoding sacB gene (31, 806) and the E. coli bgl operon, which encodes a ß-glucoside-specific EIIBCA (bglF), a 6-P-ß-glucosidase (bglB) (504) (Table 4), and a carbohydrate-specific outer membrane porin (20), were the first two PTS-related transcription units for which experimental evidence for regulation by antitermination has been obtained. The B. subtilis sacB leader region contains an imperfect 28-base-long inverted repeat called sacR (833). When part of sacR was deleted or when a point mutation was introduced, sacB was expressed constitutively, and sacR was therefore proposed to function as a transcription terminator (31, 806). A sucrose-inducible sacB'-'lacZ fusion also became constitutive when sacR was mutated, which further supported the concept of transcription termination (806). Finally, S1 nuclease mapping revealed that in the absence of sucrose, mainly small transcripts terminating within sacR were formed. The presence of the inducer sucrose did not change the total number of transcripts but increased the number of long transcripts extending beyond the terminator sequence sacR (806).


View this table:
[in this window]
[in a new window]

  		TABLE 4. Description of some well-studied PRD-containing antiterminators

An inverted repeat between the transcription initiation site and the start codon of the bgl operon of E. coli was also detected (505). In the noninduced state, more than 90% of the transcripts, which were initiated at the bgl promoter, were already terminated at the inverted repeat. Expression of a bgl-lacZ fusion containing the inverted repeat required an intact bglG (formerly bglC) gene, which is the first gene of the bgl operon, while bgl-lacZ fusions devoid of the inverted repeat exhibited constitutive, bglG-independent expression (505). Based on these results, it was proposed that the inverted repeat located between the transcription initiation site and the start codon of the E. coli bgl operon serves as a {rho}-independent transcription terminator and that the protein encoded by bglG exerts a positive effect on bgl expression by functioning as an antiterminator. In agreement with this hypothesis, bglG disruption caused drastically lowered expression from the bgl promoter (504). However, a few bglG point mutations, including the one present in the bglC9 strain described previously (682), can lead to the constitutive expression of the bgl operon.

Terminators similar to those preceding B. subtilis sacB and the E. coli bgl operon have been found in front of many other transcription units encoding sugar-specific PTS components of the glucose/sucrose class. Mutation or partial deletion of the terminator sequences preceding the B. subtilis sacPA (24), bglPH (430), or ptsGHI (448) operon also led to the constitutive expression of the corresponding transcription unit and thus confirmed their negative role in mRNA synthesis.

RAT sequences are the binding sites for PTS-regulated antiterminators. Sequencing of the bgl operon allowed the detection of a second terminator located in the intercistronic region between bglG and bglF. Interestingly, the sequences preceding the two bgl terminators are very similar. They were assumed to form weak stem-loops partly overlapping the downstream terminator (785). Insertion of 6 bp in the region upstream from the terminator preceding bglG indeed diminished the induction of a bgl-lacZ fusion (505), suggesting that the region upstream from this terminator is important for antitermination and might represent the binding site for BglG. The conserved sequences preceding the terminators were called ribonucleotidic antiterminator targets (RATs) (32). In in vitro experiments, BglG was indeed able to specifically bind to an RNA probe containing the entire E. coli bgl RAT sequence but only the 5' half of the terminator (344). In contrast, if an RNA probe containing the entire RAT sequence and the complete terminator was used, very weak binding of BglG was observed. The formation of the terminator was therefore assumed to prevent the formation of the BglG binding site, i.e., the RAT stem-loop. In agreement with this concept, BglG binding to the longer RNA probe was improved when a DNA fragment, which was complementary to the 3' part of the inverted repeat and which therefore prevented formation of the terminator, was included in the assay mixtures. These results suggested that there is competition between the formation of the terminator and the RAT stem-loop (344). Under noninducing conditions, BglG cannot bind to its mRNA, and the terminator will therefore be formed. Calculations revealed that the formation of the terminator releases much more free energy than the formation of the RAT stem-loop. Under inducing conditions, the binding of the "activated" BglG is assumed to favor the formation of the RAT stem-loop. BglG thereby prevents the formation of the terminator partly overlapping the RAT sequence and allows the synthesis of long transcripts (344).

In B. subtilis, sequences very similar to those of the E. coli bgl RATs precede the above-mentioned terminator sacR (833) and the terminators located in front of the B. subtilis 1,3-1,4-ß-D-glucan endoglucanase-encoding bglS (formerly licS) gene (585, 785), the sacPA operon (164), which encodes a sucrose-specific EIIBC (sacP) and an endocellular sucrase (sacA) (32), and ptsG (847), which encodes the glucose-specific EIIBCA (Table 4). RAT sequences are also present in front of PRD-containing antiterminator-controlled genes from other organisms (964). An extensive genetic analysis of the RAT sequence preceding the sacB gene confirmed its importance for sacB induction. Expression of sacB'-lacZ fusions containing mutations at positions assumed to be important for the formation of the RAT stem-loop was barely inducible with sucrose (32).

The antiterminators regulating the expression of the four above-mentioned B. subtilis transcription units have been identified. SacY regulates sacB transcription (832), SacT controls the expression of sacPA (164), LicT is necessary for bglS but also for bglPH expression (430), and GlcT controls ptsGHI transcription (847) (Table 4). They all exhibit significant sequence similarity to BglG of E. coli. The new family of antiterminators was called the BglG/SacY family, as the B. subtilis sacB gene and the E. coli bgl operon, which are regulated by SacY and BglG, respectively, were the first transcription units for which this termination/antitermination mechanism was established.

The RAT sequences preceding sacB and sacPA differ in only three positions (Table 4), and consequently, cross talk has been observed: SacY also controls sacPA expression, and SacT regulates the transcription of sacB (32, 146). In gel shift experiments, not only SacT but also SacY could bind to RNA fragments containing the RAT sequence of sacPA (23). In vitro interactions of the RNA binding domains of LicT, SacY, and GlcT with their respective RAT sequences on bglPA, sacB (168), and ptsGHI (448) mRNA have been demonstrated by carrying out surface plasmon resonance experiments. The dissociation constants determined in these experiments varied from 3 µM (SacY) to 10 nM (LicT). The S. carnosus GlcT homolog regulates the expression of glcA and glcB, which encode a glucose- and a ß-glucoside-specific EIICBA, respectively (131). The RAT sequence in the leader region of glcAB of S. carnosus strongly resembles the RAT sequence of B. subtilis ptsGHI. As a consequence, S. carnosus GlcT interacts with the RAT sequence of B. subtilis ptsGHI, as was shown by mutant complementation assays and surface plasmon resonance experiments (413). The high affinity between LicT and its bglPA RAT allowed the determination of the solution structure of the LicT catalytic domain:mRNA complex (964). The antiterminator interacts mainly with the minor groove of the double-stranded RNA formed by two internal loops and the stem in between. In summary, the above-described results make it clear that RAT sequences are the binding sites for BglG/SacY-type antiterminators.

In addition to regulating sacB expression, SacY controls its own synthesis by regulating the expression of the sacS locus containing the two genes sacX and sacY (993). The sacX gene encodes an EIIBC with strong similarity to sucrose-specific EIIs. It is preceded by a RAT sequence and a terminator, which are separated by about 100 bp, whereas in all other transcription units regulated by BglG/SacY-type antiterminators, the RAT sequence and the terminator are overlapping or located next to each other. Nevertheless, both regulatory elements are functional, as mutations within the RAT sequence prevent the induction of sacXY expression, while a partial or complete deletion of the terminator sequence causes constitutive expression (885). It was therefore proposed that the formation of the RAT stem-loop preceding sacXY is stabilized by binding SacY, which in turn prevents the formation of the distant terminator via not-further-specified long-range effects (885).

Phosphorylation of BglG/SacY-type antiterminators by PTS proteins. After RAT sequences were identified as target sites for proteins of the BglG/SacY family, the signal controlling the affinity of the antiterminators for the RAT elements remained to be determined. The early studies on SacY-regulated sacB and BglG-controlled bgl operon expression had already suggested that these antiterminators might be controlled by the PTS. In the late 1970s, it was discovered that the inactivation of B. subtilis EI causes "constitutive" expression of sacB (i.e., synthesis of full-length mRNA) (264, 635). In fact, it was this property of the PTS that was used to successfully clone the B. subtilis ptsHI operon (283). A promoterless kanamycin resistance gene put under the control of the sacB promoter conferred sucrose-inducible kanamycin resistance after integration into the B. subtilis genome. After transposon mutagenesis was carried out, a strain expressing the kanamycin resistance gene in the absence of sucrose was isolated. The mutant had the transposon inserted into the ptsHI operon (283), thus confirming that ptsHI inactivation leads to "constitutive" expression from the sacB promoter.

Another study from the same year showed that the inactivation of E. coli bglF, the gene located downstream from bglG, caused "constitutive" expression of the bgl operon (i.e., synthesis of full-length mRNA) (504). Because bglF encodes a ß-glucoside-specific EIIBCA (79, 785), it was concluded that BglG/SacY-type antiterminators might be controlled by a phosphorylation cascade formed by the general PTS proteins EI and HPr and those sugar-specific EIIAs and EIIBs, the synthesis of which is regulated by the antiterminator. In vitro phosphorylation of BglG by P~EIIBCABgl was indeed reported (18). However, the regulation of antiterminators by PTS-catalyzed phosphorylation turned out to be very complex. First, EI and HPr were later found to be able to phosphorylate most antiterminators in the absence of EIIA and EIIB. Second, the activity of some antiterminators depends on functional EI and HPr, and the inactivation of ptsI therefore does not lead to the "constitutive" expression of the genes controlled by these antiterminators but, on the contrary, lowered or prevented their expression. Third, BglG/SacY-type antiterminators are usually phosphorylated at more than one of the four conserved histidyl residues in their PRDs. Extensive genetic, biochemical, and structural studies carried out mainly with the four B. subtilis antiterminators were necessary to understand in detail how BglG/SacY-type antiterminators are regulated.

In principle, two types of phosphorylation with antagonistic effects on the antiterminator activity have to be distinguished. P~EIIB-requiring phosphorylation in PRD1 inactivates the antiterminator (site of negative regulation). The absence of phosphorylation in PRD1 leads to full-length expression of the corresponding transcription units. By contrast, phosphorylation by EI and HPr in PRD2 stimulates the activity of antiterminators (site of positive regulation). The absence of phosphorylation in PRD2 leads to CCR of the corresponding genes and operons via a CcpA-independent mechanism (Fig. 8).

Antiterminator-dependent induction is regulated via P~EIIBs. "Constitutive" expression from the promoter of an antiterminator-controlled operon following inactivation of the related EII has been observed not only for E. coli bglF (504) but also for several B. subtilis transcription units as well as for the lac operon of L. casei (289). In B. subtilis, disruption of sacX (147), bglP (430, 454), or ptsG (847) leads to inducer-independent expression of sacB, bglPH, and ptsGHI, respectively. As mentioned previously, B. subtilis ptsI mutants exhibit "constitutive" sacB transcription (264, 283, 635), and the deletion of ptsH was reported to lead to the synthesis of full-length ptsGHI mRNA (847). It therefore appeared that a functional phosphorylation cascade composed of the general PTS proteins EI and HPr and sugar-specific EIIAs and EIIBs leads to the inactivation of the corresponding antiterminator.

Interestingly, SacX, the inactivation of which also results in "constitutive" sacB expression (147), contains EIIB and EIIC but no EIIA domain, and it was therefore concluded that EIIB, the last link in the PTS phosphorylation chain (Fig. 1), regulates the induction of antiterminator-controlled transcription units. Experiments with the L. casei lac operon confirmed the importance of P~EIIB for antiterminator-mediated induction, as the inactivation of lacE as well as lacF (encoding EIICBLac and EIIALac, respectively) led to the synthesis of full-length lacTEGF mRNA (289). A different mechanism had originally been proposed for the regulation of E. coli BglG, which was based on the erroneous assumption that His-306 of E. coli EIIBCABgl would be the phosphorylation site in the EIIB domain (784). The replacement of His-306 with nonphosphorylatable amino acids prevents ß-glucoside transport but, in contrast to bglF disruption, does not lead to the synthesis of full-length bgl mRNA. It was therefore proposed that EIIA and not EIIB of EIIBCABgl would regulate BglG activity, and EIIAGlc was also suggested to phosphorylate and inhibit BglG (783). This concept proved to be wrong when the phosphorylation site in the EIIB domain of E. coli EIICBGlc was identified as being Cys-421 (547), which is equivalent to Cys-24 in EIIBCABgl. A strain containing a bglF(Cys24Ser) mutation not only was unable to transport ß-glucosides but also expressed a bgl-lacZ fusion in the absence of an inducer (ß-glucoside) (120), confirming that E. coli BglG is also regulated via the EIIB domain of the corresponding PTS permease.

P~EIIBs are necessary for the phosphorylation of a histidine in PRD1. To determine which of the usually four conserved histidyl residues in the PRDs would be the target for P~EIIB-mediated regulation, extensive genetic analyses were carried out by replacing the histidyl residues in the B. subtilis antiterminators SacY and LicT with various amino acids and by measuring the effect of the mutations on reporter gene fusions in different genetic backgrounds. Replacement of His-99 in PRD1 of SacY with a Tyr (883) or Ala (884) causes constitutive expression of the sacB gene. In contrast, replacing the three other conserved histidyl residues in SacY with Tyr has little effect on the antitermination activity (883). Similarly, replacement of His-100 in B. subtilis LicT (884) or His-101 in L. casei LacT (288), the equivalents of His-99 of SacY, with alanine leads to the synthesis of full-length bglPH or lacTEGF mRNA, respectively. These results strongly suggest that phosphorylation of the first conserved histidyl residue in PRD1 by the corresponding P~EIIB prevents induction by PRD-containing antiterminators. Attempts to mimic phosphorylated antiterminators by replacing the first phosphorylatable histidine in PRD1 with negatively charged Asp or Glu were only partly successful. Putting a negative charge at position 99 by replacing His-99 with Glu or Asp indeed inactivates SacY and abolishes sacB induction (884). Similarly, the His101Asp replacement in L. casei LacT prevents lac operon expression (288). However, although His100Asp LicT exhibited significantly lower activity than His100Ala LicT, it was not completely inactive (884), indicating that in LicT, the His100Asp replacement is not completely phosphomimetic. Moreover, replacement of His-104 with an Asp in B. subtilis GlcT even led to strongly "constitutive" expression of a ptsG'-'lacZ fusion (35), suggesting that the mutant protein does not at all resemble GlcT phosphorylated at the first conserved histidine in its PRD1.

The original concept for the regulation of BglG/SacY-type antiterminators implied that P~EIIBs directly phosphorylate and thereby inactivate their corresponding antiterminator (18). The finding that the first conserved histidine in PRD1 of several antiterminators is phosphorylated in vitro by P~His-HPr therefore came as a surprise (413, 781, 883, 884). For instance, His-99, the site of negative regulation in B. subtilis SacY, is efficiently phosphorylated in vitro by EI and HPr in the absence of SacX (883). Because P~His-HPr is also necessary for the phosphorylation of EIIBs, it was not clear whether P~EIIBs directly phosphorylate their corresponding antiterminators or merely stimulate the P~His-HPr-mediated phosphorylation. It proved difficult to distinguish between the two possibilities, as it was not easy to obtain P~EIIB preparations that were free of HPr (EIIBs are often fused to their membrane-integrated EIICs, and HPr usually sticks to membranes). Because the inactivation of EI as well as SacX leads to constitutive expression of the SacY-controlled sacB gene, it was proposed that in the cells, efficient P~His-HPr-mediated phosphorylation of SacY at His-99 would require the presence of P~SacX. GlcT of S. carnosus also becomes phosphorylated in vitro by EI and HPr in the absence of an EIIB, and mass spectrometry with proteolytic fragments obtained from phosphorylated and dephospho-GlcT revealed that phosphorylation occurs primarily at His-105 (413), the equivalent of His-99 in SacY. In the case of S. carnosus GlcT, phosphorylation in PRD1 was assumed to stimulate the antitermination activity. In contrast, in experiments with GlcT from B. subtilis, P~His-HPr barely phosphorylated PRD1 but mainly transferred its phosphoryl group to His-210 in PRD2 (781) (for PRD2 phosphorylation, see the next section). In fact, the first conserved histidine in PRD1 of B. subtilis GlcT has been shown to become phosphorylated by P~EIIBGlc. When the soluble B. subtilis EIIBAGlc domains were synthesized without the membrane-integrated EIICGlc (to which they are normally fused), HPr-free P~EIIBAGlc could be prepared, which could phosphorylate wild-type GlcT but not GlcT(His104Ala). The antiterminator became phosphorylated in PRD2 only when HPr was added to the phosphorylation mixture containing P~EIIBAGlc and GlcT(His104Ala), thus confirming the nearly complete absence of HPr from the P~EIIBAGlc preparation (781). It therefore seems that phosphorylation of some antiterminators (B. subtilis SacY and S. carnosus GlcT) at PRD1 is catalyzed by P~His-HPr, and this reaction is stimulated by the corresponding P~EIIB, while phosphorylation in PRD1 of other antiterminators (GlcT of B. subtilis) occurs via the corresponding P~EIIB.

Based on genetic studies with licT, it was suggested that P~EIIB domains might also sequester their corresponding antiterminator (884). A similar mechanism is operative for the E. coli transcription regulator Mlc, which is sequestered by the unphosphorylated EIIB domain of EIICBGlc (458, 588, 856) (see "REGULATION OF CARBON METABOLISM IN GRAM-NEGATIVE ENTERIC BACTERIA"). Preliminary evidence for the formation of a complex between an antiterminator and an EII was obtained for S. carnosus GlcT, which exists as an active dimer and an inactive monomer. Dimerization is favored by P~His-HPr-mediated phosphorylation at His-105. Membrane fragments containing EIICBAGlc (and possibly P~EIICBAGlc) specifically interacted only with the dimers in a GlcT monomer/dimer mixture (413). Calorimetric studies had established that the binding of sugar molecules to their respective EIIs, which leads to the dephosphorylation of the EIIB domain, causes extensive structural changes (546). In the case of S. carnosus EIICBAGlc, the structural changes accompanying the binding of glucose were assumed to lead to a release of active P~GlcT dimers, thus allowing the expression of the glcAB operon in response to the presence of a substrate.

Unlike antiterminators from B. subtilis and L. casei, the negative regulation site of BglG from E. coli was originally thought to be located in PRD2, and His-208 was reported to be phosphorylated by P~EIIBCABgl (121). However, the phosphorylation mixtures for BglG phosphorylation contained not only P~EIIBCABgl but also EI and HPr. Like most antiterminators, BglG is phosphorylated by EI and HPr in PRD2 (287) (see the next section), and Chen et al. (121) probably misinterpreted the P~His-HPr-dependent modification as being P~EIIBCABgl-dependent phosphorylation. In fact, the negative regulation site of E. coli BglG has been identified as His-101, and phosphorylation at this amino acid requires the presence of P~EIIBCABgl (285).

In summary, it can be concluded that induction via most BglG/SacY-type antiterminators is regulated by P~EIIB-controlled or -mediated phosphorylation of the first conserved histidyl residue in PRD1, which inhibits the activity of the antiterminator. In the presence of the corresponding inducer, i.e., the carbohydrate transported by (or, in the case of B. subtilis, SacX probably bound to) the EII regulating the antiterminator, the antiterminator will barely be phosphorylated at its first conserved histidine in PRD1, because phosphorylation of the inducer is probably faster than phosphorylation of the antiterminator (482) or rephosphorylation of the EIIB by its corresponding P~EIIA (Fig. 8). For B. subtilis SacY, the presence of sucrose in the growth medium indeed significantly reduced the extent of its in vivo phosphorylation (359). Antiterminator molecules that are not phosphorylated at the first conserved histidine in PRD1 are able to bind to their RAT sequence, thus preventing the formation of the terminator and allowing the synthesis of full-length mRNA of their target gene or operon. If the inducer is absent, P~EIIB will be formed, which leads to phosphorylation of the first histidine in PRD1 and/or sequestration of the corresponding antiterminator. Both effects prevent the binding of the antiterminator to its RAT sequence and therefore the induction of the corresponding gene or operon. According to genetic analyses with licT of B. subtilis (884) and bglG of E. coli (285), phosphorylation at the second conserved histidine in PRD1 plays a detectable but minor role in the induction process. By contrast, replacing either His-101 or His-159 in L. casei LacT with alanine leads to "constitutive" expression of the lac operon, suggesting that in this antiterminator, the two conserved histidines in PRD1 contribute equally to the induction process (288).

Some antiterminators need to be activated by phosphorylation in PRD2. Based on the effect of ptsI or ptsH mutations on the activity of BglG/SacY-type antiterminators, two classes of transcription attenuators can be distinguished. The first class includes B. subtilis SacY and GlcT. Mutations in ptsI or ptsH result in the "constitutive" expression of their target genes sacB (147, 264, 635) and ptsGHI (847), respectively. In these mutants, neither P~His-HPr nor P~EIIB is formed, which prevents the inactivating phosphorylation at PRD1 of the antiterminator in the absence of an inducer. The second class of antiterminators includes B. subtilis SacT and LicT and E. coli BglG. The synthesis of the full-length mRNA of their target genes sacPA (24) and bglPH in B. subtilis (429, 430, 454, 482) and bglGFB in E. coli (287) is almost completely prevented when ptsH or ptsI is inactivated. These results were surprising, as we had learned in the previous section that the inactivation of, for example, the EIIBCABgl-encoding bglP leads to the synthesis of full-length mRNA from the LicT-controlled bglPH promoter (430, 454). Because EI and HPr are essential for the phosphorylation of EIIBCABgl, which in turn inactivates LicT by phosphorylating it at His-100 in PRD1, the disruption of ptsH or ptsI should also lead to the synthesis of full-length bglPH mRNA. The observed dependence of LicT, SacT, and BglG activity on functional EI and HPr therefore suggested that the second class of antiterminators is subject to dual control by PTS proteins: positive regulation by P~His-HPr and negative control by their corresponding P~EIIBs. Because the specific replacement of His-15, the site of EI-catalyzed phosphorylation in HPr, with an alanine also prevented LicT activity (429), it was likely that EI and HPr exert their positive effect by phosphorylating this class of antiterminators. Indeed, SacT (23), LicT (482), and BglG (287) are phosphorylated by PEP, EI, and HPr. In LicT, all four conserved histidines are phosphorylated by the general PTS proteins, with the main site of phosphorylation being His-207 in PRD2 (482, 884). Because the P~EIIB-dependent negative effect is mediated via the first conserved His in PRD1 (see the previous section), it was likely that the positive effect of phosphorylation by EI and HPr is mediated via PRD2. Experiments with SacY/LicT hybrid proteins confirmed this assumption. When PRD2 of the EI/HPr-independent SacY was replaced with PRD2 of LicT, the resulting hybrid protein was EI and HPr dependent (884). This concept was further supported by an extensive genetic analysis, during which the conserved histidines in PRD2 of B. subtilis LicT were replaced with various amino acids and the effect of the mutations on bglP'-lacZ expression was measured in different pts genetic backgrounds. Replacement of His-207 or His-269 or both conserved histidines in PRD2 with Ala led to a complete loss of LicT antitermination activity, i.e., prevented the synthesis of full-length mRNA from the bglPH promoter even in the presence of its inducer salicin or in a bglP background. In contrast, in a bglP mutant, if either one of these two histidines or both were replaced with an Asp, which probably mimics a phosphorylated histidine, expression from the bglPH promoter remained "constitutive" irrespective of whether ptsHI were functional or deleted (884). These results demonstrate that PEP-, EI-, and HPr-mediated phosphorylation of PRD2 in antiterminators of the second class stimulates their activity. This is also true for E. coli BglG, where His-208, the only conserved phosphorylatable histidine in PRD2, was first claimed to be the site of negative regulation by P~EIIBCABgl (121) but was later identified as the site of positive regulation by P~His-HPr (285, 287). Interestingly, E. coli FruB, a fusion protein composed of HPr- and EIIAFru-like domains (267), can replace HPr in BglG phosphorylation and activation (287).

CCR mediated by dephosphorylation of PRD2 of antiterminators. Surprisingly, although the activity of B. subtilis SacY and GlcT does not depend on EI and HPr, they are phosphorylated by PEP, EI, and HPr in their PRD2 (781, 883). It seems that during the course of evolution, SacY and GlcT developed a PTS-independent antitermination activity but retained the ability to become phosphorylated in PRD2. On the contrary, GlcT of S. carnosus, which is probably also EI/HPr independent, is not phosphorylated in its PRD2, as only one major P~His-HPr-dependent phosphorylation site, His-105 in PRD1, could be detected (413). Interestingly, EI/HPr-independent antiterminators seem to regulate the expression of transcription units that are insensitive to CCR, whereas the EI/HPr-dependent LicT and SacT control operons that are preceded by cre's recognized by the P-Ser-HPr:CcpA complex (429). It was therefore assumed that phosphorylation of LicT and SacT by PEP, EI, and HPr might serve as an additional, CcpA-independent CCR mechanism.

The first evidence for antiterminator-mediated CCR came from studies with a B. subtilis ccpA mutant. In this mutant, expression of a bglP'-lacZ fusion was not completely relieved from CCR. The residual CCR (about 20%) disappeared when the terminator preceding the bglPH operon was deleted, suggesting that bglPH is regulated by two CCR mechanisms: one that is CcpA dependent and one that is terminator dependent (429). Interestingly, the residual CCR also disappeared when a ptsH1 mutation (which prevents the formation of P-Ser-HPr) was introduced into the ccpA mutant. It was therefore assumed that the CcpA-independent CCR mechanism operative for the bglPH operon was based on diminished phosphorylation in PRD2 of LicT due to the formation of P-Ser-HPr (a poor substrate for PEP-dependent phosphorylation by EI) during growth on a rapidly metabolizable carbon source. This assumption was supported by studies with B. subtilis mutants producing LicTs in which either one of the two conserved histidines in PRD2 was replaced with Asp. Expression of a bglP'-lacZ fusion remained inducible in both mutants, but ß-galactosidase activity was less severely repressed by glucose than in a strain producing wild-type LicT (909). Unequivocal proof for the antiterminator-mediated CCR mechanism came from studies with strains producing EI/HPr-independent mutant LicTs, which were called LicT(Pia) (Pia stands for PTS-independent antitermination). Starting from a strain carrying a bglPH'-aphA3 (aphA3 confers kanamycin resistance to bacteria) and a bglP'-lacZ fusion as well as the LicT-inactivating ptsGHI deletion, several spontaneous mutants that had gained the ability to express the LicT-dependent bgl fusions despite the absence of functional EI and HPr were isolated (483). The isolated mutants were all affected in licT, and the various mutations apparently render LicT activity independent of functional EI and HPr. The licT(Pia) mutants can be divided into two classes: mutations affecting PRD1 (they lead to constitutive expression from the bglPH promoter) and mutations affecting PRD2 (they remain inducible). The latter type of mutant LicTs resembles the naturally EI/HPr-independent SacY and GlcT. Interestingly, the residual CCR of the bglPH operon detectable in ccpA mutants disappeared when wild-type licT was replaced with one of the licT(Pia) alleles (483), thus confirming that it is the poor phosphorylation of PRD2 in LicT during the uptake of rapidly metabolizable carbon sources that triggers CcpA-independent CCR.

Based on the results obtained with the various LicT(Pia) proteins, distinct amino acids were proposed to be important for the transfer of the activating signal from PRD2 via PRD1 to the RNA binding domain (483), which was called CAT (coantiterminator) (964). When the crystal structures of the two PRDs of the constitutively active His207Asp, His269Asp mutant LicT (909) and of inactive wild-type LicT (292) were solved, a more detailed understanding of the structural changes involved in LicT activation was possible. In fact, the structures of the PRD2 of the two crystallized LicT forms are completely different, while the amino acid backbones of their PRD1s are almost identical. In inactive wild-type LicT, the two PRD2s of the LicT dimer make no contact to each other but are in contact with the two PRD1s. The four phosphorylatable histidines in the PRD2s are exposed to the surface and easily accessible for phosphorylation. In the activated His207Asp, His269Asp mutant LicT, a 180° swing movement of the C-terminal domain leads to extensive structural changes, placing the aspartyl residues (which replace the four histidines) in the center of PRD2, thus rendering them inaccessible from the outside. Two major rotations are responsible for the rearrangement of the PRD2s. Both take place in the interdomain region between PRD1 and PRD2, with one occurring at residues Leu-165 and Asn-166 and the other occurring at Met-169. As a consequence, the two PRD2s of His207Asp, His269Asp mutant LicT are in close contact, whereas most interactions with PRD1 are suspended. These alterations are probably responsible for the changes in the interdomain region connecting PRD1 and CAT, which in turn are thought to cause rearrangements of the CAT domain, ultimately leading to the activation of LicT (292). The monomer/dimer transition does not seem to be important for LicT activation, as the RNA binding domain CAT and active His207Asp, His269Asp mutant LicT as well as inactive dephospho-LicT form dimers (964). Nevertheless, the monomer/dimer transition might be important for the regulation of other antiterminators (56).

The structure of the RNA binding domain of the two antiterminators SacY and LicT has been intensively studied by NMR (510) and X-ray crystallography (910). The solution structure of the RNA binding domain of LicT (first 55 amino acids) attached to its mRNA has also been solved (964). However, the structure of an entire BglG/SacY-type antiterminator has so far not been determined.

Based on the above-described results, the following model for CCR mediated by EI/HPr-dependent antiterminators can be proposed. In the presence of a carbon source rapidly transported via the PTS, the antiterminators will have to compete with the corresponding sugar-specific EIIA for the common phosphoryl donor P~His-HPr. P~His-HPr probably transfers its phosphoryl group primarily to EIIAs, thus leaving the antiterminators in the unphosphorylated, less active state. In in vitro experiments, P~His-HPr indeed phosphorylates EIIAs much faster than antiterminators (482). In addition, the ATP-dependent phosphorylation of HPr at Ser-46 in low-G+C gram-positive bacteria, which is triggered by the presence of rapidly metabolizable carbon sources, drastically slows the PEP-dependent phosphorylation of HPr at His-15 and therefore further diminishes the activation of LicT by P~His-HPr-dependent phosphorylation. The proposed complex mechanism of induction and CCR of operons controlled by EI/HPr-dependent antiterminators is outlined in Fig. 8. The model requires that during PTS sugar uptake, the corresponding EIIB domain is present mainly in the dephospho form and therefore does not phosphorylate the first conserved His in PRD1 of LicT. On the other hand, P~His-HPr must be present in sufficient amounts to allow the phosphorylation in PRD2, which implies that PTS operons, the expression of which is regulated by an EI- and HPr-dependent antiterminator, encode low-capacity transport systems, as strong autorepression would otherwise occur. Probably in order to prevent autorepression, the operons for the B. subtilis high-capacity glucose and sucrose PTS are regulated by EI- and HPr-independent antiterminators.

Distribution of BglG/SacY-type antiterminators in bacteria. In a several-years-old review on PRD-containing transcription regulators, the BglG/SacY antiterminator family was reported to contain 13 members (844). In addition to the above-mentioned antiterminators, it included LicT of Bacillus amyloliquefaciens, AbgG of Clostridium longisporum (87), ArbG of Pectobacterium chrysanthemi (formerly Erwinia chrysanthemi) (208), BglR of L. lactis (42), CasR of Klebsiella oxytoca (446), and SurT of G. stearothermophilus (473). Owing to the many genome sequences completed in the meantime, a simple BLAST search yields numerous new potential family members. A few of these novel PRD-containing antiterminators have been studied in more detail. BvrA of L. monocytogenes controls the expression of the bvrABC operon encoding the antiterminator, a ß-glucoside-specific EIIBCA, and a presumed ADP-ribosylglycohydrolase (81), respectively. In L. monocytogenes, repression of several PrfA-controlled virulence genes by the ß-glucosides cellobiose and salicin requires an intact bvr locus, while repression by arbutin, glucose, or fructose persists in a bvrAB mutant. BglG of L. plantarum was reported to control the expression of the bglGPT operon encoding the antiterminator BglG, a ß-glucoside-specific EII, and a 6-P-ß-glucosidase (513), while LicT of S. mutans regulates an esculin-specific PTS (141). C. acetobutylicum ScrT was proposed to control the expression of the scrAKB operon encoding a sucrose-specific EII, a fructokinase, and a sucrose-6-P hydrolase, respectively (860). In the actinobacterium Corynebacterium diphtheriae, a BglG/SacY-type antiterminator-encoding gene is located downstream from the EIIBCAGlc-encoding ptsG gene (631). BglT of Pectobacterium carotovorum probably controls the expression of an operon that encodes a ß-glucoside-specific PTS and strongly resembles the arb operon of P. chrysanthemi (19). BglG/SacY-type antiterminators are especially abundant in clostridia. With nine BglG/SacY homologs, Clostridium difficile seems to be the organism that contains the highest number of PRD-containing antiterminators, followed by L. plantarum, which contains five such proteins. In fact, BglG/SacY-type antiterminators are present in most firmicutes including bacilli, clostridia, enterococci, lactobacilli, lactococci, listeriae, leuconostocs, staphylococci, streptococci, etc. BglG/SacY-type antiterminators occur less frequently in gram-negative bacteria. So far, in addition to E. coli, PRD-containing antiterminators have been detected in proteobacteria such as P. carotovorum (BglT) (19), P. chrysanthemi (ArbG) (208), Klebsiella oxytoca (CasR) (446), Klebsiella aerogenes (BglG) (690), Shigella flexneri, Shigella sonnei, Shigella boydii, Photorhabdus luminescens, and Yersinia enterocolitica. Interestingly, a gene coding for a BglG/SacY-type antiterminator is also present in the pathogenicity island of a uropathogenic E. coli strain (