Trigger enzymes: bifunctional proteins active in metabolism and in controlling gene expression


*E-mail; Tel. (+49) 551 393781; Fax (+49) 551 393808.


All regulatory processes require components that sense the environmental or metabolic conditions of the cell, and sophisticated sensory proteins have been studied in great detail. During the last few years, it turned out that enzymes can control gene expression in response to the availability of their substrates. Here, we review four different mechanisms by which these enzymes interfere with regulation in bacteria. First, some enzymes have acquired a DNA-binding domain and act as direct transcription repressors by binding DNA in the absence of their substrates. A second class is represented by aconitase, which can bind iron responsive elements in the absence of iron to control the expression of genes involved in iron homoeostasis. The third class of these enzymes is sugar permeases of the phosphotransferase system that control the activity of transcription regulators by phosphorylating them in the absence of the specific substrate. Finally, a fourth class of regulatory enzymes controls the activity of transcription factors by inhibitory protein–protein interactions. We suggest that the enzymes that are active in the control of gene expression should be designated as trigger enzymes. An analysis of the occurrence of trigger enzymes suggests that the duplication and subsequent functional specialization is a major pattern in their evolution.


There is only one constant in the life of any organism: change. To be successful in competition with their fellows, organisms need to detect these changes and respond to them rapidly and appropriately while preventing the waste of energy and resources. In bacteria, the main level at which these regulatory events occur, is transcription. Therefore, several mechanisms allowing a tight control of transcription have evolved.

In each case, regulatory events require two tasks, the perception of the environmental or internal signal, and its transduction to the transcription machinery. Many transcription factors integrate signal perception and the regulatory output, i.e. transcription regulation, in a single molecule, among them the Lac repressor or the cAMP receptor protein from Escherichia coli. In many other cases, signal recognition and the regulatory output are embodied in two distinct proteins, subunits or domains that interact with each other for signal transduction. The paradigm of this type of regulation are the two-component regulatory systems composed of a sensory protein that can autophosphorylate under certain conditions and act as a kinase to modulate the activity of transcription factors (Mascher et al., 2006). Similarly, the activity of many sigma factors of the bacterial RNA polymerase is controlled by interactions with anti-sigma factors, which may in turn be controlled by anti-anti-sigma factors (Helmann, 1999). A conserved module that involves a protein kinase and a phosphatase might even serve to control different output modules in many bacteria, including two-component systems and sigma/anti-sigma factors (Pané-Farréet al., 2005). The separation of the two tasks in signal transduction is very advantageous in evolution as it allows shifts in specificity by gene duplication and specialization. Indeed, transcription factors are still evolving (de Lorenzo and Pérez-Martin, 1996; Galvão and de Lorenzo, 2006).

There are proteins in the cell that are well informed about specific conditions, usually about metabolite availabilities. These proteins are the enzymes that recognize their substrates and that often undergo structural alterations upon interaction with the substrate or during their reaction. This makes the enzymes perfect candidates to share their knowledge with transcription regulators, or to regulate gene expression themselves. Moreover, these enzymes are an excellent substrate for the evolution of sensory components of signal transduction systems. Such proteins that exert a second, unrelated function in addition to their primary task, are called ‘moonlighting proteins’ (Jeffery, 1999). The most prominent biochemical pathway for which the ‘moonlighting’ of its enzymes can be regarded as a universal principle, is glycolysis. Glycolytic enzymes are involved in the control of DNA replication, in mRNA processing, the regulation of eukaryotic transcription and in apoptosis (Kim and Dang, 2005; Carpousis, 2007; Jannière et al., 2007). A plethora of enzymes involved in transcription regulation from a variety of biochemical pathways has recently been discovered, suggesting that this regulatory principle is much more common than hitherto acknowledged (Wray et al., 2001; Bächler et al., 2005; Tang et al., 2005; Shen et al., 2006; Commichau et al., 2007; Domain et al., 2007; Hullo et al., 2007).

The scope of this review is bacterial enzymes that contribute to the regulation of gene expression. We suggest that these enzymes be collectively designated as trigger enzymes to emphasize their role in signal transduction. These trigger enzymes can exert different functions. They might act as transcription factors by direct binding to either DNA or RNA or they might modulate the activity of transcription factors either by covalent modification or by protein–protein interactions (see Table 1). We will also discuss some ideas concerning the evolution of trigger enzymes from monofunctional enzymes and thereafter to dedicated monofunctional transcription regulators.

Table 1.  A compilation of trigger enzymes in bacteria.
OrganismEnzyme, catalysed reactionEffector/signalTargetReference
  • a. 

    These trigger enzymes are the best-studied representatives of a large class of sugar permeases that control PRD-containing transcription factors. For a comprehensive compilation refer to Stülke et al. (1998) and Deutscher et al. (2006).

  • b. 

    ?, unknown.

DNA-binding (transcription factor)
 E. coli, B. subtilisBirA, biotin-protein ligaseBiotinbir operonRodionov et al. (2002)
 E. coli, S. typhimuriumPutA, bifunctional proline dehydrogenase and Δ1-pyrroline-5-carboxylate dehydrogenaseProlinePromoter region of the put operonOstrovsky de Spicer and Maloy (1993)
 E. coli, S. typhimuriumNadR, nicotinamide mononucleotide adenylyltransferaseNMN/NADH2Promoter region of the NAD+-biosynthetic genesRaffaelli et al. (1999)
 E. coliPepA, aminopeptidasePyrHcarAB operonCharlier et al. (2000)
 B. subtilisCitB, aconitaseIronIron-responsive elementsAlén and Sonenshein (1999)
Covalent modification of a transcription regulatora
 E. coliBglF, β-glucoside permease (EIIB subunit)SalicinBglGChen et al. (2001)
 B. subtilisLevE, fructose permease (EIIB subunit)FructoseLevRMartin-Verstraete et al. (1998)
 B. subtilisPtsG, glucose permease (EIIB subunit)GlucoseGlcTSchmalisch et al. (2003)
Modulation of activity of a DNA-binding protein
 Listeria monocytogenesGmaR, O-GlcNAc transferase?bMogRShen et al. (2006)
 E. coliPtsG, glucose permease (EIIB subunit)GlucoseMlcTanaka et al. (2000)
 E. coliDhaL, DhaK, dihydroxyacetone kinase (nucleotide and substrate binding subunits)DihydroxyacetoneDhaR
DhaK antagonizes
Bächler et al. (2005)
 E. coliMalY, βC-S lyase MalK, ATP binding subunit of the maltose transporter
Aes, esterase
Maltodextrins, maltotrioseMalTJoly et al. (2004)
 E. coliTktA, transketolase?MarRDomain et al. (2007)
 E. coliPyrH, UMP kinasePyrimidinesPepAKholti et al. (1998)
 B. subtilisCysK, cysteine synthetase ACysteine?CymRHullo et al. (2007)
 B. subtilisGlnA, glutamine synthetaseGlutamineTnrAWray et al. (2001)
 B. subtilisRocG, glutamate dehydrogenaseGlutamateGltCCommichau et al. (2007)
Unknown mechanism/regulators that evolved from enzymes
 B. subtilisPyrRUMP/PRPPpyr operonTomchick et al. (1998)
 T. thermophilusMlcGlucoseglc and mal operonsChevance et al. (2006)
 S. pyogenesLacD.1Dihydroxyacetone
Phosphate (?)
RopBLoughman and Caparon (2006; 2007)
 B. subtilisRibR?rib operon (mRNA)Higashitsuji et al. (2007)

Trigger enzymes active as DNA-binding transcription factors

In bacteria, transcription is most commonly controlled by regulator proteins that bind specific DNA sequences in the promoter region. This binding results in either transcription repression or activation. The recognition of a specific sequence requires the presence of a protein domain that is able to recognize and bind a DNA motif. As DNA recognition motifs are usually not present in enzymes, the acquisition of such a domain is a prerequisite for the evolution of an enzyme to a DNA-binding trigger enzyme.

In proteobacteria, a bifunctional enzyme catalyses the two step degradation of proline to glutamate, i.e. the two consecutive oxidations of proline and Δ1-pyrroline-5-carboxylate (see Fig. 1). In E. coli, Salmonella enterica ssp. Typhimurium and other enterobacteria this enzyme, PutA, is also involved in transcription regulation of the divergent putA and putP genes encoding the bifunctional proline utilization protein and the proline transporter respectively (Ostrovsky de Spicer and Maloy, 1993; Ling et al., 1994). This regulation is dependent on the availability of proline.

Figure 1.

Localization of PutA determines its role in proline metabolism. In the presence of proline, the bifunctional enzyme PutA catalyses the two-step reaction from proline to glutamate. This reduced form of PutA (PutAred) is localized to the membrane. The putP and putA genes, encoding the proline transporter PutP and the enzyme PutA, respectively, are expressed in the presence of proline. In the absence of proline, the oxidized PutA protein (PutAox) binds to the intergenic region of the putA and putP genes and represses their transcription. P5C, Δ1-pyrroline-5-carboxylate.

PutA is able to bind directly to multiple sites in the promoter region of the putA and putP genes, thereby causing transcriptional repression in the absence of proline (Ostrovsky de Spicer et al., 1991; Ostrovsky de Spicer and Maloy, 1993; Fig. 1). On the other hand, the enzymatically active form that links reduction of the cofactor FAD to proline oxidation is a peripheral membrane protein (Brown and Wood, 1993; Fig. 1). This enzymatically active, membrane-bound form of PutA is unable to bind DNA and to repress transcription (Ostrovsky de Spicer and Maloy, 1993). Several studies addressed the identity of the molecular inducer that causes release of PutA from the DNA, resulting in transcription of the put genes. In contrast to many other repressors, two factors contribute to induction: proline and FAD must be available. The proline-dependent reduction of the cofactor FAD directs PutA in its reduced form to the membrane and concomitantly causes loss of DNA-binding activity, indicating that the two locations and the two biochemical activities of the protein are mutually exclusive (Zhang et al., 2004). The shuttle of PutA from the DNA to the membrane is accompanied by a conformational change in the protein that is triggered by the availability of proline and FAD (Zhu and Becker, 2003). Variants of PutA capable of repressing the put genes even in the presence of the inducer proline are defective in proline binding and in enzymatic activity. Thus, a single proline binding site is involved both in enzymatic activity and induction of the put genes. This result demonstrates that enzymatic activity is essential for the release of PutA from the DNA and for put gene induction (Muro-Pastor and Maloy, 1995).

Studies on the organization of PutA revealed the presence of two large domains: The N-terminal domain has the proline dehydrogenase (PDH) activity whereas the C-terminal domain harbours the Δ1-pyrroline-5-carboxylate dehydrogenase activity (Ling et al., 1994). The conformational change that occurs in the presence of proline and FAD was mapped to the PDH domain (Zhu and Becker, 2003). The DNA-binding activity of PutA is located in the N-terminal 47 amino acids of the PDH domain that form a ribbon–helix–helix (RHH) motif (Gu et al., 2004). Interestingly, this DNA-binding motif is present in the PutA proteins of enterobacteria and their close relatives but in neither the bifunctional PutA proteins of rhizobia and Pseudomonas aeruginosa nor the separate PDH proteins of Gram-positive bacteria. The DNA-binding RHH domain is different from the well-known helix–turn–helix motif in that it consists of a β-strand followed by two α-helices. The characteristic of the RHH motif is sequence-specific DNA recognition by residues in the β-sheet formed by two RHH subunits. The RHH motif is present in distinct transcription regulators and can be linked to different protein domains, but can also act by itself (Schreiter and Drennan, 2007). The determination of the crystal structure of the PDH domain of PutA (Lee et al., 2003) allowed the analysis of the molecular events that accompany the addition of proline and reduced flavin to PutA. It turns out that the conformational change in the PDH domain occurs around tryptophan 211 and switches the activity of PutA from a DNA-binding protein to a membrane-bound enzyme (Zhu and Becker, 2005).

Similar to PutA, the trigger enzymes BirA and NadR from E. coli possess distinct DNA-binding domains (see Table 1). Interestingly, BirA seems to act as a transcription factor in many bacteria and even in some archaea (Rodionov et al., 2002).

Trigger enzymes engaged in post-transcriptional regulation via protein–RNA interaction

The rather uniform structure of DNA requires DNA-binding enzymes to recognize defined sequence motifs. In contrast, RNAs can adopt a variety of structures, and they can bind and interact with virtually any molecule, including metabolites and proteins (Ellington and Szostak, 1990). Thus, it is not surprising that proteins with very different structures are able to bind RNA and, indeed, many enzymes interact with RNAs and exert regulatory effects by this interaction (for an excellent review see Cieśla, 2006).

There is so far only one example of a bacterial trigger enzyme with RNA-binding activity, i.e. aconitase. Aconitase catalyses the reversible conversion of citrate to isocitrate in the tricarboxylic acid (TCA) cycle and requires an iron-sulphur cluster for activity (Beinert et al., 1996). Under conditions of iron limitation that are frequently encountered by microorganisms, the TCA cycle cannot operate owing to the inactivity of aconitase. However, aconitase can help solve this problem by binding so-called iron-responsive elements (IREs) in the mRNAs of genes involved in iron homeostasis (see Fig. 2). This was first shown for the human enzyme, but later on also for the aconitases from such diverse bacteria as E. coli, Bacillus subtilis, and Mycobacterium tuberculosis (Kaptain et al., 1991; Alén and Sonenshein, 1999; Tang and Guest, 1999; Banerjee et al., 2007).

Figure 2.

The RNA-binding function of the aconitase depends on the iron status of the cell. In the presence of iron, the FeS cluster in the active centre of the aconitase is stable and the enzyme catalyses the isomerization of citrate to isocitrate in the TCA cycle. During iron limitation, the FeS cluster is absent from the active centre of the aconitase and the inactive enzyme gains the ability to bind specific mRNAs structures designated as iron-responsive elements. This binding results in the stabilization or destabilization of the bound mRNAs.

The determination of the structure of IRE-bound aconitase allowed us to understand the molecular basis for the two mutually exclusive activities of aconitase. In the presence of iron, the protein has a compact conformation with the iron-sulphur cluster as a ligand. In the absence of iron or under conditions of oxidative stress, the iron-sulphur cluster disassembles and the free (apo-)aconitase adopts a more open conformation with respect to two domains that are located outside of the core of the protein. This opening allows the binding of the IRE, with sequence-specific contacts of a C in the central bulge of the IRE and an AGU triplet in the terminal loop. These contacts involve the two different domains of apo-aconitase. It is worth noting that only a few bases are conserved in the IREs, and that these conserved bases are brought into the right position by secondary structure elements (Walden et al., 2006).

Based on the overall sequence similarity of the eukaryotic and B. subtilis aconitases and IREs (Alén and Sonenshein, 1999), it seems safe to assume that the B. subtilis enzyme follows the same mechanism as outlined above for the mammalian enzyme. B. subtilis aconitase binds IREs in the untranslated regions of the qoxD and feuAB mRNAs. These genes encode the iron-containing protein cytochrome aa3 oxidase and an iron uptake system respectively (Alén and Sonenshein, 1999). It has long been known that B. subtilis mutants lacking aconitase are defective in sporulation. This was attributed to the reduced accumulation of the gerE mRNA encoding a transcription regulator of late sporulation genes. Indeed, a secondary structure similar to the IRE can be formed by the untranslated region of the gerE mRNA, and the aconitase binds this sequence. Thus, aconitase might affect sporulation by stabilizing the mRNA of the GerE transcription factor (Serio et al., 2006).

Escherichia coli possesses two aconitases, of which aconitase B (AcnB) is the main enzyme involved in the TCA cycle. In contrast to the monomeric enzymes of eukaryotes and B. subtilis, AcnB has a dimerization domain. An analysis of the requirements of AcnB for RNA binding revealed that it is independent of the iron-sulphur cluster in the active centre of the enzyme. Moreover, the arrangement of the RNA in the protein seems to be different from that observed in the mammalian aconitase–IRE complex (Tang et al., 2005). This is in good agreement with the observation that the RNA elements recognized by E. coli aconitase B differ from the classical IREs (Tang and Guest, 1999).

There is a large variety of potential RNA structures that might be able to bind diverse ligands and, indeed, many eukaryotic enzymes have been shown to moonlight in RNA interactions. Thus, it would not be surprising if many bacterial enzymes turn out to be trigger enzymes involved in protein–RNA interactions. This might lead to another exciting phase of studying riboregulation after the discovery of riboswitches and small regulatory RNAs.

Trigger enzymes controlling gene expression by signal-dependent phosphorylation of transcription regulators

In many bacteria, sugars are transported by the phosphoenolpyruvate phosphotransferase system (PTS). This system transfers phosphoryl groups from phosphoenolpyruvate via enzyme I, HPr, and a sugar-specific permease to the incoming sugar. The PTS permeases are composed of three (or four) domains, with the soluble domains IIA and IIB being part of the phosphorylation chain, whereas the membrane-bound domain IIC transports the carbohydrate (Deutscher et al., 2006). However, the PTS permeases are not only involved in sugar transport; they might also control the activity of transcription activators and antiterminators by phosphorylating them in response to the availability of the respective substrate. These transcription regulators all share a duplicated so-called PTS regulation domain (PRD, Stülke et al., 1998). Phosphorylation of one of these domains by the sugar-specific permease occurs in the absence of the substrate (upon accumulation of phosphorylated permease) and results in the inactivation of the regulators. In the presence of the substrate, the phosphate groups are drained to the sugar and the regulators become dephosphorylated and, thus, regain activity (see Fig. 3). Thus, this phosphorylation event provides a link between substrate availability and the expression of the target gene(s) by controlling the activity of the transcription factors (Stülke et al., 1998).

Figure 3.

The glucose permease PtsG controls the activity of the transcription antiterminator protein GlcT in response to glucose availability. In the presence of glucose, the sugar is taken up by the glucose permease of the PTS (PtsG) and concomitantly phosphorylated. The phosphoryl group is derived from PEP and transferred via the phosphocarriers enzyme I (EI) and HPr to the EIIB component of the glucose permease. Under these conditions, the transcription antiterminator protein GlcT binds a stem-loop structure of the ptsGHI mRNA, thereby preventing the formation of a terminator structure, and the transcription of the ptsGHI mRNA can be accomplished. Inactivation of the antiterminator protein GlcT occurs in the absence of glucose. GlcT receives a phosphoryl group from the glucose permease and is unable to bind the ptsGHI mRNA. The formation of a termination structure occurs and the transcription of the ptsGHI operon is aborted.

This model was first proposed for the regulation of β-glucoside transport in E. coli by the β-glucoside permease BglF (Amster-Choder and Wright, 1990). All activities of BglF are catalysed by the same active site in the IIB domain. The presence of the substrate triggers the formation of a disulphide bond between the active site cysteine and a residue in the membrane-bound IIC domain. It is this form of BglF that can dephosphorylate the corresponding antiterminator protein BglG and phosphorylate the substrate (Chen et al., 2001). Recently, mutations in bglF that interfere with BglG dephosphorylation but not with sugar phosphorylation led to the identification of a membrane-embedded motif in BglF that is important for the dephosphorylation activity (Monderer-Rothkoff and Amster-Choder, 2007).

Unfortunately, there are conflicting reports on the interaction between BglF and the BglG antiterminator. While Görke (2003) provides compelling genetic evidence that the first of the two reiterated PRDs in BglG is phosphorylated by BglF, Chen et al. (1997) report that BglF phosphorylates the second PRD. This problem should be resolved by the in vitro analysis of BglG phosphorylation using purified BglF as the phosphate donor. In all published experiments there was HPr present in addition to BglF, and HPr by itself can phosphorylate BglG, thus masking any BglF-dependent phosphorylation of BglG. However, similar experiments have been performed with a homologue of BglG, the B. subtilis transcriptional antiterminator GlcT that controls glucose transport. This protein is phosphorylated by the B domain of the glucose permease of the PTS and by HPr on its first and second PRD respectively (Schmalisch et al., 2003; see Fig. 3). This finding is in excellent agreement with the genetic data obtained for all antiterminators controlled by phosphorylation of their PRDs (Deutscher et al., 2006). The regulatory phosphorylation of a PRD by a IIB domain was also shown for a transcription activator. In this case, the B. subtilis LevE subunit of the fructose permease inactivates the transcription activator LevR (Martin-Verstraete et al., 1998).

Regulation of PRD-containing antiterminators by sugar permeases might be important beyond the control of sugar transport. Recent studies revealed that a PRD-containing regulator might be involved in regulating the virulence of Bacillus anthracis, the causative agent of anthrax (Tsvetanova et al., 2007).

Trigger enzymes controlling the activity of transcription factors by protein–protein interactions

The fourth class of trigger enzymes controls gene expression by modulating the activity of transcription factors – either activators or repressors. The diversity of such interactions reflects the diversity of transcription factors and makes it difficult to predict these trigger enzymes from the primary sequence.

As shown in Table 1, there are many regulatory interactions between enzymes and transcription factors, and their detailed presentation would be far beyond the scope of this review. Instead, we will concentrate on interactions for which sufficient information is available. These are the relocation and inactivation of the Mlc repressor from E. coli by the glucose-specific enzyme II of the PTS, and the inactivation of the master regulator of nitrogen metabolism in B. subtilis, TnrA, by interaction with the glutamine synthetase (GS). Finally, we will briefly consider the special case of the E. coli maltose regulon.

In E. coli, the global transcription regulator Mlc represses the production of several PTS components, including the ptsG gene and the ptsHI crr operon encoding the glucose permease and the general PTS proteins respectively (Böhm and Boos, 2004). In the presence of glucose, Mlc is unable to bind its DNA targets, resulting in expression of the controlled genes. However, the mechanism of glucose sensing by Mlc has long remained enigmatic (Plumbridge, 1998; Kim et al., 1999). As observed for the control of GlcT by the B. subtilis glucose permease, the equivalent enzyme from E. coli was found to act as a trigger enzyme and to control Mlc activity in addition to its function in glucose transport and phosphorylation. However, the mechanism of control by the E. coli glucose permease is completely different from that exerted by its B. subtilis counterpart. In the presence of glucose, the glucose permease is mainly present in the non-phosphorylated state. This form of the enzyme is capable of binding Mlc, thereby sequestering it and preventing it from DNA binding. In contrast, the phosphorylated glucose permease, the idle form of the enzyme present in the absence of glucose, is unable to bind Mlc, and Mlc is free to repress the transcription from its target promoters (Tanaka et al., 2000). The differential ability of the glucose permease to bind Mlc is caused by structural changes in the region of the active site of the IIB domain upon phosphorylation (Seitz et al., 2003). It is interesting to note that the Mlc repressor of Thermus thermophilus does also control gene expression in response to the presence of glucose even though these bacteria do not possess PTS proteins. This Mlc protein binds directly to glucose, which acts as an inducer. This seems to be a remnant of the evolutionary history of Mlc and its relatedness to sugar kinases (Chevance et al., 2006; see below).

In B. subtilis, nitrogen metabolism is controlled by the global transcription factor TnrA. TnrA activates the expression of genes involved in the utilization of secondary nitrogen sources and in the uptake of ammonium at low concentrations. In contrast, it represses the expression of genes required for ammonium assimilation, i.e. the glnA gene and the gltAB operon encoding GS and glutamate synthase respectively (Wray et al., 1996; Belitsky et al., 2000; Yoshida et al., 2003). A combined transcriptome and bioinformatic analysis suggested that TnrA controls the expression of about 20 target operons (Yoshida et al., 2003). It is thus the functional equivalent of the global Ntr regulatory system in the enterobacteria (Fisher, 1999).

The DNA-binding activity of TnrA responds to the nitrogen supply of the cells. It is active if the cells are provided with poor nitrogen sources such as glutamate or nitrate, but inactive in the presence of good nitrogen sources such as glutamine (Wray et al., 1996; Nakano et al., 1998; Belitsky et al., 2000). These findings raised the question of how TnrA senses the nitrogen supply. In vitro experiments revealed that the protein is active in transcription activation without cofactors, and that none of the likely metabolite effectors affects DNA binding by TnrA (Wray et al., 2000). However, glnA mutants defective for GS exhibit constitutive TnrA activity, suggesting that GS exerts negative control over TnrA (Wray et al., 1996; Belitsky et al., 2000). Indeed, an interaction between GS and TnrA was detected, and the formation of the complex caused TnrA to lose its DNA-binding ability. The formation of the TnrA–GS complex depends on the presence of glutamine and AMP, i.e. it occurs under conditions of feedback inhibition of GS (Wray et al., 2001; Fig. 4). This provides a direct link between nitrogen supply, enzymatic activity of the GS, and TnrA-controlled gene expression. In the presence of excess glutamine, the GS is feedback-inhibited and forms a complex with TnrA, thus preventing its activity as a transcription factor. In contrast, under nitrogen limitation, GS is enzymatically active in glutamine biosynthesis, and free TnrA activates the genes for the utilization of alternative nitrogen sources (Fig. 4). This model is strengthened by the observation that mutations of the GS that interfere with feedback inhibition but not with enzymatic activity result in a defect in the interaction and concomitant inactivation of TnrA (Fisher and Wray, 2006).

Figure 4.

Control of the global nitrogen regulator TnrA by the glutamine synthetase in B. subtilis. In the absence of the preferred nitrogen source glutamine, the glutamine synthetase (GS) catalyses the formation of glutamine from ammonium and glutamate. The global transcription regulator TnrA is able to bind conserved regions in the promoter regions of its target operons (such as the amtB glnK operon), thereby activating transcription. In the presence of the preferred nitrogen source glutamine, the feedback inhibited GS binds TnrA and inhibits the DNA-binding activity of the transcription regulator. Note, that TnrA can also act as a repressor of some operons by binding to sites downstream of the promoter.

Another system in which a transcription factor is controlled by trigger enzymes is the E. coli maltose regulatory system (Böhm and Boos, 2004). Here, the transcription activator MalT is controlled by interactions with three different enzymes. In the absence of maltodextrins, MalT is sequestered by the MalK ATPase subunit of the maltodextrin ABC transporter (Panagiotidis et al., 1998). For transcription activation, MalT requires the cofactors maltotriose and ATP. Two other trigger enzymes, the esterase Aes and the βC-S lyase MalY can also bind MalT. All three trigger enzymes antagonize the binding of the coactivator maltotriose to MalT (Schreiber et al., 2000; Joly et al., 2002; 2004). The involvement of multiple trigger enzymes in the control of a single transcription factor is unprecedented and might ensure tight control of MalT activity and, thus, of the expression of the genes of the maltose regulon.

Evolution of trigger enzymes: from enzymes via trigger enzymes to regulators

Owing to their central role in metabolism and their interactions with basically all metabolites in the cell, the enzymes are the ‘best-informed’ proteins in the cell with respect to the metabolic status. This makes them perfect mediators of gene regulation and, as shown above, this potential has been used in several ways. The trigger enzymes seem to be an evolutionary intermediate as they are related to enzymes as well as to dedicated regulatory proteins. The evolution of these interesting proteins is still continuing, and we can study all stages in this evolution. Two patterns seem to be of special importance: (i) the acquisition of DNA-binding domains by enzymes, and (ii) the functional separation of enzymatic and regulatory functions via duplication events.

An example for the acquisition of a DNA-binding domain is the regulation of proline catabolism. This pathway involves two distinct enzymatic activities that can be performed by two individual proteins, as in the Gram-positive bacteria, or by a single multidomain protein as in proteobacteria. In many proteobacteria, including E. coli and the other enterobacteria, this bifunctional PutA protein has acquired a RHH domain that provides the protein with the ability to interact with specific DNA sequences and, thus, to repress transcription of the putA and putP genes (see above, Fig. 1). In contrast, in Bradyrhizobium japonicum and many other alphaproteobacteria, PutA lacks a DNA binding domain and, indeed, these proteins are strictly enzymatic, i.e. they do not act as trigger enzymes (Krishnan and Becker, 2005).

Another fascinating example for the evolution of an enzyme to a regulator is provided by the so-called ROK family comprising repressors of genes involved in sugar metabolism and sugar kinases (Titgemeyer et al., 1994). This family includes the glucose kinase that catalyses the first step in glycolysis, and the B. subtilis Xyl and the E. coli and T. thermophilus Mlc repressors. Interestingly, the Xyl repressor binds not only its inducer xylose but also glucose (Dahl et al., 1995). Mlc is an excellent example for the ongoing evolution. In T. thermophilus, glucose acts as the inducer by binding to the motif conserved in the glucose kinases (Chevance et al., 2006). The E. coli Mlc has lost its ability to bind glucose (Chevance et al., 2006), and Mlc activity is controlled by interaction with the trigger enzyme PtsG (see above, Fig. 3). This development was probably driven by the invention of the PTS for sugar transport with concomitant phosphorylation of the incoming sugars (Deutscher et al., 2006). Thus, free glucose is not available to control the E. coli Mlc protein. Obviously, this family might yield trigger enzymes although this has not been described so far. Interestingly, the glucose kinase of Streptomyces coelicolor is involved in carbon catabolite repression. This protein itself, rather than its enzymatic activity, has the regulatory activity (Kwakman and Postma, 1994). Thus, this glucose kinase might act as a trigger enzyme.

There are several cases in which the two functions of trigger enzymes have been separated by gene duplication events. This is exemplified by the B. subtilis uracil phosphoribosyltransferase. The two enzymes with this activity share a conserved active site motif. However, only one of them, Upp, is involved in nucleotide biosynthesis, whereas the other, PyrR, acts as an RNA-binding termination protein that controls the expression of the pyr biosynthetic operon. RNA-binding by PyrR is UMP dependent (Turner et al., 1998). The PyrR structure is very similar to that of other phosphoribosyltransferases. However, the dimer interface contains a unique, large, basic surface that might act as the RNA binding site (Tomchick et al., 1998).

In Streptococcus pyogenes, a unique form of carbon catabolite repression is controlled by a specialized trigger enzyme, LacD.1. This enzyme (as its paralogue, LacD.2) has tagatose-1,6-bisphosphate aldolase activity. Under conditions of high glycolytic activity, LacD.1 (but not LacD.2) binds and inactivates the transcriptional activator RopB. If the glycolytic activity is low, LacD.1 is unable to bind RopB, which can then activate transcription of the speB gene encoding a protease, thus allowing the utilization of proteins as an alternative carbon source. In contrast to LacD.1, the LacD.2 protein is involved in lactose metabolism rather than in regulation (Loughman and Caparon, 2006; 2007). The two LacD proteins share 71% identical amino acids. Thus, the duplication of an enzyme might even result in the acquisition of regulatory activity for only distantly related functions.

It is tempting to speculate that we are just beginning a new phase of discoveries that follow on from the discovery of the multiple regulatory roles of RNA.


We are grateful to Sven Halbedel, Boris Görke and Lope Andrés Flórez Weidinger for critical reading of the manuscript and helpful comments. The work in our lab was supported by grants of the Deutsche Forschungsgemeinschaft and the Fonds der Chemischen Industrie to J.S.