Biosynthesis of Vitamin B6
Chapter
45
ROBERT E. HILL and IAN D. SPENSER
Our review on the biosynthesis of vitamin B6 (25), published in 1986, and the review by Dempsey on the synthesis of pyridoxal phosphate (14) that appeared in the first edition of this volume in 1987 provide comprehensive coverage of the intensive efforts, during the period 1970 to 1985, that were devoted to genetic studies leading to the isolation and analysis of pdx mutants of Escherichia coli B, derived from wild-type E. coli B by Dempsey, and to the tracer experiments performed in our laboratory, which resulted in a detailed understanding of the biogenetic anatomy of the pyridoxine skeleton with respect to its derivation from glucose, glycerol, and several other primary metabolites. The terms vitamin B6 and pyridoxine are used interchangeably to denote the six analogs that have vitamin activity, whereas the terms pyridoxol, pyridoxal, and pyridoxamine are reserved for the specific compounds.
In this chapter, we concentrate on work that has been published during the past decade, and we refer to the older data only as necessary to place the experimental design and the results of the recent work into proper focus. Our emphasis will be on two sets of investigations: (i) the tracer experiments from our laboratory that have exploited stable isotopes in defining precursor-product relationships in pyridoxine biosynthesis and (ii) the genetic investigations of M. E. Winkler and colleagues, who have used the techniques of molecular biology in studying the pathway of biosynthesis.
The tracer studies of pyridoxol (Fig. 1, 1) biosynthesis that have been carried out to date in our laboratory have used two pdx mutants of E. coli B that had been isolated by W. B. Dempsey. Most were performed with cultures of E. coli B WG2 (11), a pdxH mutant which lacks the enzyme pyridoxol 5'-phosphate oxidase (EC 1.1.1.65 or EC 1.4.3.65). This mutant is ideally suited for such work, since under the culture conditions used, it synthesizes pyridoxol and its 5'-phosphate ester at a rate that is four to five times that of the wild type (10).
In early studies, samples of glycerol (22) and d-glucose (22, 24), labeled nonrandomly with 14C, were used as tracers. 3H was used on occasion (44), but only as a secondary tracer in conjunction with 14C as an internal standard. In more recent studies, the isotopic markers were the stable isotopes 13C or 2H(20, 23, 29).
Use as tracers of radioactive isotopes on the one hand and of stable isotopes on the other, each has advantages and disadvantages. The sole advantage of using a radioactive tracer, 14C or 14C in conjunction with 3H, is the great sensitivity of its detection by modern liquid scintillation counters, a sensitivity which is at least 3 orders of magnitude greater than the efficiency of detection of 13C or of 2H by high-field nuclear magnetic resonance (NMR) methods (18). The disadvantages of using 14C, alone or in conjunction with 3H, lie in the necessity of chemical manipulation of very small samples of the biosynthetic radioactive product. Preparation of chemical derivatives, in order to determine radiochemical purity, and the execution of sequences of controlled chemical degradation, in order to determine the precise sites of labeling, are obligatory components of biosynthetic investigations with radioactive tracers (25).
The major advantage of using substrates labeled with 13C, particularly samples in which two contiguous carbon atoms are fully enriched in 13C (so-called bond-labeled samples), lies in the fact that 13C NMR not only detects the site of labeling but at the same time confirms the identity of the labeled sample and determines its degree of chemical purity. Low levels of 13C enrichment at single sites in a product are much more difficult to detect with certainty than enrichment from bond-labeled samples, since in the former instance signal enhancement must be determined by comparison with the intensity of a natural abundance signal, whereas in the latter instance enrichment by intact incorporation of a pair of contiguous 13C enriched carbon atoms is indicated by the presence of new satellite peaks that are absent in the spectrum of a natural-abundance sample. Furthermore, the presence of such satellite peaks in the signals of the 13C NMR spectrum of a biosynthetic product, as a result of incorporation of contiguously 100% 13C-13C-enriched (i.e., bond-labeled), substrates, provides direct evidence for the transfer from substrate to biosynthetic product of an intact multicarbon unit. Substrates bond labeled with fully enriched 2H attached to fully enriched 13C may be used similarly. These methods constitute a powerful new tool for the demonstration of transfer of intact multiatom fragments in investigations of biosynthesis. Neither radioactive tracer methods nor the use of stable isotopes in conjunction with mass spectrometry (a technique that has been used in investigations of biosynthesis [e.g., reference 45], but has not been used in the case of pyridoxine biosynthesis) can show incorporation of intact multicarbon units or intact C-H bonds. Deuterium enrichment is readily detected by 2H NMR, but for technical reasons 2H NMR is less advantageous in determining sample purity and must be used in conjunction with 1H NMR. Provided that adequate incorporation can be achieved in a tracer experiment with a 13C- or 2H-labeled substrate, stable isotopes are the tracers of choice in a modern experiment, even though the sensitivity of detection of stable isotopes is much lower than that of radioactive tracers. A precondition of the application of NMR methods for the analysis of biosynthetic incorporation patterns is the reliable assignment of each spectral signal to the individual atom that gives rise to it. The major disadvantage of the method lies in the fact that very few specialized labeled, and particularly bond-labeled, materials other than the most common substrates are commercially available; for specialized investigations, therefore, the labeled substrates must be synthesized in house from labeled chemicals that are commercially available. This requirement is likely to be expensive, labor-intensive, and time-consuming.
Study of incorporation of 13C- and 14C-labeled glucose into pyridoxol will illustrate the advantages and disadvantages of the two methods. The following sections present the results of tracer experiments and their interpretation. Neither the experimental methods used to determine the sites of radioactive labeling nor the methods used for the assignment of signals in the 13C or 2H NMR spectra of pyridoxol, essential components of the work, will be discussed. The reader should refer to the original papers if this information is required.
The power of the 13C NMR method in investigations of biosynthesis is demonstrated convincingly in an experiment in which a mixture containing 20% fully 13C enriched (99% 13C per C atom) d-glucose, i.e., d-[1,2,3,4,5,6-13C6]glucose, and 80% unenriched (1.1% 13C per C atom) glucose was the sole carbon source in the incubation medium (20).
The use of such a mixture of unenriched substrate with fully enriched multiply labeled substrate with 13C at contiguous carbon atoms represents a type of isotope dilution experiment that permits the detection, by 13C NMR, of the transfer of intact multicarbon units from substrate into the biosynthetic product. The 13C NMR spectrum of the product into which such an intact unit has been incorporated exhibits characteristic signals due to 13C-13C coupling of the enriched sites. Carbon-carbon bonds formed de novo during the biosynthetic process will not exhibit such a coupling pattern, since such newly formed bonds are made largely by union of one 13C and one 12C atom, or from two 12C atoms, the probability of which increases with the mole ratio of unenriched to labeled substrate in the incubation mixture.
In the 13C NMR spectrum of the sample of pyridoxol that was isolated from our incubation with d-[1,2,3,4,5,6-13C6]glucose (20), each of the 13C signals due to C-2', C-2, C-3, C-4', C-5', and C-6 appeared as a central line straddled by a doublet. The central line is due to natural-abundance material, mainly carrier pyridoxol added to aid the isolation of the minute quantities of newly biosynthesized material. The doublet showed that each of these six carbon atoms showed direct 13C-13C coupling to only one other carbon atom. The signals due to C-4 and C-5, however, consisted of a central line straddled by a pair of doublets, a pattern expected for the central carbon atom of a 13C-13C-13C unit. This spectrum establishes biosynthetic connectivity between C-2' and C-2 but none between C-2 and C-3; it also shows that each of C-4' and C-3 had biosynthetic connectivity to C-4, that each of C-5' and C-6 had connectivity to C-5, but that there was no connectivity between C-4 and C-5. Thus, in the construction of the pyridoxol skeleton from glucose, the bonds C-2–C-3 and C-4–C-5 are the only newly formed carbon-carbon bonds (Fig. 2, 2). It follows that glucose supplies, in equimolar amounts, two C3 units and a C2 unit, which, respectively, serve as the basic building blocks of the C3 fragments, C-3,4,4' and C-6,5,5', and of the C2 fragment, C-2',2, of pyridoxol.
The experiment thus demonstrates that glucose supplies three multicarbon units, and it defines the sites into which these units are introduced into pyridoxol. The experiment does not provide evidence concerning (i) the identity of the C3 and C2 intermediates, (ii) the fragments of glucose that supply them, (iii) the sequence of events whereby these fragments are generated, and (iv) the sequence and the mechanism whereby these fragments combine to yield the pyridoxol skeleton.
The results of earlier tracer experiments with radioactive isotopes provide answers to some of these questions. In particular, the probable identities of the C3 and C2 intermediates and the way they were derived from glucose were inferred from the results of early experiments with 14C-labeled glycerol, glucose, and pyruvic acid.
The C3 and C2 units in the carbon skeleton of pyridoxol had been shown to be derived from glycerol in a very specific manner (22). Radioactivity from [2-14C]glycerol was distributed equally over three sites, C-2, C-4, and C-5, of the vitamin, each containing 33% of total label of pyridoxol (Fig. 2, 6). Furthermore, in one of the first published investigations of a biosynthetic process that employed 13C NMR (21), it was shown that the signals due to C-2', C-3, C-4', C-5', and C-6 in the spectrum of the isolated pyridoxol derived from [1,3-13C2]glycerol each showed approximately 20% enrichment (Fig. 2, 5). It thus appears that when glycerol serves as the carbon source, five of the eight carbon atoms of pyridoxol (C-2', C-3, C-4', C-5', and C-6) are derived from its primary carbon atoms and three (C-2, C-4, and C-5) are derived from its secondary carbon. The C8 skeleton of pyridoxol must therefore be constructed from three glycerol units, one of which loses a primary carbon on the route to the product.
Glycerol itself lacks the chemical and biochemical reactivity that is required to create the vitamin skeleton, and it is much more likely that the subunits of pyridoxol are derived from one or both of the two triose phosphates, dihydroxyacetone 1-phosphate and d-glyceraldehyde 3-phosphate, that are generated from glycerol in the course of metabolism and are metabolically interconvertible. When glycerol is phosphorylated on its route into the triose phosphates, it is the pro-R hydroxymethyl group that undergoes phosphorylation (5). The same triose phosphates are generated from glucose in the course of glycolysis, C-1 of glucose yielding, in the first place, the phosphate ester carbon of dihydroxyacetone 1-phosphate, and C-6 of glucose yielding the corresponding carbon atom of d-glyceraldehyde 3-phosphate.
Experiments with [1-14C]- and [6-14C]glucose (22, 24) showed that label was delivered into three sites of pyridoxol, C-2' of the C2 unit and C-4' and C-5' of the two C3 units. In each case, these three sites accounted for all of the radioactivity within the biosynthesized sample of pyridoxol (Fig. 2, 3 and 4). These three pyridoxol carbon atoms, C-2', C-4', and C-5', are thus derived from the pro-R hydroxymethyl group of glycerol, that is, from the phosphorylated terminus of one or the other of the trioses.
This conclusion was confirmed by feeding experiments with chirally labeled samples of glycerol, (S)-[1,1-2H2]- and (R)-[1,1-2H2]glycerol (20). 2H NMR analysis of the pyridoxol sample isolated from the experiment with (R)-[1,1-2H2]glycerol (Fig. 2, 7) (20) demonstrated that six glycerol deuterium atoms had been retained within pyridoxol and that they were located in pairs at the peripheral groups: one pair at the CH2OH group, C-4', another pair at the CH2OH group, C-5', and the third pair representing two of the three hydrogen atoms at the C-methyl group, C-2'. The (S) terminus of glycerol did not contribute deuterium to the biosynthetically generated pyridoxol. This observation confirmed an earlier finding (20, 44) in an experiment using 3H-14C double labeling that the hydrogen atom at C-6 of pyridoxol was not directly derived from glycerol. The inference is inescapable that this hydrogen atom is lost in the course of the biosynthetic process. Any hypothesis of the mechanism of biosynthesis must take account of this observation.
Even though the pyridoxol samples that were isolated from the experiments with [1-14C]- and [6-14C]glucose (24) contained all of their radioactivity at C-2', C-4', and C-5' (Fig. 2, 3 and 4), the three centers were not equally labeled. In the case of [1-14C]glucose, C-5' contained 26% ± 2% of the total activity, whereas C-2' and C-4' contained higher and equal level of radioactivity (36% ± 1% and 38% ± 4%, respectively) (Fig. 2, 3). Conversely, in the case of [6-14C]glucose, half (48% ± 4%) of the label resided at C-5', whereas C-2' and C-4' were again equally labeled, but each contained only one-quarter of the total activity (26% ± 2% and 27% ± 4%, respectively) (Fig. 2, 4). It was concluded that each of the C3 units C-3,4,4' and C-6,5,5' of pyridoxol was derived from glucose via a different triose phosphate, dihydroxyacetone phosphate and d-glyceraldehyde phosphate, respectively, and that the C2 unit C-2,2' was further derived from the former, via pyruvaldehyde (44; see also references 7, 26, and 39).
The fact that the three fragments were equally enriched in the experiment with d-[13C6]glucose, whereas there were differences in isotope concentration within the three sites, in the experiments with d-[14C]glucose, pinpoints another important difference between the two methods. In order to detect 13C enrichment by NMR, comparatively massive amounts of almost carrier-free labeled substrate must be used. As a consequence, the metabolic system is essentially saturated with labeled substrate and all metabolites derived from the labeled substrate readily reach equilibrium isotope concentration. In experiments with 14C-labeled samples, on the other hand, only trace amounts of labeled material are used and only very few molecules within the administered labeled sample are actually labeled. Isotope saturation is not reached, and isotope concentrations within related metabolites may differ, reflecting rates of interconversion and pool sizes.
The source of the C2 unit, C-2',2, has been identified as the CH3-CO fragment arising by decarboxylation of pyruvic acid (Fig. 2, 8) (22). It is the pro-S hydroxymethyl group of glycerol that is transformed into the carboxylic acid group of pyruvic acid. The results of incorporation experiments with pyruvate showed that the carboxyl group is lost and that the CH3-CO unit of pyruvate gives rise to C-2',2 of pyridoxol (Fig. 2, 8). This observation is entirely consistent with the generation of this unit from glucose or from glycerol via triose phosphate and pyruvic acid. Furthermore, unlabeled pyruvic acid spared the incorporation of label from glycerol into C-2' and C-2 but did not affect its entry into any other site (44). The origin of the C2 unit, C-2',2, from glycerol or from glucose by way of a two-carbon unit derived from pyruvate by decarboxylation accounts for the loss of the pro-S hydroxymethyl group from one of the three biosynthetic glycerol units.
We stated earlier that the experiment with d-[1,2,3,4,5,6-13C6]glucose (20) showed that glucose supplies three multicarbon units and defines the sites within pyridoxol into which they are introduced, but that the experiment did not provide evidence concerning (i) the identities of the C3 and C2 intermediates, (ii) the fragments of glucose that supply them, (iii) the sequence of events whereby these subunits are generated, and (iv) the sequence and the mechanism whereby the subunits combine to yield the pyridoxol skeleton. The experimental results outlined above permitted inferences to be drawn which provide answers to the first three questions. We now turn to a discussion of the remaining question.
The experiment with d-[1,2,3,4,5,6-13C6]glucose (20) showed that the C-2–C-3 and C-4–C-5 bonds were the only two new carbon-carbon bonds that were formed in the construction of the pyridoxol skeleton (Fig. 2, 2) from glucose, but it did not provide information on the sequence in which these bonds are formed. There are several reasons why it would seem likely that the C-2–C-3 bond is the first to be generated. Initial formation of the C-4–C-5 bond, by condensation of the central carbon atoms of two triose phosphates, would produce an unusual branched chain hexose. There is no precedent for the formation of such a structure. There is ample precedent, however, for the pyruvate dehydrogenase (EC 1.2.4.1)-catalyzed acyloin-type condensation of d-glyceraldehyde (Fig. 3, 14) or its 3-phosphate (Fig. 3, 12) with pyruvic acid (Fig. 3, 11), a reaction that is accompanied by decarboxylation and yields 1-deoxy-d-xylulose (Fig. 3, 15) or its 5-phosphate (Fig. 3, 13), respectively (47). This enzymatic process, which requires thiamin pyrophosphate as a cofactor (48), has been shown to take place in E. coli. It is noteworthy in this context that thiamin is required for the biosynthesis of pyridoxine (13). 1-Deoxyxylulose (Fig. 3, 15) has been implicated in the biosynthesis in E. coli of another B vitamin, thiamin (8, 9; chapter 43, this volume).
1-Deoxy-d-xylulose (Fig. 3, 15) serves as the precursor of the C5 unit, C-2',2,3,4,4', of pyridoxol.
Samples of 1-deoxy[1,1,1-2H3,(RS)-5-2H1]-d-xylulose (Fig. 4, 16) and of 1-deoxy[1,1,1-2H3,(RS)-5-2H1]-l-xylulose, synthesized by published methods (9), were administered to E. coli B mutant WG2. Deuterium NMR spectroscopy revealed that label from the d isomer, but not from the l isomer, was incorporated into pyridoxol. The product contained deuterium at C-2' and at C-4' (Fig. 4, 17), and the ratio of deuterium at these two sites corresponded to the ratio of deuterium at C-1 and C-5 of the administered substrate (Fig. 4, 16) (23). This result makes it extremely likely that 1-deoxy[1,1,1-2H3,(RS)-5-2H1]-d-xylulose is incorporated intact into the vitamin and that it serves as a basic building block of the fragment C-2',2,3,4,4' of pyridoxol in E. coli.
Furthermore, the presence of unlabeled 1-deoxy-d-xylulose partially inhibited the incorporation of 13C from d-[13C6]glucose into C-2',2,3,4,4' but not into C-6,5,5' of pyridoxol (29). 1-Deoxy-d-xylulose (Fig. 4, 15) is thereby shown to lie on the route from glucose into the C5 unit, C-2',2,3,4,4'.
A tracer experiment with a sample of 1-deoxy-d-xylulose (Fig. 4, 19), 13C bond labeled at C-2,C-3, that is at the site of union of its C2 and C3 precursors, should provide definitive proof that 1-deoxy-d-xylulose supplies this pyridoxol fragment as an intact unit. The 13C-13C bond of the substrate (Fig. 4, 19) should be transferred intact into the vitamin (Fig. 4, 20). This experiment is in progress (I. Kennedy, R. E. Hill, and I. D. Spenser, unpublished data).
The evidence which led to the conclusion that in E. coli B mutant WG2, the C3 unit, C-6,5,5', of pyridoxol is derived intact from glucose via triose phosphate has been presented above. But another source of two of these carbon atoms, C-5 and C-5', was discovered in WG3, a pdxB mutant. This mutant is a pyridoxine-requiring strain which can satisfy its pyridoxine requirement by glycolaldehyde (12). Labeled glycolaldehyde was incorporated into pyridoxal phosphate isolated from this organism (40), and the sites of incorporation of the two carbon atoms of glycolaldehyde were determined: C-5' of pyridoxal was derived from the carbon atom of the CH2OH group, and C-5 was derived from the aldehyde carbon atom (Fig. 2, 9) (19). It was shown subsequently (43) that the same process occurred also in mutant WG2, as a minor pathway contributing to the formation of the C2 fragment, C-5,5', of pyridoxol, whose major origin in this mutant is as a component of the C3 fragment, C-6,5,5', which is derived as an intact unit from glucose (Fig. 2, 2).
That the two processes were entirely distinct from one another was shown by the fact that whereas C-6 of pyridoxol in mutant WG2 originates from a terminal carbon atom of glycerol when glycerol is the sole primary precursor of pyridoxol (Fig. 2, 5), the same carbon atom arises from the central carbon atom of glycerol when C-5,5' originates from glycolaldehyde (43). This apparent contradiction was explained by the finding that in mutant WG3, C-6 of pyridoxal was supplied by the methylene carbon of [2-14C]glycine and that, furthermore, the fragment N-1,C-6 of pyridoxal arises from glycine as an intact unit (Fig. 2, 10). It was shown by 13C NMR that pyridoxal, isolated from a culture of mutant WG3 which had been incubated with bond- labeled glycine (15NH2-13CH2-CO2H), maintained the intact 15N-13C bond of the substrate at N-1,C-6 (27).
It was on the basis of this result that in our 1986 review (25), we advanced the idea that 4-hydroxythreonine (= 3-hydroxyhomoserine) (Fig. 5, 29), generated by condensation of glycine (Fig. 5, 28) with glycolaldehyde (Fig. 5, 27) in a reaction analogous to that catalyzed by threonine aldolase (EC 4.1.2.5) or serine hydroxymethylase (EC 2.1.2.1), may serve as an intermediate in pyridoxol biosynthesis. This notion was adopted by Dempsey in the first edition of this volume (14) and developed further by Lam and Winkler (31).
4-Hydroxy-l-threonine (Fig. 5, 29) is indeed an intermediate on the route from glucose to the C3N unit, N-1,C-6,5,5', of pyridoxol. A tracer experiment with a sample of 4-hydroxy-l-threonine (Fig. 4, 22), 13C bond labeled at C-2,C-3, the site of union of its precursors, glycine and glycolaldehyde, provides definite proof that 4-hydroxy-l-threonine supplies this pyridoxol fragment as an intact unit. Pyridoxol obtained from this experiment was bond labeled at C-6,5 (Fig. 4, 21) (46a). Furthermore, the presence of unlabeled 4-hydroxy-l-threonine totally suppresses the incorporation of 13C from d-[13C6]glucose into C-6,5,5' but not into C-2',2,3,4,4' of pyridoxol (29). 4-Hydroxy-l-threonine (Fig. 5, 29) is thereby shown to lie on the route from glucose into the C3 unit, C-6,5,5'.
Further, more indirect evidence to confirm the status of 4-hydroxy-l-threonine (Fig. 5, 29) as a committed precursor of pyridoxine is based on the genetic investigations of Winkler and associates (31, 49) and on recent investigations on growth requirements of a number of mutants of E. coli B (16) (see below).
The distinct dual origin of 4-hydroxy-l-threonine (Fig. 5, 29), from glucose (Fig. 2, 2) on the one hand and from glycolaldehyde plus glycine (Fig. 2, 9 and 10) on the other, must be explained.
In an attempt to rationalize this duality of origin of the C3N unit, N-1,C-6,5,5', of pyridoxol, we had postulated (25) that 1-aminopropan-2,3-diol (or its 3-phosphate) served as the ultimate intermediate on the route into the C3N fragment of pyridoxol. (S)-1-Aminopropan-2,3-diol is the decarboxylation product of 4-hydroxy-l-threonine, but it may be generated also by transamination of d-glyceraldehyde 3-phosphate, followed by phosphate ester hydrolysis. If 1-aminopropan-2,3-diol served as an intermediate, the dichotomy of the origin of the C3N unit, either from glucose by way of d-glyceraldehyde phosphate or from glycolaldehyde plus glycine via 4-hydroxy-l-threonine, would be explicable.
To test the intermediacy of 1-aminopropan-2,3-diol, we incubated mutant WG2 with dl-[1,1,2,3,3-2H5]-1-aminopropan-2,3-diol (Fig. 4, 18). Incorporation of deuterium from this labeled sample of the aminodiol into the C3 unit, C-5',5,6, of pyridoxol was not detectable (16), even though under similar experimental conditions, nonrandom incorporation of deuterium from deuterated 1-deoxy-d-xylulose (Fig. 4, 16) into pyridoxol had taken place (see above) (23). Thus, 1-aminopropan-2,3-diol is unlikely to be an intermediate of pyridoxol biosynthesis.
Another model to explain the apparent dichotomy of origin of the C3N unit, N-1,C-6,5,5', of pyridoxol, via 4-hydroxy-l-threonine, either from glucose or from glycolaldehyde plus glycine, is thus required.
Such an alternative hypothesis has been advanced by Lam and Winkler (31) on the basis of genetic studies together with a reinterpretation of published nutritional and tracer evidence. The key components of this new hypothesis lie in two propositions. First, the route from glucose into 4-hydroxy-l-threonine and then to the C3 fragment, C-6,5,5', of pyridoxol proceeds not by way of a glycolytic intermediate, as had been suggested earlier (25), but via the pentose phosphate pathway and, more specifically, via d-erythrose 4-phosphate, an intermediate of this pathway. The route from glucose to the other two fragments, C-2,2' and C-3,4,4', does involve glycolytic intermediates. Second, the glycine/glycolaldehyde route to 4-hydroxy-l-threonine and then into the C3 fragment, C-6,5,5', proceeds by way of 3-hydroxypyruvate (38) and l-serine (33, 42).
To date, no direct tracer evidence exists to support the second of the two propositions, but a considerable body of evidence can be marshaled in support of the first. This evidence is based on the genetic investigations of Lam and Winkler (31).
The route to the C3 unit, C-6,5,5', according to the picture drawn by Lam and Winkler (31), is via the pentose phosphate pathway, which leads to the formation, from C-3,4,5,6 of glucose, of d-erythrose 4-phosphate (Fig. 5, 23). Further elaboration of d-erythrose 4-phosphate, by oxidation at C-1 to yield the corresponding erythroic acid (Fig. 5, 24), followed by dehydrogenation at C-2 to the corresponding α-keto acid (Fig. 5, 25), transamination to 4-hydroxy-l-threonine 4-phosphate (Fig. 5, 26), and phosphate ester hydrolysis, yields 4-hydroxy-l-threonine (Fig. 5, 29).
Very significantly, it should be noted that intermediacy of the a-keto acid (Fig. 5, 25) accounts for the observed loss of tritium from [1,3-3H,2-14C]glycerol and of deuterium from (S)-[1,1-2H2]glycerol on route into C-6 of pyridoxol (see above).
In order to rationalize the labeling patterns of the samples of pyridoxol derived from the tracer experiments with labeled glycerol and glucose (Fig. 2, 2 to 7) and, in particular, those from [1-14C]- and [6-14C]glucose with this pathway, a number of assumptions are necessary.
First, incorporation of radioactivity from [1-14C]glucose into C-5' of pyridoxol, entering via the pentose phosphate pathway, is explicable only if glycolysis takes place at the same time, yielding triose phosphates, and if this is followed by triose phosphate isomerization and resynthesis of [1,6-14C]glucose 6-phosphate.
Second, to achieve the equimolar incorporation of each of three triose molecules into the three independently derived fragments, C-2',2, C-3,4,4' and C-6,5,5', of pyridoxol (Fig. 2, 2 and 5 to 7) when glycolysis and the pentose phosphate pathway are both implicated in generating the three fragments, saturation equilibration of label from glycerol and trioses into glucose, by reversal of the glycolytic process, must be assumed.
Third, it must be assumed that the cells do not contain significant amounts of preformed, unlabeled intermediates, since only then would the same isotope concentration be delivered into the different sites of the product by two different routes.
The skewed distribution of label from [1-14C]- and [6-14C]glucose into C-5' of pyridoxol on the one hand and into C-2 and C-4' on the other (Fig. 2, 3 and 4) is then due to the fact that the unit containing C-5' arises by way of the pentose phosphate route, whereas the unit containing C-4' and that containing C-2' are both of glycolytic origin, and is not the result of the former arising from glycolytic glyceraldehyde 3-phosphate and the latter two arising from dihydroxyacetone 1-phosphate. The simplest interpretation of the tracer results leads to the inference that the C3 unit, C-3,4,4', is derived from glyceraldehyde 3-phosphate, that pyruvic acid, the precursor of the C2 unit, C-2',2, is derived from glyceraldehyde 3-phosphate in the normal course of glycolysis, that the condensation of the two units takes place by the pyruvate dehydrogenase-catalyzed process (47, 48), which requires thiamin pyrophosphate as a coenzyme (Fig. 3). The product of this enzymatic reaction, 1-deoxy-d-xylulose (Fig. 3, 15), serves as the precursor of the C5 unit of pyridoxol, C-2',2,3,4,4', as demonstrated by the tracer experiments referred to earlier.
The sequence of reactions proposed by Lam and Winkler (31) for the conversion of d-erythrose 4-phosphate (Fig. 5, 23) into 4-hydroxy-l-threonine (Fig. 5, 29) precisely parallels the enzymatic reaction sequence from d-glyceraldehyde 3-phosphate, the lower homolog of d-erythrose 4-phosphate, to l-serine, via d-glyceric acid 3-phosphate, 3-hydroxypyruvic acid 3-phosphate, and l-serine 3-phosphate. The first step, oxidation of d-erythrose 4-phosphate (Fig. 5, 23) to d-erythroic acid 4-phosphate (Fig. 5, 24), demands a dehydrogenase analogous to glyceraldehyde 3-phosphate dehydrogenase (EC 1.2.1.9). Recently Lam and Winkler have demonstrated that crude extracts of E. coli K-12 contain the required d-erythrose 4-phosphate dehydrogenase (EC 1.2.1.?) activity (31).
The conversion of d-erythroic acid 4-phosphate (Fig. 5, 24) to the hypothetical intermediate, 2-ketoerythroic acid 4-phosphate (Fig. 5, 25), the step that is analogous to the transformation of d-glyceric acid 3-phosphate into 3-hydroxypyruvate 3-phosphate in serine biosynthesis, requires a second dehydrogenase. 3-Phosphoglycerate dehydrogenase (EC 1.1.1.95), the enzyme that acts in the serine pathway, is the product of the serA gene (41). To date, 4-phosphoerythroate dehydrogenase activity has not been demonstrated in E. coli. However, the pdxB gene, which encodes one of the enzymes that is required for pyridoxine biosynthesis in E. coli, has been isolated (1). The pdxB gene is part of the complex hisT operon (2), a sequence of approximately 3.5 kb that contains at least four genes. pdxB is the first of these, and evidence was provided to show that it encodes a polypeptide of approximately 42,000 Da (37). Moreover, the sequence of the pdxB gene has been determined and a preliminary identification of its start codon and open reading frame has been reported (37). Together, these data permitted the construction of a putative amino acid sequence for the polypeptide product. Comparison of this sequence with that of the serA product, 3-phosphoglycerate dehydrogenase, revealed a remarkable similarity between the two and led Winkler’s group (37) to propose that pdxB also encodes a 2-hydroxy acid dehydrogenase, namely, the 4-phosphoerythroate dehydrogenase that is required for the production of 2-ketoerythroic acid 4-phosphate (Fig. 5, 25). Evidence in support of this proposal was obtained with the demonstration that a synthetic sample of 4-phosphoerythroic acid served as the substrate for a preparation of the pdxB gene product that had been isolated from cells in which it had been overexpressed by a plasmid vector. It was found that when the gene product of serC, i.e., 3-phosphoserine transaminase, was also present in the incubation, the reaction proceeded only in the forward direction, i.e., to the keto acid (35).
Further support for this route to 4-hydroxy-l-threonine comes from the fact that pdxF, another gene of the pyridoxine biosynthetic pathway, is in fact identical with serC (15), which codes for 3-phosphoserine transaminase (EC 2.6.1.52). It would thus appear that serC plays a dual role, participating not only in the serine biosynthetic pathway but also in the pyridoxine pathway (31), i.e., in the transamination of 2-ketoerythroic acid 4-phosphate (Fig. 5, 25) to yield 4-hydroxy-l-threonine 4-phosphate (Fig. 5, 26).
The final enzyme of the serine biosynthetic pathway, 3-phosphoserine phosphatase (EC 3.1.3.3), is encoded by serB. To date, the corresponding phosphatase activity that would hydrolyze 4-hydroxy-l-threonine 4-phosphate (Fig. 5, 26) to 4-hydroxy-l-threonine (Fig. 5, 29) has not been found in E. coli, nor is there any evidence to equate the protein (polypeptide) products of any of the remaining identified pdx genes with that of serB. It must therefore be assumed either that 4-hydroxy-l-threonine 4-phosphate (Fig. 5, 26) is the pyridoxine precursor, which, in turn, implies that pyridoxol 5'-phosphate is the first vitamin congener to be produced, or that a nonspecific phosphatase that is not the product of a pdx gene is responsible for the production of the free 4-hydroxy-l-threonine.
If the gene products of pdxB and serC are required for the elaboration of 4-hydroxy-l-threonine, it would be expected that pyridoxine auxotrophs blocked at these sites would grow on minimal medium supplemented with 4-hydroxy-l-threonine. Such behavior has now been confirmed: synthetic samples of 4-hydroxy-l-threonine (16) have been shown to overcome the pyridoxine requirement of several pdxB and serC mutants.
pdxB and serC are not the only pyridoxine biosynthetic genes that have been identified. Dempsey’s original genetic analysis (11) of a large number of pyridoxine auxotrophs of E. coli B established that they fall into five unlinked groups (I to V), widely separated on the E. coli chromosome (3). The members of two of these linkage groups, III and IV, were shown to be mutants of single genes, and their protein products were identified. Thus, group III mutants require pyridoxine and serine for growth, lacking 3-phosphoserine-2-oxoglutarate transaminase activity, and are serC mutants. Group IV mutants, on the other hand, grow only on either pyridoxal or pyridoxamine but not on pyridoxol, lacking pyridoxol 5'-phosphate oxidase activity, and are pdxH mutants.
It was unclear from Dempsey’s work whether the remaining linkage groups (I, II, and V) contained single or multiple pyridoxine genes, since some of the strains within each group could be differentiated according to their nutritional behavior. Group I mutants, for example, were subdivided into PdxB–, PdxC–, and PdxD– phenotypes on the basis of cross-feeding studies (11), whereas in group V, the pdxK mutant was distinguished from the pdxJ mutant by the fact that it exhibited growth on minimal medium supplemented with d- or l-alanine (13).
Thus, it was not possible to deduce the exact number of true pdx genes, and hence the number of enzymes, that are required in E. coli to elaborate the pyridoxine skeleton from primary precursors. If each of the five linkage groups contained only a single gene, it follows that only four enzymes would be required for the construction of the ring system, since the fifth enzyme, pyridoxol 5'-phosphate oxidase, the pdxH product, is not involved in the biosynthesis of the ring skeleton but is involved in the modification of the C-4' side chain. If, on the other hand, more than one gene were present in some of the linkage groups, more than four enzyme-mediated steps would be implicated in the biosynthetic pathway to pyridoxol.
This problem has recently been examined by Winkler’s group (30). Pyridoxine auxotrophs of E. coli B, of E. coli K-12, and of B–K-12 hybrids, representing each of the putative pdx point mutations that Dempsey had previously identified, were subjected to complementation analysis with five minimal recombinant plasmids that had been constructed to contain one each of the pdxB, pdxA, serC, pdxH, or pdxJ genes. Each auxotroph in linkage group I (PdxB–, PdxC–, and PdxD– phenotypes) was complemented by the pdxB + plasmid, each of the group II strains (PdxA–, PdxE–, and PdxG– phenotypes) was complemented by the pdxA + plasmid, and each of the group III strains (SerC– and PdxF–) was complemented by the serC + plasmid. Furthermore, when the pdx gene in each of the recombinant plasmids was disrupted by insertional mutagenesis with mini-Mu transposons (6), the ability to complement the pdx auxotrophic strains was lost. Thus, the notion that some of the linkage groups (I to V) identified by Dempsey might contain more than one pdx gene is refuted. The genotypes of groups I, II, III, and V are shown to be pdxB –, pdxA –, serC –, and pdxJ –, respectively. Moreover, a search for new pdx genes (30), by placing insertional elements directly into the chromosome of E. coli K-12 to cause mutations (46), led to several pyridoxine auxotrophs. However, no new pdx mutants were among them, since each was complemented by one of the aforementioned recombinant plasmids. Thus, pdxB, pdxA, pdxJ, pdxH, and serC account for all the pyridoxine biosynthetic genes in E. coli. It follows that the organism uses only four dedicated enzymes, the products of pdxB, pdxA, pdxJ, and serC, to elaborate the pyridoxine skeleton from primary precursors. The fifth enzyme, derived from pdxH, catalyzes the conversion of pyridoxol to pyridoxal.
The general experimental approach employing recombinant plasmids and insertional mutagenesis in combination with sequence analysis was used by Arps and Winkler to establish also that pdxB is located in the hisT operon (see above) (2); that pdxA is part of the complex operon, pdxA-ksgA-apa-apaH (36); that pdxJ is part of a two-gene operon, pdxJ-dpJ (30); that pdxH forms a two-gene operon with tyrS (32); and that their protein products have molecular weights of 42,000, 35,000, 26,400, and 25,545, respectively. In earlier work, serC had been established as part of the serC-aroA operon (17). Thus, it appears that all the genes (pdxB, pdxA, serC, pdxJ, and pdxH) that are required for the biosynthesis of pyridoxal phosphate lie within complex operons.
pdxH and serC are known to code for pyridoxol 5'-phosphate oxidase and 3-phosphoserine transaminase (EC 2.6.1.52), respectively. A role for the pdxB product as the catalyst in the oxidation of erythroic acid 4-phosphate (Fig. 5, 24) to 2-ketoerythroic acid 4-phosphate (Fig. 5, 25) has been proposed.
Recent results confirm that the pdxH gene is the sole source of pyridoxol 5'-phosphate oxidase in E. coli and that it is activated by flavin mononucleotide and oxygen under aerobic conditions (34, 50).
Direct evidence for the role of the proteins that are derived from pdxA and pdxJ is lacking. The fact that glycolaldehyde does not support the growth of pdxA mutants suggests that the pdxA gene product catalyzes a step in the pyridoxine biosynthetic pathway that is not concerned with the elaboration of 4-hydroxythreonine. Further evidence for this conclusion comes from the nutritional experiments that have been carried out with 4-hydroxy-l-threonine (16): whereas 4-hydroxy-l-threonine supported the growth of pdxB and serC mutants, strains of pdxA mutants did not grow.
Two possibilities remain: pdxA either may be required for the biosynthesis of 1-deoxy-d-xylulose, the intermediate that supplies the C5 unit, C-2',2,3,4,4', of the vitamin, or it may be implicated in one of the steps that form the pyridoxine ring system by union of 1-deoxy-d-xylulose and 4-hydroxy-l-threonine. In our view, the latter role is the more likely, since 1-deoxy-d-xylulose is implicated also in the biosynthesis of thiamin in E. coli (chapter 43), and pdxA mutants do not require thiamin, in addition to pyridoxine.
Similar reasoning can be used to assign a function to the pdxJ gene product in the formation of the ring skeleton of the vitamin: pdxJ mutants are not supported by glycolaldehyde, nor are they thiamin auxotrophs.
Thus, either the actual or the probable roles for the protein products of pdxB, pdxA, pdxJ, pdxH, and serC can be deduced from the results of the genetic studies of Winkler and of Dempsey before him.
It has not yet been established whether it is pyridoxol itself or one of its phosphate esters that is the first vitamin B6 compound to be biosynthesized from committed precursors. Thus, the progenitor of the fragment C-2',2,3,4,4' may be either 1-deoxy-d-xylulose (Fig. 3, 15) or its phosphate ester (Fig. 3, 13), and that of the other fragment, C-6,5,5', may be either 4-hydroxy-l-threonine (Fig. 5, 29) or its 4-phosphate ester (Fig. 5, 26). For simplicity, the nonphosphorylated species are shown in Fig. 6, which illustrates a possible derivation of the pyridoxol ring system from these postulated committed precursors.
To achieve the formation of the ring structure by union of the two fragments, the C-4–C-5 bond must be generated. To accomplish this condensation, an oxidative step is formally required, and it is therefore of great interest that reducing agents inhibit the biosynthesis of pyridoxol in spinach stroma chloroplasts (28). The oxidation step is here postulated to take place at C-3 of the unit derived from 4-hydroxy-l-threonine, that is, at the carbon atom destined to become C-5 of pyridoxol. An important mechanistic function in the ring closure process is assigned to the hydroxy group that is destined to become the phenolic group at C-3 of pyridoxol.
The reaction sequence for the formation of pyridoxol from 1-deoxy-d-xylulose (Fig. 3, 15) and 4-hydroxy-l-threonine (Fig. 5, 29) presented in Fig. 6 encompasses seven steps. Yet only two genes, pdxA and pdxJ, and therefore only two enzymes, are implicated in this process. Thus, it is necessary that five of these seven steps be spontaneous reactions and only two be enzyme catalyzed. In our view, these two reactions are the dehydrogenation and the decarboxylation steps. The latter is arbitrarily shown to occur as the final step of the sequence, as a decarboxylation of a vinylogous β-keto acid, but a modified sequence in which decarboxylation of a β-keto acid occurs immediately following the dehydrogenation step is equally plausible. The remaining steps, Schiff base formation, imine-enamine equilibrium, and elimination of water from a vinylogous carbinolamine and from a site adjacent to a ketone, are all likely to be spontaneous. Finally, there is precedent for the spontaneous occurrence of an aldol condensation step following an oxidation-reduction reaction in an enzyme-catalyzed process, the formation of dehydroquinic acid from 3-deoxy-d-arabinoheptulosonic acid 7-phosphate, catalyzed by 3-dehydroquinate synthase (EC 4.6.1.3) (4). Thus, every step of the postulated sequence takes place with appropriate activation and is based on biochemical precedent. It remains to be seen whether this hypothetical reaction sequence for the construction of the pyridoxine skeleton (Fig. 6) will be substantiated by experimental evidence.
Our investigations of pyridoxine biosynthesis are supported by Public Health Service grant GM50778 from the National Institute of General Medical Sciences.
References
1. Arps, P. J., and M. E. Winkler. 1987. Structural analysis of the Escherichia coli K-12 hisT operon by using a kanamycin resistance cassette. J. Bacteriol. 169:1061–1070.
2. Arps, P. J., and M. E. Winkler. 1987. An unusual genetic link between vitamin B6 biosynthesis and tRNA pseudouridine modification in Escherichia coli K-12. J. Bacteriol. 169:1071–1079.
3. Bachmann, B. J. 1987. Linkage map of Escherichia coli K-12, edition 7, p. 807–876. In F. C. Neidhardt, J. L. Ingraham, K. B. Low, B. Magasanik, M. Schaechter, and H. E. Umbarger (ed.), Escherichia coli and Salmonella typhimurium: Cellular and Molecular Biology, vol. 2. American Society for Microbiology, Washington, D.C.
4. Bartlett, P. A., and K. Satake. 1988. Does dehydroquinate synthase synthesize dehydroquinate? J. Am. Chem. Soc. 110:1628–1630.
5. Bentley, R. 1970. Molecular Asymmetry in Biology, vol. II, p. 224–229. Academic Press, New York.
6. Castilho, B. A., P. Olfson, and M. C. Casadaban. 1984. Plasmid insertion mutagenesis and lac gene fusion with mini-Mu bacteriophage transposons. J. Bacteriol. 158:488–495.
7. Cooper, R. A., and A. Anderson. 1970. The formation and catabolism of methylglyoxal during glycolysis in Escherichia coli. FEBS Lett. 11:273–276.
8. David, S., B. Estramareix, J.-C. Fischer, and M. Therisod. 1981. 1-Deoxy-d-threo-2-pentulose: the precursor of the five-carbon-chain of the thiazole of thiamine. J. Am. Chem. Soc. 103:7341–7342.
9. David, S., B. Estramareix, J.-C. Fischer, and M. Therisod. 1982. The biosynthesis of thiamine. Syntheses of [1,1,1,5-2H4]-1-deoxy-d-threo-2-pentulose and incorporation of this sugar in the biosynthesis of thiazole by Escherichia coli cells. J. Chem. Soc. Perkin Trans. 1 1982:2131–2137.
10. Dempsey, W. B. 1966. Synthesis of pyridoxine by a pyridoxal auxotroph of Escherichia coli. J. Bacteriol. 92:333–337.
11. Dempsey, W. B. 1969. Characterization of pyridoxine auxotrophs of Escherichia coli: P1 transduction. J. Bacteriol. 97:1403–1410.
12. Dempsey, W. B. 1971. Role of vitamin B6 biosynthetic rate in the study of vitamin B6 synthesis in Escherichia coli. J. Bacteriol. 108:1001–1007.
13. Dempsey, W. B. 1972. Escherichia coli mutants that require either pyridoxine or alanine. J. Bacteriol. 111:838–840.
14. Dempsey, W. B. 1987. Synthesis of pyridoxal phosphate, p. 539–543. In F. C. Neidhardt, J. L. Ingraham, K. B. Low, B. Magasanik, M. Schaechter, and H. E. Umbarger (ed.), Escherichia coli and Salmonella typhimurium: Cellular and Molecular Biology, vol. 1. American Society for Microbiology, Washington, D.C.
15. Dempsey, W. B., and H. Itoh. 1970. Characterization of pyridoxine auxotrophs of Escherichia coli: serine and pdxF mutants. J. Bacteriol. 104:658–667.
16. Drewke, C., C. Notheis, U. Hansen, E. Leistner, T. Hemscheidt, R. E. Hill, and I. D. Spenser. 1993. Growth response to 4-hydroxy-l-threonine of Escherichia coli mutants blocked in vitamin B6 biosynthesis. FEBS Lett. 318:125–128.
17. Duncan, K., and J. R. Coggins. 1986. The serC-aroA operon of Escherichia coli: a mixed function operon encoding enzymes from two different amino acids biosynthetic pathways. Biochem. J. 234:49–57.
18. Grue-Sørensen, G., and I. D. Spenser. 1989. The biosynthesis of ephedrine. Can. J. Chem. 67:998–1009. (See footnotes 6 and 7, p. 1006.)
19. Hill, R. E., P. Horsewood, I. D. Spenser, and Y. Tani. 1975. Biosynthesis of vitamin B6. Incorporation of glycolaldehyde into pyridoxal. J. Chem. Soc. Perkin Trans. 1 1975:1622–1627.
20. Hill, R. E., A. Iwanow, B. G. Sayer, W. Wysocka, and I. D. Spenser. 1987. The regiochemistry and stereochemistry of the biosynthesis of vitamin B6 from triose units. J. Biol. Chem. 262:7463–7471.
21. Hill, R. E., I. Miura, and I. D. Spenser. 1977. Biosynthesis of vitamin B6. The incorporation of [1,3-13C2]glycerol. J. Am. Chem. Soc. 99:4179–4181.
22. Hill, R. E., F. J. Rowell, R. N. Gupta, and I. D. Spenser. 1972. Biosynthesis of vitamin B6. J. Biol. Chem. 247:1869–1882.
23. Hill, R. E., B. G. Sayer, and I. D. Spenser. 1989. Biosynthesis of vitamin B6: incorporation of d-1-deoxyxylulose. J. Am. Chem. Soc. 111:1916–1917.
24. Hill, R. E., and I. D. Spenser. 1973. Biosynthesis of vitamin B6. Incorporation of terminally labelled glucose. Can. J. Biochem. 51:1412–1416. (Correction, 52:271, 1974.)
25. Hill, R. E., and I. D. Spenser. 1986. The biosynthesis of vitamin B6, p. 417–476. In D. Dolphin, R. Poulson, and O. Avramovic (ed.), Pyridoxal Phosphate: Chemical, Biochemical and Medical Aspects. John Wiley & Sons, Inc., New York.
26. Hopper, D. J., and R. A. Cooper. 1971. The regulation of Escherichia coli methylglyoxal synthase; a new control site in glycolysis? FEBS Lett. 13:213–216.
27. Iwanow, A., R. E. Hill, B. G. Sayer, and I. D. Spenser. 1984. Biosynthesis of vitamin B6: incorporation of a C-N unit derived from glycine. J. Am. Chem. Soc. 106:1840–1841.
28. Julliard, J.-H. 1992. Biosynthesis of the pyridoxal ring (vitamin B6) in higher plant chloroplasts and its relationship with the biosynthesis of the thiazole ring (vitamin B1). C. R. Acad. Sci. Paris Ser. III 314:285–290.
29. Kennedy, I. A., R. E. Hill, R. M. Pauloski, B. G. Sayer and I. D. Spenser. 1995. Biosynthesis of vitamin B6: origin of pyridoxine by union of two acyclic precursors, 1-deoxy-d-xylulose and 4-hydroxy-l-threonine. J. Am. Chem. Soc. 117:1661–1662.
30. Lam, H.-M., E. Tancula, W. B. Dempsey, and M. E. Winkler. 1992. Suppression of insertions in the complex pdxJ operon of Escherichia coli K-12 by lon and other mutations. J. Bacteriol. 174:1554–1567.
31. Lam, H.-M., and M. E. Winkler. 1990. Metabolic relationships between pyridoxine (vitamin B6) and serine biosynthesis in Escherichia coli K-12. J. Bacteriol. 172:6518–6528.
32. Lam, H.-M., and M. E. Winkler. 1992. Characterization of the complex pdxH-tyrS operon of Escherichia coli K-12 and pleiotropic phenotypes caused by pdxH insertion mutations. J. Bacteriol. 174:6033–6045.
33. Morris, J. G. 1959. The synthesis of vitamin B6 by some mutant strains of Escherichia coli. J. Gen. Microbiol. 20:597–604.
34. Notheis, C., C. Drewke, and E. Leistner. 1995. Purification and characterization of the pyridoxol-5'-phosphate: oxygen oxidoreductase (deaminating) from Escherichia coli. Biochim. Biophys. Acta 1247:265–271.
35. Pease, A. J., and M. E. Winkler. 1993. Functional and evolutionary relationships between the pdxB and serA gene products of E. coli K-12, abstr. K111, p. 280. In Abstracts of the 93rd General Meeting of the American Society for Microbiology 1993. American Society for Microbiology, Washington, D.C.
36. Roa, B. B., D. M. Connolly, and M. E. Winkler. 1989. Overlap between pdxA and ksgA in the complex pdxA-ksgA-apaG-apaH operon of Escherichia coli K-12. J. Bacteriol. 171:4767–4777.
37. Schoenlein, P. V., B. B. Roa, and M. E. Winkler. 1989. Divergent transcription of pdxB and homology between pdxB and serA gene products in Escherichia coli K-12. J. Bacteriol. 174:6084–6092.
38. Shimizu, S., and W. B. Dempsey. 1978. 3-Hydroxypyruvate substitutes for pyridoxine in serC mutants of Escherichia coli K-12. J. Bacteriol. 134:944–949.
39. Summers, M. C., and I. A. Rose. 1977. Proton transfer reactions of methylglyoxal synthase. J. Am. Chem. Soc. 99:4475–4478.
40. Tani, Y., and W. B. Dempsey. 1973. Glycolaldehyde as a precursor of pyridoxal phosphate in Escherichia coli B. J. Bacteriol. 116:341–345.
41. Tobey, K. L., and G. A. Grant. 1986. The nucleotide sequence of the serA gene of Escherichia coli and the amino acid sequence of the encoded protein, d-3-phosphoglycerate dehydrogenase. J. Biol. Chem. 261:12179–12183.
42. Umbarger, H. E., and M. A. Umbarger. 1962. The biosynthetic pathway to serine in Salmonella typhimurium. Biochim. Biophys. Acta 62:193–195.
43. Vella, G. J., R. E. Hill, B. S. Mootoo, and I. D. Spenser. 1980. The status of glycolaldehyde in the biosynthesis of vitamin B6. J. Biol. Chem. 255:3042–3048.
44. Vella, G. J., R. E. Hill, and I. D. Spenser. 1981. Biosynthesis of pyridoxol. The origin of the C2-unit, C-2',-2. J. Biol. Chem. 256:10469–10474. (Correction, 257:4008, 1982.)
45. White, R. H. 1978. Stable isotope studies on the biosynthesis of the thiazole moiety of thiamin in Escherichia coli. Biochemistry 17:3833–3840.
46. Winans, S. C., S. J. Elledge, J. H. Krueger, and G. C. Walker. 1985. Site-directed insertion and deletion mutagenesis with cloned fragments in Escherichia coli. J. Bacteriol. 161:1219–1221.
46. Wolf, E., R. E. Hill, B. G. Sayer, and I. D. Spencer. 1995. 4-Hydroxy-l-threonine, a committed precursor of pyridoxol (vitamin B6). J. Chem. Soc. Chem. Commum. 1995:1339–1340.
47. Yokota, A., and K. Sasajima. 1984. Formation of 1-deoxy-d-threo-pentulose and 1-deoxy-l-threo-pentulose by cell-free extracts of microorganisms. Agric. Biol. Chem. 48:149–158.
48. Yokota, A., and K. Sasajima. 1986. Formation of 1-deoxy-ketoses by pyruvate dehydrogenase and acetoin dehydrogenase. Agric. Biol. Chem. 50:2517–2524.
49. Zhao, G., and M. E. Winkler. 1994. An Escherichia coli K-12 tktA tktB mutant deficient in transketolase activity requires pyridoxine (vitamin B6) as well as the aromatic amino acids and vitamins for growth. J. Bacteriol. 176:6134–6138.
50. Zhao, G., and M. E. Winkler. 1995. Kinetic limitation and cellular amount of pyridoxine (pyridoxamine) 5'-phosphate oxidase of Escherichia coli K-12. J. Bacteriol. 177:883–891.
Chapter45
