Discovery of a Biotin Synthase That Utilizes an Auxiliary 4Fe–5S Cluster for Sulfur Insertion

Abstract

Biotin synthase (BioB) is a member of the Radical SAM superfamily of enzymes that catalyzes the terminal step of biotin (vitamin B7) biosynthesis, in which it inserts a sulfur atom in desthiobiotin to form a thiolane ring. How BioB accomplishes this difficult reaction has been the subject of much controversy, mainly around the source of the sulfur atom. However, it is now widely accepted that the sulfur atom inserted to form biotin stems from the sacrifice of the auxiliary 2Fe–2S cluster of BioB. Here, we bioinformatically explore the diversity of BioBs available in sequence databases and find an unexpected variation in the coordination of the auxiliary iron–sulfur cluster. After in vitro characterization, including the determination of biotin formation and representative crystal structures, we report a new type of BioB utilized by virtually all obligate anaerobic organisms. Instead of a 2Fe–2S cluster, this novel type of BioB utilizes an auxiliary 4Fe–5S cluster. Interestingly, this auxiliary 4Fe–5S cluster contains a ligated sulfide that we propose is used for biotin formation. We have termed this novel type of BioB, Type II BioB, with the E. coli 2Fe–2S cluster sacrificial BioB representing Type I. This surprisingly ubiquitous Type II BioB has implications for our understanding of the function and evolution of Fe–S clusters in enzyme catalysis, highlighting the difference in strategies between the anaerobic and aerobic world.

Introduction

Biotin (Vitamin B7) is a sulfur-containing vitamin that functions as an essential cofactor for enzymatic carboxylation, decarboxylation, and transcarboxylation in almost all organisms.¹ While the upstream biosynthetic pathway of biotin biosynthesis can vary significantly across organisms, they all share biotin synthase (BioB) as the terminal catalytic step.² BioB is a Radical S-adenosyl-l-methionine (SAM) enzyme that inserts sulfur into the precursor desthiobiotin (DTB), at sp³ carbon centers C6 and C9, generating biotin (Scheme S1). In addition to biotin synthase, several other members of the Radical SAM (RS) superfamily catalyze sulfur insertion, namely LipA,^3,4 MiaB,⁵⁻⁷ and RimO.^5,8 However, the mechanism of sulfur insertion by these enzymes is a subject of much debate.^9,10 Like most other RS enzymes, BioB utilizes a 4Fe–4S iron–sulfur (Fe–S) cluster to transfer an electron to the sulfonium moiety of SAM, reducing it by one electron, initiating homolytic cleavage to generate a highly reactive 5′-deoxyadenosyl radical (5′-dA·), which subsequently abstracts a target hydrogen atom from the substrate,¹¹ in BioB’s case from C9 of DTB.¹²

Unlike most RS enzymes, BioB has an auxiliary Fe–S cluster that “sacrifices” itself to insert one of its sulfur atoms during catalysis.¹³ This destructive approach to sulfur insertion by BioB only allows one turnover per molecule of enzyme, making it one of nature’s slowest enzymes and prompting some to argue that BioB should be classified as a substrate rather than a catalytic enzyme.^14,15 It has recently been demonstrated that lipoyl synthase, which employs the same destructive mode of catalysis as BioB,¹⁶ can be continually regenerated with Fe–S clusters supplied by NfuA in E. coli back to a catalytically active state.¹⁶ However, while the Fe–S cluster biosynthesis machinery is thought to regenerate BioB for multiple turnovers in vivo,¹⁷ no corresponding mechanism has been shown for BioB in vitro. This destructive mechanism is puzzling as it seemingly wastes cellular resources by having to rebuild and deliver a new Fe–S cluster for every turnover, a process that requires a complex cascade of enzymatic reactions and significant energy input.¹⁸ This mechanism has been partially explained by the fact that so few biotin molecules are needed as cofactors;¹⁹ therefore, the overall fitness loss associated with this catalytic mechanism may be negligible.¹⁰

Currently, biotin for pharma, food, feed, and cosmetics is produced exclusively by a costly >10-step racemic chemical synthesis.²⁰ Production of some vitamins, such as riboflavin, has successfully transitioned from chemical synthesis to a biotechnology-based process.²⁰ The major bottleneck in a biotin-producing cell factory is unquestionably BioB,²¹ and improving upon this enzymatically challenging step is key to transitioning biotin manufacturing from chemical synthesis to a more sustainable bioprocess. To discover possible alternative types of BioBs that may be useful for biological biotin production, we performed a bioinformatic analysis, highlighting the natural diversity of auxiliary cluster coordination. Contrary to previous reports that the cysteine 97 (C97) residue (E. coli BioB) is fully conserved and required for proper coordination of the auxiliary 2Fe–2S cluster,^15,22 it was reported that there were variations in this position,²³ and utilizing presently available sequencing data, we confirm that many BioBs contain alternative residues at this position, most commonly serine. Further characterization revealed that this phylogenetic clade represents a novel type of BioB that does not utilize the currently accepted 2Fe–2S sacrificial sulfur insertion mechanism. This new type of BioB, which we term Type II BioB, contains both a 4Fe–4S and a 4Fe–5S cluster. Rather than use the inorganic sulfur from a 2Fe–2S cluster, this Type II BioB contains a ligated sulfide, an auxiliary 4Fe–5S cluster, which likely acts as the sulfur source for biotin formation. Few other enzymes have been characterized with a 4Fe–5S cluster, all performing either desulfuration, sulfuration, or dehydration reactions.^24,25 Type II BioBs are highly oxygen-sensitive, even in vivo, but they are widespread in nature. Type II BioB is likely the ancestral biotin synthase from which Type I evolved in aerobes and is still in use by virtually all sequenced obligate anaerobic biotin prototrophic bacteria, including important pathogens and many found in the human gut microbiome. Though we have not yet identified the biological source of the inserted sulfur, we discuss the implications of this novel BioB mechanism in relation to fundamental biochemical understanding.

Results

Exploration of the biotin synthase family reveals divergence within the coordination sphere of the auxiliary Fe–S cluster

BioB is a ubiquitous enzyme with an interesting and complex mechanism, yet to date, only BioB from E. coli has a three-dimensional structure determined.¹⁵ To study the sequence diversity and, subsequently, the structural diversity of BioB, we computed a sequence similarity network (SNN) for the biotin synthase enzyme family using the EFI enzyme similarity tool.²⁶ The protein sequence-based clustering of BioB sequences showed that while the main cluster contained sequences similar to E. coli BioB, there are also several large, differentiated clusters of BioB sequences (Figure S1). We chose to investigate one of these larger groups of uncharacterized BioB sequences which contained no biochemically investigated homologues of BioB.

At first glance, this secondary cluster of BioB sequences all seemed to originate from anaerobic bacteria. We further characterized these by multiple sequence-wise alignment of a subset of these “Type II” sequences with previously biochemically studied and published homologues of BioBs, including ones from model organisms E. coli and S. cerevisiae (BRENDA:EC2.8.1.6)(Figure 1a). Unsurprisingly, all sequences contain the highly conserved CxxxCxxC motif, coordinating the eponymous RS 4Fe–4S cluster.²⁷ BioB’s auxiliary 2Fe–2S cluster is generally known to be coordinated by three cysteines and one arginine that are highly conserved.^15,23 In the case of E. coli, these are C97, C128, C188, and R260. These residues are known to be important for activity, with any disruption leading to loss or partial loss of activity.^22,23,28 Surprisingly, cysteine at position 97 is swapped with a serine in the aligned Type II BioBs (Figure 1a), suggesting that this group of BioBs could potentially exhibit an alternative coordination of the auxiliary Fe–S cluster.

Figure 1.

Sequence alignment of Type II BioB demonstrates divergence in predicted auxiliary cluster coordinating residues. (a) The alignment of a group of representative sequences from the Type II SNN cluster of BioBs with previously studied BioBs (BRENDA Enzyme Database) reveals that these all utilize serine, instead of cysteine, at the canonical first auxiliary Fe–S cluster coordinating position (E. coli C97). (b) Based on custom PRATT motifs centered on this cluster coordinating residue, and varying only in the single coordinating residue, we analyzed the 23,268 biotin synthase protein sequences available in UniProt. In addition to S, many biotin synthases utilize residues A, D, or G as an alternative residue. (c) While most bacteria seem to utilize the standard C, when looking at the distribution with 21 bacterial phyla with >20 BioB sequences, many of these have high percentages of alternative residues at this position, mainly utilizing serine (Type II).

To investigate the prevalence and distribution of alternative auxiliary cluster coordination, we investigated the 23,268 protein sequences annotated as biotin synthase in UniProt (release 2020_05). We formulated a sequence motif centered around the coordinating residue. The main motif features include the coordinating residue being flanked by hydrophobic residues and always being 37 ± 5 residues downstream of the CxxxCxxC RS cluster-coordinating motif (Figure S2). Initially, we only sought out cysteine (Type I) and serine (Type II) motifs; however, after mapping the motifs to the BioB sequences, we observed that many sequences did not fall into either of these two categories. Manual inspection revealed that this was because G, D, and A often occur in place of the canonical cysteine coordination in ecBioB. With the updated motif, 22,762 of 23,268 (∼98%) BioB sequences were identified as falling into one of these five motifs: Type I (Cysteine), Type II (Serine), Type G (Glycine), Type D (Aspartic Acid), and Type A (Alanine).

Interestingly, eukaryotes, specifically plants and fungi, exclusively encode Type I BioBs. Archaeal sequences have the highest percentage of alternative motifs and also sequences without any mapped motifs. Manual inspection showed that this was partially because many archaeal BioB sequences do not contain the CxxxCxxC RS motif, but instead, an alternative CxxxxxCxxC motif (e.g., A0A2I0PIW6, T2GJS2, R7PV60). Bacteria are the only domain with Type A and Type D motifs. At first glance, the Bacterial domain has by far the most Type I BioB; however, this is skewed by database bias toward often resequenced genomes such as from Proteobacteria and Actinobacteria. These two phyla represent over half of the ∼23,000 BioB protein sequences analyzed and contain only Type I BioB sequences (Figure 1c). Looking in-depth at distribution within the top 21 most commonly occurring bacterial phyla in the data set reveals that there is high variation within individual Phyla. For example, further inspection of the Phylum Firmicutes shows that variability can be explained at the Order level. Bacillales, comprised mainly of the facultative anaerobes of the genera Bacillus, Listeria, and Staphylococcus, contain only Type I BioB sequences (Figure S3). In the order Veillonellales, obligate anaerobes such as Dialister, Veillonella, and Megasphaera, contain only alternative motif BioB sequences, the majority using Type II.

A few sequences also appeared to have two motifs around the auxiliary cluster region. The BioB from Pythium insidiosum (A0A2D4BNY5) appears to be a duplicated Type I biotin synthase with a predicted phosphatase domain interjected. However, this sequence is an outlier and stems from preliminary whole genome shotgun data, and the physiological significance is uncertain. After filtering out sequences with more than one motif anchored to the same RS motif (CxxxCxxC), there were 24 BioBs with two distinct motifs, 22 of which originate from Streptomyces (of 829 unique Streptomyces spp. sequences in the data set). These Streptomyces “double” BioBs have a conserved C-terminal extension that looks like it possibly encodes additional RS and auxiliary cluster-coordinating residues (Figure S4).

Type II BioB is ubiquitous in anaerobic bacteria

We observed a general trend that putative alternative motifs are mainly found in BioBs from anaerobic organisms. To confirm this, we mapped the organism lifestyle and motif of a representative subset of the data containing varied BioB sequences coming from well-known and studied obligate aerobes, anaerobes, or facultative bacterial genera. As expected, there is a clear divide in overall sequence similarity between BioB sequences of Type I and Type II and other alternative BioB types (Figure S5). It is also obvious that the standard Type II is exclusive to obligate anaerobic bacteria, such as those from the genera Blautia, Clostridium, Fusobacterium, and Bacteroides. While Type I is used by aerobes and facultative anaerobes, there are some exceptions to this rule (Figure S5). However, upon closer inspection of these outlier sequences (i.e., Type I BioBs found in anaerobes and Type II BioB in aerobes/facultatives), it seems that these are most likely the product of recent horizontal gene transfer. Some of these organisms, such as several species of Corynebacterium (e.g., Corynebacterium diphtheriae), contain both a naturally occurring Type I BioB (Q6NKC7) and a Type II BioB predicted as recently horizontally gene transferred (Q6NHL3) from an obligate anaerobic Firmicute (Figure S6).

Type II BioB does not contain an auxiliary 2Fe–2S cluster

To elucidate if the alternative motif of Type II BioB influences the coordination of the auxiliary Fe–S cluster in biotin synthase, we selected a Type II BioB from Blautia obeum for in-depth in vitro biochemical characterization. After expression and anaerobic purification, we noted that the Type II BioB purified with a lower apparent molecular weight (Figure S7) and had a UV–vis signature different from the typical Type I E. coli BioB (ecBioB). While the typical Type I ecBioB exhibits features at 320 and 460 nm, indicating a 2Fe–2S cluster,²⁹Blautia obeum BioB (boBioB) lacked these and instead displayed a broad shoulder around 400 nm, a typical feature of a 4Fe–4S cluster (Figure S8a). These results suggested the possibility that Type II BioB coordinates and sacrifices another Fe–S cluster type, e.g., a 4Fe–4S cluster like that of lipoyl synthase.³⁰

The notion of a biotin synthase without a 2Fe–2S cluster was initially puzzling. To ensure that this group of divergent enzymes were indeed biotin synthases and not misannotations of a different but structurally similar RS enzyme, we tested boBioB for the ability to catalyze the conversion of DTB to biotin in vitro. We found that boBioB was indeed able to couple the reductive cleavage of SAM with the insertion of sulfur into the precursor, DTB, with the E. coli flavodoxin/flavodoxin reductase electron transfer system. Under these conditions, 17.5 μM boBioB (determined by ε400 = 30,000 M^–1 cm^–1, corresponding to ecBioB with two 4Fe–4S clusters³¹) produced ∼0.13 equiv of biotin per protein monomer with an apparent first-order rate constant of 0.0078 min^–1 (Figure 2a) whereas 31.7 μM E. coli BioB (determined by ε410 = 14,900 M^–1 cm^–1, corresponding to ecBioB with one 2Fe–2S cluster and one 4Fe–4S cluster³¹) produced biotin with an apparent first-order rate constant of 0.046 min^–1 (Figure 2b). After the addition of sodium sulfide, both boBioB and ecBioB exhibited elevated biotin production, where both boBioB (Figure 2a; apparent first-order rate constant of 0.019 min^–1) and ecBioB (Figure 2b; apparent first-order rate constant of 0.036 min^–1) produced ∼1 equiv of biotin per protein monomer. Previous studies of ecBioB have shown that exogenous sulfide does not get initially incorporated into biotin and is derived from the 2Fe–2S cluster;³² therefore, this increase in biotin production with the addition of sulfide is likely due to improved reconstitution and stability of the RS 4Fe–4S cluster.²⁹ Previous studies have also shown that when ecBioB is assayed under similar conditions, it forms biotin at a rate of 0.03 min^–1, which is comparable to the observed rate constants in this study.¹⁴ To assist us in determining the Fe–S cluster state of boBioB during catalysis, we measured the UV absorbance of the Fe–S-absorbing region for both boBioB and ecBioB reactions and found that the Fe–S-absorbing region decreases for both boBioB (Figure 2c) and ecBioB (Figure 2d). While the decrease in A400 in ecBioB has been previously shown to correlate with the destruction of the auxiliary 2Fe–2S cluster,¹⁴ it is yet to be determined whether the A400 decrease in boBioB is due to either cluster destruction or reduction of the enzyme after turnover completion.³³

Figure 2.

boBioB is able to catalyze biotin formation. Fraction of biotin formed per BioB monomer with (a) 17.5 μM of boBioB or (b) 31.7 μM ecBioB with 1 mM SAM, 2 mM DTB, 10 μM HpMtan, 5 μΜ flavodoxin reductase, 25 μΜ flavodoxin, and 1 mM NADPH in the presence (pink) or absence (black) of 1.25 mM Na₂S. boBioB catalyzes biotin formation with apparent rate constants of 0.0078 min^–1 (−Na₂S) and 0.019 min^–1 (+Na₂S). ecBioB catalyzes biotin formation with apparent rate constants of 0.046 min^–1 (−Na₂S) and 0.036 min^–1 (+Na₂S). Continuous UV–vis and difference spectra of the FeS cluster absorbing region during the reaction utilize the same reaction mix for (c) boBioB and (d) ecBioB. All reactions were initiated with the addition of SAM.

Crystal structures of Type II BioBs reveal a 4Fe–5S auxiliary cluster

Due to boBioB displaying a distinct UV–vis signature characteristic of a 4Fe–4S cluster and its ability to catalyze biotin formation, boBioB indicated a potentially novel mechanism. Attempting to explain this, we subjected purified Type II BioBs to sparse matrix screening crystallization by a sitting drop vapor diffusion under strictly anaerobic conditions. boBioB and Veillonella parvula HSIVP1 BioB (vpBioB) successfully formed crystals in the presence of SAM and DTB, and we collected data sets with resolutions of 1.35 and 1.81 Å, respectively (Table S1). Like ecBioB, vpBioB appears to form a dimer in the asymmetric unit (Figure S9) while boBioB packs as a monomer (Figure 3a). These Type II BioBs contain the typical TIM-type (α/β)₈ barrels found in ecBioB (1R30)¹⁵ and have very high overall structural similarity, with an RMSD of 0.927 Å (over 213 Cα) and 1.034 Å (over 228 Cα) for boBioB and vpBioB, respectively (Figure S10). As the two novel Type II BioB structures have identical active site residues, we focus on describing the relative differences between boBioB and ecBioB.

Figure 3.

(a) The structure of the monomeric boBioB bound to its substrates: DTB and SAM. boBioB folds as a TIM-type (α/β)₈ barrel, binding a 4Fe–4S and a 4Fe–5S cluster. (b) In boBioB, the C9 atom of DTB is well-positioned to react with the C5′ of SAM for hydrogen atom abstraction (4.2 Å) and the additional sulfide ligated of the 4Fe–5S auxiliary cluster (3.8 Å). (c) 2Fo–Fc Omit map (forest green) of the putative auxiliary 4Fe–5S in boBioB (slate) at 7σ (sulfide denoted S5). (d) The auxiliary 4Fe–5S cluster of boBioB is located at the entrance of the β-barrel, coordinated by C52, C138, and C198 of the β-strands. (e) In ecBioB (color: wheat; PDB: 1R30) the positions of the S43 and C97 are swapped and the 2Fe–2S cluster is coordinated by C97, C128, C188, and R260. (f) With the assistance of a nearby serine (S140) within hydrogen bonding distance (2.8 Å), histidine (H161) is well-positioned to stabilize a deprotonated sulfide for sulfur insertion into DTB.

The RS cluster of boBioB is coordinated by the standard CxxxCxxC motif, located in a loop found after α-helix 6 of the TIM barrel, identical to that of ecBioB. Similar to other RS enzymes, SAM coordinates the unique iron site of the 4Fe–4S cluster via its amino and carboxy moieties while DTB binds with the C9 carbon ∼4.2 Å from the 5′-carbon of SAM (Figure 3b).³⁴ Upon refinement of our initial model, electron density that was reminiscent of a 4Fe–4S cluster became apparent in the location of the auxiliary cluster in boBioB, which was ligated by three cysteine residues and an arginine stacking on top (Cys52, Cys138, Cys198, and R268; Figure 3c and 3d). This coordination contrasts with that of ecBioB, which contains an auxiliary 2Fe–2S cluster bound by three cysteines (Cys97, Cys128, and Cys188) and one arginine (Arg260; Figure 3e). Nevertheless, the presence of an auxiliary 4Fe–4S is consistent with our UV–visible data of purified boBioB and vpBioB. We therefore modeled this density as a 4Fe–4S cluster. Upon further refinement of this model, we found extra density near the open coordination site of the fourth Fe in the auxiliary cluster (Figure S11a). Modeling this excess density as a sulfur atom resulted in the best fit and lowest R and Rfree values (Figure 3c). To confirm our model, we collected a single-wavelength sulfur anomalous diffraction data set at 6.5 kEV (Table S1). The anomalous signal from this data set unambiguously demonstrates that the unknown ligand to the 4Fe–4S cluster is sulfide (S5; Figure S11b). Importantly, the structure of vpBioB contains the same 4Fe–4S cluster ligated by a sulfide, making the auxiliary cluster in Type II BioBs a 4Fe–5S cluster instead of the standard 2Fe–2S cluster found in ecBioB (Figure S9e). Interestingly, S5 of the 4Fe–5S cluster is positioned 3.8 Å from the C9 position of DTB, precisely poised as if it were to act as the sulfur source for producing biotin (Figure 3b), suggesting a novel mechanism of sulfur insertion by Type II BioB.

Comparing the crystal structures of boBioB and ecBioB, the only major differences are in the auxiliary Fe–S cluster coordinating pocket and residues within the active site. The auxiliary cluster coordinating residues Ser43 and Cys97 in ecBioB are swapped with Cys52 and Ser106 in boBioB, forming the structural basis of the sequence motif used to identify this group. In the anaerobic bacterial Type II BioBs aligned in Figure 1a, there seems to be a conserved cysteine at the 43 position (ecBioB numbering), indicating this swap. However, cysteine at this position is also seen in Eukaryotic Type I BioBs. In the bioinformatic characterization of the coordinating motif, we did not include the 43 position into account when selecting Type II BioB for further characterization. Inspecting this position and general area of the global alignment of BioBs, we find that virtually all Type II BioBs identified have this swap in cysteine-serine position, but other clusters have more variation of these and other residues near the active site and auxiliary cluster (Table S2). Based on AlphaFold modeling, it is unclear how the 4Fe–5S cluster would be coordinated by the different motifs found in the other types of BioB (Figure S12B). The crystallization of representative BioBs from these groups is of further interest. Interestingly, when we made the C52S/S106C variant of boBioB, the UV–vis spectrum of the purified protein is virtually identical to Wt boBioB (Figure S8a). However, this protein was unable to catalyze the formation of biotin in our hands. Attempts to crystallize this variant of boBioB also failed. Furthermore, in ecBioB, the auxiliary 2Fe–2S cluster is coordinated by R260 (E. coli; Figure 3e), and interestingly, arginine does not coordinate the auxiliary 4Fe–5S cluster in boBioB (R268). While arginine at this position is close to 100% conserved in bioinformatically predicted Type I BioBs (Table S2), it is not as conserved in Type II, occupying 92% of sequences. While the alternative motifs (G, A, D) can also be predicted by a cysteine instead of serine at the 43 position, the occupancy of R268 is even lower in some of these (Table S2). We also observed that while the ecBioB auxiliary cluster is at the “opening” of the β-barrel, it is buried ∼9 Å beneath the surface and likely the reason why ecBioB’s auxiliary 2Fe–2S cluster can survive purification in the presence of oxygen. The auxiliary cluster of boBioB is much more exposed, which further sensitizes this cluster to oxygen damage (Figure S13, see results below).

Within the boBioB active site, histidine (H161) and serine (S140), which are not found in the ecBioB active site, are within proximity for hydrogen bonding (2.8 Å), and histidine is directly adjacent to both DTB (3.0 Å) and the sulfide (3.9 Å; Figure 3f and Figure S9f). In the investigated Type I BioB sequences, the residues at the boBioB H161 and S140 positions are asparagine (99% conserved) and threonine (69% Thr, 18% Cys, and 12% Ser), respectively. Interestingly, both histidine and serine are highly conserved (99% and 100%, respectively) within the investigated Type II BioB sequences, suggesting that these residues may be important for catalysis. To determine the importance of the active site histidine and serine in the Type II mechanism, we produced the boBioB H161A and S140A variants, which both exhibit the typical 4Fe–4S cluster broad shoulder around 400 nm, and assayed them for activity (Figure S14). While the S140A variant formed biotin at a rate comparable to that of Wt boBioB (apparent first-order rate constants of 0.012 min^–1 vs 0.015 min^–1, respectively), the H161A variant exhibited a remarkable decrease in catalysis, producing biotin with a linear rate of 0.001 min^–1. We did not observe accumulation of 9-mercaptodesthiobiotin with either of these variants (data not shown), suggesting that these residues do not play an essential role in thiolane ring formation. However, based on these data, histidine is important for catalysis, and we propose that a protonated histidine, with the assistance of a nearby serine as a hydrogen bond donor, may play a role in stabilizing deprotonated sulfide on the auxiliary 4Fe–5S cluster, thereby assisting in sulfur insertion (Scheme 1;Figure 3f).

Scheme 1. Proposed Mechanism of Biotin Formation in Type II BioB.

Hydrogen atom abstraction on DTB by a 5′-dA· yields a carbon-centered radical on C9 (1) that reacts with the sulfide atom from the auxiliary 4Fe–5S cluster (2) to yield a reduced cluster and thiol-croslinked intermediate (3). Release of 5′-dA, methionine (Met), and oxidation of the the second cluster (e^–_Aux) allows the second equvilant of SAM to bind. Reduction of the RS cluster (e^–_RS) produces a second 5′-dA· that abstracts the C6 hydrogen atom (4), facilitating thiolane ring formation (5) concomitant with reduction of the 4Fe–5S producing biotin (6) and a 4Fe–4S cluster (7).

Type II BioB functions in vivo in E. coli but is highly oxygen-sensitive

Having proven the biotin synthase activity of Type II BioB in vitro, we set out to characterize its catalysis in vivo. We found that the plasmid expression of boBioB was able to produce biotin and complement a ΔbioB E. coli strain (BS1575). However, this complementation was possible only while cultivating E. coli anaerobically (Figure 4a). While ecBioB and all other Fe–S cluster enzymes are known to be highly oxygen-sensitive in vitro, they are usually not as sensitive in vivo as the intracellular molecular oxygen levels of actively respiring cells are low.³⁵ Some notable exceptions to this exist, such as hydrogenase and nitrogenase, whose Fe–S-dependent catalysis are both famously destroyed by oxygen even in vivo.³⁶ To ensure that this anaerobic requirement for catalysis was an inherent trait of Type II BioB and not just a feature of boBioB alone, we cloned and tested 12 phylogenetically diverse Type II BioBs from a wide array of organisms identified in bioinformatic analysis (Table S3). We also specifically sought to identify and include Type II BioBs received by obligate and facultative aerobes via. recent horizontal gene transfer in Corynebacterium ulcerans, Campylobacter ureolyticus, Pasteurella bettyae, Conexibacter woesei, and Patulibacter sp. This was done under the assumption that these would be oxygen tolerant or may have evolved oxygen resistance while both retaining the Type II mechanism.

Figure 4.

(a) Complementation of a biotin auxotrophic strain of E. coli expression of either ecBioB (BS2029, boBioB (BS3482), or a GFP expression control plasmid (BS4387). Growth was done on minimal media in the presence of the additives indicated, DTB, iotin (BTN) or none (na). All plates were incubated aerobically for 24 h, except one done anaerobically for 48 h (bottom right). (b) Small-scale screening of ecBioB (BS2029, pink) and boBioB (BS3377, blue) for conversion of DTB to BTN in a cell factory setting. The production testing was done under aerobic conditions, with 2 g/L glucose (left) and with high glucose concentrations (10 g/L, right) that led to depletion of oxygen and microaerobic conditions in the sealed microtiter plates during cultivation and production.

All bacterial BioBs complemented ΔbioB when tested for complementation, but only when grown anaerobically as initially observed with boBioB. Predicted Type II BioBs from archaea were not functional in E. coli, most likely due to incompatibility of Fe–S cluster biogenesis or electron transfer machinery.³⁷ This confirmed the hypothesis that Type II BioB is extremely oxygen-sensitive, much more so than Type I BioB and most other RS enzymes. This observation agrees with the hypothesized ancestral nature of Type II BioB and its prevalence in obligate anaerobic organisms. It should be noted that boBioB did allow for growth by biotin auxotrophy complementation aerobically only after greatly increasing expression levels from a high copy number plasmid. This strain was not able to grow as quickly as strains expressing its native ecBioB or with exogenously added biotin, indicating that the growth was still limited by the availability of biotin, even at a high expression of boBioB.

When tested for in vivo cell factory production of biotin from DTB, Type II BioBs were only able to produce significant levels of biotin when cultured in pseudoanaerobic cultivation, with boBioB able to produce the highest levels of biotin of any tested Type II BioB. In contrast, ecBioB produced best under aerobic cultivation conditions (Figure 4b). To better assess the potential biotechnological application of Type II BioBs and obtain a more in-depth understanding of their oxygen sensitivity in vivo, we performed dissolved oxygen-controlled bench-scale fermentations. By controlling the aeration and stirring speeds in the reactors as well as boBioB overexpression using IPTG, we identified the optimal conditions for a fed-batch bioconversion of DTB to biotin in an E. coli cell factory. While it may give the best turnover per boBioB polypeptide, a fully anaerobic fermentation did not give the best productivity (mg biotin/mg biomass/h) for Type II BioB, as the growth rate and biomass yield in an E. coli anaerobic fermentation are very low. Microaerobic fermentation was the best condition for Type II BioB as a biocatalyst (Table S4). Under these conditions, ∼50 mg/L biotin from excess DTB was produced over 48 h, resulting in a specific productivity of ∼0.04 mg biotin/OD/h. The control, ecBioB, was run under the same conditions, resulting in >2-fold higher titers and productivities (Table S4). SDS-page indicated that higher levels of boBioB protein were produced in the fermentation by increasing induction with IPTG (Figure S15). This increase in protein did not result in the improved conversion of DTB to biotin in the process, and cooverexpression of the E. coli Fe–S biogenesis machinery operons of suf or isc with boBioB did not improve in vivo production (data not shown). This points clearly to the fact that catalysis is limited, not by the supply of Fe–S clusters, but by another factor necessary for the complex reaction of the Type II BioB catalysis, such as a supplier of the 4Fe–5S auxiliary cluster-bound sulfide or by electron transfer.³⁸ We then attempted to elucidate this bioinformatically via. comparative genomics, hoping to identify a conserved “helper” gene that may be needed to supply the Type II sulfur source. We inspected the operons of Type I and Type II BioBs manually and with the EFI Genome Neighborhood tool.³⁹ We did an in-depth analysis of 100 randomly selected annotated genomes containing Type I BioB vs 100 randomly selected annotated genomes containing Type II BioB, attempting to systematically identify any significantly over-represented sulfur, cysteine, or Fe–S related genes (Figure S16).⁴⁰ We analyzed genomes to which Type II has been horizontally gene-transferred via. whole genome–genome pBLAST, searching for cotransferred genes (Figure S17).⁴¹ We trawled publicly available microarray transcriptomic expression databases, looking at well-studied anaerobic organisms to find Type II BioB coregulated genes. However, all attempts to find the missing factor have so far failed.

Discussion

This study reveals significant diversity in the BioB sequences and alternative BioB sulfur insertion mechanisms found in nature. Using exploratory bioinformatics, we found a major variable feature of BioB, namely, the first residue coordinating the auxiliary Fe–S cluster. While the widely accepted cysteine is present in all previously characterized BioBs at position 97, one of the alternative motifs shows serine in this position, and this substitution is ubiquitous among obligate anaerobic organisms. We denote this as Type II BioB and propose, through crystallographic evidence, that this type of BioB catalyzes sulfur insertion by using the additional sulfide of the auxiliary 4Fe–5S cluster.

While we were able to crystallographically show the ligated sulfide uniquely positioned for sulfur insertion in Type II BioB, the mechanism of how this occurs is still unknown. Previous studies have discovered similar findings for the sulfuration and desulfuration reactions for Thermotoga maritima LarE and Aeromonas TudS, respectively.²⁴ Both LarE and TudS harbor auxiliary 4Fe–4S clusters with an auxiliary 4Fe–5S cluster to either insert or remove a sulfur atom. However, whether the auxiliary 4Fe–5S cluster remains intact and reduced or is destroyed like Type I BioBs is still unclear. While the absorbance around the Fe–S cluster signature region decreases over time for boBioB, we cannot rule out the possibility that reduction of the resultant auxiliary 4Fe–4S cluster upon reaction completion results in a depressed UV signature.³³

A recent publication demonstrated that two enzymes from Thermococcus kodakarensis (LipS1 and LipS2), previously annotated as biotin synthases, have 4Fe–4S auxiliary clusters and work together to catalyze the formation of lipoyl groups on an octanoyl-lysyl-containing peptide substrate.⁴² However, no current crystal structure for these enzymes exists, and Mössbauer spectroscopy data from this study cannot distinguish whether a sulfur atom is ligated to the 4Fe–4S auxiliary cluster. We believe that these novel lipoyl synthases will utilize a 4Fe–5S cluster as the sulfur source, similar to Type II BioBs. This is based on the facts that LipS1 and LipS2 have an absolutely conserved His residue that is required for catalysis and a nearby hydrogen bond donor (serine) that are both structurally adjacent to the predicted binding site of the auxiliary 4Fe–4S cluster.⁴² Due to this misannotation and the similarity between LipS1 and LipS2 with Type II BioB, this opens the possibility of an alternative substrate for boBioB. Since lipoate is mainly used for metabolism in aerobic organisms, many anaerobic bacteria do not contain a lipoyl synthase gene,⁴³ and as expected, there is no gene encoding for lipoyl synthase that can be found within published Blautia obeum genomes (NC_021022.1, NZ_CYZA01000015.1). Due to most Type II BioBs tested being able to complement a ΔbioB strain, we propose that DTB is likely the endogenous substrate. Based on our findings and that of the lipoyl synthases from Thermococcus kodakarensis, we postulate that these Type II BioBs are the progenitors of the RS sulfur insertion enzymes, including Type I BioBs, LipAs, RimOs, and MiaBs.

Indeed, based on the precedence of RS enzymes MiaB and RimO,^5,6,44 we propose that the Type II BioBs catalyze biotin formation by insertion of the ligated sulfide of the 4Fe–5S auxiliary cluster, resulting in an intact 4Fe–4S auxiliary cluster and formation of biotin (Scheme 1). This strategy would give anaerobic bacteria an increased advantage, because the enzyme can be utilized for multiple turnovers, saving precious energy that is hard to come by with anaerobic metabolism. In this mechanism, the initial reduction of the RS cluster by the endogenous reduction machinery (FldA/Fpr in E. coli) after binding of SAM and DTB yields the potent 5′-dA· that abstracts a hydrogen atom from C9 of DTB (1) and concomitant electron transfer to the auxiliary cluster reducing it from the 2+ to 1+ state (2). We propose that the conserved active site histidine (H161) stabilizes the deprotonated sulfide prior to and during these steps. For the reaction to carry on, the auxiliary cluster would need to give up an electron to either the RS cluster, as in the case of DesII,⁴⁵ or to an electron carrier, as in the case of RS dehydrogenases.⁴⁶ We prefer the scenario in which the endogenous reduction machinery accepts an electron at this step. Release of 5′-dA and methionine, followed by binding of a second SAM, allows the completion of the reaction by a subsequent round of RS cluster reduction, 5′-dA· abstraction of the C6 hydrogen atom (4), and thiolane formation (5). Again, the auxiliary cluster would need to be oxidized to allow for another catalytic cycle with additional S supplied from an unknown source. The biggest difference between this mechanism and Type I BioB is the electronic structure of the auxiliary cluster. In this mechanism, the overall reaction is redox neutral due to the requirement of two electrons to split SAM and the production of two electrons derived from the formation of biotin. Overall, the electronic structure of the auxiliary cluster is unchanged. In the Type I mechanism, the iron atoms of the 2Fe–2S cluster are reduced, and one sulfur atom is lost, destroying the cluster (Scheme S1). Further mechanistic studies, particularly electron paramagnetic resonance and Mössbauer spectroscopy, are needed to fully elucidate the mechanism of sulfur insertion in Type II BioBs.

Along with the mechanism of Type II BioB, the source of the ligated sulfide of auxiliary 4Fe–5S in Type II BioB is also unknown. Our first approach was to add excess sulfide (1.25 mM sodium sulfide) to the boBioB enzyme assay. While the addition of sulfide allowed 1 equiv of biotin per protein monomer formed for both ecBioB and boBioB, it did not allow multiple turnovers (Figure 2). Using bioinformatics methods, we attempted to find candidate genes that could support multiple turnover by a Type II BioB. However, none of these methods revealed any obvious candidates for a genetic source of the Type II sulfur source. While the lack of a sulfur source and 4Fe–5S cluster reformation could prohibit multiple turnovers, this limit could also be caused by the over-reduction of the enzyme. Based on our proposed mechanism, two electrons must be shed from the enzyme for subsequent rounds of catalysis. If no electron acceptor is present, the enzyme would be reduced to the [4Fe–4S]¹⁺ state, rendering it catalytically inactive. In our in vitro reactions with boBioB, we utilized the E. coli FldA-Fpr electron transfer system (as well as in vivo complementation), but this is not the native machinery for B. obeum. Logically, it seems that this nondestructive mode of biotin production would be significantly more energy efficient and thus confer a fitness benefit over Type I catalysis, for which the cascade of Fe–S cluster biogenesis machinery needs to be mobilized to regenerate the auxiliary cluster after each turnover. However, these questions still need to be addressed to unravel this novel mechanism of sulfur insertion.

There is a profound evolutionary connection between iron–sulfur clusters and oxygen, with the use of Fe–S clusters as biochemical cofactors emerging in the time before oxygen.⁴⁷ Indeed, this also means that the archaea and anaerobic bacteria of today represent an ancient anoxic world, with genera such as Firmicutes, Clostridia, Aquificae, and Thermotogae often indicated as the earliest branch points bacterial phylogeny.⁴⁸ It is therefore likely that these previously unknown “alternative” motif BioBs are the original, ancient type of BioB, while the well-studied Type I BioBs evolved later in aerobic bacteria and eukaryotes, posing the question of why and how Type I evolved to fill the niche of the aerobic world. In this work, we focus on the Type II biotin synthase characterized by a serine-cysteine swap at the coordinating residue. However, the deeper we dove into the BioB sequence data set, the more interesting observations we made. We find it likely that there is still more natural diversity in the BioB structure and mechanism.

Methods

Bioinformatics

The initial sequence similarity network (SSN) was generated with the EFI- Enzyme similarity tool³⁹ using the full Pfam biotin synthase (BATS) family: PF06968, using UniRef50 instead of the whole family. Default parameters (Protein Fraction: 1, E-Value: 5) were used for the calculation. A sequence similarity score cutoff of 100 was used to compute the edges of the final network that was visualized in Cytoscape.⁴⁹ Multiple sequence alignments were carried out with the EMBL-EBI Clustal Omega web server, using default settings.⁵⁰ Alignments were visualized with Jalview.⁵¹ Phylogenetic trees were constructed from multiple-sequence alignments using Neighbor-joining without distance corrections⁵² and visualized with iTOL.⁵³ We accessed UniProtKB/Swiss-Prot and TrEMBL UniProt release 2020_05 to fetch all sequences annotated with “family:biotin family:synthase”, including Taxonomic tags: phylum, family, order, class, superkingdom, and genus. These resulting 23,268 biotin synthase protein sequences were scanned with the ScanProSite tool⁵⁴ using the custom PRATT motif pattern: C-x(3)-C-x(2)-C-x(30,40)-[HSRGEAKTVD]-[VILMFYWHKTA]-C/S/G/A/D-[VILMFYWHKTA]-[VAGSTDNRKH]-[VILMFYWHKTAS]-[ASGQKI]. The motif was refined over several iterations for maximum sequence coverage while keeping the approximate expected random matches frequency⁵⁵ to a minimum (∼0.002). Five individual scans were performed, one for each coordination residue (C/S/G/A/D). We identified a subset of horizontally gene-transferred biotin synthases by selecting genera with both Types I and II under the assumption that the same genus generally has the same type of BioB. Outlying BioBs were validated by simple BLAST to show that they shared more sequence similarity to BioBs from phylogenetic distant species than those from their own genus. Data availability: All BioB sequences and scripts used for handling and analysis of the BioB sequence data available at - https://github.com/DavidL-H/BioB/. Predicted models of the alternative types of BioB were generated by Alphafold.⁵⁶

Protein Expression and Purification

BioBs were amplified from genomic DNA with respective primers (Tables S11 and S12) with a KOD Xtreme Hot Start PCR kit (Millipore) according to manufacturer protocol. PCR products were purified with Agencourt Ampure XP PCR cleanup kit (Beckman Coulter) according to manufacturer protocol. Purified PCR products were cloned by ligation-independent cloning (LIC) into pSGC-HIS as previously described.⁵⁷ This incorporated BioB with an N-terminal hexahistidine tag and a 15 AA cleavable TEV-protease linker, behind a T7 promoter. All cloned BioBs were Sanger sequence verified (Genscript Biotech), resulting in BioB expression plasmids listed in Table S5. Chemical competent E. coli BL21(DE3) containing the pPH151 plasmid (Escherichia coli suf operon; used for boBioB)⁵⁸ or pDB1282 plasmid (ecBioB)^3,59 were transformed with the BioB expression plasmid. The rescued transformants were directly grown in 20 mL of LB containing 50 μg/mL kanamycin and either 34 μg/mL chloramphenicol (pPH151) or 100 μg/mL ampicillin (pDB1282) overnight. This was used to inoculate a 2 L PYREX media bottle with PA-5052, a defined autoinduction medium,⁶⁰ with 100 μM FeCL₃ and ∼2 mL antifoam solution (Spectrum). The cultures were grown with a constant sparge of 0.22 μm filtered air at 37 °C until OD 0.6–0.8 (approximately 5 h). After ∼5 h, 1.2 mM cysteine was added to the cultures (along with 0.2% l-arabinose for the pDB1282 plasmid to induce expression of Fe–S cluster biogenesis), the temperature was reduced to 22 °C, and the cultures were incubated for ∼20 h before being harvested by centrifugation at 10,000 × g. The resulting E. coli pellets were then flash-frozen and stored in liquid N₂ until purification. The resulting cell pellet was resuspended in approximately 30 mL of lysis buffer containing 50 mM HEPES, pH 7.5, 300 mM KCl, 5% glycerol, 5 mM imidazole, 14.3 mM 2-mercaptoethanol (BME), and 0.5% Triton x305 detergent. During sonication, cells were cooled in an ice bath while being subjected to cycles of 3 s on and 20 s off, for a total of 30 min of sonic disruption (80% output). The resulting cell lysate was then centrifuged for 1 h at 15,000 × g at 4 °C. Protein purification was performed with the ÄKTA express FPLC system, anaerobically in an MBraun anaerobic chamber (<1.2 ppm of O₂). The resulting supernatant was passed over a 5 mL Fast-Flow Ni-Sepharose column (GE Biosciences) that was equilibrated in lysis buffer (without detergent). After loading, the column was washed with 20 column volumes of lysis buffer (without detergent) and an on-column reconstitution was performed where the column was washed with a buffer containing 600 μM FeCl₃, 400 μM Na₂S, 57 mM BME, 100 mM HEPES pH 7.5, 300 mM KCl and 10% glycerol. Reconstituted BioB was eluted with 2 column volumes of elution buffer (lysis buffer without detergent and with 300 mM imidazole). The resulting eluent with high UV 280 nm absorbance was collected and loaded onto a HiPrep 16/60 S-200 size-exclusion column equilibrated with 20 mM HEPES, pH 7.5, 300 mM KCl, 10% glycerol, and 5 mM DTT. The resulting fractions displaying a high UV 280 nm absorbance and brown color were pooled and concentrated using an Amicon centrifugal filter unit (10,000 M.W. cut off). The final purified protein concentration was estimated by UV–visible (A₂₈₀) using the extinction coefficient calculated based on the BioB amino acid sequence (Expasy ProtParam). The final concentration of boBioB and ecBioB for in vitro assays were determined by utilizing published extinction coefficients of ecBiob containing two 4Fe–4S clusters (ε400 = 30,000 M^–1 cm^–1) and a mixed state of one 4Fe–4S cluster and one 2Fe–2S cluster (ε410 = 14,900 M^–1 cm^–1), respectively.³¹ HpMTAN was a gift from the laboratory of Vern Schramm, Ph.D. at the Albert Einstein College of Medicine, and was expressed and purified as described previously.⁶¹

In Vitro Biotin Synthase Enzyme Assay

Biotin synthase enzyme assays were performed at 25 °C, anaerobically in an MBraun anaerobic chamber (<0.1 ppm of O₂). The reaction mixture consisted of 31.7 μM E. coli or 17.5 μM B. obeum biotin synthase, 25 μM E. coli Flavodoxin (FldA), 5 μM E. coli Ferredoxin:NADP⁺ oxidoreductase (Fpr), 1 mM NADPH, 2 mM desthiobiotin (Sigma), and 10 μM H. pylori 5′-methylthioadenosine/S-adenosylhomocysteine nucleosidase (MTAN) in 50 mM HEPES and 100 mM KCL. The enzyme reaction was initiated with the addition of SAM to a 1 mM final concentration, and after 0, 2, 5, 10, 30, 90, 120, and 180 min, 20 μL aliquots of the reaction were quenched with 20 μL of a quench buffer containing 100 mM H₂SO₄, 200 μM Tryptophan, and 20 mM DTT. After completion of the assay, samples were centrifuged at 4,000 × g at 4 °C for 20 min, and supernatants were used for LC-MS quantification of biotin, using the 100 μM tryptophan present from quenching, as the internal standard. A standard curve of d-biotin (Sigma) was utilized for quantification. For boBioB variant assays, the same conditions as above were utilized, except these assays all contained 1.25 mM Na₂S, and the concentrations for boBioB S140A and H161A were 13.1 and 14.4 μM, respectively. All assays were performed in triplicate unless otherwise stated. To monitor the decrease of the FeS-absorbing region over time, the same master mix utilized for the in vitro assays was placed in an airtight cuvette with a hanging drop of SAM suspended on the lid. Singular assays were initiated by the gentle mixing of the cuvette. Continuous U/V spectra were collected every 2 min for 180 min at room temperature on an Agilent Cary 60 UV–vis spectrophotometer.

LC/MS Detection of Metabolites from in Vitro Enzymatic Assays

High-performance liquid chromatography was performed on an Agilent Technologies 1200 system (Santa Clara, CA) coupled to an Agilent Technologies 6490 triple quadrupole Mass Spectrometer. The resulting quenched assay mixture was separated on either an Agilent SB-C18 RRHT column (2.1 mm × 50 mm, 1.8 μm particle size; Figure 3) or an Agilent Eclipse XDB-C18 RRHD column (2.1 mm × 50 mm, 1.8 μm particle size; Figure S14), which was equilibrated with 99.2% Buffer A and 0.8% Buffer B (A: 0.1% formic acid, B: 100% acetonitrile). Metabolite separation was achieved by utilizing a gradient of 99.2–80% A from 0 to 2 min, 80–40% A from 2 to 2.5 min, maintaining at 40% A from 2.5 to 2.75 min, and returning to 99.2%A from 2.75 to 3 min. Detection of d-Biotin and l-tryptophan was performed with electrospray ionization in positive mode with multiple reaction monitoring (Table S6). Mass Hunter software (Agilent) was used for data collection and analysis.

Protein Crystallization, X-ray Data Collection, and Structure Determination

Heterologously expressed and purified BioB proteins were the subject of sparse matrix screening crystallization trials utilizing the sitting drop vapor diffusion method as previously described.⁵⁷Blautia obeum (A5ZUL4) and Veillonella parvula (T0TAB9) formed hexagonal rodlike crystals in the presence of SAM and DTB and diffracted well in a synchrotron X-ray source. The structures were solved using single-wavelength anomalous dispersion by collecting data at a wavelength of 1.378 Å, taking advantage of Fe absorption at this wavelength. The published boBioB structure in this paper was solved by molecular replacement using a boBioB structure previously determined (phase solved by Fe single-wavelength anomalous dispersion, not described herein). Diffraction data were collected at the Advanced Photon Source (Argonne National Laboratory, Argonne, IL). Sulfur single-wavelength anomalous diffraction data was collected at the National Synchrotron Light Source II Beamline 17-ID-1 (Brookhaven National Laboratory, Upton, NY) at a wavelength of 1.907 Å (6.5 keV). Native boBioB and vpBioB data were collected from National Synchrotron Light Source II Beamline 17-ID-2 (Brookhaven National Laboratory, Upton, NY; Table S1). boBioB crystals were incubated with TEV protease before crystallization. Crystals diffracted to generate the boBioB native data set were formed by drops of 0.4 μL of [10 mg/mL in 100 mM HEPES (pH 7.5), 1 mM SAM, 1 mM DTB, 1 μg/mL TEV protease] protein mixed with 0.4 μL of precipitant (0.1 M Bis-Tris pH 5.5, 25% (w/v)PEG 3350, 0.2 M NaCl) and equilibrated against a solution of 0.5 M LiCl. Crystals diffracted to generate the boBioB sulfur anomalous data set were formed by drops of 0.4 μL of [10 mg/mL in 100 mM HEPES (pH 7.5), 1 mM SAM, 1 mM DTB, 1 μg/mL TEV protease] protein mixed with 0.4 μL of precipitant (0.1 M Bis-Tris pH 5.5, 17% (w/v)PEG 10K, 0.1 M NH₄OAc) and equilibrated against a solution of 0.5 M LiCl. Crystals diffracted to generate the vpBioB native and anomalous data sets were formed by drops of 0.4 μL [30 mg/mL in 25 mM HEPES (pH 7.5), 0.7 mM SAM, 2 mM DTB] protein mixed with 0.4 μL of precipitant (0.1 Bis-Tris pH 5.5, 0.2 M NH₄OAc, 25% (w/v)PEG 3350) and equilibrated against a solution of 0.5 M LiCl.

Structures were solved utilizing the programs Coot, CCP4 (refmac refinement), and Phenix.⁶² The boBioB native data set contained density data that were characteristic of both SAM and reductively cleaved SAM (5′-dA and methionine). Due to poor fitting of 5′-dA and methionine, SAM was fit to the density. Coordinates and structure factors of the solved B. obeum and V. parvula structures have been deposited in the Protein Data Bank with the PDB accession codes: 8VCW and 8VDW, respectively.

Cloning of Heterologous Biotin Synthases for in Vivo Study

BioB genes were cloned using Uracil-Specific Excision Reagent (USER) cloning with the Phusion U Hot Start Polymerase (Thermo) as previously published.⁶³ Genes of interest were amplified with oligomers containing USER overhangs compatible with cloning behind an IPTG-inducible T5lacO promoter in the pBS679-based plasmid. In some cases, with an additional degenerate nucleotide sequence or a defined library of sequences, in the overhang as a ribosomal binding site library.⁶⁴ Expression libraries were included to ensure that a wide expression space was sampled and that we did not miss complementing/function heterologous biotin synthases possibly caused by insufficient or too high/toxic transcription. All DNA sources and primers for BioB and plasmid backbone amplification are listed in Table S7 and Table S8. After USER-cloning, constructs were treated with T4 ligase (Thermo) according to manufacturer protocol and transformed into BS1575 by standard electroporation. Genes indicated as codon-optimized were done so with the IDT codon optimization tool.

In Vivo Characterization of Type II BioB

All complementation testing and production of biotin were done in the biotin auxotroph: BS1575 (ΔbioA, ΔbioB, ΔbioF, ΔbioD, ΔbioH, iscR H107Y). It should be noted that this strain contains a point mutation in the DNA-binding transcriptional dual regulator IscR that improves BioB performance.²¹ Plasmids containing a heterologous biotin synthase expression plasmid were transformed into said strain and washed three times with minimal media without biotin (mMOPS; Table S9) after rescue in SOC, to remove biotin contamination. mMOPS was prepared as previously published.²¹ The transformants were then plated on mMOPS 1.5% agar plates, supplemented with 100 ng/mL ampicillin and 10 μg/mL DTB.

For small-scale cell factory production of biotin from DTB, 3 replicates (single colonies) of the strain of interest were picked from transformation plates for preculturing in 400 μL of mMOPS with necessary antibiotics (ampicillin, 100 μg/mL) in 96-well deep-well plates (DWPs). 10 nM biotin was added to the preculture for oxygen-sensitive Type II BioBs to overcome auxotrophy during aerobic cultivation. Precultures were incubated overnight at 37 °C with shaking (250 rpm) in an Innova 44. Precultures were used to inoculate 800 μL of the production mFIT media (Table S10), with ampicillin, 10 nM biotin, and 50 μM DTB in 96-well DWPs, diluting precultures 100-fold to an initial OD600 ∼0.01. DWPs were sealed with airtight aluminum seals. This, coupled with a smaller headspace due to high fill volume (800 μL) and high glucose concentrations (up to 10 g/L), led to rapid depletion of oxygen by respiring E. coli and microaerobic conditions in the majority of the production culture incubation time. mFIT is highly buffered and can better sustain the organic acid byproducts generated during the microaerobic growth of E. coli. Production cultures were incubated for 24 h at 37 °C with shaking (250 rpm). OD600 of production cultures was measured at the end of incubation, cells were spun at 5,000 × g for 5 min, and supernatants were transferred to detect biotin concentrations by biotin bioassay as previously published.²¹

Fed-Batch Bioreactor Production of Biotin

A single colony of the indicated strain (Table S4) was inoculated into 50 mL of mMOPS with relevant antibiotics and grown to full cell density (cuvette OD600 ∼2) overnight in a 250 mL baffled shake-flask. 200 mL of Fermentation Batch medium (Table S11) supplemented with antibiotics and 1 mL of a 1% (v/v) antifoam solution were added to Applikon 500 mL MiniBio Reactors with the temperature set to 37 °C, pH-controlled to pH 7 by addition of 5 M NH₄OH, and dissolved oxygen (DO) set-point to DO = 15% by agitation speed and air sparge for aerobic fermentation. For microaerobic fermentation, a DO = 0% set-point was used. 10 mL of each strain preculture was used to inoculate reactors, resulting in a starting cuvette OD600 = 0.10 for all reactors and strains/conditions. Once the CO₂ in the outlet gas reached greater than 0.4%, ∼4–6 h into the exponential batch phase, each fermentation was induced by the addition of IPTG at the indicated concentrations to induce expression of E. coli or B. obeum BioB. Following the depletion of glucose in the medium, as seen by a drop in CO₂, a fed-batch phase was initiated by the addition of the Feed Medium (Table S12) to each fermentation at a constant feed rate of 0.6 mL/hour. This transition took place between ∼8–10 h. The DO control continued to operate at DO = 15%, with the agitation increasing in all reactors to between 1250 and 1800 rpm. An additional 1 mL of 1% (v/v) antifoam solution was added at 7 and 24 h. The fermentations were terminated at the time indicated by the last sampling, ∼48 h. Culture samples were taken from the fermenter at up to five various time points after inoculation, as indicated in the fermentation profile(s). From these samples, optical density (cuvette OD600) was measured after diluting samples into the range of 0.2–2 OD. Supernatants were obtained by spinning biomass down in a microcentrifuge at 13,000 × g for 1 min. Supernatants were diluted 10-fold and stored at −20 °C for later quantification of biotin by LC-MS. The final time-point sample biotin quantification is indicated in Table S4. Glucose feed bottles were weighed before and after fermentation to accurately calculate the total amount of glucose fed. From these data, the LA yield per OD or glucose was calculated.

SDS-Page of Whole-Cell Lysates

Samples for SDS-page were taken during fed-batch fermentation, at the 24 h time point, approximately 20 h after the expression of boBioB was induced with IPTG from the plasmids/strains indicated in Table S13. Samples were normalized, corresponding to 1 mL of 1 OD culture broth. Samples were centrifuged, supernatants were removed, and pellets were resuspended in 100 μL Millipore Bugbuster 1x and incubated for 1 h at room temperature. Samples were spun down, and supernatants, the soluble protein fractions, were removed. The supernatants were centrifuged at 5 min at 13,000 × g and loaded in Bio-Rad Mini-PROTEAN TGX precast gels along with Bio-Rad Precision Plus Protein Dual Color Standards. Gels were run at 200 V for 40–60 min in Bio-Rad Tris/Glycine/SDS buffer. Gels were stained with EZBlue Gel Staining Reagent (Coomassie) according to manufacturer protocol until bands were visible.

LC-MS Quantification of d-Biotin from Biological Production

The internal standard d4-biotin (d4-BTN, ISTD) was purchased from Sigma-Aldrich. Stock solutions of the analytes and internal standards were prepared in DMSO to a concentration of 1 mg/mL. Working standard solutions of the stock solutions were then prepared in H₂O. Calibration curves in concentrations of 0.1, 0.25, 0.5, 1, 2.5, 5, 10, 25, 50, 100, 250, 500, and 1000 ng/mL were prepared in H2O containing 5 ng/mL d4-BTN. Before analysis, samples were diluted, and the ISTD was added to correct for possible technical variation. For the quantification of BTN and DTB, a two-step dilution pattern was followed to achieve a 1:1500 dilution of the original sample. First, a 1:15 dilution was created by pipetting 980 mL of H₂O and 70 mL of the original sample into an Eppendorf tube. The solution was vortex mixed. Finally, 980 mL of H₂O, 10 mL of the ISTD, and 10 mL of the 1:15 diluted sample were pipetted into a glass vial and the solution was vortex mixed. The samples were randomized after sample preparation and analyzed along with long-term quality control samples by ultrahigh performance liquid chromatography (Infinity II, Agilent Technologies) coupled to tandem mass spectrometry (6470 Triple Quadrupole, Agilent Technologies) using electrospray ionization in positive ion mode. Selected reaction monitoring was used for quantifying the analytes, and fragmentor voltages, collision energies, and cell accelerator voltages were optimized for each ion transition. The analytes were separated chromatographically before they entered the mass spectrometer. This was done using an ACQUITY UPLC HSS T3 Column (2.1 mm × 100 mm, particle size 1.8 mm, Waters Corporation) and H₂O + 0.1% (v/v) CH₃COOH as eluent A and ACN + 0.1% (v/v) CH₃COOH as eluent B with a flow rate of 0.4 mL min^–1. The elution gradient is as follows: 0–0.5 min 0% B, 0.5–1.5 min 0% to 15% B, 1.5–3 min 15% B, 3–5 min 15% to 100% B, 5–7 min 100% B. After each run, the column was re-equilibrated at 0% B for 2 min. The injection volume for each sample was 5 mL. All data were acquired and processed using the MassHunter Quantitative Analysis software (Version B.09.00, Build 9.0.647.0). The peak areas for the analytes of interest were normalized against the peak areas of the corresponding internal standards. The quality control samples were carefully monitored, and the data from the sample set were approved only when the results from these samples remained within ±10% from the long-term average concentration.

Supporting Information Available

The Supporting Information is available free of charge at https://pubs.acs.org/doi/10.1021/jacs.3c05481.

• Additional detail on the bioinformatics, crystallography, structural characterization, enzyme activities, biocatalysis, genetic constructs, and recipes in this study. (PDF)

Author Contributions

^∥ J.C.L. and D.L.-H. contributed equally to this work.

The authors declare the following competing financial interest(s): The Albert Einstein College of Medicine and Biosyntia ApS have both filed patent applications covering Type II BioB. Biosyntia ApS is engaged in the development and commercialization of E. coli cell factories to produce vitamins, including biotin.