Chapter 32

Biosynthesis of Threonine and Lysine

JEAN-CLAUDE PATTE

INTRODUCTION

In members of the Enterobacteriaceae, aspartate is used as a donor of carbon for the synthesis of the amino acids that constitute the aspartate family, namely, threonine, methionine, diaminopimelate, and lysine. These syntheses are performed by a rather complex pathway, with several branch points. Establishment of the pathway and studies of its regulation in Escherichia coli led to a scheme that was taken as a model of regulation for amino acid branched pathways. This extremely sensitive regulatory scheme is created by the existence of isoenzymes that catalyze the same biosynthetic reaction, but the activity and/or synthesis of which are regulated by different amino acids (157). In summary, at a given branch point (aspartate, aspartate semialdehyde, homoserine), there are as many isozymes as there are amino acids deriving from this biosynthetic step (124). Two comments have to be made: (i) this apparently very logical pattern also exists in the case of another branched pathway in E. coli, the aromatic pathway; (ii) in fact, in other species, quite different mechanisms have been selected, and the solution selected in E. coli is just one from among several (46).

Another feature that makes the aspartate pathway distinct from others is that in addition to the variations in regulation, differences in the pathway itself exist among species. The reactions leading to threonine appear to be conserved, but in methionine (36) and lysine biosynthesis (129), a large diversity occurs. In mammals, the aspartate pathway is not functional. In fungi, lysine is synthesized by a completely different pathway called the α-aminoadipate pathway. Moreover, even though all known bacteria derive their lysine from aspartate (a pathway conserved in plants), different reactions are used by different species (see below). Finally, gene organization differs widely among bacteria. It would be of interest to know the selective pressure exerted by growth conditions during evolution that led to such diversity. Some hypotheses were recently proposed by Paulus (129), taking into account two points. First, diaminopimelate is not a constituent of proteins but rather is part of the bacterial cell wall, and its presence and concentration in the peptidoglycan differ from one species to another. Special regulatory mechanisms must ensure that its biosynthesis is balanced with protein synthesis. Second, the syntheses of other metabolites are also frequently found as an additional branch in the pathway. In parallel with secondary metabolites (see below), this is exemplified particularly by the role of dipicolinic acid in the sporulation of Bacillus subtilis (129).

THE COMMON PATHWAY

The first three enzymes catalyze the synthesis of intermediates common to lysine, threonine, and methionine (aspartate kinase and aspartate semialdehyde dehydrogenase) or to threonine and methionine (homoserine dehydrogenase) syntheses (Fig. 1). They are part of what has been called the common pathway (40). E. coli K-12 has three isofunctional aspartokinases and two isofunctional homoserine dehydrogenases. In fact, homoserine dehydrogenase activities are carried by larger polypeptides that also exhibit aspartokinase activity. These proteins were the first clearly established multifunctional enzymes. Threonine and isoleucine regulate the expression of the aspartokinase-homoserine dehydrogenase I, and threonine inhibits both activities. Methionine represses the synthesis of aspartokinase-homoserine dehydrogenase II. Lysine inhibits the activity and represses the synthesis of aspartokinase III.

Fig. 1Common biosynthetic pathway to lysine, threonine, and methionine. The coordinates for the gene positions are those of the E. coli physical map.

Source:

Aspartokinase-Homoserine Dehydrogenase I

Stadtman et al. (157) recognized first a threonine-sensitive aspartokinase activity. Later, a homoserine dehydrogenase activity, also regulated by threonine, was found (125). Genetic (a unique mutation leading to the simultaneous loss of inhibition by threonine of the two activities [44]), kinetic (inhibition of one activity by the substrates of the others; protection of both activities by NADPH against thermal denaturation [128]), and biochemical (copurification to homogeneity [173]) evidence indicated that the two activities are carried by the same polypeptide chain. Subsequent sequence analysis of the protein (48) and of the corresponding gene, thrA (95), confirmed this conclusion, whereas genetic experiments demonstrated the bicistronic organization of the gene (170). These proteins have been studied mainly by Cohen’s group, and their properties have been reviewed in detail (see references 40, 41, 42, and 45, where most references may be found). Only a brief summary will be given here. The kinetic parameters are as follows (41, 42): K_m for Asp, 1.5 mM; K_m for ATP, 0.18 mM; K_m for aspartate semialdehyde, 0.12 mM; and K_m for NADPH, 0.04 mM. Recent studies were performed on the detailed kinetic mechanisms of the reactions. They are consistent with (i) a random addition of l-aspartate and ATP and an ordered release of ADP and aspartyl phosphate (1) and (ii) a preferred-order random kinetic mechanism in which the dominant pathway involves association of NADPH prior to aspartate semialdehyde and dissociation of homoserine prior to NADP⁺ (182). Threonine inhibits aspartokinase activity in a competitive manner, whereas homoserine dehydrogenase is inhibited noncompetitively; inhibition curves are sigmoïd in both cases, displaying homotropic cooperative effects. The concentrations necessary to inhibit 50% of the maximal activity are 0.3 mM for the kinase and 0.5 mM for the dehydrogenase. Threonine binding to the pure protein is sigmoidal at high K⁺ concentrations and hyperbolic at low K⁺ concentrations. Eight (two sites on each subunit of the tetramer) binding sites could be determined for threonine (88). On the basis of kinetic data (5, 86, 87, 91), it was proposed that this protein obeys an allosteric model in which the equilibrium between at least three states, R1, R2, and T, is displaced by threonine and aspartate. More recently (181), isotope exchange kinetics at chemical equilibrium indicated that l-threonine inhibited catalysis without altering substrate association-dissociation rates. Recent data have shown that serine inhibits homoserine dehydrogenase activity at low K⁺ concentrations, which may play a physiological role (77).

The molecular weight of the native enzyme was determined as approximately 350,000, and the protein has the tetrameric α ₄ structure (37, 60, 158). This structure will be discussed below. The thrA gene has been localized at min 0 (169) on the chromosome and is part of the thrABC operon.

Aspartokinase-Homoserine Dehydrogenase II

It was shown later that in E. coli K-12 there is a second bifunctional enzyme carrying the same two activities, aspartokinase-homoserine dehydrogenase II (124). The amount of this protein in the cell is much less than that of enzyme I. No ligands inhibit this activity. The kinetic parameters are as follows (41, 42, 61): K_m for Asp, 2.1 mM; K_m for ATP, 1.9 mM; K_m for aspartate semialdehyde, 0.19 mM; and K_m for NADPH, 0.15 mM. The molecular weight of the protein is approximately 185,000 (61); it is a dimer composed of two identical subunits (55).

The corresponding metL gene has been localized at min 89 on the chromosome (169), in the metBL operon, and its sequence has been determined (192). As pointed out above, the expression of this gene is regulated by methionine; it belongs to the Met regulon and is repressed by the MetJ repressor (see chapter 33).

In E. coli B, no homoserine dehydrogenase II activity was detectable, whereas aspartokinase II activity was present (124). It is now known that a protein of similar length lacks dehydrogenase activity because of a missense mutation (I. Saint-Girons, personal communication). The corresponding methionine-repressible protein found in Salmonella typhimurium (official designation, Salmonella enterica serovar Typhimurium) exhibits both activities and is present in much greater amounts than in E. coli K-12 (28).

Aspartokinase III

An aspartokinase activity sensitive to lysine inhibition was found simultaneously with aspartokinase I (157) and was subsequently named aspartokinase III. The kinetic parameters are as follows (41, 99): K_m for Asp, 4.7 mM; and K_m for ATP, 4.8 mM. Lysine inhibits the activity (50% inhibition at 0.2 mM l-lysine) in a noncompetitive manner, but certain so-called nonspecific amino acids (such as leucine, isoleucine, and phenylalanine) also inhibit this aspartokinase substantially (126). The observed synergy between lysine and these nonspecific inhibitors appears to be physiologically significant. The dimeric protein binds 4 mol of lysine (142). The four sites are nonequivalent: two have a K_d of 8 μM, while the remaining two have a K_d of 100 μM. Lysine binding is cooperative, but cooperativity disappears in the presence of one of the nonspecific amino acids, thus explaining the synergistic inhibition. Experimental kinetic data are in good agreement with theoretical values predicted for a concerted model of a class V allosteric enzyme (111). The kinetic parameters of the state and saturation functions have been determined. The results support the contention that aspartokinase III is one of the best examples of a class V allosteric enzyme. The purified enzyme has a molecular weight of 100,000 and consists of two identical subunits. It may tetramerize in the presence of lysine at high ionic strength and monomerize at low ionic strength in the absence of lysine. The existence of these multiple aggregation states made its molecular weight controversial (99, 116, 141, 179). The value of 50,000 for the monomer has been confirmed by sequence analysis of the lysC gene (32), which lies at min 91 on the chromosome (169).

Structures of the Aspartokinases and Homoserine Dehydrogenases

The structure of aspartokinase-homoserine dehydrogenase I has been studied in great detail as a model of a bifunctional protein (see review by Cohen and Dautry-Varsat [42]). It was first shown that the two activities are present on different parts of the polypeptide (176): a nonsense mutant, corresponding to the N-terminal part of the protein, has only the aspartokinase activity, still inhibited by threonine, whereas a C-terminal fragment, obtained by limited in vitro proteolysis, carries only the homoserine dehydrogenase activity, insensitive to threonine (22, 155). Subsequent experiments resulted in the isolation of other active intermediate forms (177). All of these results were consistent with the hypothesis that the protein contains different domains (176) with intersubunit and intrasubunit interactions (107), linked together by unstructured and very exposed short segments. They led Cohen and his group to propose a triglobular model of the protein, with the central domain responsible for subunit contact (64). Such a structure was in agreement with the proposed model, i.e., that such a multifunctional protein appeared during evolution by gene fusion between separate coding sequences (155).

Renaturation of the protein has been studied in vitro. Folding occurs in three successive steps. A monomolecular step leads to a monomeric species with kinase activity. An association step then leads to a dimeric species with kinase and threonine-sensitive dehydrogenase activities. Finally, a second association step leads to a tetrameric species with both activities sensitive to threonine (53, 54, 71, 175). The active species is indeed the tetramer (178). The different domains seem to acquire their native conformations rather independently. Similar experiments performed on enzyme II have shown that the E. coli enzyme is also composed of three different globular domains (7, 52, 55; see review in reference 42).

This biochemical model is now confirmed and extended by sequence comparisons between the corresponding enzymes isolated from different species that exhibit different characteristics but have strong sequence homologies with the E. coli enzymes. For example, multifunctional proteins are only found in Enterobacteriaceae, whereas other species have two separate genes encoding kinase and dehydrogenase (curiously, a bifunctional protein was recently identified in Daucus carota [184]). Table 1 summarizes these data; in Fig. 2, regions of homologies are schematized. One special point must be made with regard to the lysC gene, encoding aspartokinase II, of B. subtilis (and other gram-positive species). Paulus and his group showed that this protein has an α _2β2 structure, but the β polypeptide results from an in-frame initiation of translation internal to the lysC coding sequence and is thus a C-terminal part of the α polypeptide. Even though the exact biological role of this β subunit is not yet completely established, it strongly interacts with the α subunit (34; reviewed in reference 129).

Fig. 2Structural homology among aspartokinases and homoserine dehydrogenases from different species. , polypeptide sequence; , gap introduced for optimal alignment (only gaps longer than five residues are indicated); , region highly conserved among most sequences (residues 5 to 50, 125 to 145, 190 to 225, 320 to 355, 390 to 410, 465 to 480, 625 to 690, and 790 to 810). Residue numbers are chosen arbitrarily as those of the ThrA protein. Arrows point to regions of interest: a, D-P-R sequence (kinase active site?); b, site of proteolysis between domains 1 and 2; c, serine 352, mutated in desensitized enzymes from Serratia and Corynebacterium spp. (absent in enzyme II of E. coli); d, site of proteolysis between domains 2 and 3; e, ADP binding site. The length of the interdomain segments is arbitrarily estimated (taken between conserved sequences). The approximate lengths of the four domains are given. The data are summarized in Table 1.

Source:

Table 1Summary of data from Fig. 2

In summary, it appears that at least four domains can be identified (Fig. 2). All aspartokinases are composed of two domains (see also the discussion in reference 129). The first one (domain 1) ends around residue 290, just after a highly conserved region that contains a D-P-R sequence (32), a region homologous to γ-glutamyl kinase, proposed to be involved in the enzyme active site (118). It exhibits noninhibitable aspartokinase activity. The second domain, the central domain of the multifunctional proteins (domain 2), starts at residue 310 and stops at residue 460. It must be noted that the β subunit of the B. subtilis aspartokinase enzyme starts exactly at the beginning of this domain (34). An additional sequence is found in the yeast enzyme also at this precise position (136) (Fig. 2). This domain 2, which is also homologous between species but to a lesser extent than domain 1, appears not to have any catalytic activity. Nevertheless, it plays a central role in the inhibitory effect of the ligands, as shown by the altered interactions between subunits upon modification of its structure. Site-directed mutagenesis experiments performed on the Serratia thrA gene (whose product is 83% homologous to the E. coli enzyme) indicated recently the important role of a specific residue, Ser-352, on these interactions, modifying the feedback properties of the enzyme (118, 119). The mutation of a serine located at a similar site in the lysC aspartokinase of Corynebacterium glutamicum also led to desensitization toward threonine and lysine (92, 118). Moreover, this short region diverges in the mutifunctional enzyme II of E. coli (which is insensitive to any ligand), whereas it is conserved in the E. coli lysC gene.

A third domain (domain 3) of the bifunctional protein carries the homoserine dehydrogenase activity. It starts at residue 470. The proteins with only aspartokinase activity end exactly at this position, whereas the proteins with homoserine dehydrogenase activity alone begin at this point (29, 32). One may find just downstream a sequence characteristic for an ADP binding site for NADP (188) which is known to be usually localized at the beginning of a structural domain. A fourth domain (domain 4) is found in the homoserine dehydrogenases of several species, but not in the bifunctional proteins. It is likely for several reasons that this domain plays a role in the allosteric regulation of homoserine dehydrogenase activity (see the discussion in reference 29): (i) most of the mutations leading to desensitization of homoserine dehydrogenase activity have been localized in this region (3, 138); (ii) this domain is absent in the yeast protein, which is not inhibited by any ligand (171); and (iii) this domain is also absent in the multifunctional proteins, in which it is postulated that the allosteric interactions are shared with the aspartokinase domain through the central domain (see above).

A hypothetical model may be proposed for the evolutionary building of these complex molecules. Two separate catalytic domains (1 and 3) first appeared, to which were grafted regulatory domains (2 and 4). Then, in E. coli, gene fusion led to a multifunctional protein that lost its C-terminal regulatory domain 4, which had become redundant in the presence of domain 2. Duplications allowed the appearance of three isozymes, followed in the case of aspartokinase III by the loss of the homoserine dehydrogenase domain (another possibility being that a first duplication, before fusion, led to lysC, followed by fusion and a second duplication for multifunctional enzymes I and II [32]).

No apparent homology exists between the regulatory domains 2 and 4. The domain at which the regulatory ligands actually bind is not clearly established.

Aspartate Semialdehyde Dehydrogenase

Aspartate semialdehyde dehydrogenase catalyzes the reversible substrate-dependent reduction of NADP in the presence of phosphate or arsenate (Fig. 1, reverse reaction). This reaction is formally similar to that catalyzed by glyceraldehyde-3-phosphate dehydrogenase. The K_ms for the substrates could be determined only in the reverse reaction and are as follows (80): for aspartate semialdehyde, 0.2 mM; for NADP⁺, 0.09 mM; for P_i, 9 mM; and for HAsO₂,6 mM. No amino acids that inhibit the activity were found (note that the Bacillus brevis enzyme differs from all other aspartate semialdehyde dehydrogenases described in its sensitivity to inhibition by threonine and leucine [137]).

The enzyme has been purified (9, 80). Its molecular weight is 77,000, made up of a dimer of identical subunits. Its gene has been localized on the E. coli chromosome at min 76, and its sequence has been shown to specify a 40,000-M _r polypeptide (79). Homology is observed between the Asd protein from E. coli and other sequences from different species (35).

The substrate binding site of the E. coli protein has been labeled with 2-amino-4-oxo-5-chloropentoate. This analog inactivates the enzyme with pseudo-first-order kinetics and half-of-the-sites reactivity. A single group is labeled at the active site (10). Sequence data (79) and site-directed mutagenesis (94) indicate that Cys-135 is the target of the modification. This residue is conserved in proteins from other species (35).

THREONINE BIOSYNTHESIS

The enzymes catalyzing the two steps leading from homoserine to threonine are the products of the thrB (homoserine kinase) and thrC (threonine synthase) genes (Fig. 3).

Fig. 3Threonine biosynthetic pathway. Gene coordinates are as in Fig. 1.

Source:

Homoserine Kinase

Homoserine kinase catalyzes the ATP-dependent phosphorylation of homoserine. Its activity requires Mg²⁺, and K⁺ stimulates the enzyme (168). Homoserine kinase activity is inhibited by threonine (K_i of 0.6 mM) in a competitive manner with respect to homoserine. The K_ms for homoserine and ATP are 0.15 and 0.2 mM, respectively. Homoserine is an inhibitor at high concentrations. The binding of ATP is strictly Michaelian. The kinetics follow a preferred-order random mechanism, with ATP preferentially binding before homoserine. A two-site model with separate catalytic and inhibitory sites for homoserine has been proposed (153). The sequence of the gene is known (47) and is in agreement with the known M _r of the protein (two identical 29,000-Da subunits [26]). Homology is found among homoserine kinases from different species; however, this homology is low, and no phylogenic tree can be constructed (29). In Pseudomonas aeruginosa, a protein with true homoserine kinase activity (that complements E. coli thrB mutants) has been isolated; it has no sequence homology with any other homoserine kinases (39).

Threonine Synthase

Threonine synthase catalyzes the rather complex conversion of homoserine phosphate to threonine (65). This complexity is evidenced by experiments with the B. subtilis enzyme, which catalyzes deamination of homoserine phosphate or threonine to α-ketobutyrate in vitro as well as in vivo by mechanistically similar reactions (150). It catalyzes a pyridoxal phosphate-mediated g-elimination and replacement reactions, with formation of carbanions during the catalytic cycle (23, 63). The apparent K_m for homoserine phosphate is 0.23 mM (164). The nucleotide sequence of the thrC gene has been determined (123) and found to encode a polypeptide of 47,000 Da. Homology between the different threonine synthases exists, the enzymes of gram-positive bacteria being far distant from those of E. coli and other species (29). Moreover, similarities between threonine synthase and other proteins catalyzing similar reactions (such as tryptophan synthase, d-serine dehydratase, and threonine dehydratase) have been described (120, 121), suggesting the existence of a conserved site for hydroxyamino acids which is also found in homoserine kinases (108).

THE BIOSYNTHETIC BRANCH LEADING TO LYSINE

As mentioned in Introduction, different types of reactions have been selected by evolution in different bacterial species to synthesize lysine from aspartate semialdehyde. The two first steps, dapA and dapB, as well as the last one, lysA, are common. Three pathways between these steps have been determined: (i) the succinylase variant involving succinylated intermediates, used in Enterobacteriaceae, described in Fig. 4; (ii) the acetylase variant, in which the acetyl residue is used instead of succinyl as the blocking group, employed by certain Bacillus species (183); and (iii) the dehydrogenase variant, in which the intermediate tetrahydrodipicolinate, common to all three pathways, is converted in a single step to the ultimate lysine precursor, meso-diaminopimelate, used by Bacillus sphaericus (113) (in this species, lysine replaces diaminopimelate in the peptidoglycan). The simultaneous existence of two pathways (succinylase and dehydrogenase) has been detected in Bacillus macerans (4) and C. glutamicum (151).

Fig. 4Biosynthetic pathway for diaminopimelate and lysine. Gene coordinates are as in Fig. 1.

Source:

In E. coli, mutants blocked at each enzymatic step are diaminopimelate auxotrophs (or lysine auxotrophs for lysA mutants, which accumulate diaminopimelate in the medium), except dapF mutants (see below) and some leaky mutants. These diaminopimelate auxotrophs undergo lysis when grown in a complex medium (which normally lacks diaminopimelate), because of its requirement for cell wall synthesis. This property has been used for the construction of strains that could be considered safe hosts for protein production in large fermentors, as they could not survive in the environment (56). Analogs of substrates for some of these enzymes, especially the racemase, are also studied as potential antibacterial agents (76, 100, 104).

Dihydrodipicolinate Synthase

The reaction (154, 191) involves the condensation of aspartate semialdehyde with pyruvate, followed by a cyclization reaction, the removal of two water molecules resulting in the formation of the seven-carbon 2,3-dihydrodipicolinate. The protein (172) is encoded by the dapA gene, localized at min 53 on the chromosome (24). The sequence of this gene has been determined (145). The calculated M _r of the product is 31,300. As the molecular weight of the purified protein was evaluated as 134,000 (154), the native protein is thus a tetramer.

Lysine inhibits the activity at a relatively high concentration (50% inhibition at 1 mM [191]). No inhibition was observed on the proteins isolated from gram-positive bacteria, whereas a strong inhibition is described for the enzyme from plants. The apparent K_ms for the substrates are 0.57 and 0.55 mM for pyruvate and dl-aspartate semialdehyde, respectively. Binding of pyruvate to the enzyme, forming a Schiff base with the ε-amino group of Lys-161, is followed by the binding of aspartate semialdehyde (98). A ping-pong mechanism is proposed. Homologies with enzymes from other species have been found; interestingly, a striking homology exists also with the N-acetylneuraminate synthase of E. coli, an enzyme that catalyzes a similar reaction involving condensation of the C-3 of pyruvate with the aldehyde group at C-1 of N-acetylmannosamine (131).

Dihydrodipicolinate Reductase

The enzyme, first described by Farkas and Gilvarg (62), is specified by the dapB gene at min 0.5 on the chromosome (24). The pure protein, with a native molecular weight of 115,000, has a K_m for dihydrodipicolinate of 0.01 mM; no specific inhibitors of the enzyme are known (165). The gene sequence has been determined (13) and found to encode a polypeptide of M _r 29,000, suggesting a tetrameric structure for the protein. Homologies exist between proteins of different species (131), with a conserved ADP-binding pocket (188).

Tetrahydrodipicolinate Succinylase

The enzyme (75) is the product of the dapD gene (mutations at this locus have been erroneously attributed to the following step [25]) localized at 3.5 min clockwise from tonA (24, 144). Its sequence has been determined (144) and found to encode a polypeptide of M _r 30,000, translated from a monocistronic mRNA. In E. coli, succinyl coenzyme A (succinyl-CoA) is a substrate of the reaction, as it is for acylation of homoserine in the first step of the methionine-specific pathway. It must be noted that in gram-positive species, both reactions use acetyl-CoA instead of succinyl-CoA (see reference 129 for a discussion). Mutants requiring either (i) lysine plus methionine or (ii) succinate have been described (93); this phenotype reflects an impaired synthesis of succinyl-CoA.

N -Succinyl Diaminopimelate Aminotransferase

The aminotransferase of E. coli has been partially purified and characterized (130). It is assumed to be the product of the dapC gene, mutations in the last locus of which lead to a Dap^– phenotype, localized at min 3.5 counterclockwise from tonA (24).

N -Succinyl Diaminopimelate Desuccinylase

The enzyme (74, 97) has been purified (104) and found to have a dimeric or tetrameric structure. Co(II) or Zn is required for its activity, Co being more active. The K_m for the N-succinyl-l-diaminopimelate has been determined as 0.4 mM, with high substrate specificity.The corresponding gene is dapE, localized at min 53. Its sequence has been determined (12) and shown to specify a polypeptide of M _r 41,100. The coding sequence is expressed from two promoters, one being in front of an open reading frame cotranscribed with dapE; this open reading frame does not appear to be involved in diaminopimelate biosynthesis (12). Strong homology has been observed with acetylornithine deacylase, product of the argE gene (20). It has been shown recently (190) that overproduction of succinyl diaminopimelate deacylase suppresses a mutation in heat shock gene grpE, suggesting that chaperone proteins may interact with succinyl diaminopimelate deacylase in vivo.

Diaminopimelate Epimerase

This activity was identified by Antia et al. (2). The sequence of the corresponding dapF gene, localized at min 85, has been determined (143). It encodes a polypeptide of M _r 30,200. The protein has been purified (189) as a monomer of M _r 34,000. The activity does not require pyridoxal phosphate, which is unusual for an epimerase and makes the enzyme analogous to proline racemase and hydroxyproline 2-epimerase. The SH residue of Cys-63 has been identified as necessary for enzyme activity (82). A unique stable mutant, obtained by reverse genetics, is known (139). This mutant accumulates ll-diaminopimelate in the cytoplasm but does not require meso-diaminopimelate for growth, indicating that racemization is possible by a nonspecific reaction. The specificity for the substrate is not absolute, as it has been shown recently (142) that l-lanthionine (a sulfur analog of ll-diaminopimelate) is racemized in vivo by the dapF gene product, leading to incorporation of meso-lanthionine in the peptidoglycan.

Diaminopimelate Decarboxylase

The enzyme (57) has been purified and shown to have an M _r of approximately 200,000 (187). The nucleotide sequence of the lysA gene, at min 61, has been determined (160); it specifies a polypeptide of M _r 46,000, suggesting a tetrameric structure for the native protein. Only meso-diaminopimelate (and not the ll isomer) is a substrate; activity requires pyridoxal phosphate. Members of a second class of mutants with lesions at the same locus, formerly designated lysB, require lysine or pyridoxal phosphate for growth; they were identified as K_m mutants for pyridoxal phosphate (25). The pyridoxal phosphate binding site appears different from the cofactor binding site on other amino acid decarboxylases (109). The LysA sequences are quite homologous among different species; a surprisingly high degree of homology has been found with mouse ornithine decarboxylase but not with other amino acid decarboxylases of bacterial origin (109).

REGULATION OF GENE EXPRESSION

Regulation of Threonine Biosynthesis

The three genes thrA, thrB, and thrC are organized in this order in an operon, localized at min 0 on the chromosome (169, 170). The nucleotide sequence of the entire operon has been determined (47, 68, 95, 123). The full-length mRNA is composed of 4,860 nucleotides. Only one nucleotide separates thrA and thrB coding sequences; thrB and thrC are contiguous. Transcription starts from a strong promoter, 190 bases upstream of the thrA ATG. An internal promoter at the 3' end of thrA allows the formation of thrB transcripts in addition to those initiated at the major promoter, but with a much lower efficiency (148). More recently (105), translational coupling between thrA and thrB has been demonstrated.

Expression of the operon depends on the intracellular concentrations of both threonine and isoleucine, via a multivalent repression mechanism (67), isoleucine limitation being more efficient for derepression of expression (43). By isolation of derepressed mutants, it was shown that a cis-acting locus, first called an operator, was involved (70, 147), as well as trans-acting mutations. Those mutations appeared to be localized in the threonyl- or isoleucyl-tRNA synthetase (11, 90, 115), or they were responsible for a physiological decrease in the regulatory amino acid intracellular pool.

The sequence of the mRNA leader (where so-called operator mutations were localized) indicated that this leader presented all of the characteristics required for a typical attenuation mechanism (68, 69): existence of a rho-independent signal of transcription termination; existence of several secondary structures in the mRNA, which are mutually exclusive; and existence of a leader peptide, in which numerous Thr and Ile codons are present (the sequence is -Thr-Thr-Ile-Thr-Thr-Thr-Ile-Thr-Ile-Thr-Thr-). When this leader is translated, the presence of the ribosomes allows a termination of transcription. On the contrary, when ribosomes stall on the Thr-Ile-rich sequence, because of lack of charged tRNAs, this termination does not occur and RNA polymerase reads through the coding sequences. A more detailed study of derepressed mutants gave results in complete agreement with the proposed mechanism (106).

One study has suggested the existence of a trans-acting element that might reduce expression of the thr, ilv, and leu genes by binding to consensus sequences upstream of their promoters (89). These sequences exhibit some similarity to sequences found upstream of aroH, trpR, and the trp operon. Indeed, mutations were found in a locus, designated ileR, lying at min 99.5, that resulted in increased expression of the thr and ilv genes. The ileR gene has been cloned, and its sequence has been determined (185). The effect of the ileR mutations was partially suppressed in cells overproducing the TrpR protein. Whether the effect of the locus can be modulated by the effectors that regulate the thr operon, threonine and isoleucine, has not been reported. Thus, it cannot yet be defined as a regulatory element or a repressor. Nevertheless, the possibility that such DNA binding proteins exist in the cell could put constraints on gene expression that would require compensatory adjustments in transcript initiation if such binding were not to have a detrimental effect. That the TrpR protein is maintained at such low levels in normal cells prevents it from having any significant effect on thr operon expression.

From an evolutionary point of view, it must be noted that the genetic organizations of the three genes differ markedly from one species to another, especially as the threonine aspartokinase is usually either absent or not linked to homoserine dehydrogenase in a multifunctional protein (39, 66, 122; for a discussion, see reference 29).

Regulation of Lysine Biosynthesis

The main characteristic of the lysine genes in E. coli is that they are widely distributed over the chromosome. There is no multigenic operon. Although many points concerning the regulation are yet unclear, the term "lys regulon" is usually used to describe the whole set of genes. In contrast, different clusters or operons have been found in gram-positive species (35, 49,131).

It has been clearly established, either directly or by the use of fusions with ' lacZ (many enzymatic activities being difficult to measure in crude extracts), that the expression of several genes depends on the intracellular pool of lysine; this is true for lysC (157), asd (16), dapB (13), dapD (144), and lysA (19). In contrast, expression of dapA (27), dapE, and dapF (C. Richaud, personal communication) appears constitutive (for dapE, discrete variations that had been observed when enzyme activities were measured were not observed when dapE ' -'galK constructions were used [12]; however a potential Lrp binding site has been recently identified in front of the upstream dapE promoter [P. Plateau, personal communication]). Results are yet to be reported for dapC. However, even though regulation by lysine has been demonstrated, the molecular mechanism that correlates variations of lysine concentration with variations of gene expression remains to be established and differs quite certainly from one gene to another.

The expression of LysC specific activity varies widely (over 100-fold between repression values and lysine-limited chemostat values) with the lysine intracellular pool. Excess arginine interferes to a certain extent with lysine repression (31). Although an adenylylation process (117) has been suggested (and not confirmed in vivo), LysC synthesis varies mostly in parallel with enzyme specific activity. The allelic state of the rel gene also plays a role in derepression (127). Because of the lack of selective pressure (desensitized mutants are selected in preference; see below), derepressed mutants were obtained only after numerous unsuccessful attempts (15). No trans-acting mutants were identified. All cis-acting mutants have been localized in the lysC gene (15). Their sequences have been determined recently, and the mutations have been identified (J. C. Patte, unpublished data): they all lie in the 317-base leader region of the lysC mRNA (33) and consist of base changes (one duplication is also observed). These data must be compared with those obtained by the Paulus group for B. subtilis (discussed in reference 129): this lysC gene also contains a leader sequence partly homologous to that of E. coli, mutations in which lead to a derepressed phenotype. However, those data do not indicate the molecular level at which lysine plays its role in regulation.

Expression of the asd gene varies in response to the concentrations of lysine, threonine, and methionine (16), but lysine is a much more important effector (50-fold variation) than the other two (only 2-fold derepression) and therefore is included in the lysine regulon. Operon fusions with ' lacZ have indicated that regulation occurs at the transcriptional level. Observation of the sequence (79) has led to the conclusion that a typical multivalent attenuation mechanism is not involved (such a mechanism exists for the Streptococcus asd gene [30]). The molecular mechanism of regulation therefore remains unknown. It has been shown that another metabolite, glucose 6-phosphate, modifies asd expression (17). Here again, no molecular mechanism has been proposed.

Variations in the lysine pool also greatly affect dapB expression (20-fold derepression [14]). Mutations in the relA gene (127) and in the lysS gene (18) indicated that ppGpp on the one hand and lysyl-tRNA on the other must play a role in this regulation. However, again examination of the nucleotide sequence upstream of the dapB gene did not allow proposal of a molecular mechanism; some sequence similarities with the asd promoter sequence have been suggested tentatively (13). Binding of an unidentified cytoplasmic protein to the DNA promoter region has been observed in gel retardation experiments (J. Bouvier and P. Stragier, personal communication). Similarly, examination of the coding sequence upstream of the dapD gene gave no indication of the molecular mechanism by which lysine modifies the expression of the gene (144).

The only example in which a clear mechanism has been established is the regulation of the lysA gene. Expression of this gene depends on the intracellular concentration of both diaminopimelate, which acts as an inducer, and lysine, which acts as a corepressor. This double regulation reflects the special situation of the diaminopimelate decarboxylase, as a catabolic enzyme for diaminopimelate and an anabolic enzyme for lysine (19). Moreover, ppGpp must play a role in lysA expression, which is modified in relA mutants (127) (lysX mutations have been found and are likely localized in relA). Mutants with a Lys phenotype were obtained outside the lysA structural gene. They had lesions in the lysR gene, specifying a positive regulatory protein for lysA expression (162). The lysR gene is located next to lysA and transcribed divergently (65 bp lie between the two sites of transcription initiation). Genetic evidence has shown that LysR (with diaminopimelate) binds to this intergenic sequence, allowing activation of the lysA promoter (160) and repression of the lysR promoter (autoregulation). It has been proposed that lysine blocks LysR binding, but this point remains to be firmly established. The lysR sequence has been determined and found to encode a protein of M _r 34,000 (161). A characteristic helix-turn-helix sequence for DNA binding has been demonstrated. Henikoff et al. (81) have shown that several regulatory proteins share several characteristics with LysR and constitute the so-called LysR family. This family comprises at the present time (for a review, see reference 149) more than 50 members in several prokaryotic species (none was found in archaebacteria). Most of the proteins are transcriptional activators, and a few act as repressors. The gene organizations with the gene under regulation are mostly identical, with a divergent transcription unit and a short (25- to 70-bp) intergenic sequence. Multiple regions of homologies are found when the sequences of the proteins are compared (149): (i) a DNA-binding domain with a helix-turn-helix motif (residues 1 to 65); (ii) domains involved in coinducer recognition and/or response (residues 100 to 173 and 196 to 206); and (iii) a domain required for both DNA binding and coinducer response (residues 227 to 253). These similarities must be of biological significance as mutations leading to trans suppression of lysR mutants have been isolated, and some have been localized in the xapR or the cysB gene, both of which are members of the LysR family (Patte, unpublished data).

METABOLIC FLUXES

As pointed out in Introduction, the complexity of the branched pathway, with a multiplicity of isozymes specifically regulated, has evolved as an efficient mechanism for adapting the flux of metabolic intermediates to environmental conditions. Different approaches have elucidated the relative roles of the selected regulatory mechanisms. Detailed studies have been made with bacteria such as corynebacteria that are used for the industrial production of amino acids.

In E. coli, the flux of threonine synthesis has been studied in vitro by Shames et al. (152), using purified enzymes and different aspartate-derived antimetabolites. Without addition, no accumulation of the intermediates was observed. In the presence of an inhibitor, there was an accumulation of the intermediate immediately preceding the inhibited reaction.

The role of the multiple isozymes was investigated in mutants lacking one or several isofunctional activities. A thrA metL double mutant (in which aspartokinase I and II activities have been lost, aspartokinase III being the only activity left) can grow in minimal medium, but its growth is inhibited by the presence of lysine unless threonine and methionine are simultaneously added (16) (a parallel situation exists for a metL lysC double mutant). Similarly, a metL mutant affected in homoserine dehydrogenase activity is conditionally auxotrophic for methionine in the presence of threonine and isoleucine (which repress and inhibit homoserine dehydrogenase I). Such a situation must be very rare in nature, as the MetL enzyme is very poorly expressed in E. coli K-12 (and no homoserine dehydrogenase II is present in E. coli B [124]).

The respective roles of feedback inhibition and repression have also been studied, selecting, for example, mutants resistant to growth inhibition by analogs of threonine (β-hydroxynorvaline [43]) or lysine (S-aminoethyl cysteine). In the threonine pathway, the main regulation acts on the bifunctional ThrA protein. As already noted, feedback inhibition is not complete, and derepressed or desensitized mutants could be equally well isolated. The LysC step is also a crucial point of regulation for lysine biosynthesis. However, in this case, the feedback inhibition is 100%. Only desensitized lysC mutants could be isolated in a first stage (with the exception of permeation mutants), and accumulation of lysine in the growth medium results from such mutations (16).

The biological importance of feedback inhibition on the thrB and dapA enzymes has been questioned, in view of the high K_i of the ligands. However, it must play a role in vivo, in parallel with some channelling, as desensitized lysC mutants overproduce mainly lysine, and not threonine, and vice versa.

The existence of the multifunctional ThrA and MetL proteins is difficult to reconcile with the existence of a unique and independent aspartate semialdehyde dehydrogenase. The existence of multienzymatic complexes that could facilitate the successive reactions without dilution of the intermediate has been searched for intensively, but all experiments gave negative results.

Finally, in vivo determination of the limiting steps in the lysine pathway was made possible by the use of mutants and by gene amplification with multicopy plasmids carrying one of the genes of the pathway: the LysC activity is the first point of limitation of the flux, followed by the DapA step (51).

METABOLISM OF THREONINE, LYSINE, AND OTHER METABOLITES OF THE PATHWAY

Biodegradation of threonine is covered in chapters 22, 36, and 94, and that of lysine is discussed in chapter 25. As already stressed, meso-diaminopimelate in E. coli and S. typhimurium, and lysine in some other bacteria, are essential components of the bacterial cell wall. Mutant auxotrophs for diaminopimelate undergo lysis when grown in the presence of lysine and in the absence of diaminopimelate. Cell wall synthesis is detailed in chapter 68. However, it must be stressed here that meso-diaminopimelate utilization is not entirely specific and depends on the lysine biosynthetic pathway, as recent data have indicated that ll-diaminopimelate (accumulated in a dapF mutant [112]) and meso-lanthionine (when synthesized endogenously [142]) are incorporated in the peptidoglycan.

Homoserine lactone was found recently to intervene as a signal in numerous cell density-dependent processes. The biosynthetic pathway of this compound remains to be established. A simple hypothesis is that homoserine is a precursor. Homoserine lactone or related compounds were identified in all enteric bacteria except E. coli and S. typhimurium (163); one cannot exclude the possibility that adequate inducing conditions were not used in the experiments.

PERMEATION

Both threonine and lysine appear to be accumulated within the cytoplasm by both specific permeases and multiligand transport systems. In this latter case, the molecules involved are well documented in E. coli as well as in S. typhimurium and are discussed in chapters 74, 75, and 76. For threonine (and also for l-leucine, l-isoleucine, l-valine, l-alanine, and l-serine), the periplasmic LITV protein is used. The LAO binding protein system functions in the transport of lysine, arginine, and ornithine basic amino acids (for a review, see reference 78).

The specific transport of threonine, also used for homoserine, has been described (166), but the corresponding protein and gene are still unknown. In contrast, a novel membrane-associated threonine permease was recently identified. This enzyme is the product of the tdcC gene, which is part of the tdcABC operon and is discussed in chapter 22.

Lysine permease is encoded by the lysP gene (159). The product of translation is predicted to be a hydrophobic protein of 489 residues which exhibits sequence similarity to a family of amino acid transport proteins, especially the aromatic amino acid permease (AroP) and the PheP protein. Mutations in the lysP gene affects expression of the cadA gene, which specifies lysine decarboxylase. For this reason, the locus was formerly also called cadR. This enzyme is probably also involved in the inducible decarboxylation of lysine.

Diaminopimelate is transported in E. coli by the cystine transport system and is inhibited by several diamino-dicarboxylic acid analogs (8).

AMINOACYL-tRNA SYNTHETASES

Threonyl-tRNA Synthetase

thrS, the gene for threonyl-tRNA synthetase (EC 6.1.1.3), has been cloned, and its sequence has been determined (110). It is an α ₂ dimer with an M _r of 147,800. It belongs to group II of synthetases (50, 59), members of which have strong homologies with the prolyl and seryl enzymes. Cocrystallization with a cognate tRNA has been unsuccessful, but interactions have been studied in solution (167). Phosphorylated forms of the protein have been described, but no biological role has been assigned to these modified enzymes (180). thrS lies at 38 min on the chromosome; it is the first of a cluster of three genes, with infC and rplT, exhibiting a rather complex pattern of transcription with multiple promoters (110, 134, 186). The most interesting characteristic is the mechanism by which its expression is regulated. It is autoregulated negatively at the translational level, as clearly established by Springer et al. (156). The threonyl-tRNA synthetase protein acts as a translational repressor by binding to the leader region of the thrS mRNA (approximately 160 nucleotides), preventing ribosome binding to its loading site in front of the coding sequence (114). This binding is made possible by the existence of well-defined secondary structure domains within the leader mRNA (which shares sequence and structural similarities with tRNA^Thr) with an accessible threonine anticodon sequence(6, 146); the specificity of binding has been established. The uncharged tRNA^Thr competes with this leader sequence structure for binding of the synthetase, thus leading to translation of the coding sequence, when the tRNA is predominantly uncharged(146).

Lysyl-tRNA Synthetases

E. coli has two distinct genes encoding lysyl-tRNA synthetase activity (EC 6.1.1.6), lysS and lysU (83), and is thus one of the few exceptions to the rule of one synthetase per amino acid, the other being the threonyl-tRNA synthetase of B. subtilis (135). The lysS gene is located at min 62.1 on the chromosome (prfB-lysS operon [58]), and lysU is located at min 93.5 (174). The sequences are highly homologous (85% identity for amino acids and 80% identity for nucleotide sequences). Codon usage analysis indicates that the lysS message is more efficiently translated than that of lysU (38, 102). The M _r of each of the proteins is approximately 57,500. They belong to class II of synthetases (50, 59); the proteins have domains of strong homology with aspartyl- and asparaginyl-tRNA synthetases and, surprisingly, with the E. coli asparagine synthetase, especially in a cluster of glycines that plays a role in the activity and stability of the proteins as shown by site-directed mutagenesis (72). The two lysS and lysU genes are regulated differently. Expression of lysS (formerly called herC) is constitutive, whereas expression of lysU was shown to depend on multiple external factors (high temperature, pH, anaerobiosis, and presence of leucine, serine, and different peptides [84]). Thus, a null lysS mutant exhibits cold-sensitive lethality (96), as lysU expression is high enough for phenotypic complementation only at high temperatures; it is suppressed by a multicopy plasmid carrying lysU. A null lysU mutant exhibits no clear-cut phenotype except for slow growth (45% of normal) at 44°C (38). It has been shown recently that lysU belongs to the lrp regulon (73, 103), the Lrp protein binding to sites upstream of the lysU promoter. Such regulation largely explains the role of the multiple effectors, but perhaps not the thermoinducibility which could act at a translational level (85).

As with many tRNA synthetases, the lysyl-tRNA synthetases of E. coli and S. typhimurium synthesize dinucleoside oligophosphate Ap4A in vitro as well as in vivo after triggering by different stresses (101); their role in such a synthesis has been studied (21, 132, 193), though the biological role of this metabolite is unclear (for a review, see reference 133).

ACKNOWLEDGMENTS

I express my gratitude to G. N. Cohen for his constant interest for these many years. I thank all colleagues who provided information before publication, and I thank J. De Moss and H. Rickenberg for help in preparation of the manuscript. The work performed in my laboratory has been supported by grants from the Centre National de la Recherche Scientifique and the Université d’Aix-Marseille II.