Advances in Synthesis of Biotin and Assembly of Lipoic Acid

Abstract

Although biotin and lipoic acid are two universally conserved cofactors essential for intermediary metabolism, their synthetic pathways have become known only in recent years. Both pathways have unusual features. Biotin synthesis in Escherichia coli requires a methylation that is later removed whereas lipoic acid is assembled on the enzymes where it is required for activity by two different pathways.

INTRODUCTION

Biotin and lipoic acid (Figure 1A) share many similarities [1]. Both cofactors must be covalently attached to their cognate enzyme subunits to play their roles in central metabolism. Attachment is via an amide linkage to a specific lysine residue. Once attached, they act as part of “swinging arms” that shuttle intermediates between active sites in a covalent form of substrate channeling. The protein species modified by biotin or lipoate attachment are rare. E. coli has a single biotinylated protein and three lipoylated proteins whereas mammals have four each of biotinylated and lipoylated proteins.

Figure 1.

Biotin and lipoic acid. A. Shown are the structures of biotin, lipoic acid, reduced lipoic acid (dihydrolipoic acid) and the octanoate precursor. Note that the biotin structure is larger to aid legibility of the numbering system. B. The structures of the lipoyl and biotinoyl domains of E. coli. Note that the “thumb” structure that protrudes from the E. coli biotinoyl domain is only found in acetyl-CoA carboxylase subunits, the biotinoyl domains of other enzymes lack this protrusion and thus more closely resemble lipoyl domains. C. Superimposition of the E. coli BirA biotin protein ligase (black, PDB entry 2EWN) superimposed on the E. coli LplA lipoate ligase (gray, PDB entry 2ART) with insertions corresponding to the BirA biotin and biotinoyl adenylate binding loops colored cyan and green, respectively. Lipoyl adenylate (cyan) and biotinoyl adenylate are depicted in stick representation. The lipoyl adenylate binding loop is not visible due to its disorder. The C-termini have been deleted for clarity.

The ca. 60–80 residue protein domains to which biotinoyl and lipoyl moieties are attached have very similar structures [1]. (Figure 1B). Both biotinoyl and lipoyl domains are flattened ß-barrels, comprised of two 4-stranded anti-parallel ß– sheets. The lysine residue earmarked for cofactor attachment is found in an exposed position in a tight type I ß–turn. At the opposite end of the domains the N- and C-terminal residues of the domain are found in close together in space. Indeed, depending on the pair of domains chosen for overlay, the backbone atoms can superimposed to 1 Å and these proteins define a protein family (PF00364). Indeed, mutation of residues close to the target lysine are sufficient to allow a biotinoyl domain to accept lipoic acid [2]. Both biotinoyl and lipoyl domains are tethered to their cognate enzyme subunits by long (25–30 residue) polypeptide chains, characteristically rich in alanine, proline and charged amino acids that form flexible but extended linkers. The biotinoyl domains are at the C-termini of biotinylated proteins whereas the lipoyl domains are found at the N-termini of lipoylated proteins. Although only lipoyl linkers have been studied in detail, a lipoyl linker can be substituted for a biotinoyl linker [3]. In the original swinging arm concept only the cofactor plus the lysine side chain was thought to swing. However, has become clear that entire domains swing between active sites [4]. A useful way to think of the domain-linker arrangement is as a tadpole: the modified lysine being the nose and the linker, the tail. Note that these domain structures are highly conserved throughout biology. Foreign domains often are modified when expressed in diverse host cells. Biotin and lipoic acid can be attached by ligases called biotinprotein ligase and lipoate ligase, respectively [1]. Both ligase reactions proceed via an active site-bound acyl-adenylate intermediate that is attacked by the ε-amino group of the lysine to be modified. All biotin proteins are modified via a ligase whereas lipoate ligases function only when environmental lipoate is available [1,5]. The backbones of of the E. coli biotin and lipoate ligases can be superimposed to about 2.8 Å (Figure 1C). Together with the enzymes of lipoate assembly (discussed below) they comprise a new protein family (PFAM 03099.13). In vivo assembly of lipoyl moieties proceeds via pathway in which the lipoate precursor, octanoate, is first attached to the target proteins and then converted to lipoate via sulfur insertion [5]. This will be discussed below.

Recent progress in biotin synthesis.

Biotin is composed of two fused heterocyclic rings plus a valerate side chain (Figure 1). The cofactor is an essential vitamin for mammals and birds; only bacteria, archaea, plants and some fungi synthesize biotin. Commercial biotin is made by chemical synthesis, but industrial interest in obtaining biotin by bacterial synthesis has resulted in detailed studies of the assembly of the biotin heterocyclic rings by DuPont and Shiseido [1]. Both companies determined the origin of the biotin carbon atoms by feeding ¹³C-labeled precursors to E. coli [6,7]. These analyses showed that the biotin carbon atoms are derived from acetate, alanine and CO₂. The valerate side chain plus the first two carbons (carbons 6 and 7) of the first ring were known to come from pimelic acid, a seven-carbon α, ω-dicarboxylic acid (which can replace biotin in some bacteria). The ¹³C-labeling studies showed that six of the seven carbon atoms (carbons 2–7, Figure 1A) are derived by head-to-tail condensations as in fatty acid synthesis whereas the valerate carboxyl (C1) is derived from CO₂. The differing labeling patterns of the pimelate carboxyl carbons indicated that free pimelate was not a pathway intermediate and argued that pimelate was made by a fatty acid synthesis pathway in which the acetate-derived carboxyl (now C7 of the ring) was linked to a thiol, such as that of acyl carrier protein (ACP). Straightforward use of the membrane lipid fatty acid synthesis pathway seemed precluded by the requirement that the hydrophobic enzyme active sites must tolerate a carboxyl group. E. coli genetics had defined the genes responsible for heterocyclic ring synthesis. The remaining genes, bioC and bioH, were placed early in the pathway. BioH was annotated as an acyltransferase, a thioesterase or an esterase. The enigma was BioC, which appeared to be an Sadenosylmethionine (SAM) dependent methyltransferase, although no biotin carbon atoms originate in methionine. This enigma was solved by the realization that BioC could methylate the free carboxyl of malonyl-ACP and thereby disguise the carboxyl by eliminating its charge and providing a methyl group [8,9] (Figure 2). This was demonstrated both in vivo by feeding molecules that should (or should not) be intermediates in pimelate synthesis and by enzymology [8,9]. When pimelate methyl ester synthesis is complete BioH removes the methyl group which prevents nonproductive further elongations of the pimelate chain and frees the carboxyl for ligation to the cognate proteins [10]. About 75% of bacteria with sequenced genomes use the BioC-BioH pathway (although other esterases can replace BioH). In Bacillus subtilis and related bacteria pimelate synthesis proceeds via fatty acid synthesis however free pimelate is an intermediate [11] which an atypical acyl-CoA synthetase must convert to pimeloyl-CoA [12,13]. The α–proteobacteria have a third pathway in which a fatty acid synthetic enzyme has become a dedicated biotin synthetic enzyme [14].

Figure 2.

Synthesis of the biotin pimeloyl moiety in E. coli.

The pathway proceeds from A to D. Methyl groups are red. A. BioC-catalyzed methyl transfer from S-adenosyl-L-methionine (SAM) to malonyl-ACP initiates the pathway. B. Malonyl-ACP methyl ester enters the fatty acid synthetic cycle as the primer. Two rounds of the fatty acid chain elongation cycle results in pimeloyl-ACP methyl ester. C. BioH cleaves the methyl group from pimeloyl-ACP methyl ester is give pimeloyl-ACP which is a substrate for BioF. D. BioF begins assembly of the biotin rings by catalyzing the pyridoxal phosphate-dependent decarboxylative condensation of pimeloyl-ACP and Lalanine. Abbreviations: SAH, S-adenosyl-homocysteine; KAPA, 7-keto-8-aminopelargonic acid (also called 8-amino-7-oxononanois acid). The pathway is given in greater detail in [35].

Recent progress in assembly of the lipoyl moiety

Although lipoic acid was discovered long ago, only recently have we understood its synthetic pathway. Lipoic acid is an octanoic acid in which one hydrogen atom each of C6 and C8 is replaced by sulfur. The resulting thiols readily form the disulfide called lipoic acid. Early work showed that labeled octanoic acid fed to E. coli was converted to protein bound lipoic acid [15,16] but mechanistic details were sparse. The prevalent scenario was that free octanoic acid was converted to free lipoic acid which was attached to the cognate proteins by lipoate ligase. However, when E. coli mutants lacking the ligase were found to have normal levels of lipoyl proteins [17,18] this scenario was eliminated. The next candidate as the substrate for sulfur insertion was the fatty acid synthetic intermediate, octanoyl-ACP, and this seemed likely when the first in vitro synthesis of lipoic acid was reported [19]. However lipoyl-ACP, the putative intermediate in the reaction, could not be detected. A clue to this absence came from the lipB class of lipoate requiring E. coli strains [18]. LipB transfers either lipoyl or octanoyl moieties from ACP to lipoyl domains (octanoyl transfer is the physiological reaction). It was found that octanoate could replace lipoate in supporting growth of lipB strains and this replacement was lipoate ligasedependent [20]. These finding indicated that a remarkably atypical pathway assembles lipoyl proteins. The ligase attached octanoate to the cognate lipoyl domains and these octanoylated proteins, not octanoyl-ACP, were the substrates for sulfur insertion by LipA. The combination of octanoate and ligase activity bypassed the normal LipB pathway in which octanoyl moieties were transferred from octanoyl-ACP to the lipoyl domains. That is, unlike biotin, lipoic acid is not made and sunsequently attached, but rather is assembled on its cognate domains [20].

The Bacillus subtilis genome challenged the generality of the E. coli LipB-LipA lipoyl assembly pathway and led to discovery of another widely distributed and more complex pathway. It began with the realization that the genome encoded no good candidate for a LipB homologue. Instead, B. subtilis encoded three putative lipoate ligase homologues. A molecular genetic approach identified a B. subtilis gene that allowed growth of a lipB mutant strain in the absence of lipoate or octanoate [21]. This gene (LipM) encoded one of the putative lipoate ligase homologues which like E. coli LipB had octanoyl transferase activity although the two proteins have very dissimilar sequences. Another of the ligase homologues (LplJ) was indeed a ligase [22] that left the third homologue, LipL, as a serious loose end because deletion of the gene resulted in a lipoate requirement. Upon biochemical analysis a conundrum arose. The B. subtilis LipM octanoyl transferase was inactive with the B. subtilis LD although it was active with the E. coli LD. Moreover, the B. subtilis LD was active with the E. coli LipB transferase [23]. Two hypotheses for this conundrum were that the LipM substrate was an intermediate in pyruvate dehydrogenase assembly or that a protein somehow couples LipM to LD. The latter hypothesis was favored because it could provide a role for LipL and this was tested using the glycine cleavage system of C1 metabolism because GcvH, the lipoylated protein, is a small, discrete protein. GcvH was indeed the missing link between LipM and the LD. LipM transfers octanoate to the GcvH lysine residue to form the substrate for LipAcatalyzed sulfur insertion. Following LipA action LipL attacks the GcvH lipoamide with a cysteine thiol to form a lipoyl-LipL intermediate which in turn is attacked by the LD lysine residues to give the lipoylated 2-oxoacid dehydrogenases [23] (Figure 3A). The biochemistry predicted that loss of GcvH should result in both lipoate auxotrophy and block glycine cleavage and this is the case. Hence, the 120 residue GcvH has two discrete functions in metabolism [24]. GcvH is also involved in lipoyl assembly in two other bacteria [25], in yeast [26] and in mammals [27].

Figure 3.

The lipoyl relay pathway of lipoyl-protein assembly.

A. This pathway has been demonstrated in the Firmicute bacteria, B. subtilis [21–23] and Staphyloccus aureus [25]. Lipoyl moieties are assembled on the GcvH protein of the glycine cleavage system and then transferred to the lipoyl domains (LDs) of the 2-oxoacid dehydrogenases. This pathway requires LipM, a distinct octanoyl transferase to transfer octanoate from octanoyl-ACP to GcvH via an acyl-enzyme intermediate. The LipA radical SAM enzyme then inserts two sulfur atoms into octanoyl-GcvH to produce dihydrolipoyl-GcvH. The amide linkage of dihydrolipoylGcvH is then attacked by the lipoyl amidotransferase LipL which transfers the dihydrolipoyl moiety to the 2-oxoacid dehydrogenase LDs. GcvH is required for 2oxoacid dehydrogenase lipoylation in these bacteria in addition to its role in glycine cleavage [24]. B. For comparison the straightforward E. coli pathway which illustrates the two-step sulfur insertion catalyzed by LipA is illustrated.

A major recent advancement in study of the LipA sulfur insertion enzyme was the use of small octanoylated peptides in place of the large domains. These peptides facilitated both analytical and crystallographic studies of this unusual radical SAM enzyme. Spectral studies had shown that LipA contains two [4Fe-4S] clusters, the SAM radical [4Fe-4S] cluster and a second cluster (called the auxiliary cluster) thought to be the source of the lipoate sulfur atoms [28]. Indeed ]³⁴S-labeled LipA gave rise to ³⁴S-lipoyl groups and af ³⁴S-LipA and ³²S-LipA mixture showed that both sulfur atoms come from the same polypeptide [29]. Use of an octanoylated peptide showed that LipA first inserts sulfur at C-6 and this intermediate stays bound to LipA consistent with both sulfur atoms being derived from the same LipA polypeptide [30]. Reductive cleavage of second SAM molecule results in sulfur insertion at C-8 [30]. The first LipA crystal structure showed the expected two [4Fe4S] clusters per monomer in a common partial (β6α6) TIM barrel with the clusters located at opposite ends within the barrel [31]. One cluster is the classical radical SAM reduction cluster whereas the auxiliary cluster has an unusual and essential serine ligand. Subsequent trapping of the C6 substrate radical and its analysis by electron paramagnetic resonance demonstrated juxtaposition of the octanoyl chain and the auxiliary cluster [32] which has now been directly observed in a crystal structure of LipA in a complex with an octanoyl-peptide substrate [33]. In the presence of one equivalent of SAM to trap the enzyme at the C6 stage the serine ligand dissociates from the cluster, an iron ion is lost and a sulfur atom of the auxiliary cluster is covalently attached to C6 of the octanoyl chain [33]. Hence we now have a dramatic demonstration that the auxiliary cluster is the source of the lipoate sulfur atoms (Figure 4). Is LipA is an enzyme? Enzymes are catalytic by definition and in vitro even the best LipA preparations are less than catalytic. Without auxiliary cluster repair LipA would be a substrate rather than an enzyme. From the relative numbers of E. coli LipA molecules and substrate molecules, a convincing argument could be made that LipA is catalytic [5]. However, it was unknown how cluster repair was accomplished until a very recent report that NfuA, a (Fe-S) cluster carrier protein, or IscU, a(Fe-S) assembly scaffold protein, can support LipA catalysis by efficiently reconstituting the auxiliary cluster [34].

Figure 4.

A simplified LipA reaction mechanism. A molecule of SAM is reductively cleaved to give an 5’-adenosyl radical (5’-Ado), This radical abstracts a hydrogen from C6 of the octanoyl moiety to give a C6 carbon radical. The C6 radical attacks a sulfur atom adjacent to the Fe that was liganded to the atypical serine ligand (the other iron atoms are liganded to cysteine). A reduced iron atom is lost from the cluster followed by a second reductive cleavage of a second SAM molecule to an 5’Ado radical that abstracts a hydrogen from C8 of the octanoyl moiety resulting in a C8 radical that attacks a second sulfur atom. Protonation of the doubly cross-linked complex releases the dihydrolipoyl moiety plus a degraded Fe-S cluster shown as a {2Fe-2S] cluster in this scheme although the exact species is unknown. A more complete scheme is found in [33].

Conclusions

Although the E. coli pimelate synthetic pathway is known, two other pimelate synthetic pathways await elucidation. In lipoyl assembly the detailed mechanism of LipA remains to be explored. It seems likely that substrate binding engenders a conformational change that releases the serine ligand of the auxiliary cluster to allow C6 of the octanoyl chain access to the first sulfur atom, this is consistent with the observation that substitution of cysteine or alanine results in significantly less uncoupled SAM turnover than the wild type protein [31]. Since uncoupled SAM turnover is the function of the RS [4Fe-4S] cluster, this suggests that the two clusters communicate. The enzyme gymnastics necessary for the second sulfur insertion at C8 also remain unknown.