Chapter 46

Biosynthesis of Biotin and Lipoic Acid

EDWARD DeMOLL

INTRODUCTION

The structure of biotin was determined by Melville et al. (36) in 1942; however, knowledge of the exact stereochemistry of the active form awaited resolution of the X-ray structure (63) in 1966. The biotin biosynthetic pathway has been studied in a variety of microorganisms and in plants, and the pathway from pimeloyl coenzyme A (CoA) is apparently identical in all (Fig. 1). The discussion in this chapter will include primarily work done with Escherichia coli. Because of its close proximity to the λ insertion site in the E. coli chromosome, the biotin operon was one of the first operons to be examined genetically. However, to date the biotin biosynthetic pathway has not been completely elucidated. The steps leading up to synthesis of pimeloyl-CoA, as well as the final step, are not fully characterized.

Fig. 1Synthesis of biotin from pimeloyl-CoA. The genes that encode the various catalytic activities are indicated for each reaction: bioF, pimeloyl-CoA ? 7-KAP; bioA, 7-KAP ? DAPA; bioD, DAPA ? dethiobiotin; bioB, dethiobiotin ? biotin. Although it is known that cysteine is the metabolic origin of the sulfur atom, the mechanism by which the sulfur atom is inserted is still unknown.

Source:

The biosynthetic pathway of lipoic acid from octanoic acid consists solely of two reactions in which sulfur is inserted into the carbon chain at C-6 and C-8. However, as with the analogous reaction in the biotin biosynthetic pathway, the mechanisms, substrates, and cofactor requirements have not been fully characterized. Although the lipoic acid biosynthetic pathway is undoubtedly widespread among microorganisms, published studies have been confined primarily to E. coli and yeast.

Both the biotin and lipoic acid molecules are attached by homologous enzymes to lysine residues in their respective apoenzymes. The intact biotin holoenzymes function as carboxyl group transfer enzymes, whereas the lipoate-containing enzymes function as redox and group transfer proteins. These functions are discussed elsewhere in this book.

BIOSYNTHESIS OF BIOTIN

Pimeloyl-CoA

The pathway of pimeloyl-CoA synthesis is presently unknown. The main question is how the odd-numbered pimeloyl moiety is produced when fatty acid synthesis generates even-numbered molecules. Some conclusions can be drawn from labeling studies where E. coli was grown with ¹⁴CO₂ and isovaleric acid as carbon sources (35). These experiments indicate that pimeloyl-CoA, but probably not pimelic acid, is part of the normal biosynthetic pathway. They also suggest that pimeloyl-CoA may be synthesized via three molecules of malonyl-CoA. Two C₁ units from each of two malonyl-CoA molecules would be lost as CO₂ during chain elongation, as is usual in fatty acid biosynthesis, whereas the third malonyl-CoA molecule would be preserved, thus yielding pimeloyl-CoA. Their results also argue against the origin of pimeloyl-CoA being due to the loss of a C₁ unit from a C₈ precursor. Although it has not been proven, the incapacity of bioC or bioH mutants to excrete any of the known components of the biotin biosynthetic pathway indicates that the ability to synthesize pimeloyl-CoA is encoded by those genes (19). No information concerning possible enzymatic mechanisms was deduced by comparison of the nucleotide sequences of the bioC and bioH genes with those in the database at the times of sequence submissions (38, 39).

7-KAP

The reaction in which 7-keto-8-aminopelargonic acid (7-KAP) is synthesized has been characterized in cell extracts of E. coli (24). Alanine and pimeloyl-CoA are condensed to form 7-KAP in a pyridoxal phosphate-dependent reaction in which the C-1 of alanine is lost as CO₂. The enzyme has been partially purified, and its native molecular weight was estimated to be approximately 45,000 (20). More recently the bioF gene sequence was reported to encode a protein with a molecular weight of 41,599 (39). Thus, the BioF protein is likely to be monomeric.

DAPA

In the synthesis of 7,8-diaminopelargonic acid (DAPA), S-adenosyl-l-methionine (SAM) acts in the unusual role of an amino donor. The methionine nitrogen of SAM is transferred to 7-KAP in a pyridoxal phosphate-dependent transaminase reaction. Initial studies of this reaction were carried out by Pai (42). He determined that methionine and pyridoxal phosphate were required to catalyze the conversion of 7-KAP to DAPA by crude extracts of E. coli. Eisenberg and Stoner (25) demonstrated, also with crude extracts, that it was not methionine, but instead SAM, that was required. Later they purified 980-fold the enzyme that catalyzes this reaction. They confirmed the specificity for SAM and pyridoxal phosphate (26) with the purified enzyme. The native protein was shown to be a dimer of 47,000-dalton subunits. The predicted molecular weight per monomer based on the nucleotide sequence of the bioA gene is 47,403 (39). The expected product of SAM deamination, S-adenosyl-2-oxo-4-methylthiobutyric acid, proved to be labile and was not detected. Direct incorporation of the methionine nitrogen of SAM into the 3'-nitrogen of biotin was later demonstrated by DeMoll et al. (17) with mass spectrometry methods.

Dethiobiotin

Genetic evidence that dethiobiotin is an intermediate in the biosynthetic pathway from DAPA to biotin was reported by Rolfe and Eisenberg (57) and Pai (40). In those studies it was shown that the ability to use DAPA in lieu of biotin could be lost by mutation. Later Pai (41) and Eisenberg and Krell (22) showed that crude extracts of E. coli could catalyze the synthesis of dethiobiotin from DAPA and bicarbonate. Krell and Eisenberg (34) purified to greater than 90% the enzyme, dethiobiotin synthetase, which catalyzes this reaction. They found that the enzyme is a dimer of identical subunits of 24,500 daltons. Krell and Eisenberg also showed that the preferred substrates are DAPA, Mg^2+· ATP, and CO₂. In the reaction CTP has about one-half the activity of ATP, Mn²⁺ has about 35% of the activity of Mg²⁺, and diaminobiotin has approximately 37% of the activity of DAPA. That the active enzyme is a homodimer was confirmed during studies that led to the crystallization of the enzyme (2). These investigators also reported the sequence of the bioD gene and determined that it encodes for a protein of 225 amino acids and a molecular weight of 24,009. They noted that the nucleotide sequence of the bioD gene originally published by Otsuka et al. (39), which had predicted a protein of 218 amino acids, was in error. Since the error was determined to be in the omission of a base, the sequence of Otsuka et al. would be out of frame for approximately the last 5% of the sequence.

Biotin from Dethiobiotin

Dittmer et al. (18) first reported the possibility that dethiobiotin was converted to biotin by yeast. Pai and Lichstein (48, 49) confirmed that this conversion also occurred in E. coli. Parry (51) and Parry and Kunitani (52) studied incorporation of sulfur into biotin in Aspergillus niger by using specifically labeled forms of ³H-dethiobiotin. The system of Parry assigns numbers to consecutive carbon atoms of dethiobiotin, beginning with the methyl carbon, which is designated C-1. Numbering continues along the carbon chain so that the two tertiary carbons of the ureido ring are C-2 and C-3, the first methylene carbon of the side chain is C-4, and so on. The results demonstrated that no loss of ³H at either C-2, C-3, or C-5 occurs (Fig. 2), so the formation of an unsaturated intermediate is not likely. These results were confirmed in E. coli by employing stable isotopes (28). Parry and coworkers (51, 52, 62) were also able to show that only a single hydrogen was lost at both the C-1 and C-4 positions. Additionally, incorporation of sulfur into C-4 of biotin is stereospecific. The 4-pro-S hydrogen is lost with retention of configuration. A hydroxylated intermediate on either C-1 or C-4 of dethiobiotin was shown to be unfeasible (27). The bioB gene, which encodes the protein that catalyzes this reaction, is homologous to the lipA gene (30, 54), which encodes the enzyme activity that catalyzes the insertion of sulfur into lipoic acid. It is therefore likely that the two reactions are catalytically similar. In a recent study of biotin biosynthesis in plant cells, Baldet et al. (3) reported the formation of an intermediate between dethiobiotin and biotin. That this intermediate exists is not surprising in light of the suspected similarity between the insertion of sulfur in the biosynthesis of both biotin and lipoic acid. The mechanism of sulfur insertion is discussed later in the section on the biosynthesis of lipoic acid.

Fig. 2Possible mechanism of sulfur insertion in biotin and lipoic acid biosynthesis. Sulfur is predominantly inserted into the methyl carbon first, probably following abstraction of a hydrogen atom. A second hydrogen atom is then abstracted from the methylene carbon (except in the low proportion of cases in which the methylene abstraction occurred first) and a second atom of sulfur is inserted in the case of lipoic acid. In the case of biotin, the sulfur already attached to the nascent biotin molecule inserts intramolecularly. All three sulfur donors originate in cysteine and are probably sulfhydryl groups, as is the form of sulfur in the intermediate compound in biotin synthesis.

Source:

The first clear evidence that implicated cysteine as the metabolic origin of the sulfur atom of biotin in E. coli was presented by DeMoll and Shive (15). These studies also eliminated methionine from consideration as the source of the sulfur atom. However, they did not rule out the possibility that cysteine was first being decomposed to H₂S before incorporation into biotin. By employing mass spectrometry methods, DeMoll et al. (17) were able to cast doubt on the possibility of H₂S being the source of the biotin sulfur. The mass spectrometry experiments showed that the incorporation of ³⁴S from either SO₄ ^2– or l-[sulfane[³⁴S]thiocystine into cysteine matched the incorporation of ³⁴S into biotin. Thiocystine was chosen as a source of sulfur because it was thought that, by feeding this compound, labeled with ³⁴S in the sulfane sulfur, to E. coli, the internal sulfur pools could be differentially labeled with ³²S and ³⁴S. It was previously thought that it was impossible to differentially label the sulfur pools in E. coli because of the strict regulation that E. coli exerts on its sulfur intake. E. coli will only transport sulfur in the form that requires the least amount of energy to convert that compound to cysteine (56). By providing thiocystine as the sulfur source, the problem of how to get two different labels into E. coli was circumvented, since sulfur was introduced into E. coli in two forms, but in a single compound. It was predicted that once inside the cell, thiocystine would decompose so as to generate H₂S primarily from reduction of the sulfane sulfur. This would cause the H₂S pool to be labeled predominately with ³⁴S, whereas the label in cysteine would be approximately two-thirds ³²S (the predominant form in natural abundance sulfur) and one-third ³⁴S (via de novo synthesis from H₂ ³⁴S). By analyzing E. coli grown with l-[sulfane-³⁴S]thiocystine as the sole sulfur source, White (67) was able to show that the percentage of ³⁴S in the H₂S pool is approximately twice that in cysteine. Consequently, if the sulfur of biotin were derived from H₂S rather than from cysteine, the ³⁴S/³²S ratio in biotin in the previously mentioned studies (17) would not have matched that in cysteine, but would have been twice that actually seen. These experiments did not rule out the possibility that cysteine in glutathionine might be the sulfur source.

Recently a report by Ifuku et al. (33) of biotin synthesis from dethiobiotin catalyzed by cell extracts indicated that several compounds might be required for biotin synthesis from dethiobiotin. In their summary they listed fructose-1,6-bisphosphate, Fe²⁺, SAM, NADPH, and KCl as being required in this reaction. Their results, however, show that NADP⁺ functioned as well as NADPH (3.0-fold versus 2.7-fold over background) and that methionine functioned as well as SAM (1.5-fold versus 1.7-fold over background). The results are difficult to interpret in another respect, since their assay for biotin synthesis from dethiobiotin employed the Lactobacillus plantarum bioassay. It has been shown that levels of dethiobiotin above approximately 1 μM allow biotin biosynthesis by and therefore growth of L. plantarum (7, 16), even in the absence of biotin. In the presence of biotin and dethiobiotin, the L. plantarum bioassay can overvalue the actual biotin concentration by 30-fold (16). Thus, the dethiobiotin alone or in combination with endogenous biotin in their assays for biotin synthesis from dethiobiotin would have been more than sufficient to cause growth of the assay organism. These conditions could account for many of the surprising results in that study.

Biotin biosynthesis from dethiobiotin by cell extracts of E. coli overexpressing the bioB gene product has been observed (W. C. Bowman and E. DeMoll, unpublished results). However, this reaction may be followed only if the endogenous biotin, both bound and unbound, is first removed prior to the assay, so that the background level of biotin does not overwhelm the amount formed in the reaction. In addition, an E. coli mutant with the entire biotin operon deleted was used as the assay organism to ensure that the biotin requirement could be only met by biotin and not the dethiobiotin in the cell-free assay. The cell-free assay is oxygen sensitive. In air at 37°C, the reaction halts within 2 h, whereas under anaerobic conditions, the reaction proceeds at a linear rate for at least 6 h. This oxygen sensitivity is likely due to oxidation of the suspected sulfur donor, cysteine, since exposure of the enzyme to air for 2 h prior to the reaction did not affect synthesis. It was also observed that the reaction is completely inhibited by EDTA. Gel filtration high-pressure liquid chromatography showed that the purified bioB gene product exists as a monomer at neutral pH.

There is an additional report of biotin synthesis in an extract (59). However, endogenous biotin was not removed prior to assay, and since there was no control without dethiobiotin, it is difficult to assess the extent of true synthesis. The authors also estimated that they purified the bioB gene product to approximately 90%. Electron paramagnetic resonance spectroscopy and elemental analysis of the BioB preparation indicated the presence of an iron-sulfur center.

Biotin Transport

Pai (44) studied biotin transport in E. coli and found that the ability to transport biotin is regulated by the external biotin concentration. High biotin levels (21 nM) repress the ability of E. coli to transport biotin. This repression is gradually lost when cells are resuspended in biotin-free medium. Pai also showed that this regulation of biotin uptake is independent of the regulation of biotin biosynthesis.

BIOSYNTHESIS OF LIPOIC ACID

The biosynthesis of lipoic acid from octanoic acid was demonstrated by Parry (50) and later by White (65). Octanoic acid can be derived directly from the fatty acid biosynthetic pathway, so the pathway of lipoic acid biosynthesis consists solely of two steps in which two sulfur atoms are introduced into the carbon skeleton at C-8 and C-6. White (68) showed that the sulfur atom in lipoic acid is derived from cysteine. As with the incorporation of the sulfur atom into biotin, the insertion of sulfur evidently occurs without the formation of either unsaturated or hydroxylated intermediates. Parry (51) synthesized several different isotopically labeled forms of octanoic acid and studied their incorporation into lipoic acid. His results indicate several features of the system. During the incorporation of sulfur into octanoic acid, no loss of ³H label on either C-5 or C-7 occurs, thereby showing that an unsaturated intermediate is unlikely. Furthermore, loss of ³H at C-6 is stereospecific. The pro-R ³H is the only hydrogen atom lost upon sulfur addition, and knowledge that the original form of C-6 was R demonstrates that sulfur addition proceeds with inversion of configuration. Stable isotope studies by White (65) confirmed these results. Later White (66) showed that this inversion did not proceed via hydroxylated intermediates. In these studies White also demonstrated that 8-mercaptooctanoic acid is a likely intermediate in the synthesis of lipoic acid; insertion of sulfur occurs first primarily at C-8 and then at C-6. The incorporation of sulfur into both lipoic acid and biotin evidently occurs by a direct insertion into saturated carbon atoms. In both cases insertion of sulfur occurs first predominantly into methyl groups (C-1 of dethiobiotin and C-8 of octanoic acid) and then into methylene groups (C-4 of biotin, with retention of configuration, and C-6 of 8-mercaptooctanoic acid, with inversion of configuration). Two different sulfur atoms are inserted in lipoic acid biosynthesis, whereas a single sulfur atom inserts into both carbon atoms in biotin biosynthesis. There is evidence to indicate that a single enzyme in each case, the bioB gene product in biotin biosynthesis (Bowman and DeMoll, unpublished results) and the lipA gene product in lipoic acid biosynthesis (31), catalyzes both sulfur insertions for each molecule. Reed and Cronan (54) suggest the possibility that an additional gene product may be needed in the case of sulfur insertion into lipoic acid, although their results would also be in agreement with a mechanism in which there would be no exclusive preference toward insertion at one carbon atom over the other. The data of White (66), which show that 6-mercaptooctanoate was found to be incorporated into lipoic acid to a significant degree, also agree with there being a preference for sulfur insertion first at C-8, but that insertion first at C-6 occurs occasionally. That the lipA and bioB genes are homologous plus the results of the biochemical data suggests that the same catalytic mechanism may be working in all four sulfur insertions. A free radical mechanism consistent with the data would involve an initial abstraction of a hydrogen atom most often initially from a methyl group. This would be followed by insertion of sulfur at the methyl group. Next, a second hydrogen atom would be abstracted from the methylene group. This would be followed by insertion of the sulfur already attached in the case of biotin, thus forming a stable five-membered ring. In the case of lipoic acid, where insertion of the attached sulfur cannot lead to the formation of a stable five-membered ring, a second molecule of cysteine would provide an additional sulfur atom to be inserted at C-6. Perhaps the fact that the second sulfur insertion is an intramolecular insertion into a sterically restricted site in the case of biotin biosynthesis and an intermolecular insertion into a less rigid structure in the case of lipoic acid biosynthesis accounts for retention of configuration at the chiral carbon in the former instance and inversion of configuration in the latter instance.

This free radical mechanism would be similar to that proposed for isopenicillin N synthase (IPNS). That enzyme catalyzes the formation of isopenicillin N (IPN) and 2 mol of H₂O from δ-(l-α-aminoadipyl)-l-cysteinyl-d-valine (ACV) and O₂. The gene that encodes IPNS, pcbC, has been cloned and overexpressed in E. coli (58). The protein is a single polypeptide of 336 amino acids. It contains two cysteine residues, which are near the active site, and probably one iron atom in the Fe²⁺ form (4). IPNS catalyzes the formation of the two rings of IPN. What is relevant to this discussion is that one of the ring closings involves the insertion of a cysteinyl sulfur into a saturated carbon atom to form a five-membered ring. This reaction has obvious similarities to the incorporation of sulfur into biotin and lipoic acid. Interestingly, IPNS is similar in size to the bioB and lipA gene products; also the conserved cysteine clusters in the bioB and lipA gene products suggest a metal binding site similar to the one in IPNS. However, the nucleotide sequence of pcbC (58) suggests no homology with either bioB or lip (30, 54). On the basis of the ability of IPNS to form a variety of products from synthetic analogs of ACV, a free radical mechanism that involves the enzyme-bound Fe²⁺ has been proposed for IPNS-catalyzed IPN synthesis (4). The insertion of sulfur in the biotin and lipoic acid (54) biosynthetic reactions does not require O₂ as the IPNS-catalyzed reaction does, and at least in the case of biotin, O₂ is inhibitory. Therefore, a primary question yet to be addressed is the mechanism by which the free radical would be generated under anaerobic conditions.

GENETICS OF BIOTIN BIOSYNTHESIS

Most of the genes of the biotin biosynthetic pathway are located in the biotin operon. In E. coli and Salmonella typhimurium (official designation, Salmonella enterica serovar Typhimurium), the operon is organized as shown in Fig. 3 (12, 13, 14, 57, 60). Transcription of the biotin operon is divergent (12, 29); it proceeds on one strand from the operator region (O) through bioA and on the opposite strand from the operator region through bioBFCD. Situated at different loci are bioH, which apparently encodes an activity responsible for catalyzing one of the steps in pimeloyl-CoA biosynthesis, and birA, which encodes the bifunctional protein that acts both as repressor of the biotin operon and the biotin holoenzyme synthetase. The sequences of all of the genes involved in biotin biosynthesis in E. coli have been determined. This list includes the bioH (38) and birA (32) genes, as well as all of the genes in the biotin operon (39). A permeability gene, bioP, has been mapped on the E. coli chromosome (21). A detailed review of work on the biotin operon prior to the appearance of most of the nucleotide sequence information has been published (21).

Fig. 3Organization of the biotin operon in E. coli and S. typhimurium (see text for details).

Source:

Early work on the regulation of biotin biosynthesis in E. coli by Pai and Lichstein (45) demonstrated that synthesis of biotin as well as other components of the biotin biosynthetic pathway was repressed by their excretion into the medium. Later they demonstrated that this repression was due solely to biotin and not other components of the pathway (46). This repression was prevented by chloramphenicol, thereby demonstrating that regulation of biotin synthesis required protein synthesis and was not due to feedback inhibition (47). Later E. coli mutants that did not regulate biotin biosynthesis were isolated (10, 43). The mutants isolated by Campbell et al. (10) also could only grow in the presence of high levels of biotin. These two phenotypes were indicative of the later-discovered bifunctional role of the birA gene product (5, 6, 9). Eisenberg et al. (23) later proved this bifunctionality with a preparation of the birA gene product that had been purified approximately 3,500-fold. With the purified protein Prakash and Eisenberg (53) were able to prove conclusively that the true corepressor of biotin regulation is biotinyl-5'-AMP. This must be bound in a complex with the repressor before the repressor can bind to the operator and shut off biotin biosynthesis. Evidence from study of the purified BirA protein (23) and from the crystal packing (8) suggested that the native BirA protein is a homodimer; however, later experiments showed that the protein is monomeric at concentrations orders of magnitude higher than those needed for DNA binding (1). These later studies also showed that two molecules of the BirA protein bind, probably in a cooperative fashion, to the operator region of the DNA.

The enzymatic reaction (reactions 1 and 2) and the binding of the repressor/corepressor to the operator (reaction 3) carried out by the birA gene product, BirA, are shown below.

biotin + ATP + BirA ↔ biotinyl-5'-AMP·BirA + PP_i(1)

biotinyl-5'-AMP·BirA + apoenzyme ↔ holoenzyme + AMP + BirA(2)

biotinyl-5'-AMP·BirA + operator ↔ biotinyl-5'-AMP·BirA·operator(3)

Under conditions in which there is sufficient biotin, all of the apoenzymes would have been converted to the catalytically competent holoenzyme forms. Consequently, an accumulation of the corepressor/repressor complex (biotinyl-5'-AMP·BirA) would occur. Shortly biotinyl-5'-AMP would reach a concentration high enough to allow binding to the operator and thus repression of biotin biosynthesis.

GENETICS OF LIPOIC ACID BIOSYNTHESIS

The lipA and lipB genes are located at 14.5 min on the E. coli chromosome (11, 64). Both of these genes have been cloned and the nucleotide sequence has been determined (30, 54, 64). The lipA sequences reported differed slightly in the two sequencing reports (30, 54); however, the predicted gene products differed significantly. Because of this, Reed and Cronan (54) experimentally determined the translation start site and found that it was 40 codons upstream of the site proposed by Hayden et al. (30). Thus, the lipA gene product has a predicted molecular mass of 36 kDa. There is an open reading frame between the lipA and lipB genes, but no product or function has been assigned to this region (54). The lipB gene product has a predicted molecular weight of 21,339 (54). A gene designated lplA (lipoate-protein ligase) has been cloned and characterized (37). Mutants with lesions in this gene were originally detected by their resistance to an analog of lipoic acid that contained selenium in place of sulfur (55). The gene product of this reaction catalyzes the ATP-dependent attachment of lipoic acid to the ε-amino groups of specific lysine residues on transacylase enzymes. The mechanism of attachment apparently occurs in a manner analogous to that of the birA gene product as discussed above, whereby lipoyl-AMP is first formed and then in a second step transfer of the lipoyl moiety to lysine occurs. In fact, the lplA and birA genes share 16% amino acid identity (37). Studies of lplA mutants indicated that E. coli also contains an additional gene that encodes a catalytic activity similar to that of the lplA gene product (37). There is no conclusive proof of the function of the lipB gene product; however, there is evidence that it is involved in a lipoyl-protein ligase reaction that is genetically and biochemically distinct from the lplA gene product (37, 54).

In other studies Smith et al. (61) examined expression of the lip locus of S. typhimurium and found that the transcriptional activity of this locus is not regulated by the level of lipoic acid. These data are consistent with those of Morris et al., who observed no significant level of regulation of lipoic acid biosynthesis by the lplA gene product, even though there is homology of that protein with the DNA binding region of the BirA protein (37).