Biotin, a universal and essential cofactor: synthesis, ligation and regulation

ABSTRACT

Biotin is a covalently attached enzyme cofactor required for intermediary metabolism in all three domains of life. Several important human pathogens (e.g. Mycobacterium tuberculosis) require biotin synthesis for pathogenesis. Humans lack a biotin synthetic pathway hence bacterial biotin synthesis is a prime target for new therapeutic agents. The biotin synthetic pathway is readily divided into early and late segments. Although pimelate, a 7-carbon α,ω-dicarboxylic acid that contributes 7 of the 10 biotin carbons atoms, was long known to be a biotin precursor, its biosynthetic pathway was a mystery until the Escherichia colipathway was discovered in 2010. Since then, diverse bacteria encode evolutionarily distinct enzymes that replace enzymes in the E. coli pathway. Two new bacterial pimelate synthesis pathways have been elucidated. In contrast to the early pathway, the late pathway, assembly of the fused rings of the cofactor, was long thought settled. However, a new enzyme that bypasses a canonical enzyme was recently discovered as well as homologs of another canonical enzyme that functions in synthesis of another protein-bound coenzyme, lipoic acid. Most bacteria tightly regulate transcription of the biotin synthetic genes in a biotin-responsive manner. The bifunctional biotin ligases which catalyze attachment of biotin to its cognate enzymes and repress biotin gene transcription are best understood regulatory system.

Biotin is an essential cofactor required throughout biology that is synthesized in bacteria, fungi and plants by mixtures of conserved and diverse pathways and is required for virulence by some pathogenic bacteria.

INTRODUCTION

Biotin (also called vitamin H or B7) is an essential cofactor of biotin-dependent enzymes including carboxylases, decarboxylases and transcarboxylases, all of which participate in various aspects of intermediary metabolism, including gluconeogenesis, fatty acid synthesis and amino acid metabolism (Walsh 2006, Tong 2013). The chemical structure of biotin consists of a tetrahydroimidizalone ring fused with an organosulfur-containing tetrahydrothiophane ring that bears a valeric acid substituent (Fig. 1A). Biotin must be covalently attached to its cognate enzyme proteins (e.g. carboxylases) for function and attachment is mediated by formation of an amide linkage between the biotin carboxyl group and the ε-amino group of a specific lysine residue in the conserved biotin accepting domains of the apoenzymes (see below). This reaction is catalyzed by an enzyme called biotin protein ligase (BPL) in bacteria or (less often in recent years) holocarboxylase synthetase in animals (Chapman-Smith and Cronan 1999).

Figure 1.

(A) The E. coli biotin biosynthetic pathway and (B) the proposed reaction mechanism of the new biotin biosynthetic enzyme BioU that bypasses BioA. The biotin biosynthetic pathway can be divided into two stages: the generation of the pimelate moiety catalyzed by BioC-BioH pathway in E. coli and the conserved biotin ring assembly pathway catalyzed by BioF, BioA, BioD and BioB, respectively.

The biotin biosynthetic pathway can be conveniently divided into two stages: the generation of the pimelate moiety (a 7-carbon atom α,ω-dicarboxylate) which contributes 7 of the 10 biotin carbon atoms and the pathway for assembly the biotin rings (Fig. 2). The two stages differ in that pimelate synthesis proceeds by several different pathways in bacteria. Moreover, a given step in a pathway can be carried out by different enzymes in diverse bacteria. In contrast, the enzymes of ring assembly are (with a recently discovered exception) highly conserved. Our knowledge of the pimelate moiety biosynthesis was a long-standing puzzle until the BioC-BioH pathway was defined in Escherichia coli (Lin, Hanson and Cronan 2010). Based on extant genome annotations, this seems to be the dominant pathway in bacteria (albeit performed by diverse proteins). Since 2010, two other pathways in Bacillus subtilis(and close relatives) and in the α-proteobacteria, have been reported that can functionally replace the BioC-BioH pathway.

Figure 2.

The B. subtilis BioW reaction generates pimeloyl-CoA for incorporation into KAPA followed by assembly of the biotin rings.

It has long been known that mammals lack the ability to synthesize biotin and must acquire biotin from their diet and/or intestinal microflora. Severe biotin deficiency due to BPL mutations and/or lack of biotinidase, the enzyme that releases biotin from ingested proteins, lead to human clinical abnormalities such as skin and neurological disorders plus growth retardation (Baumgartner and Suormala 2016). Although de novo biosynthesis of biotin occurs in bacteria, archaea, fungi and plants, accumulation of biotin by intake from the environment is energetically more favorable than synthesis. Depending on the pathway, as many as 19 molecules of adenosine triphosphate (ATP) and at least 6 enzymes are required to generate a molecule of biotin (Feng, Zhang and Cronan 2013). The major energetic cost is the two S-adenosyl-methionine molecules (hence six ATP equivalents) cleaved in formation of the tetrahydrothiophene ring.

In the second stage of the biotin biosynthetic pathway, the pimelate moiety contributes two of the ring carbon atoms plus those of the valeryl sidechain. The remainder of the coenzyme is constructed from L-alanine, CO_2, sulfur atoms donated by a [Fe-S] cluster and a nitrogen atom from S-adenosyl-L-methionine (SAM). This process proceeds through sequential catalysis by four highly conserved enzymes encoded by the bioF, bioA, bioD and bioB genes (Fig. 1). Once complete, the biotin molecule is attached to the cognate biotin-requiring enzymes of intermediary metabolism by BPL. Furthermore, in some bacteria, the BPL not only functions to generate enzyme-bound biotin but acts as a biotin-responsive transcriptional repressor of biotin operons and biotin transporters (Beckett 2007, Cronan 2008, Chakravartty and Cronan 2013, Henke and Cronan 2014, Henke and Cronan 2016) (see below).

BIOTIN BIOSYNTHETIC PATHWAYS

Diversity in synthesis of the pimelate moiety

The best characterized pathway of pimelate moiety synthesis is that of E. coli which proceeds by a modified fatty acid synthetic pathway altered by methylation and demethylation enzymes called BioC and BioH (Lin, Hanson and Cronan 2010). Since then, the different pathways of B. subtilis and the α-proteobacteria have been defined and will be reviewed.

The BioC-BioH pathway

The biotin molecule is derived from CO₂, acetate and L-alanine. This was first defined in E. coli by Nuclear Magnetic Resonance (NMR) analyses of the patterns of incorporation into biotin of various metabolic precursors labeled with ¹³C (Ifuku et al. 1994; Sanyal, Lee and Flint 1994). These studies demonstrated that the pimelic acid moiety is formed from three acetate units incorporated in a head to tail manner as seen in fatty acid synthesis (Ifuku et al. 1994, Sanyal, Lee and Flint 1994) with the seventh pimelate carbon originating as CO₂. The challenge was how to make a seven-carbon dicarboxylic acid using a pathway designed to make the long chain monocarboxylic fatty acids found in phospholipids. In 1963, Lezius and coworkers (Lezius, Ringelmann and Lynen 1963) suggested that the pimelate moiety could be formed by condensation of three molecules of a malonyl thioester in two decarboxylative Claisen-like condensations to obtain pimeloyl-CoA. However, the hydrophobicity of the active sites evident in crystal structures of fatty acid synthetic enzymes (White et al. 2005) argued that the postulated free carboxyl group of the primer malonyl thioester would not be tolerated. This dilemma is avoided in E. coli and many other bacteria by a pathway in which the free carboxyl group of malonyl-acyl carrier protein (ACP) is methylated by BioC, a SAM-dependent methyltransferase. Methylation eliminates the charge of the free carboxyl group, and introduces a methyl group that mimics that of the canonical acetyl-thioester primer. This methylation deceives the fatty acid biosynthetic enzymes into utilizing malonyl-ACP methyl ester as a primer (Lin, Hanson and Cronan 2010) (Fig. 1A). Following two reaction cycles of canonical fatty acid synthesis, the methyl group of the pimeloyl-ACP methyl ester is removed by BioH, a short-chain fatty acid esterase, to give pimeloyl-ACP which is condensed with L-alanine by BioF to begin assembly of the fused rings of biotin (Fig. 1).

The first cycle of fatty acid chain elongation synthesizes the C5 dicarboxylate. A second cycle elongates this to the C7 dicarboxylate to give pimeloyl-ACP methyl ester. The BioH esterase removes the methyl group, producing pimeloyl-ACP (Fig. 1A). This cleavage prevents further elongation by fatty acid biosynthesis and frees the carboxyl group required for ligation to the cognate intermediary metabolism proteins (Lin, Hanson and Cronan 2010, Agarwal et al. 2012).

BioH (EC. 3.1.1.85) is a member of the α/β-hydrolase-fold superfamily containing the classical catalytic triad: Ser-His-Asp and a pentapeptide motif (Gly-X-Ser-X-Gly) that includes the key serine nucleophile. The cocrystal structure of an inactive form of E. coli BioH (erroneously annotated as Shigella flexneriBioH) with pimeloyl-ACP methyl ester demonstrates how the enzyme determines the appropriate acyl chain length.

Seven evolutionarily distinct pimeloyl-ACP methyl esterases encoded by diverse bacterial species have been identified: BioH, BioG, BioJ, BioK, BioV, BtsA and BioUh (Rodionov, Mironov and Gelfand 2002, Shapiro, Chakravartty and Cronan 2012, Feng et al. 2014, Bi et al. 2016, Hang et al. 2019, Zeng et al. 2020). All are restricted to specific bacterial species except BioH and BioG which are widely distributed among bacterial species (Rodionov, Mironov and Gelfand 2002; Bi et al., 2016). An archaeal enzyme with a proposed Ser-His-Asp catalytic triad and a GSSXGG motif that complements an E. coli ∆bioH strain has also been reported (Chow et al. 2018). (Note that we have designated BioU as BioUh with the ‘h’ standing for hydrolase to distinguish it from another BioU that catalyzes a novel reaction late in the pathway-see below.) The pimeloyl-ACP methyl esterases are encoded by genes that are sometimes located immediately upstream of bioC within biotin gene clusters and in other cases are free-standing genes that lack genome context (as does E. coli bioH). To date, all known esterases that cleave pimeloyl-ACP methyl ester are members of the α/β-hydrolase family and contain the signature Ser-His-Asp catalytic triad. The species-specific genes remain a puzzle (Rodionov, Mironov and Gelfand 2002). The open reading frames of the hydrolases encoded in gene clusters often overlap with both the downstream bioC and upstream bioF genes indicating that the arrangements are long standing. Remarkably, there are bacteria that encode two esterase genes (bioH and bioG) and others that encode a functional BioG-BioC fusion protein (Shapiro, Chakravartty and Cronan 2012). One esterase, BioJ of Francisella novicida (a surrogate of Francisella tularensis, the causative bacterium of tularemia), is required for pathogenesis in mice (Feng et al. 2014). This and other examples demonstrate that bacterial pathogens are ‘on their own’ for biotin. Mammals cannot synthesize biotin and mammalian tissues and blood contain insufficient free biotin to support bacterial growth. The BtsA of Moraxella catarrhalis, a gram-negative, human-restricted opportunistic bacterial pathogen, has also been shown to be required for invasion of lung epithelial cells and survival within macrophages (Zeng et al. 2020). Another possible example of a hydrolase required for pathogenesis is BioC of mycobacteria (Hu and Cronan 2020). In the nonpathogenic Mycobacterium smegmatis inactivation of bioC results in biotin auxotrophy. The analogous Mycobacterium tuberculosis protein has the same in vitro activity but has not been tested due to biosafety regulations. If the M. tuberculosis bioC is required for biotin synthesis, it would be good target for development for anti-tuberculosis drugs as validated for M. tuberculosis BioA (Bockman, Mishra and Aldrich 2020).

High quality crystal structures of E. coli BioH (Sanishvili et al. 2003, Agarwal et al. 2012) and Haemophilus influenzae BioG (Shi et al. 2016) are available. These are useful in modeling other pimeloyl-ACP methyl esterases. In contrast, crystal structures of BioC proteins are lacking. BioC homologues from a series of diverse bacteria (including E. coli) form insoluble inclusion bodies when expressed in E. coli, the exception being Bacillus cereus BioC (Lin and Cronan 2012, Haushalter et al. 2017). Bacillus cereus BioC, although soluble and stable, has not yet been crystallized. BioT, a protein from the rickettsia bacterium Ehrlichia sp. that functionally replaces E. coli BioC might also be a good candidate. BioT seems a soluble and stable protein that may be more amenable to crystallization (Zeng et al. 2020). BioT shares only 28% sequence identity with E. coli BioC and that only in the methyltransferase domain (Hang et al.2019). Hence, another possibility is the M. smegmatis BioC which has a sequence highly diverged from other BioC proteins (Hu and Cronan 2020).

The BioW pathway

Bacilli are gram-positive spore-forming bacteria that fall into two groups commonly called the B. cereus group and the B. subtilis group (Alcaraz et al. 2010). Surprisingly, the pimelate synthetic pathways of the two groups are remarkedly different (Rodionov, Mironov and Gelfand 2002). The B. cereus organisms make biotin by the BioC-BioH (E. coli) pathway whereas the B. subtilis group lacks bioC and bioH genes and have two novel genes in their biotin operons, bioW and bioI, that bracket the conserved ring assembly genes (Bower et al. 1996). In a previous review, we predicted that BioI would be the source of pimelate as pimeloyl-ACP (Lin and Cronan 2011). This prediction was based on reports that BioI, a cytochrome P450 enzyme (CYP107H), could cleave fatty acid chains and fatty acyl-ACPs to give traces of pimelate (Cryle and De Voss 2004) and pimeloyl-ACP (Stok and De Voss 2000, Cryle 2010), respectively. Moreover, a beautiful cocrystal structure of BioI bound to acyl-ACP with the heme iron atom poised over the C7-C8 bond of the pimeloyl acyl chain (Cryle and Schlichting 2008) seemed to complete the picture. In this scenario, BioW, which was thought to be a pimeloyl-CoA synthetase, played a secondary role perhaps in scavenging pimelic acid from the environment. This seemed plausible since the major source of free pimelate in the environment is cow and horse urine and B. subtilis is a soil bacterium (first called the ‘hay bacillus’). However, subsequent genetic and biochemical studies demonstrate that the prediction was incorrect: bioW is an essential biotin synthetic gene whereas bioI is nonessential and is an enigma (Manandhar and Cronan 2017, Manandhar and Cronan 2018). The bioW versus bioI question was settled by genetic analysis. Deletion of bioI, the last gene of the operon, had no effect on biotin synthesis; growth proceeded normally in the absence of biotin (Manandhar and Cronan 2017, Manandhar and Cronan 2018). In contrast, disruption of bioW, the first gene of the operon, resulted in biotin auxotrophy (the disruption cassette contained a promoter to drive the downstream genes). The nonessential nature of bioI was further demonstrated by anaerobic growth of a wild type B. subtilis strain in the absence of biotin (Manandhar and Cronan 2017). Cytochrome P450 enzymes require oxygen for activity (Meunier, de Visser and Shaik 2004).

Although BioW was considered a pimeloyl-CoA synthetase, there were misgivings because the protein is half the size of the canonical acyl-CoA synthetases and lacks all of the sequence motifs characteristic of this well-studied enzyme family. Given its atypical nature, the BioW reaction mechanism was studied in detail (Manandhar and Cronan 2013). Remarkably, despite its small size, BioW has the same mechanism as the classical acyl-CoA synthetases. BioW requires ATP, CoA plus a metal ion and the reaction proceeds through the standard two half-reactions; formation of pimeloyl-adenylate followed by transfer of the pimelate moiety to CoA with release of adenosine monophosphate (AMP) and pyrophosphate (Fig 2).

The B. subtilis enzyme is specific for pimelate whereas shorter (C5) and longer (C8) dicarboxylates are converted to the adenylate. Rather than catalyzing transfer to CoA, the C5 and C8 adenylates are hydrolyzed to the acid plus AMP in proofreading reactions similar to those of the pre-transfer proofreading activity of some aminoacyl-ribonucleic acid (RNA) synthetases (Manandhar and Cronan 2013). Aquifex aeolicus BioW also shows proofreading of the C5 dicarboxylate whereas the C8-adenylate is formed but cannot be transferred to CoA (Estrada et al. 2017). The crystal structures of both the B. subtilis (Wang et al. 2017) and A. aeolicus (Estrada et al. 2017) enzymes have been determined and are in excellent agreement.

The establishment of B. subtilis BioW as a bona fide acyl-CoA synthetase raised the question of the origin of the pimelate substrate and argued that pimeloyl-CoA synthesis must proceed through a free pimelate intermediate. The possibility of free pimelate as an intermediate was tested by ¹³C labeling studies modeled on those done in E. coli. In E. coli, the labeling pattern of the pimelate carbons showed that the carboxyl group consumed in KAPA synthesis came from acetate whereas the free carboxyl originated as CO₂ (Ifuku et al. 1994, Sanyal, Lee and Flint 1994). Since pimelate is a rotationally symmetrical molecule this showed that it is synthesized without a free pimelate intermediate.

In contrast, when B. subtilis was grown with acetate labeled with ¹³C in either carbon atom and the biotin synthesized was subjected to NMR analyses, the free carboxyl group and that incorporated into KAPA had the same labeling pattern (Manandhar and Cronan 2017). Therefore, pimelate synthesis in this bacterium proceeds through free pimelate. Moreover, the alternate patterns of acetate incorporation indicated that the pimelate acyl chain was assembled by head to tail incorporation, the pattern seen in fatty acid synthesis. The implication that fatty acid synthesis is involved was supported by inhibition of biotin synthesis by cerulenin, an antibiotic that inhibits the major acyl chain elongation enzyme (Manandhar and Cronan 2017). The overall mechanism of pimelate synthesis in B. subtilis remains undetermined. The most straightforward process would be use of malonyl-ACP as primer in place of the usual acetyl-CoA. After two elongation cycles, the pimeloyl-ACP would be cleaved by a thioesterase to provide free pimelate for BioW-catalyzed conversion to pimeloyl-CoA (the obligate BioF substrate—see below). However, the identity of the postulated thioesterase remains unknown.

BioZ, the α-proteobacterial pimelate moiety synthesis pathway

A third pimelate synthesis pathway is found in the α-proteobacteria typified by the plant pathogen Agrobacterium tumefaciens and the animal pathogen Brucella abortus together with the symbiotic nitrogen-fixing bacteria, Mesorhizobium japonicum and Sinorhizobium fredii. These bacteria lack all of the pimelate synthesis genes discussed above and synthesize the pimelate moiety by a pathway distinctly different from those of E. coli and B. subtilis. Instead, α-proteobacteria have a gene encoding a putative 3-ketoacyl-ACP synthase (condensing enzyme-like) called bioZ clustered with the biotin ring-forming genes (Sullivan et al. 2001). BioZ proteins have about 35% sequence identity with the E. coli and Streptomyces coelicolor FabH 3-ketoacyl-ACP (acyl carrier protein) synthase III (KAS III) proteins which catalyze the initial elongation/condensation in the fatty acid synthetic pathway. Hence, these proteins are often annotated as FabH proteins, although a valid fabH gene is found elsewhere in these genomes within a cluster of fatty acid and phospholipid synthesis genes. In fatty acid synthesis, FabH condenses a molecule of acetyl-CoA with a molecule of malonyl-ACP to form acetoacetyl-ACP plus CO₂.

The presence of FabH homologs in biotin synthesis gene clusters was first reported by Ronson and coworkers (Sullivan et al. 2001) who named these proteins BioZ. The amino acid sequences of BioZ proteins indicate conservation of the canonical FabH Cys-His-Asn catalytic triad. These investigators confirmed the role of BioZ in biotin synthesis by the biotin auxotrophy resulting from inactivation of bioZ in M. japonicum. The M. japonicum biotin gene cluster contained genes that replaced function of the E. coli enzymes required for assembly of the fused heterocyclic rings of biotin. Therefore, BioZ seemed likely to function in the earliest phase of biotin synthesis, formation of the pimeloyl-thioester intermediate (Sullivan et al. 2001). This was shown to be the case when expression of BioZ allowed growth of an E. coli bioH strain (Sullivan et al. 2001) and of a ∆bioC ∆bioH strain in the absence of biotin (Hu,Y and Cronan 2020).

Deletion of bioZ in A. tumefaciens resulted in biotin auxotrophy in confirmation of the M. japonicum auxotrophy (Hu and Cronan 2020). Moreover, upon replacement of active site C115 nucleophile of the chromosomal BioZ with serine (the C115S mutation) inactivated the protein and resulted in biotin auxotrophy. A clue to BioZ function was provided by two brief papers in the early 1970s in which Ogata and coworkers reported that resting cell suspensions of A. radiobacter IAM 1562 and related bacteria synthesized dethiobiotin (DTB, the last intermediate of the pathway) from glutarate (Ogata, Izumi and Tani 1970; Ogata, Izumi and Tani 1972). L-lysine supplementation also stimulated DTB production, although the levels obtained were only about one-fourth of those obtained with glutarate. These results were specific to glutarate and lysine; supplementation with other dicarboxylic acids (including pimelate) and other amino acids failed to stimulate biotin synthesis. Glutarate was further implicated by the demonstration that radioactive glutarate was incorporated into DTB (Ogata, Izumi and Tani 1972). Given the reaction catalyzed by FabH enzymes, a plausible pathway was BioZ-catalyzed elongation of glutaryl-CoA by a decarboxylating Claisen condensation with malonyl-ACP to form pimeloyl-ACP (Fig. 3). This was shown to be the case by a series of in vitro experiments (Hu and Cronan 2020). The next question was the source of glutarate. In Pseudomonas putida, glutarate is formed during degradation of L-lysine and if a similar pathway existed in A. tumefaciens, a mechanism was needed to convert glutarate to glutaryl-CoA. Agrobacterium tumefaciens encodes a predicted type III acyl-CoA transferase (annotated as CaiB) that seemed a likely candidate for conversion of glutarate to glutaryl-CoA via transfer of CoA from succinyl-CoA. The putative transferase was purified and shown to catalyze transfer of CoA from succinyl-CoA to glutarate (Hu and Cronan 2020). Finally, disruption of the caiB gene resulted in biotin auxotrophy (Hu and Cronan 2020).

Figure 3.

Current pathway of pimeloyl-ACP synthesis in α-proteobacteria. BioZ catalyzes a decarboxylative Claisen condensation to form 3-keto-pimeloyl-ACP which then enters fatty acid synthesis. Further elongation is halted by reaction with BioF and L-alanine to form KAPA as in Fig. 2.

Pimeloyl-ACP synthesis in α-proteobacteria uses an unusual primer, glutaryl-CoA. Since glutaryl-CoA is generally associated with degradative pathways such as that for L-lysine (Thompson et al. 2019), this raises the question of how an essential biotin synthesis pathway can depend on L-lysine catabolism, a degradative pathway. Two factors provide a rationale. First, A. tumefaciens like most bacteria requires only trace amounts of biotin. Agrobacterium tumefaciens and E. coli biotin auxotrophs grow well when supplemented with 2 nM biotin (Feng et al. 2014). Since E. coli requires only a few hundred biotin molecules per cell (Cronan 2001), A. tumefaciens growth seems likely to require a similarly small number. A second factor is that lysine is an abundant protein residue and A. tumefaciens and other plant associated α-proteobacteria contain lysine-modified phospholipids (Sohlenkamp and Geiger 2016). Hence, glutaryl-CoA resulting from degradation of lysine liberated in turnover of proteins and phospholipids could readily provide the modest glutaryl-CoA levels required for pimeloyl-ACP synthesis.

THE (ALMOST) TOTALLY CONSERVED PATHWAY FOR ASSEMBLY OF THE BIOTIN RINGS

BioF, 8-amino-7-oxononanoate synthase (KAPA synthase)

In contrast to the generation of the pimelate precursor, the second stage of biotin biosynthetic pathways (Fig. 1A) is highly conserved among biotin biosynthetic organisms even those lacking the early stage genes, as in many yeast strains (see below). The pathway begins with the activated pimelate moiety. It can be pimeloyl-CoA or pimeloyl-ACP depending on substrate specificity of the first enzyme of ring formation called KAPA synthase (7-keto-8-aminopelargonic acid synthase or formally AON synthase) encoded by the bioF gene. Escherichia coli ∆bioC ∆bioH strains grow without biotin when provided with either B. subtilis BioI which makes pimeloyl-ACP or BioW plus pimelate to produce pimeloyl-CoA (Manandhar and Cronan 2017, Manandhar and Cronan 2018). Hence, in vivo E. coli BioF can utilize either of the pimelate thioesters for KAPA synthesis and this has been demonstrated by enzymatic analyses in vitro (Manandhar and Cronan 2018). In contrast, B. subtilis BioF can accept only pimeloyl-CoA (Manandhar and Cronan 2018). Therefore, the B. subtilis bioI gene presents an enigma in that its cognate BioF, the next enzyme in the pathway, cannot utilize the product of the enzyme it encodes. Moreover, bioI expression in Corynebacterium glutamicum strains that also express the E. coli bioF and bioB genes resulted in biotin prototrophy (Ikeda et al. 2013). Therefore, BioI is a functional enzyme in its native host and in foreign hosts.

Why does BioI, a functional enzyme, not suffice for pimelate synthesis with the native BioF? The answer may lie in nitrogen assimilation. Unlike E. coli, B. subtilis can efficiently reduce nitrate or nitrite to ammonium, the form of nitrogen required in biosynthetic reactions (Nakano and Zuber 1998). Although long considered an obligately aerobic bacterium, B. subtilis grows well without O₂ when nitrate or nitrite are provided as alternate electron-acceptors (Nakano and Zuber 1998). Therefore, in accepting electrons, nitrite or nitrate provide ammonium, a valuable metabolic commodity. However, the responsible enzymes, nitrate reductase and nitrite reductase, function only under anaerobic conditions (Nakano and Zuber 1998) where BioI cannot function (Manandhar and Cronan 2018). In soil, the natural environment of B. subtilis, nitrification results in nitrate and nitrite which accumulate to high concentrations. A possible scenario for the emergence of BioW is that an ancestor(s) of the present-day B. subtilis was a strict aerobe where BioI sufficed and BioF accepted acyl-ACP substrates. When the ability to obtain valuable and essential ammonium from nitrate and nitrite was developed, the progenitor cells required an anaerobic means to synthetize the pimeloyl thioester required for biotin synthesis, conditions where BioI is nonfunctional. However, this scenario fails to explain why BioI remains a functional enzyme since B. subtilis BioF ignores the BioI product in favor of that produced by BioW. One possibility is that it is an evolutionary relic that will be lost, since BioI is found only in B. subtilis and its very close relatives. In B. subtilis, a terminator blocks 85% of bioI transcription from the upstream bio operon genes (Bower et al. 1996, Perkins et al. 1996) implying that high levels of BioI may be deleterious to growth.

Other BioW-encoding bacteria (e.g. Staphylococcus aureus and Aquifex aeolicus) lack BioI although they grow in the presence of oxygen. In vitro B. subtilis BioI hydroxylates and cleaves free fatty acids to generate a variety of products (Cryle and De Voss 2004). If this occurs in vivo, some products could be toxic. Fatty acids linked to ACP show specificity for pimelate generation (Stok and De Voss 2000). However, the enzyme showed only a very marginal production of pimeloyl-ACP, a level that was much less than catalytic. These considerations suggest that BioW might be a better-behaved enzyme than BioI in vivo.

Recent studies indicate that the E. coli BioF KAPA synthase can utilize either pimeloyl-CoA or pimeloyl-ACP as a substrate whereas B. subtilis BioF KAPA synthase can utilize only pimeloyl-CoA (Manandhar and Cronan 2018). We have proposed that the B. subtilis enzyme may lack the ‘positive patch’ of basic residues required to dock the strongly acidic ACP moiety of pimeloyl-ACP. ACP-enzyme interactions are predominately hydrophilic in nature, with almost all being salt bridges between acidic residues of helix II and enzyme arginine or lysine residues (Cronan 2014). The E. coli and M. smegmatis BioF crystal structures show that the protein is shaped like a half-open hand with the active site (identified by the pyridoxal 5-phosphate bound to Lys-236) located at roughly the palm/fingers interface. Attempts to obtain cocrystals of E. coli BioF with pimeloyl-ACP have not yet been successful. However, in silico docking studies suggest that a properly oriented ACP molecule could readily fit into the lumen of the half-open hand of E. coli BioF. Moreover, five basic residues (four Arg and one Lys) are present on the luminal surface and could readily form salt bridges with ACP. The crystal structure of M. smegmatis BioF presents a similar array of basic residues consistent with its use of pimeloyl-ACP. However, these workers were unable to bind pyridoxal 5-phosphate to the protein either in solution or by crystal soaking. An unpublished structure of F. tularensis BioF (PDP 4IW7) may present a similar picture, but the distribution of basic residues is obscured because the protein is much more basic (pI 8.6) than the other BioF proteins of known structure (E. coli, pI 6.6; M. smegmatis pI 5.7). When modeled on E. coli BioF, the B. subtilis protein seems to lack the basic patch. Therefore, it may be possible to distinguish BioF proteins that accept pimeloyl-ACP (e.g. E. coli and M. smegmatis) from those specific for pimeloyl-CoA by the presence or absence of a positive patch. However, no crystal structure of B. subtilis BioF is available and given that the protein expressed in E. coli had to be refolded from inclusion bodies (Manandhar and Cronan 2018), this BioF seems a poor candidate for crystallization. A BioF that seems certain to be specific for pimeloyl-CoA is that of Desulfosporosinus orientis because it is the C-terminal domain of a BioW-BioF fusion protein. Both domains have complementation activity in E. coli (Manandhar and Cronan 2018). A BioF recently reported to be specific for pimeloyl-CoA is that of Ralstonia eutropha H16 (recently renamed Cupriavidus necator), a gram negative ß-proteobacterium able to grow on hydrogen and carbon dioxide as sole carbon and energy source and thus valuable in industry. The biotin auxotrophy of a R. eutropha ∆bioF mutant could be complemented by expression of the cognate BioF or that of B. subtilis, but not by E. coli BioF. Since pimeloyl-CoA is a rather poor E. coli BioF substrate (Hu and Cronan 2020), the pimeloyl-CoA levels in R. eutropha may not be sufficient for activity. These results argue that pimeloyl-CoA not pimeloyl-ACP is the R. eutropha BioF acyl thioester substrate (Eggers et al. 2020).

Note that M. tuberculosis encodes two putative bioF genes in the genome named bioF and bioF2 (Salaemae et al. 2011). The bioF gene is encoded in an operon with bioA and bioD and thus was the leading candidate for a role in mycobacterial biotin biosynthesis. This is the case, the bioF gene complements the M. smegmatis ∆bioF strain whereas bioF2 does not (Fan et al. 2015). Furthermore, M. tuberculosis KAPA synthase was shown to have a significant role in survival and virulence (Sassetti, Boyd and Rubin 2003). This enzyme can use both L- and D-forms of alanine as an amino donor. This is not the case in other bacterial BioF proteins where the D-isomer is an inhibitor (Bhor et al. 2006). The D-form of DAPA derived from D-alanine is an inhibitor of BioA activity rather than a substrate.

BioA, 7,8-diaminononanoate synthase (DAPA synthase)

The next reaction in assembly of the biotin rings is transamination of KAPA to generate DAPA (7,8-diaminopelargonic acid) formally called 7,8-diaminononanoate (DAN). This reaction is catalyzed by DAPA (7,8-diaminopelargonic acid) synthase or DAN synthase (Fig. 1A), which is encoded by the bioA gene and active as a homodimer. DAPA synthase (EC. 2.6.1.62) belongs to aminotransferase subclass III of enzymes that use pyridoxal phosphate as a cofactor. SAM is generally the amino donor in the reaction although L-lysine is reported to serve this role in B. subtilis (Van Arsdell et al. 2005). Numerous crystal structures of DAPA synthase with different ligands and inhibitors have been reported. Each subunit contains an N-terminal domain that is important for dimerization and a central domain that binds pyridoxal phosphate. Deletion of the M. tuberculosis bioA gene demonstrated that the enzyme is required for successful infection of mice (Woong Park et al. 2011). The mechanism and structure of the M. tuberculosis BioA have been thoroughly studied and has led to identification of several DAPA synthase inhibitors and thus this enzyme is considered to be a promising target drug for tuberculosis therapy (Bockman, Mishra and Aldrich 2020).

BioU, discovery of the first new biotin ring biosynthesis enzyme in many years

Sakaki and coworkers (Sakaki et al. 2020) noted that the genomes of two halophilic archaea, Haloferax meditteranei and Natronoccus occlitus, encode biotin biosynthetic operons containing a novel gene between bioB and bioF. They additionally noted that these archaeal genomes lacked bioA genes and hypothesized that the protein encoded by these novel genes somehow functionally replaced BioA. However, the proteins lacked a pyridoxal-phosphate binding motif and instead contained a canonical NAD(P)H binding motif suggesting a reductive activity. Similar genes were found in cyanobacteria which provided a well-developed genetic system to test the function of this gene they called bioU. Indeed, deletion of the bioU of Synechocystis sp, PCC608 resulted in biotin auxotrophy. Introduction of bioA genes from E. coli or from other cyanobacteria complemented the ∆bioU mutant showing that BioU somehow replaced BioA function. Moreover, expression of the cyanobacterial and archaeal BioU proteins complemented an E. coli ∆bioA strain.

The first hypothesis of Sakaki and coworkers was that BioU catalyzed a reductive amination of KAPA to DAPA. They proceeded to incubate BioU with NADH, KAPA and a variety of candidate amino donors. They observed a brief burst of NADH oxidation and loss of KAPA but could not isolate DAPA. Control experiments performed in the absence of any added amino donor gave the same result (Sakaki et al. 2020). This raised the possibility that BioU itself might be the amino donor. Mass spectral analyses confirmed the covalent attachment of dehydrated and reduced KAPA. Peptide analysis and a crystal structure identified the site of attachment as Lys124 (Fig. 1B). Replacement of Lys124 with alanine or arginine resulted in an inability to form the BioU-KAPA conjugate.

How could the BioU-KAPA conjugate replace BioA? These workers reasoned that perhaps BioU could also carboxylate the conjugate to form BioU-DAPA carbamate which would be released from the enzyme by NAD⁺ oxidation and subsequent hydrolysis thereby providing the intermediate formed in the first half reaction of the next enzyme DTB synthase (BioD). BioD would then perform its second half reaction on the BioU-donated intermediate to give DTB. This was the case. Coupled assays containing BioU-DAPA-carbamate, NAD⁺ and E. coli BioD synthesized DTB in a BioD-dependent manner.

Moreover, BioU expression failed to complement an E. coli ∆bioD mutant strain for growth in the absence of biotin. This was expected since BioU cannot catalyze the second BioD reaction, formation of the uriedo ring (Sakaki et al. 2020). The oxidative release of DAPA-carbamate from BioU should convert the Lys124 side chain to a semialdehyde and inactivate the enzyme (Fig. 1B). The ‘used’ enzyme was inactive and the Lys 124-derived 2-amino 6-semialdehyde was demonstrated by peptide analysis. Hence, at least in vitro, BioU is a suicide enzyme and in the strict sense is not an enzyme since its reaction is stoichiometric rather than catalytic. The remaining question is whether or not the Lys 124-derived 2-amino 6-semialdehyde can be restored to a lysine sidechain. Enzymes are known that convert a semialdehyde to an amine (e.g. E. coli 4-aminobutyrate aminotransferase). Glutamate could be the amino donor with production of 2-ketoglutarate. However, in this case, the semialdehyde lies well within the BioU structure (Sakaki et al. 2020) and thus access to the aldehyde moiety seems likely to be problematical.

BioD, dethiobiotin synthase

In contrast to the preceding enzymes (with the exception of BioU), BioD (DTB synthase) catalyzes an unusually interesting step. This is the mechanistically unusual reaction of ATP (or CTP) dependent insertion of CO₂ between the N7 and N8 nitrogen atoms of DAPA to form the ureido moiety of biotin. The detailed mechanism of the enzyme has not received recent attention and we refer the reader to a prior review (Lin and Cronan 2011). The three-dimensional structures of DTB synthase from several bacteria in complexes with various ligands have been reported. Mycobacterium tuberculosis DTB synthase has been a favored enzyme due to attempts to find therapeutic inhibitors (Bockman, Mishra and Aldrich 2020). DTB synthase appears in an active homodimer in solution. It contains a classical P-loop motif (Gly-X-X-Gly-X-Gly-Lys-Thr/Ser), which is responsible for binding of the nucleotide triphosphate. Furthermore, two alternative binding mechanisms in M. tuberculosis DTB synthase have been identified a high affinity interaction by CTP and low affinity sites bound by other nucleotide triphosphates (Salaemae et al. 2015, Thompson et al. 2018). The bacterial DTB synthases of known crystal structure differ markedly in their carboxyl terminal regions (Porebski et al. 2012).

BioB, biotin synthase

The last reaction of the biotin biosynthetic pathway is catalyzed by biotin synthase (EC. 2.8.1.6), encoded by the bioB gene. BioB inserts a sulfur atom into dethiobiotin to form the tetrahydrothiophene ring of the biotin molecule. Biotin synthase belongs to the radical SAM superfamily and contains the signature Cys-X-X-X-Cys-X-X-Cys motif sequence required for [4Fe-4S] cluster binding. The single crystal structure determination of biotin synthase in complex with dethiobiotin and SAM shows a homodimer (Berkovitch et al. 2004). The active E. coli biotin synthase contains two iron-sulfur clusters: the air-sensitive [4Fe-4S]²⁺ cluster and the air-stable [2Fe-2S] cluster. Both clusters are essential for activity. Biotin synthase folds into a closed triosephosphate isomerase type (α/β)₈ barrel flanked with two helices at the N-terminus and an unstructured C-terminal region. The two clusters are at the opposite ends of the barrel, the catalytic radical SAM cluster at the open end of the barrel and the [2Fe-2S] donor cluster at the closed end. Although, the detailed molecular mechanism of biotin synthase cannot be considered as fully established, the mechanism consists of three main steps: the SAM-mediated hydrogen abstraction from the dethiobiotin, the derivation of a sulfur atom from the [2Fe-2S] cluster and the complex regeneration of the active enzyme. Furthermore, it has been noted that the regeneration of biotin synthase requires the HscA chaperone protein which commits the protein to degradation (Reyda, Fugate and Jarrett 2009). Nonetheless, biotin synthase has a short half-life but the requirement for biotin biosynthesis is very modest and the continuous expression of biotin synthase compensates for degradation.

Although BioB was reported to be a pyridoxal phosphate-dependent cysteine desulfurase which obtained the biotin sulfur atom directly from cysteine (Ollagnier-de-Choudens, Mulliez and Fontecave 2002a, Ollagnier-de-Choudens et al. 2002b), subsequent work showed that synthesis of biotin from DTB does not require pyridoxal phosphate (Abdel-Hamid and Cronan 2007). In 1999, it was reported that a BioB [Fe-2S} cluster was the likely source of the sulfur atom (Gibson, Pelletier and Turner 1999) and this was confirmed by studies in which the BioB [2Fe-2S] cluster was labeled with a sulfur isotope or with selenium and shown to be the donor of the sulfur atom (Bui et al. 1998), Tse Sum Bui 2006).

These data lead to predictions that the active BioB was consumed in the reaction and hence BioB is a suicide enzyme (as postulated for BioU). Although this was long debated (Jarrett 2005), E. coli BioB has been shown to be very modestly catalytic in vivo (Choi-Rhee and Cronan 2005) and in vitro (Farrar et al. 2010). It should be noted that [2Fe-2S] clusters are very rare in E. coli (<five enzymes) and all except that of BioB are involved in electron transfer. Assembly of [2Fe-2S] clusters has been neglected relative to the much more abundant [4Fe-4S] cluster proteins.

Similarities between BioB and LipA, the enzyme that inserts the sulfur atoms of lipoic acid

Very recently (Jin et al. 2020) reported two proteins annotated as BioBs that cooperate to form a functional LipA enzyme. This discovery was made in the archaeon Thermococcus kodakarensis, a hyperthermophile. Similar proteins are found in many other thermophilic archaea whereas mesophilic archaea tend to have classical LipAs. A working model is that each protein has a [Fe-S] cluster that donates one of the two lipoate sulfur atoms. This report strengthens the relationship of BioBs to the classical LipA proteins of lipoic acid synthesis. The classical LipAs are α/β barrel radical SAM enzymes that have two [4Fe-4S] clusters, a radical SAM cluster and an auxiliary cluster. Two sulfur atoms from the auxiliary cluster replace the chemically inert hydrogen atoms on C6 and C8 of a protein-bound octanoyl moiety to produce lipoic acid (Cronan 2016). Hence, LipA and BioB both make two C-S bonds. Like BioB, LipA has been considered a suicide enzyme (Booker, Cicchillo and Grove 2007), but recent work has demonstrated that given certain cluster repair enzymes, the protein becomes catalytic (McCarthy and Booker 2018). It would be interesting to test this repair process with the archaeal two-component lipoate ligases.

Biotechnological approaches to biotin production

Commercial biotin is made by chemical synthesis where the Hoffman LaRoche synthetic route seems to dominate. Biotin has three chiral centers and thus many clever syntheses have been developed to provide the correct enantiomer. Attempts to avoid chemical synthesis and produce biotin by biosynthesis have been performed in multiple laboratories using a variety of bacterial species. These have invariably shown BioB to be a bottleneck; much more DTB is produced than biotin (Acevedo-Rocha et al. 2019). Efforts in E. coli to overcome the bottleneck by overproduction of BioB failed due to toxicity (Bali et al. 2020). Recently this toxicity was utilized to select mutant strains that are less sensitive to BioB overproduction and also convert DTB to biotin more rapidly (a 2.2-fold increase). Three overproduction-resistant mutants were found and all had missense mutations in IscR, a deoxyribonucleic acid (DNA) binding protein that contains a [2Fe-2S] cluster. Both the [2Fe-2S] cluster containing protein and the protein lacking the cluster have regulatory roles. The relief of toxicity mutations were proposed to prevent formation of the IscR cluster resulting in derepression of the Isc operon and activation of the Suf operon both of which would increase iron-sulfur cluster biogenesis (Bali et al. 2020). These mutations also gave increased synthesis of lipoic acid and thiamine. It is not clear if the BioB effects are exerted on the [2Fe-2S] cluster, the [4Fe-4S] or both. The iron-sulfur clusters of the lipoic acid synthesis and thiamine synthesis radical SAM enzymes are all [4Fe-4S] clusters. Delineating the effects of mutating a global regulator such as IscR will not be straightforward because IscR regulates more than 40 genes, some of which have regulatory roles. This work probably explains the toxicity of BioB overproduction in wild type cells as due to disruption of iron-sulfur cluster biogenesis. Although the IscR findings are a step in the right direction, biotin production on an industrial scale probably will require at least an order of magnitude increase in production.

Regulation of biotin biosynthesis and protein biotinoylation

There are only two routes to acquire biotin; scavenge from the environment or de novo synthesis based on biotin availability plus the presence of genes which encode the necessary enzymes. The bio genes are generally (but not invariably) organized into operons and thus cotranscribed. This allows the proteins to be produced in the proper stoichiometry. Although bacteria and yeast require only very modest amounts of biotin for growth, the enzymes tend to be remarkably poor catalysts (especially the enzymes late in the pathway) and thus must be synthesized in appreciable quantities. Biotin synthesis is also metabolically expensive. a major cost is SAM consumption by BioC, BioA and BioB. Hence, as a rule, organisms shut down biotin synthesis when an exogenous source of the cofactor is available.

Transcriptional regulation by the bifunctional type II biotin protein ligases (BirAs)

The first and still paradigmatic biotin synthesis regulatory system is that of E. coli (Fig. 4). Parallel investigations of biotin excretion and resistance to biotin analogs resulted in the same gene being identified called bioR and birA with the latter name having become standard (Chapman-Smith and Cronan 1999, Cronan 2008). BirA, named for biotin retention, has two functions. BirA is both a transcriptional repressor of the biotin biosynthetic operon and a BPL, the enzyme required to attach biotin to its cognate metabolic enzymes (Barker and Campbell 1981). BirA catalyzes protein biotinoylation through the biotinoyl-5’-AMP (biotinoyl adenylate) intermediate (Fig. 4). The BirA:biotinoyl-5’-AMP complex has two functions. When the biotin demand is low because the cognate metabolic enzymes are fully biotinoylated, the BirA:biotinoyl-5’ AMP complex accumulates, dimerizes and represses bio operon transcription by binding to the operon promoter region to inhibit the expression of biotin responsive genes (Beckett 2007, Cronan 2008). This also occurs when environmental biotin is available. On the other hand, in the presence of the unbiotinoylated forms of biotin-dependent enzymes the biotinoyl-5’-AMP is consumed in biotinoylation. The protein then reverts to a monomeric form and repression of the bio operon enzymes is derepressed, resulting in synthesis of biotin to replace that consumed in biotinoylation (Cronan 2008). Hence, this simple regulatory system provides remarkably sophisticated regulation. It recognizes that protein-bound biotin is the ‘bottom line’ and has developed the ability to respond to both the unbiotinoylated forms of biotin-dependent enzymes and biotin concentration (Cronan 2008). Regulation by BirA proteins of B. subtilis and S. aureus, also respond to both exogenous biotin concentration (Henke and Cronan 2014, Henke and Cronan 2016).

Figure 4.

Protein biotinylation catalyzed by BPLs. Biotin is activated by reaction with ATP to generate the biotinyl-5′-AMP (biotinoyl-adenylate) intermediate. The enzyme bound biotinyl-5′-AMP intermediate is then subjected to nucleophilic attack by the conserved lysine residue of the biotin-accepting domain (represented by BCCP) of the cognate enzymes to form the active biotinylated enzyme (Chapman-Smith and Cronan 1999).

However, BPLs are not all regulatory proteins, although they all share the same catalytic domain. Domain architectures of BPLs are placed into three classes (Beckett 2018) (Fig. 5). BPL class I contains only the core catalytic domain which is essential for protein biotinoylation. In class II BPLs, a DNA binding domain (DBD) containing a winged helix-turn-helix motif comprises the N-terminus of the protein to give a bifunctional BPL, like BirA. These two classes are present in microorganisms. BPL class III enzymes which have very long N-terminal extensions appended to the catalytic domain are present in eukaryotes and mammals including humans. The extended N-terminal region of the human BPL was shown to increase specificity for the targets to be biotinoylated (Ingaramo and Beckett 2012). It is interesting that the N-terminal extensions of the human and yeast BPLs show no significant sequence conservation.

Figure 5.

(A) The domain architectures of BPLs throughout biology. Conserved catalytic domain, the DBD and, extended DBD are arranged and classified into three classes. Class I enzymes are found in bacteria and plants whereas class II enzymes are found only in bacteria. Class III enzymes are present in fungi and mammals including humans. (B) The model represents the transcriptional regulation mechanisms and biotinylation by class I and class II BPL.

The class II BPLs of E. coli (Barker and Campbell 1981) and B. subtilis (Henke and Cronan 2014) are essential genes, and inhibitor studies indicate that this is also true for S. aureus (Paparella et al. 2018). Both B. subtilis and S. aureus BirAs follow the E. coli paradigm (Henke and Cronan 2014, Henke and Cronan 2016, Wang and Beckett 2017), the main difference is that biotin transport genes are also under BirA control. There are conflicting data on BirA control of the E. coli biotin transporter (see below).

Interestingly, BirA functions somewhat differently in various bacteria. Only biotin responsive genes are regulated by E. coli BirA whereas the B. subtilis and S. aureus BirA also regulate two genes located downstream of the bioY biotin transporter; yhfS and yhfT annotated as encoding an acetyl-CoA acetyltransferase and a fatty acyl-CoA ligase, respectively (Satiaputra et al. 2018). Moreover, even in bacteria unable to synthesize biotin such as the zoonotic pathogen Streptococcus suis, BirA represses biotin uptake at the transcription level by binding to the bioY promoter region (Ye et al. 2016).

Although most organisms including humans have a single BPL-encoding gene, a few bacteria have two such genes. The first example was the intracellular pathogenic bacterium, F. novicida (studied as a less pathogenic surrogate of the human pathogen F. tularensis) which encodes two BPLs: BirA which is similar to that of E. coli BPL and thus has a regulatory role and BplA which lacks the DBD and hence lacks regulation ability (Feng et al. 2015). BplA is essential for pathogenesis whereas BirA regulates biotin biosynthesis (Feng et al. 2015). This finding demonstrates an alternative adaptation mechanism linked to a biotin requirement for virulence. Another bacterium with two BPL enzymes is Lactococcus lactis which lacks a biotin synthetic pathway but has a unique biotin scavenging pathway with redundant genes. This bacterium has two BPL enzymes (BirA1 and BirA2) both of which catalyze protein biotinoylation and two biotin-specific S unit BioY class II energy-coupling factor (ECF) transporters (see below). The regulatory targets are two BioY biotin transporters (BioY1 and BioY2) (Zhang et al. 2016). BirA1 represses the expression of BioY1 whereas BioY2 is constitutively expressed. This might be a mechanism for bacterial survival in response to the fluctuation of biotin levels in the environment or gastrointestinal tract. Structural and biochemical studies suggests that BPLs might be good targets for a novel antibacterial agent production and compounds active as inhibitors against M. tuberculosis (Feng, Zhang and Cronan 2016, Tiwari et al. 2018) and S. aureus BPLs (Paparella et al. 2018) have been reported.

Escherichia coli BirA remains the best studied class II BPL (Ingaramo and Beckett 2011, Beckett 2018). BirA super repressor mutants (strains which repress biotin synthesis at low biotin concentrations) have been shown to owe this property to increased interactions between the two BirA-biotin-AMP monomers of the dimers (He et al. 2018) as was originally proposed (Chakravartty and Cronan 2012). The dynamics and allosteric behaviors of E. coli BirA have demonstrated interactions between the wing of the N-terminal winged helix-turn-helix, which structures the biotin binding site of the protein and explains why loss of the N-terminal domain results in an inactive enzyme (Chakravartty and Cronan 2013). Other mutagenesis studies have demonstrated interactions between the C-terminal domain and the central catalytic domain that give insight into the importance of the C-terminal domain (Laine et al. 2008, Ingaramo and Beckett 2011). Some attributes may be specific to the E. coli protein in that deletion of the N-terminal winged helix-turn-helix domain of B. subtilis BirA fails to inactivate the catalytic domain whereas the E. coli and S. aureus BirAs (Henke and Cronan 2014, Henke and Cronan 2016) are inactivated by such deletions. Biotinoyl-adenylate analogs have recently been shown to inhibit the M. tuberculosis BPL resulting in death of the bacterium (Ikeda et al. 2013). Moreover, down-regulated expression of this BPL eliminated the pathogen efficiently from mice during acute and chronic infection of mice with M. tuberculosis (Tiwari et al. 2018).

It should be noted that E. coli BirA is key to several technological innovations. First, proteins of interest fused to either a natural biotin domain or to small peptides can be biotinoylated either in vivo or in vitro and purified and tracked using either avidin/streptavidin reagents or anti-biotin antibodies (Cronan 1990, Schatz 1993). This is superior to chemical biotinoylation in vitro because it results in a single biotin attached at a known location. The tag can be placed at either terminus or within a surface exposed loop. The tagged proteins can then be stably bound to streptavidin in a tetrameric display as in the major histocompatibility complex tetramers produced by NIH (Yang et al. 2004).

Another innovation uses a BirA protein having a mutation in the binding site for biotin and biotinoyl-5’-AMP. The mutation causes release of the normally tightly bound biotinoyl-5’-AMP which acts as a transient (due to hydrolysis) chemical biotinoylation reagent. The mutant BirA is fused to a protein of interest that when expressed in cells can be induced to biotinoylate interacting and proximate proteins. The biotinoylated proteins are then recovered by binding to immobilized avidin/streptavidin and identified by mass spectroscopy. This approach called BioID is much utilized in the cell biology community for identification of candidate protein-protein interactions in living cells (Gingras, Abe and Raught 2019; Sears, May and Roux 2019).

Regulation by BioR and BioQ

Some bacteria such as C. glutamicum, M. smegmatis and A. tumefaciens do not contain BirA or any other class II BPL homolog and use a class I BPL for protein biotinoylation (Fig. 5). In these bacteria, alternative DNA binding proteins called BioQ (Brune et al. 2012) and BioR (Feng, Zhang and Cronan 2013) regulate biotin synthesis and transport. BioQ belongs to the TetR family of transcription factors whereas BioR is a member of the GntR family. Both proteins act as repressors to decrease transcription of the biotin responsive genes in response to cellular biotin levels. Moreover, analysis of their promoters suggests that each protein could regulate its own transcription. Both BioQ and BioR are have species specificity. BioQ is found in many mycobacterial species such as M. abcessus, M. gilvum and M. smegmatis but not in M. tuberculosis and Mycobacterium bovis. The absence of BioQ in M. tuberculosis might be of advantage in survival and pathogenesis or an alternative strategy to control the production of intracellular biotin. Conversely, BioR is present in α-proteobacteria such as A. tumefaciens, Brucella melitensis, Mesorhizobium loti and is also found in Paracoccus denitrificans.

Recently, the crystal structure of M. smegmatis BioQ was elucidated (Wei et al. 2018, Yan et al. 2018). Mycobacterium smegmatis BioQ requires acetylation of lysine 47 to activate binding between the protein and DNA operator sequence required to repress biotin responsive genes (Wei et al. 2018, Yan et al. 2018). This modification plays an important role in regulatory switching and is the first case of protein post-translational modification in the regulation of biotin biosynthesis (Wei et al. 2018). Another striking recent finding is that inactivation of the M. smegmatis pyruvate carboxylase (Pyc) blocked biotin synthesis in a BioQ-dependent manner (Lazar et al. 2017). This is attributed to pyruvate carboxylase-dependent synthesis of the unknown regulatory ligand of BioQ. Note that BioQ repression of bioB transcription is the most severe, the other biotin genes show considerably less repression. The BioR regulatory ligand remains unknown although it is suspected of being a biotin-related molecule. The E. coli and B. subtilis BirAs exert stronger (perhaps 5-fold) better repression than BioQ and BioR in their native settings.

Regulation of biotin synthesis in R. eutropha (a β-proteobacterium) presents a puzzle. Biotin synthesis is stimulated by overproduction of the biotin-acceptor protein AccB as seen in the BplII bacteria, E. coli (Cronan 1988) and B. subtilis (Henke and Cronan 2014). However, this bacterium has only a BplI ligase (Eggers et al. 2020) and lacks the other biotin synthesis regulators BioR and BioQ. AccB forms a complex with AccC that is required for acetyl-CoA carboxylase activity. Overproduction of AccC together with AccB results in loss of most of the stimulation given by overproduction of BioB alone (Eggers et al. 2020). This competition between biotinoylation and complex formation was first seen in E. coli (Abdel-Hamid and Cronan 2007) and argues that biotinoylation must precede complex formation.

Active transport of biotin in bacteria

The most abundant vitamin uptake systems in bacteria belong to the ECF transporter, subclass of ATP binding cassette (ABC) transporters. These systems use ATP as the energy source to transport solutes across cell membranes. ECF transporters, which share three common subunits, can be classified into two classes. The homo or hetero-dimeric ABC contains an ATPase (called the A unit), a transmembrane protein (the T unit), and a high affinity substrate binding protein (the S unit) (Rodionov, et al. 2009, Eitinger, et al. 2011, Erkens, et al. 2011). ECF transporter class I has all three of these basic components encoded in the same operon. Each S unit interacts with a dedicated A₂T or A₁A₂T complex whereas class II ECF transporters have a single A₁A₂T complex with multiple S units that are encoded separately from the other genes. Recently, ECF transporters have been thought to be an alternative antimicrobial target. This is due to the absence of these transporter type in humans and their involvement in survival, growth and virulence of the various pathogenic bacteria (Bousis, et al. 2019)

The Rhodobacter capsulatus biotin transporter (BioM₂NY) belongs to the class I of ECF transporter family and is the best characterized and paradigm system. In this system, BioM (A unit) is an ATPase with a typical Walker A, Q loop signature (LSGGQ) plus Walker B and H motifs (Eitinger, et al. 2011, Hohl, et al. 2012). The R. capsulatus BioN (T unit) contains 4–5 transmembrane helices with an additional C-terminal cytoplasmic and two amphipathic helices that resemble the glutamate-alanine-alanine (EAA) loop of classical ABC transporters. The biotin specific BioY (S unit), composed of six transmembrane proteins, is a central unit of this biotin transporter. The BioM ATPase plays a role in substrate affinity whereas the role of BioN stabilization of the BioM₂NY complex (Hebbeln, et al. 2007). Although the detailed mechanism of biotin transport via BioM₂NY remains an active field, a model has been proposed: ATP binds to induce a conformational change of the BioM dimer leading to biotin capture by the BioY S unit. Subsequent hydrolysis of ATP results in release of free biotin into the cytoplasm, the nucleotide-free complex then returns to the resting state (Finkenwirth, et al. 2015). Surprisingly, in the absence of any other ECF-transporter components, the S unit BioYs (denoted Solitary BioYs) mediate biotin uptake by an unknown mechanism. However, the transport flux is low and it may depend on oligomerization of membrane proteins (Finkenwirth, et al. 2010, Fisher, et al. 2012, Finkenwirth, Kirsch and Eitinger 2013). On the other hand, about one-third of biotin specific S BioY units are known or predicted to interact with the shared A₁A₂T complex of ECF transporter class II (Hebbeln, et al. 2007). At present, the biotin transport mechanism of class II of the ECF transporters has not been elucidated.

Biotin transport by facilitated diffusion

Recently, the long unknown E. coli biotin transporter has been identified as YigM (Ringlstetter 2010). Based on its genomic location and phenotype, the yigM gene is bioP, a transporter gene originally discovered by Pai (Pai 1973). Subsequent isolation of a strain called birB defective in biotin transport by the same selection used by Pai provided more accurate mapping (Campbell et al. 1980). Later, Eisenberg isolated bioP mutant strains that mapped identically to birB by resistance to a biotin analog (Eisenburg et al. 1975).

BioP, a 34 kDa protein, is predicted to contain 10 transmembrane helices and belong to the carboxylate/amino acid/amine family of secondary active transporters (Ringlstetter 2010). BioP mediates biotin uptake with high affinity with what seems a facilitated diffusion mechanism. However, the molecular mechanism of this transporter is not established (Ringlstetter 2010). Recently, an E. coli strain having deletions of both bioP and the biotin synthetic gene bioH has been constructed and is a useful tool for validating and investigating novel biotin transporters (Finkenwirth, Kirsch and Eitinger 2013, Finkenwirth, Kirsch and Eitinger 2014). Transcription of bioP may (Ringlstetter 2010) or may not be (Pai 1973) under BirA control.

Biotin synthesis in the archaea

There is scant information on biotin synthesis in the archaea (note, however, the contributions in the BioH and BioU sections above). Bioinformatic analyses by Rodionov and coworkers (Rodionov, Mironov and Gelfand 2002) first reported that although most archaea lack recognizable biotin synthetic genes, all have BPLs and many have bioY-like biotin transporters. Chow and coworkers (Chow et al. 2018) have recently provided an update that considers archaeal taxonomy. Of the 94 archaeal genomes in the MicrobesOnline database, 47 and 41 have plausible bioB and bioD genes, respectively, whereas less than half that number have putative bioA and genes.

The most complete sets of biotin synthetic genes are those of Methanocaldococcus (formerly Methanococcus) jannaschii which has a putative bioW, bioF, bioD and bioA operon with bioB being encoded elsewhere in the genome and the Thaumarchaeon, Nitrososphaera gargensis. Nitrososphaera gargensis encodes bioABD genes that functionally replace the cognate E. coli genes and a promiscuous E. coli ∆bioH-complementing esterase EstN1 (Chow et al. 2018). The N. gargensis bioF gene failed to complement an E. coli bioF strain. Since archaea lack ACP, it seems likely that the N. gargensis BioF cannot use pimeloyl-ACP as a substrate. Note that both M. jannaschii and N. gargensis are biotin prototrophs (Jones et al. 1983, Chow et al. 2018), and each has a strong BPL candidate. It is an open question of how the pimeloyl thioester is made in these archaea although they may differ based on the presence of BioW in M. jannaschii and EstN1 in N. gargensis.

Note that archaea are often grown in media with a complex biotin-containing supplement (e.g. yeast extract) that precludes detection of biotin auxotrophy. A straightforward approach to this shortcoming would be to add avidin or streptavidin to the medium to bind and sequester the biotin. Avidin and streptavidin specifically bind biotin very tightly (Kd of ∼10 ^-14 M). Both proteins are very resistant to heat, denaturants, extreme pH values and proteolysis (Green 1975; Dundas, Demonte and Park 2013) and thus should be stable in growth media. The concentration of avidin or streptavidin needed to sequester all biotin from a medium could be determined by the concentration required to block growth of E. coli or B. subtilis biotin auxotrophs in the presence of the chemically undefined components of the archaeal medium.

Other than BioU the BPLs are the only archaeal biotin metabolism enzymes to be purified and studied in detail. Crystal structures of the Pyrococcus horikoshii (Bagautdinov et al. 2005; Bagautdinov et al. 2008) including substrate complexes are available. The mechanism of the ligase reaction has also been investigated (Daniels and Beckett 2010). An unpublished crystal structure of the M. jannaschiiBpl is also available (PDB 2EJ9).

In Methanosarcina barkeri, there is an interesting genomic juxtaposition of the bpl gene with the genes encoding the two subunits of pyruvate carboxylase, PycA and PycB, a biotin-requiring enzyme that carboxylates pyruvate to oxaloacetate (Mukhopadhyay et al. 2001). The bpl and pycC genes are cotranscribed together with a downstream gene in a pycA pycB bpl operon (Lopez Munoz, Schonheit and Metcalf 2015). Genome sequencing of a mutant strain that acquired the ability to grow on pyruvate showed a frameshift in the bpl gene resulting in premature termination of the protein. When this bpl mutation was constructed in a wild type strain, the pyruvate carboxylase was no longer active. However, the strain grew as well as the wild type on a standard medium. Growth of the bpl mutant was explained by an alternate pathway of oxalacetate synthesis from pyruvate via phosphoenolpyruvate (Lopez Munoz, Schonheit and Metcalf 2015). Cotranscription with pycA and pycB argues that the major role of Bpl is to biotinoylate PycB. Indeed, the bpl gene is sometimes called pycC. Avidin blotting of SDS-gel fractionated M. barkeri extracts shows several biotinoylated bands other than PycB (Mukhopadhyay et al. 2001). Since these bands are smaller than PycB, they could be PycB degradation products. However, addition of biotin to the medium increased the biotinoylated PycB level but not the levels of the more rapidly migrating bands. If there are biotinoylated enzymes in M. barkeri other than pyruvate carboxylase, they are not required for growth in normal laboratory conditions.

Biotin synthesis in yeast

Although the budding yeast Saccharomyces cerevisiae is not a bacterium, at least some of the biotin synthetic genes are thought to be of bacterial origin (Hall and Dietrich 2007). Hence, this yeast deserves consideration in this review. Biotin synthesis in S. cerevisiae has long been a puzzle with two perplexing aspects. Most strains are biotin auxotrophs and a few are prototrophs whereas other strains are able to grow weakly without biotin supplementation, but grow much faster with biotin supplementation (Hall and Dietrich 2007, Perli, et al. 2020a, Wu, Ito and Shimoi 2005). Auxotrophy is exemplified by strain S288C (the first yeast genome sequenced), which is a confirmed biotin auxotroph completely unable to grow without biotin (Wu, Ito and Shimoi 2005, Hall and Dietrich 2007). This auxotrophy is due to lack of the genes (Bio1 and Bio6) recently recognized as required for synthesis and utilization of pimelate. A surprise is that KAPA, DAPA or DTB supplementation can replace the biotin requirement of strain S288C indicating that the last three genes of the biotin pathway are functional (Phalip, Lemoine and Jeltsch 1999, Perli et al. 2020b, Wu, Ito and Shimoi 2005). The genes of the late part of the pathway encode proteins that are homologs of BioA (yeast Bio3, 47% identical), BioD (yeast Bio4, 29% identical) and BioB (yeast Bio2, 49% identical including the [2Fe-2S] cluster ligands).

The riddle is why these genes have been preserved intact rather than drifting toward becoming nonfunctional pseudogenes given that the overall pathway is inoperative. Strain S288C and its progenitor strains have been grown with biotin for more than 70 years in many laboratories and are the source of the bulk of the strains in the Saccharomyces Genetic Stock Center. KAPA, DAPA and DTB are essentially absent in natural environments and hence it seems most unlikely that these compounds provide a rationale for maintenance of the bio2, bio3 and bio4 genes. Moreover, there are two very degenerate pseudogenes for each of bio1 and bio6 in strain S288C (Hall and Dietrich 2007). Why were the late genes spared?

Recent reports have clarified the roles of two genes early in the pathway, bio1 and bio6. Since these genes are immediately adjacent on the chromosome, their functions were often analyzed together. For example, in crosses between strains S288C and A364a, a biotin prototroph closely related to S288C, the tetrads segregated 2:2 indicating a single gene (or in this case a small chromosomal segment) (Hall and Dietrich 2007). One of the biotin genes known to be absent in strain S288C was bio6. However, introduction of bio6 on a plasmid failed to allow biotin-independent growth of S288C suggesting that a closely linked gene also required for biotin synthesis was present. This led to discovery of bio1. Upon introduction of bio1 from A364A, biotin-independent cousin of strain S288C, resulted in growth but only when expressed together with the A364A bio6 (Hall and Dietrich 2007).

More recent work used laboratory evolution to convert an industrial biotechnology strain CEN.PK113–7d (a mosaic of laboratory and industrial strains) that grew very poorly in the absence of biotin supplementation into strains that were fully biotin prototrophic (∼36-fold increase in growth rate). The resulting strains showed as many as 43 copies of the bio1-bio6 region scattered about the genome (Bracher et al. 2017). When overexpressed from a multicopy plasmid, bio1 alone allowed robust growth of the parental CEN.PK113–7d strain in the absence of biotin, indicating that the bio6 gene of this strain is functional and hence the multiple bio1 copies are responsible for prototrophy. However, the genomic complexity of this strain is not readily transferred to other strains which limits its utility.

A more facile approach to a biotin prototrophic CEN.PK113–7d strain used chromosomal expression of a foreign bio1 ortholog gene in single copy (Wronska et al. 2020). This ‘super’ bio1 was derived from a prototrophic Saccharomycotina yeast chosen for rapid growth in the absence of biotin. Similar expression levels of the native bio1 gene failed to give rapid growth. Hence the native bio1 gene seems to encode a weakly active protein.

Functions of Bio6 and Bio1

Establishment of the function of Bio6 in the biotin Bio1synthetic pathway necessarily defined the function of Bio1. At first the Bio6 sequence confused the issue of its place in the pathway because it has significant homology to Bio3, a gene that complements a bioA mutant of E. coli (Wu, Ito and Shimoi 2005). However, upon construction of bio3∆ and bio6∆ deletion strains it was found that in the bio6∆ deletion strain biotin could be replaced by KAPA, DAPA and DTB whereas only DAPA and DTB could rescue the bio3∆ deletion strain. These data showed that Bio6 and Bio3 have different functions and argued that Bio6 might have KAPA synthase activity (Wu, Ito and Shimoi 2005). This is reasonable since the overall structure of the E. coli BioF KAPA synthase is very similar to the structure of E. coli BioA DAPA synthase (Schneider et al. 2012). This, together with their common pyridoxal phosphate coenzyme and weak amino acid sequence homology, suggests that the two enzymes might be evolutionarily related and possibly were derived from a common ancestor. Although the assignment of Bio6 as a KAPA synthase is not yet supported by enzymatic analysis or the ability to complement a ∆bioF E. coli strain, it seems likely to be a correct assignment. If so, Bio1 must act earlier in the pathway: at the level of pimelate synthesis (Hall and Dietrich 2007).

Indeed, Bio1 is often called pimeloyl-CoA synthase, although there is no biochemical evidence supporting this assignment (note, that pimelate fails to rescue biotin auxotrophy in yeast; (Ohsugi and Imanishi 1985, PerWu, Ito and Shimoi 2005, li et al. 2020). Bio1 proteins contain the motifs of 2-oxoglutarate-iron(II)-dependent dioxygenases (Hall and Dietrich 2007, Wronska et al. 2020). These oxidoreductases catalyze reactions that incorporate oxygen atoms from molecular oxygen (O₂) into the substrates while decarboxylative oxidation of 2-oxoglutarate gives succinate and carbon dioxide. Consistent with these annotations S. cerevisiae has been shown to be unable to synthesize biotin when grown anaerobically (Wronska et al. 2020).

The 2-oxoglutarate-iron(II)-dependent dioxygenases catalyze a huge diversity of reactions rivaling those catalyzed by heme-containing enzymes such as the cytochrome P450s (Gao et al. 2018, Herr and Hausinger 2018, Islam et al. 2018). These reactions include hydroxylations and N-demethylations which proceed via hydroxylation (the dominant reaction of these enzymes) but also ring formation, rearrangements, desaturations and halogenation (Gao et al. 2018, Herr and Hausinger 2018, Islam et al. 2018). Given the remarkably diverse range of oxidative reactions catalyzed by these enzymes, it is difficult to posit how the Bio1 reaction might proceed. The most probable substrate (aside from 2-oxoglutarate and oxygen) seems likely to be long chain acyl-CoAs, the products of the yeast fatty acid synthase (Schweizer and Hofmann 2004). These would provide the carbon atoms and a thioester for KAPA synthesis. If so, Bio1 would not be a pimeloyl-CoA synthetase analogous to B. subtilis BioW, but rather act in general terms like the B. subtilis BioI cytochrome P450 which cleaves long chain acyl-ACPs to produce pimeloyl-ACP.

In one possible scenario, yeast Bio1 would oxidize carbons 7 and 8 of the long chain acyl-CoA to give vicinal hydroxyl groups which would be oxidatively cleaved to form two aldehydes. Further oxidation of the aldehydes would produce pimeloyl-CoA plus a monocarboxylic acid originating from the methyl end of the long chain acyl-CoA as proposed for the B. subtilis BioI cytochrome P450 (Cryle 2010). The ability of a single enzyme to sequentially oxidize two neighboring carbon atoms has been reported for both CtrZ and CrtW of the astaxanthin synthesis pathway(Tian and DellaPenna 2001). Another possibility is desaturation between carbons 7 and 8 followed by oxidation to the epoxide and collapse of the epoxide in a reaction similar to that of AsqJ of Aspergillus nidulans (Li et al. 2020).

It has been proposed that biotin biosynthesis in S. cerevisiae was almost completely lost and then reacquired by two rare genetic events. The postulated events are horizontal gene transfer from bacteria (hence the conserved sequences) and gene duplication followed by retooling one of the copies to perform a missing step (Hall and Dietrich 2007; Helliwell, Wheeler and Smith 2013). Horizontal gene transfer is consistent with the atypical clustering of biotin synthetic genes in yeast. Although there is very little clustering of genes of related function in S. cerevisiae (Hall and Dietrich 2007), bio3, bio4 and bio5 (a transporter found in some strains) are contiguous as are bio1 and bio6.

Transcription of the S. cerevisiae biotin synthetic and biotin transport genes responds to biotin concentration, although the degree of regulation is modest (Pirner and Stolz 2006). Cultures grown at low biotin concentrations have a six-fold greater level of bio2 transcription than cultures grown with high biotin concentrations whereas the bio3 and bio4 genes showed only a two- to three-fold variation with biotin concentration. The genome sequence of the strain used lacked bio1 and bio6 (Ralser et al. 2012) one of the parents of this strain was S288C, hence endogenous biotin synthesis was not a complicating factor. Transcriptional regulation was exerted by the Bpl1 ligase. Bpl1 mutants that retained some partial activity acted as though they were grown on low biotin concentrations even when grown at high biotin concentrations. Hence, a wild type Bpl1p activity is necessary for normal biotin sensing. The role of Bpl1 is not clear but the authors speculate that biotinoyl-AMP could be involved (Pirner and Stolz 2006).

Biotin synthesis in the fission yeast, Schizosaccharomyces pombe, has been little studied. This yeast is reported to require biotin for growth (McVeigh and Bracken 2018). The S. pombe bio2 gene complements an S. cerevisiae bio2 mutant. Thus S. pombe has a functional biotin synthase (Phalip, Lemoine and Jeltsch 1999). However, a search of PomB (asewmbase.org/gene/SPBC30D10.07c) produced no other biotin synthesis genes, only a Bpl1 ligase gene.

CONCLUSIONS

The synthesis of biotin is an important and fertile research area. Not only is there unexplained diversity and species specificity in the enzymes that remove the methyl group of pimeloyl-ACP methyl ester, we now have a new finding in the ring assembly steps, BioU, plus the BioB homologs that comprise a two component lipoyl synthase (a caveat to annotations). The diverse esterases that act on pimeloyl-ACP methyl ester might argue that bioC plus the genes of the late pathway entered a naive cell by horizontal gene transfer and that a previously existing gene encoding an α/β-hydrolase was mutated to hydrolyze pimeloyl-ACP methyl ester. This scenario has support from the observation that bacterial genomes encode many α/β-hydrolases and from the ability to convert an α/β-hydrolase of unknown physiological function to full BioH activity via a small number of amino acid substitutions (Flores et al. 2012). This scenario also explains the differences in substrate specificity of these esterases (Shapiro, Chakravartty and Cronan 2012). However, this scenario cannot accommodate the findings that esterase genes are often located in biotin operons immediately upstream of bioC and their sequences often overlap bioC (as well as the upstream gene). Such overlap indicates early integration into these operons. Given that a single scenario cannot explain these observations, it seemed possible that the esterases encoded within an operon might differ from those encoded by a freestanding gene such as E. coli BioH which is much more active than the other enzymes of the biotin pathway. Perhaps, relative to BioH, the ‘operon’ esterases could be less active, more poorly expressed or both to become attuned to the activities of the other biotin enzymes but this was not the case (Cao et al. 2017).

Protein biotinoylation is a rare posttranslational modification. Escherichia coli has only a single biotinoylated protein, a subunit of acetyl-CoA carboxylase, whereas humans have only four biotinoylated protein species. Some plants have five or six such proteins and some bacteria have three or four. Consideration of the lengthy biotin synthesis pathways that require multiple enzymes and considerable metabolic expenditure (up to 19 ATP equivalents per molecule) argues that nature might well have dispensed with biotin. Why have biotinoylated enzymes persisted in all three domains of life? The answer seems likely to be the essentially of the reactions catalyzed by biotinoylated enzymes. These reactions are found in the very innards of metabolism and cannot be readily bypassed. For example, acetyl-CoA carboxylase is required to make malonyl-CoA, the building block of fatty acids and several other important molecules including biotin. To replace acetyl-CoA carboxylase, another pathway to make malonyl-CoA would have to be waiting in the wings but what would be the driving force for development of this bypass pathway? The situation compounds in organisms having more than one essential biotin-containing enzyme. Some of this conundrum could be ameliorated by horizontal gene transfer, but again, a driving force for acquisition would be required. Given this situation, nature seems to be ‘stuck’ with biotin.

A second conundrum is that the known synthesis pathways for the biotin precursor pimelate require malonyl-CoA which is produced by acetyl-CoA carboxylase, a biotin-requiring enzyme. One possibility to avoid this ‘chicken or egg’ predicament would be the availability of a malonyl-CoA synthetase, a source of malonic acid and a malonate transporter. Such systems are known (An and Kim 1998; Milke and Marienhagen 2020) and the Rhizobium leguminosarum sv trifolii matABC system is the best studied. This system, which is composed of three proteins, malonyl-CoA decarboxylase (MatA), an ATP-dependent malonyl-CoA synthetase (MatB) and a malonate transporter (MatC). However, the purpose of this system is to convert malonate to acetyl-CoA to be used as a carbon source: malonyl-CoA is made and decarboxylated to acetyl-CoA. Since malonate is not abundant in the environment, this pathway seems an unlikely source of malonyl-CoA. Kirschning (Kirschning 2020) and Visser and Kellogg (Visser and Kellogg 1978) have argued that it is questionable that biotin was likely to be present and have catalytic roles in a prebiotic world. As far as biotin, other molecules more electrophilic than CO₂ likely existed (e.g. guanidine, cyanic acid, cyanamide and urea). Moreover, biotin-free CO₂ transfer is well established, for example the carbamoyl-phosphate synthetase reaction of pyrimidine biosynthesis. CoA is also considered as unlikely to be required in a prebiotic world because the business end is only a thiol group which are abundant in prebiotic conditions (e.g. hydrothermal vent effluents) (Kirschning 2020). However. others have argued that CoA is a product of the RNA world (Jadhav and Yarus 2002; Visser and Kellogg 1978).

Given the arguments of Kirschning (2020), it seems quite possible that biotin could have been assembled without the present-day fatty acid synthetic pathway. However, it does seem perverse that many eukaryotic organisms (including mammals) lack the ability to make biotin and several other cofactors (Helliwell, Wheeler and Smith 2013). In several cases, partial pathways remain arguing recent loss of essential pathway components. In a few cases (e.g. the S. cerevisiae biotin pathway), pathway genes seem to have been reacquired by horizontal gene transfer. The efficacy of inhibitors of biotin synthesis in coping with infections by pathogenic bacteria argues that host-acquired biotin is not readily available to bypass the inhibitor. Similar data are seen when the pathogen has been mutated to biotin auxotrophy. Such biotin auxotrophs are very defective pathogens. Humans have no recognizable vestiges of a biotin synthetic pathway and thus specific inhibitors should be innocuous to the host.

Conflicts of Interest

Work from this laboratory referenced in this review and preparation of this review was supported by grant AI15650 from the National Institute of Allergy and Infectious Diseases, NIH, USA.