Knowns and Unknowns of Vitamin B₆ Metabolism in Escherichia coli

ABSTRACT

Vitamin B6 is an ensemble of six interconvertible vitamers: pyridoxine (PN), pyridoxamine (PM), pyridoxal (PL), and their 5′-phosphate derivatives, PNP, PMP, and PLP. Pyridoxal 5′-phosphate is a coenzyme in a variety of enzyme reactions concerning transformations of amino and amino acid compounds. This review summarizes all known and putative PLP-binding proteins found in the Escherichia coli MG1655 proteome. PLP can have toxic effects since it contains a very reactive aldehyde group at its 4′ position that easily forms aldimines with primary and secondary amines and reacts with thiols. Most PLP is bound either to the enzymes that use it as a cofactor or to PLP carrier proteins, protected from the cellular environment but at the same time readily transferable to PLP-dependent apoenzymes. E. coli and its relatives synthesize PLP through the seven-step deoxyxylulose-5-phosphate (DXP)-dependent pathway. Other bacteria synthesize PLP in a single step, through a so-called DXP-independent pathway. Although the DXP-dependent pathway was the first to be revealed, the discovery of the widespread DXP-independent pathway determined a decline of interest in E. coli vitamin B₆ metabolism. In E. coli, as in most organisms, PLP can also be obtained from PL, PN, and PM, imported from the environment or recycled from protein turnover, via a salvage pathway. Our review deals with all aspects of vitamin B₆ metabolism in E. coli, from transcriptional to posttranslational regulation. A critical interpretation of results is presented, in particular, concerning the most obscure aspects of PLP homeostasis and delivery to PLP-dependent enzymes.

PLP HOMEOSTASIS IN THE ESCHERICHIA COLI CELL

Pyridoxal 5′-phosphate (PLP) is one of the six interconvertible vitamers whose ensemble goes by the name of vitamin B₆. The other B₆ vitamers are pyridoxine (PN), pyridoxal (PL), pyridoxamine (PM), and the 5′-phosphate derivatives PNP and PMP. PLP is a highly versatile catalyst that acts as a cofactor for several enzymes, mostly by reacting with amino acids. The catalytic power of PLP arises from the ability of its aldehyde group to form imines with the amino group of substrates and to the action of its pyridine ring as an electron sink that withdraws electrons from the substrate. PLP is bound to enzymes through a Schiff base linkage with a lysine residue present at the active site. Despite the high variability in the structure and function of PLP-dependent enzymes, many of the residues interacting with the coenzyme are conserved (1). Almost 1.5% of all genes in most prokaryotic genomes encode PLP-dependent enzymes (2). They catalyze many different transformations, such as transamination, decarboxylation, racemization, side chain cleavage, β- and γ-elimination, and replacement reactions (1). Using the B₆ database (2), a tool for the description and classification of vitamin B₆-dependent enzyme activities and of the corresponding protein families (http://bioinformatics.unipr.it/B6db), all known and putative PLP-binding proteins found in the E. coli MG1655 proteome (57 in total) were retrieved, and they are summarized in Table 1. Twenty enzymes are involved in amino acid biosynthesis, while 10 are responsible for the biosynthesis of other molecules, such as coenzymes, polysaccharides, and polyamines. Many are catabolic enzymes, and a putative PLP carrier protein (PLPHP, see below) and two PLP-dependent transcriptional regulators (YdcR and YjiR, see below) are also present. Importantly, in E. coli, 26% of these PLP-dependent proteins are essential for autotrophic growth, as the corresponding deletion strains are unable to grow in minimal medium (https://ecocyc.org/). This underlines the importance of PLP for bacterial life.

TABLE 1.

PLP-binding proteins in E. coli

Protein name	Activity	UniProt ID	EC no.	Biological process	Growth in minimal mediuma
IlvE	Branched-chain-amino-acid aminotransferase	P0AB80	2.6.1.42	Branched-chain amino acid biosynthesis	No
IlvA	Threonine deaminase	P04968	4.3.1.19	Branched-chain amino acid biosynthesis	No
CysK	Cysteine synthase A	P0ABK5	2.5.1.47	Cys biosynthesis	No
CysM	Cysteine synthase B	P16703	2.5.1.47	Cys biosynthesis	Yes
TrpB	Tryptophan synthase, β	P0A879	4.2.1.20	Trp biosynthesis	No
LysA	Diaminopimelate decarboxylase	P00861	4.1.1.20	Lys biosynthesis	No
ArgD	N-acetylornithine aminotransferase/N-succinyldiaminopimelate aminotransferase	P18335	2.6.1.11	Arg and Lys biosynthesis	Yes
ThrC	Threonine synthase	P00934	4.2.3.1	Thr biosynthesis	No
MetB	O-succinylhomoserine(thiol)-lyase	P00935	2.5.1.48	Cys and Met biosynthesis	No
MetC	Cystathionine β-lyase/l-cysteine desulfhydrase/alanine racemase	P06721	4.4.1.28	Met biosynthesis	No
AlaC	Glutamate-pyruvate aminotransferase	P77434	2.6.1.2	Ala biosynthesis	Yes
AlaA	Glutamate-pyruvate aminotransferase	P0A959	2.6.1.2	Ala biosynthesis	Yes
TyrB	Tyrosine aminotransferase	P04693	2.6.1.57	Asp, Tyr, Leu, Phe biosynthesis	Yes
HisC	Histidinol-phosphate aminotransferase	P06986	2.6.1.9	His biosynthesis	No
MalY	Cystathionine β-lyase	P23256	4.4.1.13	Met biosynthesis	Yes
AspC	Aspartate aminotransferase	P00509	2.6.1.1	Phe biosynthesis	Yes
LtaE	l-threonine aldolase	P75823	4.2.1.48	Gly biosynthesis	Yes
Alr	Alanine racemase 1	P0A6B4	5.1.1.1	d-Ala biosynthesis	Yes
DadX	Alanine racemase 2	P29012	5.1.1.1	d-Ala biosynthesis	Yes
AvtA	Valine-pyruvate aminotransferase	P09053	2.6.1.66	Ala and Val biosynthesis	Yes
SerC	Phosphoserine/phosphohydroxythreonine aminotransferase	P23721	2.6.1.52	Ser, Lys, and PLP biosynthesis	No
SpeA	Arginine decarboxylase, biosynthetic	P21170	4.1.1.19	Polyamine biosynthesis	Yes
SpeC	Ornithine decarboxylase	P21169	4.1.1.17	Spermidine and putrescine biosynthesis	Yes
SelA	Selenocysteine synthase	P0A821	2.9.1.1	Selenocysteine biosynthesis	Yes
WecE	dTDP-4-dehydro-6-deoxy-d-glucose transaminase	P27833	2.6.1.33	Polysaccharide biosynthesis	Yes
ArnB	UDP-4-amino-4-deoxy-l-arabinose aminotransferase	P77690	2.6.1.87	Polysaccharide biosynthesis	Yes
PabC	Aminodeoxychorismate lyase	P28305	4.1.3.38	Folate biosynthesis	Yes
BioF	8-amino-7-oxononanoate synthase	P12998	2.3.1.47	Biotin biosynthesis	No
BioA	Adenosylmethionine-8-amino-7-oxononanoate aminotransferase	P12995	2.6.1.62	Biotin biosynthesis	No
HemL	Glutamate-1-semialdehyde aminotransferase	P23893	5.4.3.8	Tetrapyrrole biosynthesis	No; also in LB
TdcB	Catabolic threonine dehydratase	P0AGF6	4.3.1.19	Ser and Thr catabolism	No
DsdA	d-serine ammonia-lyase	P00926	4.3.1.18	Ser and Thr catabolism	Yes
Kbl	2-Amino-3-ketobutyrate coenzyme A ligase	P0AB77	2.3.1.29	Thr catabolism	Yes
AstC	Succinylornithine transaminase	P77581	2.6.1.81	Arg and ornithine metabolism	Yes
GlyA	Serine hydroxymethyltransferase	P0A825	2.1.2.1	Gly, Ser, and tetrahydrofolate metabolism	No
GadA	Glutamate decarboxylase A	P69908	4.1.1.15	Glu metabolism	Yes
GadB	Glutamate decarboxylase B	P69910	4.1.1.15	Glu metabolism	Yes
GcvP	Glycine decarboxylase	P33195	1.4.4.2	Gly metabolism	Yes
TnaA	Tryptophanase/l-cysteine desulfhydrase	P0A853	4.1.99.1	Trp metabolism	Yes
CadA	Lysine decarboxylase 1	P0A9H3	4.1.1.18	Lys metabolism	Yes
LdcC	Lysine decarboxylase 2	P52095	4.1.1.18	Lys metabolism	Yes
YgeX	2,3-Diaminopropionate ammonia-lyase	P66899	4.3.1.15	d-Ser catabolism	Yes
DcyD	d-cysteine desulfhydrase	P76316	4.4.1.15	d-amino acid metabolism	Yes
GabT	4-Aminobutyrate aminotransferase	P22256	2.6.1.19	Gamma-aminobutyrate metabolism	Yes
PuuE	4-Aminobutyrate aminotransferase	P50457	2.6.1.19	Gamma-aminobutyrate metabolism	Yes
PatA	Putrescine aminotransferase	P42588	2.6.1.82	Putrescine catabolism	Yes
YbdL	Methionine-oxo-acid transaminase	P77806	2.6.1.88	l-kynurenine metabolism	Yes
EpmB	Lysine 2,3-aminomutase	P39280	5.4.3.2	Posttranslational modification	Yes
GlgP	Glycogen phosphorilase	P0AC86	2.4.1.1	Glycogen metabolism	Yes
MalP	Maltodextrin phosphorylase	P00490s	2.4.1.1	Glycogen metabolism	Yes
SufS	l-cysteine desulfurase	P77444	2.8.1.7	Iron-sulfur cluster assembly	Yes
CsdA	Cysteine sulfinate desulfinase	Q46925	2.8.1.7	Iron-sulfur cluster assembly	Yes
YdcR	MocR-like transcription factor	P77730		Regulation of transcription	Yes
YjiR	MocR-like transcription factor	P39389		Regulation of transcription	Yes
PLPHP	PLP homeostasis protein	P67080		PLP binding	Yes
YhfX	Putative	P45550			Yes
YhfS	Putative aminotransferase	P45545			Yes

^aGrowth of corresponding knockout strain; from https://ecocyc.org/. Different colors correspond to different functions, as specified in “Biological process” column.

Besides its role as a cofactor in catalytic processes, PLP plays other roles inside the bacterial cell. E. coli is a facultative anaerobe which thrives in either the presence or absence of oxygen; therefore, it has to counteract reactive oxygen species (ROS). It has been demonstrated that increasing concentrations of PN and PLP increase O₂ decay in vitro (3). Moreover, these B₆ vitamers are more efficient quenchers than sulfur-containing antioxidants and quench O₂ at a rate comparable to that of vitamins E and C (3). Of all the B₆ vitamers, PN shows the highest antioxidant activity, being able to scavenge up to eight ·OH radicals under high hydroxyl radical concentrations (4).

PLP also acts as a cofactor and effector molecule for a group of bacterial transcriptional regulators, called MocR transcription factors (MocR-TFs), chimeric proteins formed by the joining of a helix-turn-helix DNA binding domain and the protein scaffold of fold type I PLP-dependent enzymes (5, 6). A member of the MocR subfamily, PdxR, plays a fundamental role in the transcriptional regulation of vitamin B₆ biosynthesis in Corynebacterium glutamicum (7), Streptococcus pneumoniae (8), Listeria monocytogenes (9), and Bacillus clausii (10). In these bacteria, PdxR activates the transcription of the pdxST operon, which encodes the subunits of the PLP synthase complex (Fig. S1) that synthetizes PLP directly from glutamine, ribulose 5-phosphate, and glyceraldehyde 3-phosphate (11). As detailed below, in E. coli, PLP is synthetized through a different and more complex pathway.

The total number of vitamin B₆ molecules per cell was measured in the HMS174 E. coli strain using an enzyme assay and reported to be about 188,000, with 42% of these in the PLP form, 51% in a combination of the PNP and PMP forms, and 7% as PL. The levels of the PN and PM vitamers were under the detection limit. The same work estimated that about 60% of intracellular PLP is bound to proteins (12). PLP contains a very reactive aldehyde group at its 4′ position that easily forms aldimines with primary and secondary amines and reacts with thiols. Therefore, PLP can have toxic effects, and nonspecific binding to cell components should be kept to a minimum. Presumably, most PLP is bound either to the enzymes that use it as a cofactor or to PLP carrier proteins, such as PLPHP, formerly named YggS. The latter is a ubiquitous PLP-binding protein, with no catalytic activity, and is structurally homologous to the fold type III PLP-dependent enzymes, of which bacterial alanine racemase is the archetype (13). Although the metabolic role of this carrier protein has not been entirely characterized, the phenotype of E. coli yggS knockout cells reflects an imbalance in PLP homeostasis (14). Besides PLPHP, two essential proteins involved in PLP biosynthesis and salvage pathways (see below), PdxH and PdxK, have also been proposed to act as PLP-carrier proteins. PdxH tightly binds PLP deriving from its own catalytic activity at a secondary site, distinct from the active site (15). Our group has recently demonstrated that this secondary site coincides with an allosteric site that regulates the activity of the enzyme through PLP feedback inhibition (16). In a similar fashion, PLP produced during the E. coli PdxK catalytic turnover binds to a lysine residue positioned near the active site, thus inhibiting its activity (17). All these PLP carriers may act as a protein-bound reservoir of PLP, protecting it from the cellular environment but at the same time making it readily transferable to PLP-dependent apoenzymes (Fig. 1). It has been already demonstrated in vitro that both PdxH and PdxK are able to transfer the tightly bound PLP to apoenzymes (17, 18).

FIG 1.

Scheme of PLP homeostasis in E. coli. This figure summarizes all the different factors contributing to PLP homeostasis in the E. coli cells. The pool of free PLP is in equilibrium with other B₆ vitamers, and PLP bound to specific PLP-binding protein. PLP supply processes are in blue, whereas consuming processes are in red.

In the absence of an external source of vitamin B₆, the main supply of PLP comes from its de novo biosynthesis (Fig. 1). E. coli synthesizes PLP through the so-called deoxyxylulose-5-phosphate (DXP)-dependent pathway, using the primary metabolites erythrose-4-phosphate, glyceraldehyde-3-phosphate, and pyruvate (Fig. 2). This pathway consists of two branches. In one branch, the d-erythrose-4-phosphate sugar, coming from the pentose phosphate pathway, is converted to 3-amino-1-hydroxyacetone 1-phosphate through three reactions. In the other branch, a single enzyme condenses the glycolytic intermediates pyruvate and d-glyceraldehyde-3-phosphate to yield DXP. Finally, DXP and 3-amino-1-hydroxyacetone 1-phosphate are condensed in a reaction catalyzed by PNP synthase (PdxJ) that produces PNP, which is then oxidized to PLP by PdxH. In E. coli, as in most organisms, PLP can also be obtained from PL, PN, and PM, imported from the environment or recycled from protein turnover, via a salvage pathway (Fig. 1). This pathway interconverts the different forms of vitamers using the following enzymes: pyridoxine (pyridoxamine) 5'-phosphate oxidase (PdxH), nonspecific and specific phosphatases (such as YbhA (19)), the kinase PdxK, which is also able to phosphorylate PL, PN, and PM (20), and the kinase PdxY, which is specific for PL (21). Recently, PL reductase, which converts PN to PL, has been proposed as an important component of the salvage pathway (22) (Fig. 2).

FIG 2.

Scheme of PLP biosynthesis and salvage pathways. Primary metabolites, erythrose-4-phosphate, glyceraldehyde-3-phoshpate, and pyruvate, used as starting substrates, are in green, whereas intermediate metabolites are in black. Enzymes are in blue.

If B₆ transporters of the ECF (energy-coupling factor) family have been identified in other bacteria (23), little is known about the transport of B₆ vitamers across the E. coli cellular membrane. It was suggested that PLP is not imported into E. coli cells, while dephosphorylated vitamers are imported (24). Dated evidence has shown that in Salmonella typhimurium and in lactic bacteria, PN and PL, but not PM, can enter the cell by facilitated diffusion. Then, phosphorylation of the 5′ hydroxyl group by pyridoxal kinase efficiently traps the vitamers inside the cell (25, 26). The identity of the transmembrane protein involved in the transport of B₆ vitamers across the membrane is not known; nor is it known whether a specific protein plays such a role.

Factors contributing to the lowering of the intracellular PLP pool are dilution by growth and chemical damage (Fig. 1). In fact, PLP is one of the more damage-prone metabolites that face spontaneous chemical reactions (27). No evidence has been reported about the enzyme catabolism of PLP in E. coli. Two catabolic pathways were identified in Pseudomonas, Arthrobacter, and Mesorhizobium loti, which through either 8 or 5 steps, respectively, give as final products acetate, ammonia, carbon dioxide, and either succinic semialdehyde or 2-(hydroxymethyl)-4-oxobutanoate (28); however, the complete pathway has not yet been identified in other microorganisms whose genome has been sequenced. It has been reported that in rat, the oxidation of PL to 4-pyridoxic acid is catalyzed by a NAD⁺-dependent aldehyde dehydrogenase (29). In E. coli, a similar activity could be linked to AldB, which catalyzes the NADP-dependent oxidation of diverse aldehydes, such as benzaldehyde (30).

It should be noted that if PM seems absent from the E. coli cell (12), PMP may derive from aminotransferase reactions.

GENES INVOLVED IN VITAMIN B6 METABOLISM

Genes involved in vitamin B₆ metabolism are spread all over the E. coli chromosome. Only pdxH and pdxY are cotranscribed (Fig. 3). At the same time, almost all genes encoding enzymes involved in the PLP biosynthesis and salvage pathways are part of complex operons. Because bacterial operons often contain genes with related functions, it has been proposed that the inclusion of PLP biosynthetic genes in complex, multifunctional operons may act genetically to integrate PLP biosynthesis into several branches of intermediary metabolism (31). In the following paragraphs, all the operons and genes related to PLP metabolism are described.

FIG 3.

Schematic representations of operons involved in PLP homeostasis. Numbers indicate the genome positions in kb. The blue arrows represent mRNAs. The position of promoters is indicated.

Genes Involved in PLP Biosynthesis

epd

The epd (or gapB) gene located between 3,072,672 and 3,073,691 kb on the bacterial chromosome encodes d-erythrose-4-phosphate dehydrogenase (UniProt P0A9B6). This tetrameric enzyme catalyzes the first reaction in the de novo PLP biosynthesis, namely, the NAD-dependent oxidation of d-erythrose-4-phosphate (Fig. 2). This enzyme is homologous to glyceraldehyde 3-phosphate dehydrogenase (GAPDH, encoded by the gapA gene), a key enzyme of glycolysis and gluconeogenesis, with which it shares more than 40% amino acid sequence identity (32). However, Epd shows an efficient nonphosphorylating erythrose-4-phosphate dehydrogenase activity and a low phosphorylating glyceraldehyde-3-phosphate dehydrogenase activity; in contrast, GAPDH shows a highly efficient phosphorylating glyceraldehyde-3-phosphate dehydrogenase activity and a low nonphosphorylating erythrose-4-phosphate dehydrogenase activity (33). The low nonphosphorylating erythrose-4-phosphate dehydrogenase activity of the highly abundant gapA-encoded GAPDH allows growth of an epd mutant strain (34).

The epd gene is cotranscribed either with pgk and fbaA or with only pgk (Fig. 3A) (35). The pgk and fbaA genes are essential genes encoding the glycolytic enzymes phosphoglycerate kinase (UniProt P0A799) and fructose-bisphosphate aldolase (UniProt P0AB71), respectively. The expression of epd might also be disjointed from that of pgk and fbaA, as these two genes are also transcribed from three promoters located downstream of epd (35).

pdxB

The pdxB gene (located between position 2,436,715 and 2,437,851 kb on the E. coli chromosome) encodes 4-phosphoerythronate dehydrogenase (UniProt P05459), which catalyzes the second reaction of the PLP biosynthetic pathway, consisting in the NAD-dependent oxidation of 4-phosphoerythronate to 2-oxo-3-hydroxy-4-phosphobutanoate (Fig. 2). This protein is homologous to d-3-phosphoglycerate dehydrogenase (Fig. S2), encoded by serA, which catalyzes the first step of the l-serine biosynthetic pathway (36).

In E. coli, pdxB is the first gene of a multifunctional operon, which contains three other genes: usg, truA, and dedA (Fig. 3B). The usg gene encodes a putative semialdehyde dehydrogenase (UniProt P08390), as suggested by its sequence similarity to aspartate β-semialdehyde dehydrogenase (37). The truA gene product is tRNA pseudouridine synthase I (UniProt P07649), which catalyzes the modification of specific uridine residues in the anticodon stem and loop of certain tRNA species. The resulting pseudouridine residues improve the efficiency of translation by stabilizing the structure of tRNAs and ensuring reading frame maintenance (38). The last gene of the pdxB operon is dedA, encoding an uncharacterized inner membrane protein (UniProt P0ABP6), a member of the DedA family.

Of these four genes, only pdxB is absolutely required for growth of bacteria on minimal medium, as the pdxB mutant is auxotrophic for PN (39). Although these genes form an operon, their functions appear to be unrelated to each other. However, the close association of pdxB and truA is evolutionarily conserved among different enterobacterial species (39).

Two different transcription units were identified: the first unit includes all four genes forming the operon, and it is transcribed from the P_pdxB promoter, whereas the other unit contains the last three genes and is transcribed from an internal promoter (P_usg) located near the end of the pdxB coding region (40). The P_pdxB promoter shares the same –10 region with an overlapping divergent promoter which controls the expression of the flk gene encoding a flagellar regulator (UniProt P15286) (36, 41). However, a functional association between the pdxB and flk genes is yet to be found.

serC

The serC gene (located between position 957,653 and 958,741 kb on the E. coli chromosome) codes for phosphoserine/phosphohydroxythreonine aminotransferase (UniProt P23721), an enzyme involved in both PLP and l-serine biosynthesis by using different substrates (42). In the PLP biosynthesis pathway, SerC adds an amino group to 2-oxo-3-hydroxy-4-phosphobutanoate (produced by PdxB) to form 3-hydroxy-4-phospho-l-threonine (Fig. 2); in the serine pathway, SerC adds an amino group to 3-phosphohydroxypyruvate to form 3-phosphoserine (Fig. S2). In both transamination reactions, the amino-group donor is l-glutamate, which is then converted into 2-oxoglutarate. The enzyme does not show any activity with nonphosphorylated substrates (43). Interestingly, since SerC needs PLP as a cofactor for its enzyme activities, PLP is involved in its own biosynthesis (43).

The serC gene forms an operon with the aroA gene, which encodes the enzyme 3-phosposhikimate 1-carboxyvinyltransferase (UniProt P0A6D3) (44). This enzyme is involved in the chorismate pathway, which leads to the biosynthesis of aromatic amino acids. Mutants of serC are auxotrophic for l-serine and PN, while aroA mutants require aromatic amino acids; thus, both genes are essential for growth in minimal medium (42). The serC-aroA operon is transcribed from the P_serC promoter located upstream of serC (Fig. 3C). As a consequence, the number of encoded enzymes may be simultaneously regulated at the transcriptional level. However, serC can also be transcribed as a monocistronic mRNA from the same promoter. In addition, a Rho factor-independent transcription terminator located between serC and aroA is responsible for transcriptional attenuation, which leads to an increased expression of serC with respect to aroA (45).

The arrangement of serC and aroA genes in a single operon does not seem to be accidental, since it can guarantee a coordinated regulation of PLP, serine, and chorismate biosynthetic pathways. In fact, it should be pointed out that these pathways are connected by three important features. First, PLP and chorismate pathways share erythrose 4-phosphate as the starting substrate. Second, equimolar amounts of serine and chorismate are required for the biosynthesis of the iron chelator enterochelin. Third, PLP, serine, and aromatic compounds are also linked in numerous steps and branches of intermediary metabolism. For example, chorismate is the precursor of tetrahydrofolate, which captures one-carbon units when l-serine is cleaved to glycine by the PLP-dependent enzyme serine hydroxymethyltransferase (45).

pdxA

4-Hydroxythreonine-4-phosphate dehydrogenase (UniProt P19624), encoded by the pdxA gene (encompassing position 52,427 to 53,416 kb of the E. coli chromosome), catalyzes the fourth step in the PLP biosynthesis pathway (Fig. 2). PdxA is a NAD⁺-dependent dehydrogenase responsible for the oxidation of 4-hydroxythreonine-4-phosphate to 3-amino-1-hydroxyacetone-1-phosphate (46). A pdxA mutant is able to grow on minimal medium only if supplemented with PN or PL.

The pdxA gene is part of a complex operon, which can give rise to different transcription units: lptD (from either the impp1 or impp2 promoters), lptD-surA-pdxA (from the impp3 promoter), surA-pdxA-rmsA-apaGH, pdxA-rsmA-apaGH, pdxA-rmsA, rmsa-apaGH, and apaGH (47) (Fig. 3D).

The first gene of the operon, lptD (lipopolysaccharide transport), encodes an essential protein of the outer membrane (UniProt P31554), required for lipopolysaccharide assembly (48). Considering that the decrease of lptD expression causes an increase of sensitivity to organic solvents (49), and lptD mutants show an increase in the membrane permeability (50), LptD is important for both organic solvent tolerance and outer membrane integrity. The surA gene, downstream of lptD, codes for a periplasmic peptidyl-prolyl isomerase (UniProt P0ABZ6), which is a chaperone required for folding and assembly of several outer membrane proteins, including OmpA, OmpF, and LamB. It has been reported that surA mutants are unable to survive in stationary phase at high pH (51). surA expression increases in cultures of high density, as well as at low or high pH conditions. Another gene forming the operon is rsmA (also denoted as ksgA), whose gene product is a methyltransferase (UniProt P06992) that catalyzes the methylation of two adjacent adenosines in the 16S rRNA of the 30S ribosomal subunit. Interestingly, the KsgA protein binds specifically to its own mRNA, and it may regulate its own translation (52). It has been shown that the mutation of rsmA induces resistance to kasugamycin, an inhibitor of translation initiation, without substantial growth defects. The penultimate gene of the operon is apaG, encoding a DUF525 domain-containing protein (UniProt P62672), whereas the last gene is apaH, coding for a diadenosine tetraphosphatase (UniProt P05637). Both apaG and apaH belong to a group of genes involved in stress-activated mutagenesis (53). In particular, ApaH hydrolyzes diadenosine tetraphosphate (Ap₄), produced as a side product of aminoacyl-tRNA synthetases, as well as molecules related to Ap₄. Recently, it was reported that ApaH removes caps containing Ap₄ and Gp₄ at the 5′ end of RNA molecules, generating a diphosphorylated RNA product (54).

Concerning the regulation of all these genes, the lptD-surA-pdxA-rsmA-apaG-apaH operon contains at least eight promoters. In the P_lptD region, located upstream of lptD (Fig. 3D), three promoter sequences were identified: impp1, impp2, and impp3. The first two depend on the σ^D factor and are responsible for the constitutive expression of the lptD gene in equal measure (55), whereas the impp3 sequence is σ^E-dependent and controls the transcription of the polycistronic lptD-surA-pdxA mRNA. Specifically, the σ^E factor is the sigma factor of RNA polymerase responding to heat shock or other environment stresses that cause an accumulation of misfolded polypeptides (56). P_surA is another promoter in this operon recognized by the σ^E factor that leads to the formation of the surA-pdxA-rmsA-apaGH cotranscript. Interestingly, surA is always cotranscribed with pdxA, and insertion mutants in surA cause PN auxotrophy (41). The transcription of pdxA also occurs from the P_pdxA promoter, which drives the formation of the pdxA-rsmA-apaGH and pdxA-rmsA transcripts. Two other promoters are present downstream of pdxA: the P_rmsA promoter, located at the end of the pdxA gene (47), and a promoter within the rmsA gene, from which the transcription of both the apaG and apaH genes initiates (57).

dxs

The essential dxs gene (position 438,315 to 440,177 kb, in the genome) encodes 1-deoxy-d-xylulose-5-phosphate synthase (UniProt P77488), which catalyzes the thiamine diphosphate-dependent condensation between pyruvate and d-glyceraldehyde-3-phosphate (GAP) to yield DXP (Fig. 2). This metabolite is also an intermediate of the isoprenoid and thiamine biosynthesis pathways (58, 59).

The transcript containing dxs is the xseB-ispA-dxs-yajO mRNA, which originates from the P_xseB promoter (58) (Fig. 3E). The xseB gene product is the small subunit of exonuclease VII (UniProt P0A8G9). Like dxs, the ispA gene is involved in the isoprenoid biosynthesis. It is an essential gene encoding the enzyme geranyl diphosphate/farnesyl diphosphate synthase (UniProt P22939), which catalyzes two sequential reactions in the polyisoprenoid biosynthetic pathway (60). The yajO gene can also be transcribed as a single transcriptional unit from its own promoter. It is functionally linked to dxs, since it encodes an enzyme that catalyzes the synthesis of DXP from ribulose 5-phosphate (1-deoxyxylulose-5-phosphate synthase; UniProt P77735), providing an alternative, albeit less efficient, route to supply DXP (61).

pdxJ

Pyridoxine 5′-phosphate synthase (PdxJ; UniProt P0A794) is a homooctameric enzyme that catalyzes the complex intramolecular condensation reaction between DXP and 3-amino-1-hydroxyacetone 1-phosphate to form PNP (Fig. 2) (62, 63).

The pdxJ gene, located in the E. coli chromosome between 2,700,998 and 2,701,729 kb, can be transcribed from two different promoters, which give two different transcription units. The P_rnc promoter directs transcription of five genes (rnc-era-recO-pdxJ-acpS), while transcription of the pdxJ-acpS mRNA initiates from P_pdxJ (Fig. 3F) (64). The rnc and era genes are both important for efficient ribosomal processing and maturation. The rnc gene encodes RNase III (RNase III; UniProt P0A7Y0), an endonuclease specific for double-stranded RNA. In E. coli, the main substrates for RNase III are the rRNAs transcribed from the seven E. coli rRNA operons leading to rRNA processing. However, this endonuclease is considered a global regulator because, by cleaving transcripts at either the 3' or 5' untranscribed regions of genes, but also within coding regions, it plays a general role in posttranscriptional gene expression control. Moreover, RNase III is involved in the defense against viral infection (65). Despite the important role that RNase III plays in rRNA processing, it is not essential for cell viability. In the absence of RNase III, the full-length ribosomal operon transcript is processed by other ribonucleases to generate 16S and 23S rRNAs, albeit more slowly (66). In contrast, era is essential for viability. It encodes a GTPase (UniProt P06616) that interacts with RNase III at the level of the precursor rRNA, carrying out the maturation of the 30S ribosomal subunit. RecO (UniProt P0A7H3) is involved in RecA-mediated homologous recombination (67). The downstream acpS gene is also essential for viability (64). It encodes the holo-acyl-carrier-protein synthase (AcpS; UniProt P24224) involved in lipid synthesis by transferring the 4-phosphopantetheine moiety of CoA to the acyl-carrier protein, activating this important component of fatty acid biosynthesis (68).

Three major transcripts are found in this unit (Fig. 3F). The abundant 1.9-kb rnc-era and the less-abundant 3.7-kb rnc-era-recO-pdxJ-acpS mRNAs are expressed from P_rnc, whereas the 1.3-kb pdxJ-acpS mRNA is transcribed from P_pdxJ (67). Expression studies carried out using various genetic fusions showed that pdxJ is mainly transcribed from P_rnc but is translated 2.5-fold more efficiently from the 1.3-kb pdxJ-acpS mRNA than from the 3.7-kb rnc-era-recO-pdxJ-acpS transcript. The expression of the rnc-era-recO-pdxJ-acpS operon is negatively autoregulated at the level of mRNA stability by RNase III, which cleaves a stem-loop in its own 5' untranscribed region, causing the mRNA to become vulnerable to degradation by other ribonucleases. In contrast, no apparent effect of RNase III is observed on the P_pdxJ-dependent 1.3-kb pdxJ-acpS mRNA (67).

pdxH

The pdxH gene (located at position 1,717,351 to 1,718,007 kb of the E. coli chromosome) encodes the flavin mononucleotide-dependent enzyme (UniProt P0AFI7) catalyzing the oxidation of PNP to PLP, as the final step of PLP biosynthesis, and the oxidation of PMP to PLP in the PLP salvage pathway (Fig. 2). In both reactions, the final electron acceptor is oxygen, which is reduced to H₂O₂ (31). pdxH mutants are PL auxotrophs (31) but do not grow as the wild-type strain even in the presence of a high concentration of this vitamer (22). pdxH is an essential gene in both the presence or absence of oxygen, suggesting that an unidentified compound acts as the electron acceptor in the absence of O₂ (31).

Two different transcription units are found in this operon (pdxH-tyrS-pdxY and tyrS-pdxY) (Fig. 3G), making it the sole case where genes involved in PLP biosynthesis and salvage pathways are cotranscribed. Indeed, downstream from the tyrS essential gene, encoding tyrosyl-tRNA synthetase (UniProt P0AGJ9), responsible for the attachment of l-tyrosine to tRNA(Tyr), is the pdxY gene encoding pyridoxal kinase 2 (UniProt P77150; see next paragraph). Importantly, tyrS mutations are lethal. About 20% of tyrS transcripts originate from the P_pdxH promoter, whereas the remaining 80% originate from the P_tyrS promoter. This latter promoter is about four times stronger than P_pdxH, explaining the difference in steady-state amounts of the pdxH-tyrS and tyrS-pdxY transcripts (31). A possible, independent pdxY promoter has also been identified by a genome-wide study using gfp transcriptional fusions (69) (Fig. 3G). Without this last promoter, pdxY transcription would originate only from the P_pdxH and P_tyrS promoters. Therefore, tyrS and pdxY would always be cotranscribed. Blockage of transcription from P_pdxH, by an insertion in pdxH, reduced the amount of the tyrS transcript by about 20%, while it reduced PL kinase activity by about 25%. This suggests a low level of coupling between the transcription of pdxH and that of pdxY. About 92% of tyrS transcripts terminate at a terminator located between tyrS and pdxY, and only about 8% read through into pdxY. Therefore, this terminator attenuates pdxY expression compared to tyrS. Nevertheless, cotranscription of pdxY and tyrS may provide a point of genetic integration that coordinates incorporation of amino acids into proteins with PLP coenzyme supply (21).

Evolutionary Considerations on the DXP-Dependent Pathway

A phylogenetic and comparative genomics analysis carried out in 2001 indicated that the DXP-dependent pathway of PLP biosynthesis is mainly limited to E. coli and gammaproteobacteria (70). This information concerning the distribution of vitamin B₆ biosynthesis pathways should be reviewed in light of the current increased availability of genomic data. Archaea, fungi, plants, and many bacteria rely on the action of the PLP synthase complex (coded by the Pdx1 and Pdx2 genes in plants, and by the pdxST operon in bacteria), which directly produces PLP from glutamine, either ribose or ribulose 5-phosphate, and either glyceraldehyde 3-phosphate or dihydroxyacetone phosphate (Fig. S1) (11). Since the γ division of proteobacteria represents the most recent lineage of prokaryotic evolution, it has been suggested that, in the primordial environment where these bacteria lived, PLP was present in the environment, so that the Pdx1-Pdx2 function was lost from their genome (11); however, when, following changes in the environment, PLP became scarcer and the need to produce it arose, gammaproteobacteria did not evolve a novel PLP biosynthesis pathway from scratch but mostly drew on the existing metabolisms to create a pathway that produces PNP (36). The salvage pathway evolved before this event (70), and therefore PNP oxidase was already present to convert PNP into PLP. On the other hand, PdxJ is evolutionarily distinct but highly similar to Pdx1 in structural and mechanistic terms, suggesting that it is an example of convergent evolution (11). The function of Dxs is not only limited to PLP biosynthesis but is also involved in isoprenoid biosynthesis (58). The first branch of PLP biosynthesis probably derives from the serine biosynthesis pathway by gene recruitment (36). In fact, at a glance, the intermediates of the two routes are very similar (Fig. 2 and Fig. S2). Moreover, the two pathways share the same enzyme, SerC, and as mentioned before, pdxB and serA are homologous (36). We also want to point out that strong evidence supporting this hypothesis derives from the observation that a ΔpdxB strain is still able to slowly grow at 30°C in the absence of vitamin B₆ supplementation (42), suggesting that PLP can be synthesized by different routes. In 2010, Kim et al., by overexpressing different E. coli genes, characterized a serendipitous pathway that leads to PLP synthesis in the absence of PdxB. This pathway uses an intermediate of serine biosynthesis, 3-phosphohydroxypyruvate, and three enzymes to bypass the reactions catalyzed by PdxB and SerC in PLP biosynthesis. It consists of NudL, a putative CoA pyrophosphohydrolase, which hydrolyzes 3-phosphohydroxypyruvate to 3-hydroxypyruvate (3HP) that is spontaneously decarboxylated to produce glycolaldehyde; LtaE, a low-specificity threonine aldolase whose physiological function is unknown, which condensates glycolaldehyde and glycine to a mixture of 4-hydroxythreonine and 4-hydroxy-allo-threonine; and finally, a promiscuous homoserine kinase (ThrB), which produces 4-phosphohydroxythreonine (71). Recently, a ΔpdxB strain, naturally selected upon 10 days growth in minimal medium containing glucose as the sole carbon source, was identified to contain a 4-step bypass pathway that restores PLP synthesis, producing levels of vitamers similar to those of the wild-type strain. The bypass pathway consists of multiple phosphatases, which produce erythronate from the normal substrate 4-phosphoerythronate, 3-phosphoglycerate dehydrogenase (SerA) that is capable of oxidizing erythronate to (3R)-3,4-dihydroxy-2-oxobutanoate (then converted to 4-hydroxythreonine by SerC), and lastly, ThrB, the kinase producing 4-phosphohydroxythreonine (72). It is worth noting that these serendipitous pathways tap into the serine biosynthesis pathway.

Genes of the Salvage Pathway

pdxK and pdxY

The pdxK gene, located at position 2,536,386 to 2,537,237 kb of the E. coli chromosome, is one of the few genes involved in PLP metabolism transcribed as a single transcription unit (Fig. 3H). It encodes pyridoxal kinase 1 (UniProt P40191), the enzyme that catalyzes the ATP-dependent phosphorylation of PL, PN, and PM (Fig. 2) (73). As mentioned above, pyridoxal kinase 2 (encoded by pdxY) was suggested to specifically phosphorylate PL. Therefore, in E. coli, two kinases able to phosphorylate B₆ vitamers are present. This finding derived from an in vivo study carried out on a series of E. coli pdxB mutant strains unable to carry out the de novo synthesis of PLP (73). In particular, a pdxB-pdxK mutant grew in minimal medium supplemented with PL but did not grow on medium supplemented with PN, whereas a pdxB-pdxY mutant grew when either PL or PN was present. In contrast, growth of the pdxB-pdxK-pdxY triple mutant was observed only if precursors that bypass the need for pdxB were present, but not on PL-, PN-, or PM-containing media. These experiments clearly demonstrated that PdxK and PdxY are the only physiologically significant PL, PN, and PM kinases in E. coli, with PdxY being specific for PL (73). The reason why two kinases of B₆ vitamers are present in the E. coli cell is unknown. It is worth noting that in vitro enzyme activity measurements using recombinantly expressed and purified PdxY showed that it is endowed with only 1% of the kinase activity measured with PdxK (20). For this reason, it has been suggested that PdxY may play the role of kinase in a different metabolic pathway and, at the same time, have enough PL kinase activity to support growth of E. coli in a medium supplemented with PL when both de novo PLP synthesis and PdxK activity are abolished. However, the crystal structure of recombinant PdxY, obtained from a purification procedure without addition of exogenous ligands, showed that both active sites of the enzyme dimer were filled by very tightly bound ligands. In one active site, a mixture of PL and PLP was present, while in the other active site, a covalently attached form of PL was visible. The presence of these tightly bound ligands made it unlikely that the protein is involved in a metabolic pathway different from the PLP salvage pathway (74). A possibility is that PdxY may only be active as a component of a multiprotein complex.

The pdxK promoter, mapped in a genome-wide study that identified several transcription start sites in E. coli (75), has yet to be characterized in detail. Downstream of pdxK, and transcribed in the opposite direction, yfeK encodes an uncharacterized protein (UniProt Q47702).

ybhA

The ybhA gene (located between 797,613 and 798,431 kb of the E. coli genome) encodes a phosphatase (UniProt P21829) belonging to the superfamily of haloacid dehalogenase (HAD)-like hydrolases. In vitro, YbhA has phosphatase activity with PLP, erythrose-4-phosphate, fructose-1,6-bisphosphate, flavin mononucleotide, thiamine-pyrophosphate, glucose-6-phosphate, and ribose-5-phosphate. Nevertheless, among the examined substrates, the purified enzyme shows the lowest Michaelis-Menten constant for PLP (76). In vivo, the overexpression of YbhA attenuates the inhibitory effect of PL on bacterial growth. Conversely, overexpression of YbhA in cells grown in minimal medium leads to a reduced growth rate and decreased levels of intracellular PLP. This growth defect can be rescued by addition of PL in the growth medium (19). Therefore, YbhA has been indicated as a specific phosphatase for PLP in E. coli (Fig. 2), whose activity balances the activity of the pyridoxal kinases PdxK and PdxY. Although its second preferred substrate is fructose 1,6-bisphosphate (76), YbhA does not seem to function physiologically as fructose 1,6- bisphosphatase. This proves that the primary role of YbhA is in PLP homeostasis and not in the gluconeogenesis process (19). This role was confirmed by the work of Kim et al., in which they found mutations in the ybhA gene in 9 of the 11 naturally selected ΔpdxB strains able to grow after 10 days in minimal medium, suggesting that in this background, the loss of a PLP phosphatase would be advantageous (72).

The pgl gene, encoding the enzyme of the oxidative pentose phosphate pathway 6-phosphogluconolactonase (UniProt P52697), is transcribed in the opposite direction from ybhA. Therefore, ybhA constitutes a single transcription unit (Fig. 3I).

pdxI

Recently, another component of the E. coli salvage pathway was identified. It is the NADPH-dependent PL reductase encoded by pdxI (UniProt P25906), which converts PL into PN (Fig. 2). This enzyme, also known as pyridoxine 4′-dehydrogenase, belongs to the aldo keto reductase family (AKR family) and was initially identified in a large-scale metabolomics screen for enzymatic activities of uncharacterized proteins (77). Evidence that PdxI is a PL reductase comes from a recent paper from Ito and Downs (22). These authors noticed that in a strain lacking pdxI, as well as two other genes involved in PLP production (pdxJ and pdxH), the amount of PN inside the cell was significantly lower and deduced from this observation that the externally supplemented PL was not converted into PN as in the wild-type strain (22). Kinetic characterization of B₆ vitamer fluxes suggested that PdxI converts PL into PN and that the reduction of PL to PN is favored over the phosphorylation of PL to PLP. Therefore, following this suggestion, in E. coli, the conversion of externally supplemented PL into PLP would preferably take place through a pathway involving PdxI (PL→PN→PNP→PLP) rather than a pathway involving PdxK and PdxY kinases (PL→PLP).

The downstream gene ydbD encodes an uncharacterized protein probably involved in detoxification of methylglyoxal (UniProt P25907). The promoters of these two genes were only predicted, and it is still unknown if the pdxI gene is transcribed as a single transcription unit or together with ydbD (Fig. 3J). A microarray analysis indicated that transcription of pdxI is strongly activated by NsrR (78), but a subsequent chromatin immunoprecipitation with microarray technology (ChIP-chip) experiment indicated an NsrR-binding site on the ydbD promoter (79). The transcriptional regulator NsrR controls the expression of genes encoding enzymes that reduce or oxidize NO in response to NO endogenously generated or produced by host cells.

PLP-Binding Proteins

yggS

The yggS gene (located at 3,095,098 to 3,095,802 kb) encodes a PLP-binding protein involved in PLP homeostasis renamed PLPHP (PLP homeostasis protein; UniProt P67080). PLPHP is widely distributed in bacteria and eukaryotes. The E. coli yggS mutant strain accumulates PNP, but not the other vitamers, and shows a toxicity ring when grown on solid medium containing an excess of exogenous PN (14). Moreover, with respect to the wild-type strain, the yggS deletion causes accumulation of metabolites, such as l-valine and 2-ketobutyrate, in the growth medium (13). Recently, it was proposed that elevated levels of PNP caused by yggS deletion perturb threonine and isoleucine/valine biosynthetic pathways, causing accumulation of metabolites such as l-valine that prevent bacterial growth (80).

Physical clustering associations, resulting from a comparative genomic analysis, linked yggS to cell division and murein synthesis genes, as well as to PLP salvage pathway genes, and the glyA gene, coding for the PLP-dependent serine hydroxymethyltransferase enzyme (14). Nevertheless, in E. coli, yggS may be the first gene of a polycistronic operon (yggSTU-rdgB-hemW; Fig. 3K), although no correlation has been found between yggS and the downstream genes (13). Moreover, no experimental evidence of transcription of the yggS operon is present. The yggT gene encodes an integral membrane protein (UniProt P64564) probably involved in K⁺ uptake. In fact, under hyperosmotic conditions, overexpression of yggT compensates for the growth defects of E. coli mutant cells lacking the major K⁺ uptake systems (81). The yggU gene product is a small, uncharacterized protein (UniProt P52060). RdgB, encoded by the downstream gene (rdgB, also known as yggV) is a nucleoside triphosphate pyrophosphatase (UniProt P52061), which hydrolyzes dITP and XTP, mutagenic products of purine nucleotide deamination. RdgB may therefore remove misincorporated bases, such as xanthine or hypoxanthine, from the pool of DNA precursors (82). Transcription of rdgB, together with the downstream gene hemW, is also mediated by the alternative sigma factor σ³², which governs the immediate response to temperature stress (83). Moreover, under anaerobiosis, rdgB expression is repressed by FNR (fumarate and nitrate reduction regulatory protein), for which a binding site was identified upstream of the gene. The hemW (also known as yggW) gene codes for a heme chaperone (UniProt P52062). HemW is a radical SAM-family protein that appears to be able to transfer heme into enzymes of respiratory chains (84).

In the opposite direction with respect to the yggS-containing operon, there is the yggR gene encoding a putative transporter annotated as the type II/IV secretion system family protein (UniProt P52052). Because the yggS and yggR start codons are only 17 bp apart, the two genes probably share a regulatory region in their respective promoters.

MocR-Like Transcription Factors

All bacteria, except Chlamydiae, possess MocR-like transcription factors (MocR-TFs), formed by the joining of a helix-turn-helix DNA binding domain and the protein scaffold of fold type I PLP-dependent enzymes (6). MocR-TFs control several different metabolic processes involving vitamin B₆ and amino acids (5). E. coli has been reported to contain two MocR-TFs (85) of unknown function, encoded by ydcR and yjiR genes (Table 1). Unlike other well-studied MocR-TFs, i.e., Salmonella enterica serovar Typhimurium PtsJ, B. clausii, L. monocytogenes PdxR, and several others, in which the regulatory gene is adjacent to the target gene (5), in the case of E. coli MocR-TFs, no putative target gene involved in vitamin B₆ biosynthesis and salvage pathways is located in close proximity.

The ydcR gene is located between 1,510,003 and 1,511,409 kb in the E. coli genome. Downstream and in the same direction, there is the ydcSTUV-patD operon (Fig. 3L). The ydcSTUV genes encode a putative ABC transporter for putrescine transport, though the major described putrescine transporter in E. coli is PuuP (86). In particular, YdcS constitutes the periplasmic binding protein; YdcT is the ATP-binding component, which may provide the driving force for putrescine transport, while YdcU and YdcV are the predicted inner membrane components of the ABC transporter. PatD, whose gene is transcribed from the ydcS promoter, encodes a γ-aminobutyraldehyde dehydrogenase, belonging to one of the two pathways of putrescine degradation. The ydcSTUV-patD operon is subject to the control of the stationary phase sigma factor σ^S, promoting transcription initiation from two σ^S-binding sites within the ydcS promoter. Moreover, expression is induced under nitrogen-limited growth conditions by the transcriptional regulator Nac, whereas putrescine does not exert any control (87). These observations suggest that the ydcSTUV-patD operon is induced in response to several stresses in E. coli and do not indicate any possible involvement of the YdcR MocR-TF.

The second gene encoding a MocR-TF in E. coli is yjiR, located between 4,570,162 and 4,571,574 kb, in an unexplored region of the E. coli chromosome (Fig. 3M). In fact, the adjacent yjiS gene encodes an uncharacterized small protein (UniProt P39390).

In summary, considering all the operons containing genes involved in PLP metabolism, it may be assumed that the PLP biosynthesis and salvage pathways are strictly associated with primary metabolism of E. coli. In fact, as reported in Table 2, they are generally cotranscribed with genes whose products are involved in glycolysis, biosynthesis of amino acids, isoprenoids, fatty acid, RNA and DNA metabolism, and response to stress.

TABLE 2.

Genes associated with vitamin B₆ metabolism and biological processes in which they are involved

Vitamin B6 gene	Gene in operon	Biological process	Essentiala
epd	pgk	Glycolysis and gluconeogenesis	Yes
	fba	Glycolysis and gluconeogenesis	Yes
pdxB	usg	Amino acid biosynthesis	No
	truA	RNA modification	No
	dedA	n.r.b	No
serC	aroA	Aromatic amino acid biosynthesis	Yes
pdxA	lptD	Membrane organization/response to stress	Yes
	surA	Protein folding	No
	rmsA	rRNA processing and modification	No
	apaG	Response to stress	No
	apaH	Response to stress/RNA decapping	No
dxs	xseB	DNA catabolism	No
	ispA	Isoprenoid biosynthesis	Yes
	yajO	Isoprenoid biosynthesis	No
pdxJ	rnc	RNA processing	Yes
	era	Ribosome biogenesis	Yes
	recO	DNA repair/response to stress	No
	acpS	Fatty acid biosynthesis	Yes
pdxH	tyrS	Translation	Yes
yggS	yggT	Response to stress	No
	yggU	n.r.	No
	rdgB	Nucleotide metabolism	No
	hemW	Oxidation-reduction process	No

^aEssential means that the corresponding knockout strain does not grow in minimal medium without supplementation (https://ecocyc.org/). The essential genes are highlighted in green.

^bn.r., not reported.

STATE OF THE ART ON PLP METABOLISM REGULATION

Transcriptional Regulation

Despite the importance of PLP metabolism in E. coli, the transcriptional regulation of the involved genes is poorly understood. It is known that three global regulators are involved in the regulation of epd and serC expression. As the epd gene product is homologous to GAPDH, it is not a surprise that Cra and CRP are involved in its transcriptional regulation. The catabolite repressor-activator (Cra) protein controls the direction of carbon flux by activating genes encoding enzymes involved in oxidative and gluconeogenic carbon flow and by repressing genes concerned with the fermentative carbon flow (88). On the other hand, the cyclic AMP receptor protein (CRP) is an important transcription factor that regulates transcription initiation of many genes required for energy production, amino acid metabolism, nucleotide metabolism, and ion transport systems (89). In particular, the expression of epd is negatively affected by Cra, by binding at the −6/+12 region of the epd promoter, and is activated by the CRP-cAMP regulator, which binds at the –81/–60 region. Both regulators are also involved in the transcriptional regulation of gapA (90), which can replace epd in its absence. The serC-aroA operon is repressed by the CRP-cAMP complex and is activated by the leucine-responsive global regulator Lrp, through their binding at specific, conserved sequences of the promoter. Lrp regulates the expression of more than 40 genes in E. coli, including several operons involved in amino acid biosynthesis (91). The combined effect of CRP and Lrp is to maximize the expression of the serC-aroA operon in cells growing in glucose-containing minimal medium, when a rapid synthesis of vitamin and metabolites is beneficial to cell growth, and to attenuate it in amino acid-rich media, where l-serine and B₆ vitamers are present (45).

It has been reported that the expression of the operons pdxA-rsmA-apaGH and pdxA-rms from P_pdxA is positively regulated by the growth rate through the Fis protein (41). Although the exact location of the binding site of this regulator on the P_pdxA promoter is unknown, the interaction with Fis was confirmed in a genome-wide analysis (92). Fis is a nucleoid-associated protein, whose deletion affects the transcription of 21% of E. coli genes. In the case of pdxA, it has been suggested that Fis is responsible for the positive growth rate regulation (41).

So far, no specific transcriptional regulators of the genes involved in PLP biosynthesis and salvage pathways have been found in E. coli. On the other hand, in S. typhimurium, the expression of the pdxK gene is repressed from the MocR-like regulator PtsJ, whose gene is divergently transcribed from pdxK itself. PLP acts as an effector molecule of this regulator, because its binding to PtsJ induces a protein-conformational change that increases affinity for DNA and reinforces repression (93). As stated above, in several bacteria which synthesize PLP through the so-called DXP-independent pathway, the transcription of the pdxST operon encoding PLP synthase is regulated by another MocR-TF (PdxR) (7–10). A similar transcriptional regulation of PLP biosynthesis has never been identified in E. coli, where the function of the two MocR-TFs present in the genome (Table 1) has never been identified.

We analyzed the promoter region of all the genes and operons involved in PLP biosynthesis and the salvage pathway (Supplemental Material) to check for the presence of common motifs that may correspond to transcriptional regulation binding sites. A 13-bp conserved sequence was identified in all the 18 promoter regions analyzed, with pdxJ, pdxK, tyrS, and yjiR showing the highest scores (Fig. 4). In most cases (11 out of 18), this consensus sequence encompasses the promoter region (−35 hexamer, −10 hexamer, or +1 start site) or is located downstream of the transcription start site, suggesting that it may correspond to a repressor-binding site. The actual role of this sequence in transcriptional regulation is far from being elucidated; however, the fact that it is present in the promoter region of both E. coli MocR-TFs (YjiR being one of the targets with the highest scores indicated by the MEME tool) encourages an experimental investigation.

FIG 4.

Alignment of promoter regions containing the putative motif involved in the regulation of genes linked with PLP homeostasis. The promoter regions were analyzed using the MEME suite webserver to identify a common motif. The orientations (+ or –) of the promoter region and the score are reported. The logo of the sequence highlights conservation of the identified motif.

Translational Regulation

Although transcription represents the main point of gene expression control in bacteria, regulation at the translational level also occurs (94, 95). Since translational regulation is important especially in the expression of genes forming operons, in which open reading frames (ORFs) are translated at different rates, this mode of regulation is expected to play a significant role in the expression of genes involved in the PLP biosynthetic pathway, such as the epd, pdxB, pdxA, and pdxJ genes, which are members of complex operons. However, the actual translational control of these genes and the molecular mechanism through which it may occur have not been characterized yet. The epd, pdxB, and pdxA genes contain infrequent and rare codons, suggesting that the mechanism of codon usage could negatively affect translational rates and lead to low levels of the corresponding proteins (35, 36, 47). Moreover, two structural features related to the epd translation initiation region (TIR) cause the poor expression of its ORF. The ribonucleotides located upstream of the ribosome binding site (RBS) can form a stem-loop structure, and the distance between the RBS and the AUG start codon is unusually short, and it is evolutionarily conserved among epd ORFs (35). Furthermore, Roa et al. highlighted the fact that pdxA lacks a Shine-Dalgarno sequence, the typical RBS of bacteria. This may cause a minor affinity of the ribosome for the translational initiation site (47).

Finally, the start codons of pdxB, pdxA, and pdxJ ORFs are close to a conserved motif, ACGT(G/T)AAAATCC, named the “PDX box” (36, 47, 64). Notably, other PDX boxes were found within the pdxJ coding sequence. It has been proposed that this consensus motif might play a role in the regulation of translation, being recognized by a specific regulatory protein, or may form particular secondary structures affecting ribosome activity. Definitely, the translational regulation mediated by PDX boxes may represent a mechanism through which the expression of pdxB, pdxA, and pdxJ is coordinated in the cell.

To summarize, the structural properties of epd, pdxB, pdxA, and pdxJ coding DNA sequences could limit their translation, allowing a differential expression with respect to other coding sequences present in the same polycistronic unit (Fig. 3).

Cellular Levels of Proteins Involved in PLP Homeostasis

The combined effect of regulation at transcriptional and translational levels determines the cellular content of proteins. Usually, proteins involved in a specific process show similar abundance (96, 97). This seems to be not the case of the proteins involved in vitamin B₆ metabolism. In fact, we analyzed the results from two quantitative proteomic studies, carried out on E. coli lysates from cells collected at either the exponential (96) or stationary (97) phase, and although different experimental setups have been used in combination with liquid chromatography-tandem mass spectrometry methodologies, we compared the copy number per cell of proteins involved in PLP metabolism as determined in the two studies with respect to two housekeeping proteins, GyrB and RecA (98) (Table S1). Among the enzymes of PLP biosynthesis, the concentrations of SerC and PdxJ are the highest in both proteomic analyses. In particular, in the exponential phase, SerC is 10-fold more abundant than the housekeeping proteins and 8 times more represented than PdxJ. This can be easily rationalized by the fact that SerC is a key enzyme in l-serine synthesis, as mentioned above. In the exponential phase, the levels of PdxB, PdxH, PdxK, and PdxY are within the same order of magnitude. In contrast, PdxA and Epd are the less abundant PLP biosynthesis proteins; this is probably a direct consequence of translational control, negatively affecting the protein synthesis rate (see previous paragraph). The level of proteins decreases in the stationary phase compared to the exponential phase, with the exception of PdxK and PdxH. This might be because the salvage pathway, in contrast with de novo biosynthesis, is essential during the stationary phase when PLP is recycled from protein turnover.

Enzyme Regulation

Significantly, the activity of some PLP biosynthesis and salvage enzymes is regulated by PLP itself. For example, PLP produced from the oxidation of PNP catalyzed by PNP oxidase (PdxH) tightly binds to an allosteric site causing the inhibition of the enzyme. An exhaustive characterization of this phenomenon has been recently carried out by our group through kinetics and equilibrium binding experiments (16). A structural and functional connection between the active site and the allosteric site has been proposed. PLP bound at the allosteric site causes the complete inactivation of the enzyme. However, this tightly bound PLP can be transferred to PLP-dependent enzymes. In fact, when the PdxH-PLP complex was added to a PLP-dependent apoenzyme diluted in an E. coli cell extract, the rate of reactivation of the apoenzyme was severalfold faster than when free PLP was added (99).

In vitro studies of the purified recombinant PdxK showed that the enzyme is inhibited by both its own substrate and product, PL and PLP, respectively. The mechanism of inhibition consists of the formation of a Schiff base between either PLP or PL and the Lys229 residue located close to the active site (17, 100). Concerning the binding of PLP, it was demonstrated that the inhibition occurs rapidly during the catalytic turnover of the enzyme, when both PLP and MgADP are present. The tightly bound PLP is protected from dephosphorylation by either a specific PLP phosphatase or alkaline phosphatase and cannot easily be removed. However, it can be transferred to PLP-dependent apoenzymes with consequent reactivation of PdxK (17). In contrast, inactivation by PL takes place before ATP is added to assay enzyme activity. PL initially binds very rapidly and noncovalently to the enzyme. In the absence of ATP, this noncovalent complex is slowly converted into a covalent Schiff base linkage between PL and Lys229. If ATP is added, the enzyme is able to catalyze a single turnover conversion of this covalently bound PL to PLP (100).

The crystal structure of the other PL kinase, PdxY, showed that the active site is filled with a density that fits either PL or PLP. In particular, in one of the monomers, the ligand appears to be covalently attached to a cysteine residue present at the active site as a thiohemiacetal, while in the other monomer it is not covalently attached (74). It is very likely that as consequence of this binding inside the bacterial cell, the purified PdxY appeared almost inactive (20).

It is worth noticing that PLP is also difficult to separate from the Pdx1 component of PLP synthase of the DXP-independent pathway of PLP biosynthesis. Indeed, it has been proposed that it is bound to the protein through a Schiff base (101).

In all these enzymes, PLP plays a dual role: (i) modulation of enzyme activity, which is of physiological importance to keep the PLP concentration at an appropriate level, and (ii) storage of PLP, which is not released into the cellular environment, preventing spurious reactions with nonspecific substrates, but that can be transferred to PLP-dependent apoenzymes.

CONCLUSION

Vitamin B6 metabolism and PLP homeostasis are poorly understood although they are crucial aspects of bacterial metabolism, considering the importance of this vitamin in bacterial physiology and virulence. The elucidation of vitamin B₆ metabolism in bacteria may also be relevant to human health, since it may indicate novel targets of antimicrobial intervention. Moreover, the proteins involved in vitamin B₆ metabolism also play pivotal roles in many human disorders, and many aspects of bacterial vitamin B₆ metabolism are similar in humans, as in the case of PNP oxidase and PLPHP (102, 103). Therefore, E. coli could be a useful model organism to acquire information on the mechanisms of regulation of vitamin B₆ relevant to human health.