Biotin protein ligase as you like it: Either extraordinarily specific or promiscuous protein biotinylation

Biotin (vitamin H or B7) is a coenzyme essential for all forms of life. Biotin has biological activity only when covalently attached to a few key metabolic enzyme proteins. Most organisms have only one attachment enzyme, biotin protein ligase (BPL), which attaches biotin to all target proteins. The sequences of these proteins and their substrate proteins are strongly conserved throughout biology. Structures of both the biotin ligase‐ and biotin‐acceptor domains of mammals, plants, several bacterial species, and archaea have been determined. These, together with mutational analyses of ligases and their protein substrates, illustrate the exceptional specificity of this protein modification. For example, the Escherichia coli BPL biotinylates only one of the >4000 cellular proteins. Several bifunctional bacterial biotin ligases transcriptionally regulate biotin synthesis and/or transport in concert with biotinylation. The human BPL has been demonstrated to play an important role in that mutations in the BPL encoding gene cause one form of the disease, biotin‐responsive multiple carboxylase deficiency. Promiscuous mutant versions of several BPL enzymes release biotinoyl‐AMP, the active intermediate of the ligase reaction, to solvent. The released biotinoyl‐AMP acts as a chemical biotinylation reagent that modifies lysine residues of neighboring proteins in vivo. This proximity‐dependent biotinylation (called BioID) approach has been heavily utilized in cell biology.

biotin ligase-and biotin-acceptor domains of mammals, plants, several bacterial species, and archaea have been determined.These, together with mutational analyses of ligases and their protein substrates, illustrate the exceptional specificity of this protein modification.For example, the Escherichia coli BPL biotinylates only one of the >4000 cellular proteins.Several bifunctional bacterial biotin ligases transcriptionally regulate biotin synthesis and/or transport in concert with biotinylation.The human BPL has been demonstrated to play an important role in that mutations in the BPL encoding gene cause one form of the disease, biotin-responsive multiple carboxylase deficiency.Promiscuous mutant versions of several BPL enzymes release biotinoyl-AMP, the active intermediate of the ligase reaction, to solvent.The released biotinoyl-AMP acts as a chemical biotinylation reagent that modifies lysine residues of neighboring proteins in vivo.This proximity-dependent biotinylation (called BioID) approach has been heavily utilized in cell biology.
Biotin attachment to its cognate proteins is a two-step reaction 1 resulting in the formation of an amide linkage between the biotin carboxyl group and the ε-amino group of the targeted lysine residue (Figure 1).The first step of the reaction is an attack of the biotin carboxylate on the α-phosphate of ATP, producing biotinyl-5 0 -AMP (Bio-AMP), which remains stably bound to the enzyme.The second step is the transfer reaction, where the nucleophilic attack of the Bio-AMP mixed anhydride by the ε-amino group of the specific substrate lysine residue, resulting in covalent biotinylation of the protein.The BPL reaction is closely analogous to those catalyzed by amino acyl-tRNA synthetases 1 and the proteins share structural elements. 2The biotin moiety attached to BPL target proteins remains bound throughout the life of the protein, although it can be cleaved from peptides released by proteolysis of biotinylated proteins. 3e discovery of biotin protein ligase (BPL) by Lane and colleagues [4][5][6][7] followed the realization that protein-bound biotin was involved in mammalian carboxylation reactions. 86][7] The bacterium required biotin for growth and when starved for biotin, high levels of unbiotinylated acceptor protein accumulated.Very active extracts prepared from starved cells allowed definitive proof that biotinylation proceeded through the synthesis of Bio-AMP, that Bio-AMP could replace biotin and ATP in the biotinylation of the acceptor protein.6][7] Note that BPL is also known as holocarboxylase synthetase [EC 6.3.4.10], which is somewhat misleading, as some biotin-dependent enzymes are decarboxylases or transcarboxylases.In early work, these enzymes were sometimes named after the enzyme-substrate used to assay BPL activity (e.g., holopyruvate carboxylase synthetase) in the belief that each biotin-dependent enzyme was modified by a specific ligase.However, genetic studies in microorganisms and humans indicate each organism generally encodes a single BPL that modifies each of the biotin-requiring enzymes.Nonetheless, some bacteria have two BPL enzymes, one that does most of the biotinylation and a second that regulates biotin synthesis and/or biotin transport.BPL enzymes fall into three groups 9 (Figure 2A,B).The Group I BPLs are the minimal enzymes and have only the catalytic (200 residues) and C-terminal (50 residues) domains (Figure 2A,B).BirA proteins, the regulatory BPLs that comprise Group II, have an N-terminal DNA-binding domain of about 60 residues attached to a minimal BPL by a linker (Figure 2A).The Group III BPLs have very long N-terminal The biotin protein ligase (BPL) reaction follows the two-step reaction first demonstrated for aminoacyl-tRNA ligases by Berg. 1 The first step of the BPL reaction is an attack of the biotin carboxylate on the α-phosphate of ATP to produce biotinyl-5 0 -AMP (Bio-AMP).Bio-AMP remains stably associated with hydrogen bonding with the enzyme.The second step is the transfer reaction.Nucleophilic attack of the BPLbound Bio-AMP mixed anhydride by the ε-amine of the specific substrate lysine results in covalent biotinylation of the attacking lysine residue plus AMP.
F I G U R E 2 Domain structure of biotin protein ligase (BPL) enzymes.(A) The Group II BPL Escherichia coli BirA and (B) The three BPL groups.BirA proteins from diverse bacteria have structures similar to that of E. coli.The monofunctional Group I BPL proteins lack the blue N-terminal domain, whereas the Group III proteins have a long N-terminal extension in place of the Group II DNA-binding domain.The C-terminal domain is essential for activity, as shown by mutagenesis of E. coli BirA 10 and by deletion of the human Group III BPL C-terminal domain. 11A slice through the linker that connects the BirA N-terminal and catalytic domains can produce a Type I BPL, as demonstrated for the Bacillus subtilis BirA. 12The BirA structure is the most recent (PDB 1HXD) and was rendered using Chimera.extensions attached to a Group I BPL (Figure 2B).The first two groups are found in bacteria, archaea, and plants, whereas the Group III BPLs are found in mammals, fungi, and insects.The InterPro database places the catalytic domains in PF03099, whereas the C-terminal domains are placed in PF02237.Since the DNA-binding domains recognize different DNA sequences, the residues involved in DNA binding have disparate amino acid sequences but show structural homology.Moreover, the linkers coupling the DNA-binding domain to the catalytic domain are of differing lengths.
The conservation of the catalytic and C-terminal domains throughout biology argues that nature developed enzymatic protein biotinylation only once.6][7] Indeed, a bacterial protein has been used to assay the activity of the human enzyme in screening for the human disorder, multiple carboxylase deficiency. 13The Escherichia coli BirA was reported to biotinylate acceptor proteins from other bacteria, yeast, and plants 14 and as discussed below, the Bpls of yeast, humans, and plants can functionally replace E. coli BirA function.This conservation argues that the proteins that become biotinylated should also have conserved structures.
The first defined biotin-acceptor domain characterized was that of the Propionibacterium transcarboxylase 1.3S subunit.Deletion studies and point mutations showed that the last 82 residues of the 123-residue protein comprise a domain that is readily biotinylated in E. coli. 15,16Further shortening by even two residues resulted in the loss of biotin-accepting activity, as did substitutions of the residues that immediately neighbor the target lysine residue. 17Crystal and NMR structures of the domain have been reported, PDB 1DCZ and PDB 1O78, respectively.
Subsequent work focused on the E. coli AccB biotin-accepting domain.The domain was first delineated by-products of variable lengths 18 due to inadvertent proteolysis during purification to give a family of C-terminal biotinylated domains called BCCPs.BCCP is now named AccB because it is a subunit of acetyl-CoA carboxylase. 19The N-terminal half of AccB is unstructured and thus a prime target of proteolysis, whereas the C-terminal biotin domain is tightly structured and very protease-resistant.Classical protein chemistry gave the amino acid sequence of an 82-residue protease fragment. 20The first biotin-acceptor structure 21 was that of a crystal structure of a 79-residue proteolytic fragment of E. coli AccB, 20 a form like the 82-residue protein that was too short to be a BirA substrate. 22,23e functional biotin-accepting AccB domain was obtained only after the sequence of the encoding gene was known. 22,24A domain that accepted biotin, AccB-87, was constructed, and the protein was purified. 22The resulting AccB-87 NMR structures confirmed the prior x-ray structure. 25,26The biotin domain structure is a long β-hairpin structure of four pairs of antiparallel β-strands that wrap around a central hydrophobic core (Figure 3A,B).The domain structure is separated into two quasi-symmetrical halves divided by a hairpin β turn located opposite the N-and C-terminal residues.The lysine residue targeted for biotinylation is on the tip of the hairpin.
Investigations of other biotin-acceptor protein domains from diverse organisms show that all have structures very similar to AccB-87 (Figure 3C), except they lack the protruding "thumb" structure that disrupts the symmetry of the AccB domain.Thus far, such thumb domains are only found in bacterial and plant acetyl-CoA carboxylases.The thumb contacts the biotin moiety of the biotinylated lysine residue. 21,26The biotinylated form of AccB-87 is much more resistant to chemical modification and proteolysis than the unbiotinylated (apo) form. 28The increased stability is largely lost in the biotinylated form when the thumb structure is deleted. 28The thumb is essential for in vivo function of E. coli AccB. 29| E. coli BirA, THE PARADIGM BPL, DISCOVERY AND DEMONSTRATION OF THE BIFUNCTIONAL ENZYMATIC AND REGULATORY ACTIVITIES BPL studies became largely focused on BirA, the sole BPL of E. coli.
BirA was first identified by genetic studies of the regulation of biotin synthesis.At about the same time, Pai at the University of Alberta and Campbell et al. at Stanford University isolated E. coli mutants that overproduced biotin and had elevated levels of biotin biosynthetic enzymes.These strains were called BioR (biotin Regulation) by Pai 30,31 and birA (biotin retention A) by Campbell et al. 32 Campbell et al. went on to show that a strain lacking a functional biotin synthesis gene grew well on 2 nM biotin, whereas a birA derivative of this strain required 4.1 μM biotin. 32The defect in these strains was thought to be either in biotin transport or attachment to biotin-requiring enzymes.Eisenburg et al. 33 later isolated similar strains by selection for resistance to a toxic biotin analog.The three laboratories mapped their genes on the E. coli genetic map, and each found their mutations were closely linked to the same marker located far from the biotin synthesis genes.An important finding was that of Prakash and Eisenberg, 34 who demonstrated regulation of biotin operon transcription in an in vitro coupled transcription-translation system and showed that Bio-AMP, rather than biotin, was the BirA regulatory ligand.Barker and Campbell 35,36 then published two seminal papers demonstrating that birA and bioR are alleles of the same gene that encoded BirA, a bifunctional protein having both ligase activity and transcriptional regulatory activity.They cloned the birA gene and demonstrated that partially purified BirA bound the biotin operon regulatory region (operator) of the biotin operon promoters. 36Later, BirA was purified to homogeneity and used to confirm that repressor activity, ligase activity, Bio-AMP synthesis, and binding of both biotin and Bio-AMP were all properties of the same protein. 37e model that emerged from these studies (Figure 4) was that when the biotin supply is limited, any Bio-AMP synthesized is rapidly consumed in biotinylation of the acceptor protein (apo-AccB), and hence no significant levels of the BirA-Bio-AMP repressor complex accumulate. 39,40Transcription of the biotin synthetic genes increases due to lack of repression, which greatly increases the levels of the biotin synthetic enzymes and thereby accelerates biotin synthesis.When sufficient biotin has been made to biotinylate all apo-AccB, Bio-AMP accumulates in the BirA active site, resulting in dimerization of the protein to give the repressor species. 35,36,40The BirA-Bio-AMP dimers then bind the operator sequence that controls the two out-reading biotin operon promoters and represses transcription (Figure 4).The model of Figure 4 accounts for the regulation of the biotin synthetic gene transcription.A prediction from the properties of the BirA protein and the Bio-AMP ligand was that the level of repression should be sensitive not only to biotin levels but also to the level of apo-AccB.This is because modification of the unbiotinylated AccB would consume the Bio-AMP required for repression of transcription.This prediction was articulated by Barker and Campbell 36 but was not tested until Cronan 39 expressed a gene encoding a biotin-acceptor protein from an unrelated bacterium to provide a gratuitous regulator.
Expression of this protein resulted in a 5-to 20-fold increase in biotin operon transcription over a several hundred-fold range of biotin concentrations.This showed upon increased expression of the biotinacceptor protein, higher biotin concentrations were required to give levels of repression comparable to those found in wild-type strains. 39ter, when the structure of AccB and its biotin-acceptor domain had been determined, biotin operon transcription by expression of either the full length or the biotin-acceptor domain gave the results first seen with the foreign biotin-acceptor protein. 41As discussed below, small synthetic peptides can act as biotin acceptors.In conclusion, the expression of the E. coli biotin synthetic operon is regulated by a simple, yet remarkably sophisticated mechanism.The rate of transcription of the operon responds not only to the intracellular biotin concentrations, but also to the level of the AccB biotin-acceptor protein that requires covalent attachment of biotin.Biotinylated AccB is the physiologically important parameter ("where the rubber meets the road") because, in the absence of biotinylated AccB, E. coli cannot synthesize the phospholipid acyl chains required for membrane function. 24,42,43

| ALLOSTERIC PROPERTIES OF BirA
Studies in my laboratory and that of D. Beckett (University of Maryland) have demonstrated that E. coli BirA is a dynamic and highly allosteric protein.BirA is composed of three domains: the N-terminal DNA-binding domain, the central catalytic domain widely conserved in BPL proteins, and the essential C-terminal domain of unclear function (Figure 2A).Although in crystal structures, the three domains appear to be independent entities linked only by flexible loops, this is not the case.5][46][47] The first indication that the seemingly distant C-terminal domain of BirA modulates events at the active site was the finding that mutations in the C-terminal domain resulted in as much as a 25-fold decrease in affinity for ATP relative to the wild-type enzyme. 10Also, deletion of the C-terminal domain of the mammalian BPL resulted in an inactive enzyme 11 and a Clostridium acetobutylicum BirA homolog lacking a C-terminal domain is devoid of ligase activity 48

(see below).
In crystal structures, the N-terminal DNA-binding domain of E. coli BirA appears remote from the catalytic domain (Figure 2A).However, deletion of the DNA-binding domain surprisingly resulted in defective catalysis. 49The affinity of the N-terminally deleted protein for biotin was decreased 100-fold, whereas the affinity for Bio-AMP was decreased 1000-fold. 49Later investigations showed strong interactions of the central catalytic domain with both the DNA-binding and C-terminal domains.These interactions not only affect catalysis but also the ability of BirA to form the dimers required for DNA binding, as demonstrated by isolation of BirA superrepressor mutant strains. 47Super repressor BirA proteins repress the biotin operon transcription in vivo at biotin concentrations well below those needed for repression by wild-type BirA. 47The mutant proteins were all single amino acid substitutions mapping within the central catalytic domain, although one mapped to the C-terminal domain. 47The mutant proteins formed significantly tighter dimers than those formed by the wild-type protein. 50Surprisingly, none of the mutations mapped to the dimerization interphase, and all bound biotin normally (Bio-AMP binding could not be measured because it is inextricably intertwined with dimerization). 50 mentioned above, deletion of the N-terminal DNA-binding domain resulted in very large decreases in affinity for both biotin and Bio-AMP. 49To probe this puzzling finding, a series of deletions within the BirA-DNA-binding domain were constructed. 51In vivo, each of these deletion strains, including the original deletion of Xu and Beckett, 49 were deficient in growth at physiologically relevant low biotin concentrations, whereas high biotin concentrations allowed good growth. 51Strikingly, the smallest deletion, 14 residues, had the same growth behavior as overlapping deletions of 60 or more residues.The 14-residue deletion removed only the wing of the winged helix-turn-helix (wHTH) DNA-binding domain.In wHTH transcription factors, the wing often aids DNA binding by making contacts in the DNA minor groove. 52Hence, it was not surprising that the wing-less BirA protein was deficient in the regulation of bio-operon expression.
However, the effects on catalysis remained to be explained and argued that the wing may play a structural role or contribute to residue side chains involved in catalysis. 51A search of the protein structure databases for wHTH wings that closely matched the BirA wing disclosed that the wing of OmpR, another E. coli wHTH transcription factor, was a good structural match.Indeed, the substitution of the OmpR wing for the BirA wing largely restored growth at low biotin concentrations to that of the wild-type BirA.As expected from the The regulatory system of biotin synthesis in Escherichia coli.BirA is represented by green ovals, biotin is represented by black operon circles, the AMP moiety by red pentagons and AccB by dark blue ovals.The biotin synthesis operon is shown below each panel.The arrows indicate transcription from the leftward and rightward bio promoters, which are controlled by a common operator sequence.Panel A. In the presence of excess biotin, BirA synthesizes Bio-AMP, which remains protein bound and dimerizes.The BirA-Bio-AMP complex binds the operator DNA and represses transcription.Derepression (greatly increased transcription) can be triggered either by a limited supply of biotin (Panel B) or by a large excess of the AccB biotin-acceptor protein (Panel C).If unbiotinylated AccB levels are high and sufficient biotin is present, the protein functions as a biotin protein ligase.When the unbiotinylated AccB has been converted to the biotinylated form, Bio-AMP is no longer consumed and the homodimeric BirA-Bio-AMP complex accumulates and the bio operator becomes fully occupied, resulting in repression of transcription of the biotin biosynthetic genes.Biotin synthesis and transport in Bacillus subtilis and Staphylococcus aureus have been shown to be regulated by very similar mechanisms. 12,38iffering sequences of the OmpR wing, the foreign wing could not restore DNA binding. 51In prior work, 47 a BirA mutant was isolated that had a mutation in a wing residue, T25S.The original isolate also contained a BirA mutation, G115S, which is the birA1 mutation isolated by Campbell et al. that defined the gene. 32The G115S protein is extremely defective both in vivo and in vitro, 53 but somehow, the T25S wing mutation largely overcame the G115S defects, resulting in good growth and bio-operon repression. 51The G115S mutation maps in the biotin binding loop, which argued that the T25S wing interacts with the defective G115S loop and largely restores its function.Two other loop mutations, one being the R118G mutation used for BioID promiscuous biotinylation (see below), were also repaired by the T25S mutant wing.Hence, growth at low biotin concentrations and repression were restored to all three of the biotin-binding loop mutant proteins.The straightforward interpretation of these results is that the wHTH wing stabilizes or aids the organization of the biotin/Bio-AMP binding loop to form a functional active site.This explains why the original N-terminal deletion of Xu and Beckett 49 had such a dramatic phenotype: the biotin/Bio-AMP binding loop could not be properly organized.Note that a shortcoming of our knowledge of BirA action is the lack of a crystal structure of dimeric BirA-Bio-AMP bound to the bio operon operator DNA sequence.This deficiency is not for lack of effort.In several labs, significant quantities of BirA and numerous DNA sequences were tipped down the drain after failing to form BirA-DNA complexes.

| STUDIES OF OTHER BACTERIAL BPLs
Three other bacterial BPLs have been studied in detail.These are the Group II enzymes of Bacillus subtilis and Staphylococcus aureus and the type I enzyme of the thermophile Aquifex aeolicus.Pero et al. demonstrated that B. subtilis BirA functioned both as a repressor of biotin synthesis and a BPL. 54They selected mutants resistant to toxic biotin analogs as previously utilized in E. coli by Eisenburg et al. 33 and obtained strains that overproduced biotin, thereby diluting the toxic analogs. 54These mutations were single amino acid substitutions in the N-terminal domain of the protein, which had some sequence similarity to the N-terminus of E. coli BirA.These workers also showed that expression of the wild-type B. subtilis BirA supported the growth of a B. subtilis mutant strain. 54Although B. subtilis is a Firmicute and thus only very distantly related to E. coli, its BirA has regulatory properties very similar to those of E. coli BirA.Subsequent studies showed that B. subtilis bound the promoter of the cognate biotin synthesis operon and the promoters of two biotin transporters with similar binding affinities. 12DNA binding required both biotin and ATP and proceeded via the synthesis of Bio-AMP.In vivo, the amplitude of regulation by biotin and acceptor protein levels was very similar to those of E. coli. 12surprise was that B. subtilis BirA, unlike E. coli BirA, does not require its N-terminal DNA-binding domain for normal ligase activity.12 Several N-terminal B. subtilis BirA deletion proteins were constructed with endpoints based on the modeling of B. subtilis BirA on the crystal structure of S. aureus BirA 55 (PDB 4DQ2).Two deletions that eliminated the DNA-binding domain had essentially normal ligase activities, whereas two other deletions that cut into the predicted central catalytic domain were either partially or totally defective in ligase activity.12 The ligase activities of the purified B. subtilis BirA N-terminal deletion proteins were tested in biotinylation assays in vitro.The two N-terminal deletion proteins gave full complementation of an E. coli birA mutant and showed Bio-AMP synthesis and biotin transfer activities indistinguishable from those of the wild-type protein.12 In contrast the deletions that cut into the predicted central domain had significantly reduced Bio-AMP synthesis abilities.12 To test if another Firmicute BirA had similar properties, the BirA of S. aureus, a relative of B. subtilis was altered. Th crystal structure of S. aureus BirA (PDB 4DQ2) has the expected N-terminal wHTH DNA-binding domain.55 However, the protein was reported to dimerize in the absence of either biotin or Bio-AMP with a K d of 29 μM, a concentration much lower than that of E. coli BirA.56 Moreover, S. aureus BirA was reported to bind the biotin operon promoter in the absence of biotin and Bio-AMP, leading to questions about the biological function of the protein.The addition of Bio-AMP gave only a six-fold increase in binding affinity.56 The reported electrophoretic mobility shift assays showed that the unliganded S. aureus BirA bound the cognate bio operon operator DNA.This high level of DNA-binding activity in the absence of biotin or Bio-AMP would prevent S. aureus BirA from being an effective regulator of biotin operon transcription since DNA binding, hence repression, would occur without regard for the cellular levels of biotin, Bio-AMP and acceptor protein.Thus, even if the cellular concentration of biotin was low, engendering a requirement for biotin biosynthesis (and/or biotin transport), the unliganded protein would repress the expression of the biotin biosynthesis operon and transporters.38 Therefore, S. aureus BirA would obstruct rather than regulate biotin synthesis and transport.
To address these seemingly perverse properties of S. aureus BirA, advantage was taken of the fact that the S. aureus BirA-DNA-binding sites are very similar to those of the previously studied B. subtilis BirA. 38The chromosomal B. subtilis birA gene was replaced with S. aureus birA which put expression of the B. subtilis biotin synthesis operon under control of S. aureus BirA.This construct responded to biotin and acceptor protein levels in a parallel manner to that seen when controlled by B. subtilis BirA. 38However, repression by S. aureus BirA was somewhat weaker than in the cognate B. subtilis system. 38 vitro experiments were complicated by the fact that S. aureus BirA copurified with bound Bio-AMP.Upon removal of Bio-AMP, physiologically reasonable ligand binding constants were obtained for S. aureus BirA. 38ong the Group II BPLs of known structure, S. aureus BirA had the highest amino acid sequence identity to B. subtilis BirA (31%), and thus, it was expected to also tolerate the loss of the N-terminal domain.This was not the case.All three N-terminally deleted S. aureus BirA proteins failed to restore ligase activity to an E. coli BirA mutant strain at low biotin concentrations. 38That is, the removal of the S. aureus BirA N-terminal domain resulted in the defective catalysis first seen for E. coli BirA.Deletion of the wing structure had the same effect as deletion of the N-terminal domain as previously seen with E. coli BirA. 38e report 56 that S. aureus BirA dimerized at very low protein concentrations in the absence of either biotin or Bio-AMP was reinvestigated, and the prior report was demonstrated to be incorrect. 53e S. aureus BirA protein is monomeric in the absence of ligands and the addition of biotin or Bio-AMP increases dimerization. 53However, the S. aureus BirA differs from those of E. coli and B. subtilis in that biotin alone increases dimer formation. 53This property may, in part, explain the differing reports.The later report demonstrated that the S. aureus BirA preparations used were free of biotin and Bio-AMP by direct means (mass spectrometry), 53 whereas the prior report used indirect methods, the results of which were not reported. 56e only well-characterized bacterial Group I BPL is that of the hyperthermophile A. aeolicus. 57,58The protein is monomeric and is composed of two domains, the N-terminal catalytic domain and the C-terminal domain seen in other BPLs.The structure of these domains is very similar to the analogous domains from the Group II BirAs discussed above.The activity (k cat /K M of A. aeolicus BPL is 300-fold lower than that of E. coli BirA. 57However, a caveat is that the enzyme is assayed at 65-70 C, whereas the bacterium grows best at 85-95 C. The A. aeolicus BPL has no detectable activity at 37 C, which precluded complementation of the E. coli BirA mutant strains and biotinylation of the E. coli biotin-acceptor protein.It is interesting that the crystal structure of A. aeolicus BPL has fewer disordered loops than the E. coli and S. aureus BirA proteins. 58This could be due to the low temperature of crystallization relative to the growth temperature of the bacterium.The one loop that remained was the biotin-binding loop which, as seen in the BirAs, became ordered in the presence of ligands. 58A very similar Group I BPL of known crystal structure (albeit a dimeric protein) is that of another hyperthermophile, the archaeon Pyrococcus horikoshii OT3. 59Curiously, the P. horikoshii BPL is reported to modify the E. coli biotin-acceptor protein but the converse reaction, E. coli BirA biotinylation of the P. horikoshii biotin-acceptor protein, failed. 9However, the reaction was done at 20 C, whereas P. horikoshii OT3 grows at 95 C. Hence, the acceptor protein may not have been in its native conformation.It is often difficult to find conditions in which hyperthermophile and mesophile proteins are compatible.

| BACTERIA HAVING TWO BPLs
Typically, an organism (mammals, plants, bacteria, or archaea) encodes a single BPL that modifies all of the biotin-requiring enzyme proteins.
Nevertheless, there are three examples of bacteria that encode two BPL proteins.The first of these was Francisella novicida, a Gramnegative bacterium often studied as a surrogate for the extremely pathogenic Francisella tularensis.F. novicida encodes two BPL enzymes, a Group II BirA resembling that of E. coli BirA and a Group I BPL, called BirA and BplA, respectively. 60Both enzymes complemented the growth of an E. coli birA mutant strain and the F. novicida BirA regulated its cognate biotin operon and weakly regulated E. coli biotin operon expression.However, F. novicida BirA had poor ligase activity, whereas BplA had robust activity. 60F. novicida deletion mutants showed that either BirA or BplA could support growth.The weak ligase activity of F. novicida BirA argued that its primary function was regulation rather than biotinylation because BplA does the great bulk of biotinylation. 60The presence of BplA argues that in F. novicida, regulation of bio operon transcription by the supply of apo-AccB would not take place because the very active BplA would modify AccB and thereby short circuit that mode of regulation. 60Hence, F. novicida BirA primarily functions to monitor the intracellular concentration of biotin and would perform this task only at high intracellular biotin concentrations because only then could it form the key regulatory ligand, Bio-AMP. 60The low affinity of F. novicida BirA for biotin seems an advantage because it would prevent the regulatory system from starving the BplA ligase for biotin.
The Firmicute bacterium Lactococcus lactis also has both a BirA and a Group I BPL, which were called BirA1 and BirA2, respectively. 61 lactis differs from the bacteria discussed thus far in that it requires biotin for growth because it lacks biotin synthesis genes.L. lactis has two biotin transporters, the only sites for regulation.Indeed, the BirA1 (but not BirA2) bound the promoter of one of the transporters.61 Both L. lactis proteins have modest ligase activities relative to that of E. coli BirA, so both enzymes might be needed to give sufficient biotinylation for growth.61 It is interesting that Streptococcus suis, a close relative of L. lactis, has only a single BPL, a BirA-like enzyme that binds the promoter of the sole biotin transport gene.61 Another Gram-positive bacterium, the industrially important C. acetobutylicum, encodes a Group I BPL (BplA) plus a curious truncated BirA homolog called BirA 0 (the 0 denotes truncation) that lacks the C-terminal domain (the last 60 residues).48 The question was why BirA 0 , a seemingly inactive ligase, has been preserved when the bacterium has the BplA ligase.Like L. lactis, C. acetobutylicum lacks biotin synthesis, and thus, any regulation would be targeted to the three putative biotin transporters deduced by Rodionov et al. 62 BplA is a functional ligase as shown by restoration of growth of an E. coli birA mutant strain and in vitro Bio-AMP synthesis assays whereas BirA 0 lacked detectable activity in both assays.48 Fortuitously, the three C. acetobutylicum biotin transporter genes had BirA-binding sites within their promoters that were closely similar to that of the B. subtilis biotin operon promoter.This allowed testing of the BirA 0 protein for regulatory activity.Indeed, BirA 0 regulated B. subtilis biotin operon transcription in response to biotin levels. 48This was buttressed by in vitro binding assays showing that BirA 0 bound both the C. acetobutylicum bioY1 transporter promoter and the B. subtilis biotin operon promoter in a biotin-dependent manner.48 Finally, BirA 0 was converted to functional ligase proteins of modest activity by fusion to either the C-terminal 49 or 66 residues of BPL.The two fusion proteins also had modest regulatory activity in B. subtilis.48 The physiological role of BirA 0 is to regulate biotin uptake from the environment, which selects for its maintenance in the genome.Some evidence for interaction between BirA 0 and BplA suggests a means to adjust BplA activity to biotin supply.48 A plausible scenario is that some progenitors of C. acetobutylicum had a functional biotin biosynthesis pathway regulated by a classical BirA bifunctional ligase.When the ability to synthetize biotin was lost, duplication and subsequent remodeling of the bifunctional ligase gave the BplA and BioA 0 proteins.In this scenario, BplA is required for biotinylation of acetyl-CoA carboxylase and pyruvate carboxylase whereas BioA 0 regulates biotin transport.48

| PEPTIDE MIMICS OF BIOTIN-ACCEPTOR DOMAINS
[65] The fusion proteins are readily detected, quantitated, and purified by use of the many techniques based on the high-affinity binding of biotin by avidin and streptavidin.4][65] Such fusion proteins also served as markers in electron microscopic investigations. 66,67Although the C-terminus is the usual location of biotin domains in nature, these domains function as modules and can be attached to the N-terminus of a target protein 63 or even spliced within a coding sequence to give an internal fusion. 65,68However, the small peptide sequences active in E. coli discovered by Schatz 69 provide a more facile means to tag proteins.Starting from large peptideencoding oligonucleotide libraries, Schatz isolated a series of small peptide substrates (13-30 residues) that were E. coli BirA biotinylation substrates.Upon fusion to either end of a wide variety of proteins, the peptide sequences are efficient BirA biotinylation substrates, 69,70 although they have sequences strikingly different from AccB, the natural substrate (Figure 5).Certain peptide library residues were fixed to match those that bracket the AccB biotinylated lysine residue.However, most of those peptides were not biotinylated and, hence, were lost in the selection process. 69Despite the lack of sequence similarity to AccB peptide 85-14, it was shown to be as active a biotin acceptor as AccB-87. 72The lack of sequence conservation with AccB argues strongly that the peptide sequences bind to BirA differently than does AccB.Indeed, peptide 85-14 is biotinylated only by E. coli BirA.The BPLs of other prokaryotic organisms (B.subtilis and the archaeon, Methanococcus jannaschii) and from eukaryotes (the yeast, plant, insect, and human enzymes) fail to biotinylate peptide 85 fusion proteins, [73][74][75][76][77][78][79] although each of these ligases modifies AccB.Peptide sequences have been isolated that are biotinylation substrates for the yeast BPL, but not for BirA.Consistent with the E. coli results, the sequences of these peptides bear little resemblance to the natural yeast acceptor domains. 73e Schatz peptides are very useful for site-specific biotinylation of proteins. 63,70However, the biotin-accepting peptides were also used to test models of transcriptional regulation by E. coli BirA.Two models have been put forth for the mechanism whereby accumulation of the unmodified biotin domain leads to increased transcription of the biotin synthesis operon.One model is the heterodimer model in which an unmodified AccB acceptor protein is proposed to bind a monomeric BirA:Bio-AMP complex. 80,81Formation of this heterodimer would compete with the assembly of the dimeric complex required for DNA binding and repression.AccB was proposed to form extensive interactions with residues on the BirA face used for dimerization to form a heterodimeric species that blocked transcriptional repression. 80Hence, the postulated heterodimer must be sufficiently long lived to compete with the homodimerization of BirA:Bio-AMP, indicating that it should be directly detectable.However, this has not been reported, and only indirect evidence for its existence is available. 81The alternative to the heterodimer model is the original model in which AccB and BirA compete for Bio-AMP. 39,40Excess AccB would "steal" the Bio-AMP synthesized by monomeric BirA and thereby block homodimerization and repression. 39,40The findings of Schatz 69 that small peptides are efficient BirA biotinylation substrates provided a means to assess the two models.The peptides are too Although a biotin-accepting fusion protein sometimes gave a 10%-20% greater level of derepression than the other fusions, no clear pecking order was seen.An Acc-87 protein lacking the biotinylation target lysine (K122M) showed the lack of repression at high biotin concentrations as the empty vector.From Solbiati and Cronan 71 with permission.
short to form the extended interactions proposed to be involved in heterodimer formation 80 and lack certain of the residues within the peptides considered important in the interaction. 71Therefore, the extensive protein-protein interactions required by the heterodimer model would be absent and transcription would proceed.In contrast, in the original model, the peptides would compete with BirA for Bio-AMP and result in increased transcription.
Fusion proteins to two different Schatz peptides (Figure 5) and a parallel fusion to AccB-87 were constructed. 71Expression of all three fusions gave efficient and essentially identical derepression of biotin operon transcription over a 100-fold range of biotin concentrations. 71at is, the peptide fusions were as efficient as AccB-87 in bio-operon regulation (Figure 5).It might be argued that the Schatz peptide sequences are somehow able to form appropriately long-lived complexes with BirA:Bio-AMP as effective in derepression as the postulated complex with AccB.This very remote possibility cannot be totally excluded.However, the peptide sequences would have to interact with BirA:Bio-AMP with essentially the same binding strengths and kinetics as that of the natural acceptor protein.Given their small size and markedly diverged sequences, this seems most unlikely, particularly because Schatz selected only for the ability to accept biotin and not for derepression. 69Moreover, the Schatz selections were done in the presence of excess biotin (10 μM), which precluded inadvertent selection for derepression.In conclusion, the peptide fusion data strongly favor the original model in which AccB and BirA compete for Bio-AMP and demonstrate that the heterodimer model is not the switch that regulates the bio operon. 71

| EUCARYOTIC BPLs AND A HUMAN DISORDER OF BPL FUNCTION, MULTIPLE CARBOXYLASE DEFICIENCY
Most eucaryotic BPLs are much larger than the bacterial proteins except for the plant BPL. 79The human, yeast, and plant BPL genes were isolated by complementation of E. coli birA mutants with cDNA libraries, which required that the foreign BPLs would biotinylate AccB. 79,82,83This approach was independently used to isolate the genes encoding the BPLs of the yeast Saccharomyces cerevisiae 83 and humans. 82Later, the plant Arabidopsis thaliana gene was also isolated by the same approach. 79e gene encoding the human BPL was also isolated by conventional means, purification of the enzyme 84 followed by screening a cDNA library with oligonucleotides based on derived peptide sequences. 85e human BPL is a protein of 727 amino acid residues, whereas the yeast BPL is slightly smaller (690 residues).In contrast, the plant BPL (367 residues) is only slightly larger than E. coli BirA (321 residues).All three eucaryote BPLs contain central catalytic domains (including the biotin-binding site) and the C-terminal domains seen in E. coli BirA.However, the human and yeast BPLs have very long N-terminal extensions upstream of their catalytic domains, 351 and 269 residues, respectively.The human and yeast extensions have unrelated sequences, although both proteins are predicted by Alpha-Fold to have structured elements within their N-terminal sequences.
In the human protein, deletion of most extension residues is tolerated (as assayed by complementation of an E. coli birA mutant strain), although there is a small internal region where deletion weakens or abolishes activity. 11Deletion of the yeast ligase extension results in an activity loss of >3500-fold. 86The role of the human extension sequence is thought to allow discrimination in the biotinylation of the five different carboxylase substrates. 9,87,88Much of our knowledge of the human BPL data comes from studies of the disease, multiple carboxylase deficiency.Mutations in the HLCS gene encoding the human BPL (called HCS) are one cause of the disease. 89e human HLCS gene (HGNC:HGNC:4976) spans 14 exons and about 250 000 bp. Extensive mRNA splicing and perhaps three promoters generate a variety of protein forms.The best-characterized form is the full-length protein.Only those multiple carboxylase deficiency patients that respond to biotin treatment can be studied because HCS is an essential enzyme.The essentiality of HCS is shown by the absence of individuals with deletions of both copies of the gene and, more directly, by the finding that HCS knockout mice show embryonic lethality. 90Most of the biotin-responsive HCS patients have point mutations in the catalytic domain of the ligase, resulting in decreased affinities for biotin. 27,91,92However, there are a few mutations that map in the N-terminal extension that may reflect the interactions of the N-terminal extension with the catalytic domain postulated by Campeau and Gravel. 11Indeed, a mutation within this region constructed in vitro had very low ligase activity that was not remedied by increased biotin concentrations. 93Studies suggest that patients can be classified into two types.In the dominant type, biotin responsiveness is due to mutations resulting in a high K M for biotin.
Such patients do well on biotin supplementation and show essentially complete relief of clinical and biochemical symptoms. 91The other type has a normal affinity for biotin but low enzyme activity.
It was mentioned above that the E. coli AccB-87 protein has been used for screening for defects in HCS. 13 However, AccB-87 is a rather poor HCS substrate, about one tenth as active as the p-67 protein derived from the human propionyl-CoA carboxylase. 27The low activity of AccB-87 with HCS is largely due to the thumb loop.Deletion of the thumb loop gave a large increase in biotinylation. 27Note that there is a second cause of multiple carboxylase deficiency: mutations within the gene encoding biotinidase. 1,3Biotinidase is the amidohydrolase responsible for releasing biotin from biotinylated peptides resulting from proteolytic degradation.The released biotin can then be recycled for biotinylation of new biotin-dependent carboxylase proteins.As seen in the HCS K M mutants, biotinidase deficiency is readily treated by biotin supplementation.However, both forms of multiple carboxylase deficiency must be treated very soon after birth because the biotin-requiring enzymes are essential for intermediary metabolism.

| THE ORIGINS OF BioID
Many birA mutants and the first birA plasmids were isolated by Barker and Campbell 94 by use of strains in which the biotin operon promoters transcribed the lacZYA operon of E. coli.This facile selection resulted in many strains having defects in biotin operon regulation and ligase activity (see below).Otsuka et al. 95 sequenced the birA genes of several of these mutant strains and the mutant proteins encoded by two of these genes were studied by Kwon and Beckett. 96ese genes encoded the BirA G115S (the birA1 mutation) and BirA R118G (the birA91 mutation) proteins. 96Kwon and Beckett 96 reported that both mutant proteins bound biotin with lower affinity than the wild-type protein, consistent with the in vivo data of Barker and Campbell 94 and showed much less ability to bind Bio-AMP.Indeed, the reported Bio-AMP dissociation constants of BirA G115S and BirA R118G were, respectively, 3000-and 400-fold greater than that of wild-type BirA protein, whereas those of biotin were 250 and 100-fold greater (binding of ATP was normal for both mutant proteins). 96These data led to the hypothesis that if these mutant proteins made Bio-AMP, it would leak from the enzyme active site into solution and act as a promiscuous chemical biotinylation reagent in a proximity-dependent manner. 97Bio-AMP is a mixed anhydride, so it would be attacked by the ε-amino groups of protein lysine residues to give amide linkages between biotinoyl-lysine modified proteins identical to those given by chemical biotinylation reagents (e.g., biotin-Nhydroxysuccinimide).The dependence on proximity would result from the rapid hydrolysis of the mixed anhydride in free solution, a property that would limit the ability of Bio-AMP to react with distant proteins. 97Choi-Rhee et al. 97 tested this hypothesis by expressing the wild-type and mutant proteins in E. coli, arguing that if promiscuous biotinylation occurred, many proteins would be modified.Indeed, in cells expressing BirA R118G, many biotinylated protein bands were seen in addition to BCCP (AccB), the sole E. coli biotinylated protein (Figure 6A).The most prominent labeled band was BirA itself. 97This self-biotinylation was expected from prior studies with aminoacyl-tRNA synthases 99,100 and is an intramolecular reaction. 97Subsequent in vitro experiments showed that purified BirA R118G protein biotinylated bovine serum albumin and RNAse A (Figure 6A).The BirA G115S protein labeled only BCCP and had wild-type selfbiotinylation, indicating that it probably did not release Bio-AMP to free solution.This is probably due to the indirect release assay used (intrinsic BirA protein fluorescence). 96 test the proximity-dependent hypothesis, the investigators took advantage of the BirA R118G C-terminal penta-histidine tag and an anti-tag antibody. 97Although the antibody heavy and light chains were relatively poor biotin acceptors, the chains were labeled more intensely when attached to the C-terminus than when the antibody was attached to another protein in free solution.Moreover, bovine serum albumin or RNAse A became biotinylated but did not decrease biotinylation of the antibody bound to the C-terminus. 97though the report of proximity-dependent biotinylation appeared in a readily available journal, the discovery of proximitydependent biotinylation languished for 8 years until Roux et al. 101 recognized that it could be used to assay protein-protein interactions in mammalian cells and called it BioID.A plasmid encoding a fusion of a codon-optimized birA encoding R118G (often called BirA*) to the protein of interest (sometimes called the "bait") is introduced into the cell line and biotin is added.After incubation, the biotinylated proteins are captured by use of an immobilized biotin binding protein and identified by mass spectroscopy.This has given rise to 490 BioID entries in PubMed.The approach has been used successfully in mammalian cell lines, yeast, apicoplast parasites, plants, mice as referenced herein 102 and the nematode Caenorhabditis elegans. 103The most recent culmination is the use of 235 qualified bait proteins to define the intracellular locations of 4145 unique proteins in HEK293 (human embryonic kidney) cells. 104A prior research group had used 18 localization baits to determine 1911 interactions also in human embryonic kidney cells. 105m et al. 98 have estimated that 10 nm is the practical labeling radius of BioID.This estimate was obtained by fusing BirA* to several Bio-AMP diffuses from the enzyme and biotinylates lysine residues of neighboring proteins and itself (not shown).The red protein is at the perimeter of the Bio-AMP cloud, which is limited by hydrolysis and the radius of labeling is estimated to be 10 nM. 98Panel A from Choi-Rhee et al. 97 with permission.

F
I G U R E 3 The AccB biotin-acceptor domain AccB-87 of Escherichia coli and conservation of biotin-acceptor domains throughout biology.(A) The overall fold of the AccB-87 holoprotein with the biotin moiety attached to K122.The β strands are coded from red at the N-terminus to purple at the C-terminus.The protruding "thumb" seen in yellow at the left of the molecule is found in bacterial acetyl-CoA carboxylases but is absent in other biotin enzymes.(B).The structure of Panel A was rotated by 90 about the y-axis to display the quasi-symmetry of the AccB biotin domain structure.(C) Structural conservation of acceptor domains.The α-carbons of the domain structures were superimposed to illustrate the high level of conservation across species.The conserved biotin attachment on the β-hairpin loop is colored red.The thumb of E. coli AccB-87 is indicated.The domain structures of Rhizobium etli pyruvate carboxylase (yellow, PDB 2QF7), Bacillus subtilis AccB (green, PDB 1Z7T), E. coli AccB-87 (gray, PDB 1BDO), Propionibacterium freudenreichii subsp.shermanii transcarboxylase (blue, PDB 1DCZ), human acetyl-CoA carboxylase 2 (purple, PDB 2DN8), and the archaeon, Pyrococcus horikoshii BCCP (black, PDB 2EVB) are given.From Healy et al. 27 with permission.

F
I G U R E 5 Regulation of Escherichia coli bio operon transcription by biotin-acceptor peptides.(A) Alignment of the residues neighboring the biotin attachment site (asterisk) of E. coli AccB and five diverse biotin-acceptor proteins reported to be biotinylation substrates for E. coli BirA, but inactive with peptide 85 fusion proteins.The sequences of peptides 85 and ME 69,70 are also given.Residues conserved in at least four of the natural sequences are boxed (the * denotes the lysine residue that becomes biotinylated).The acceptor sequences are as follows: Ec, E. coli AccB; Bs, Bacillus subtilis AccB; Mj, Methanocaldococcus jannaschii pyruvate carboxylase; Sc, Saccharomyces cerevesiae pyruvate carboxylase 1; At, Arabidopsis thaliana AccB; and Hs, Homo sapiens pyruvate carboxylase.(B) The effects of expression of the E. coli maltose binding protein (MalE) fusions to AccB-67, Pep-85, or Pep-ME on bio operon transcription.

F I G U R E 6
Origins and cartoon of BioID.(A) The left-hand gel is an SDS gel of whole-cell extracts of an Escherichia coli strain transformed with plasmids expressing the mutant BirA proteins denoted at the bottom of the gel (WT denotes wild-type).The biotinylated bands were detected using a streptavidin conjugate following transfer to an Immobilon membrane.The right-hand gel is in vitro biotinylation of proteins by purified BirA R118G.The acceptor proteins were the endogenous biotin-acceptor protein BCCP (AccB-87), bovine serum albumin (BSA), and RNase A. (B) Cartoon of BioID labeling.The bait is the yellow protein covalently fused to the green mutant BirA ligase.The black ovals are biotin.