A Conserved and Seemingly Redundant Escherichia coli Biotin Biosynthesis Gene Expressed Only During Anaerobic Growth

Summary

Biotin is an essential metabolic cofactor and de novo biotin biosynthetic pathways are widespread in microorganisms and plants. Biotin synthetic genes are generally found clustered into bio operons to facilitate tight regulation since biotin synthesis is a metabolically expensive process. Dethiobiotin synthetase (DTBS) catalyzes the penultimate step of biotin biosynthesis, the formation of 7,8-diaminononanoate (DAPA). In Escherichia coli, DTBS is encoded by the bio operon gene bioD. Several studies have reported transcriptional activation of ynfK a gene of unknown function, under anaerobic conditions. Alignments of YnfK with BioD have led to suggestions that YnfK has DTBS activity. We report that YnfK is a functional DTBS, although an enzyme of poor activity that is poorly expressed. Supplementation of growth medium with DAPA or substitution of BioD active site residues for the corresponding YnfK residues greatly improved the DTBS activity of YnfK. We confirmed that FNR activates transcriptional level of ynfK during anaerobic growth and identified the FNR binding site of ynfK. The ynfK gene is well conserved in γ-proteobacteria.

Abbreviated summary

The YnfK protein was precited to catalyze the dethiobiotin synthase. We have shown that this is the case. The close relationship with the well-studied BioD dethiobiotin synthase is shown by the superimposition of the AlphaFold model (in blue) of YnfK with the known crystal structure of BioD (in yellow). Single residue mutations made in YnfKare shown in pink.

Introduction

Biotin, also known as vitamin H, is an essential cofactor required by all forms of life (Sirithanakorn and Cronan, 2021). De novo biotin biosynthetic pathways are widespread in microorganisms, fungi and plants, but are absent in mammals and birds (Rodionov et al., 2002; Feng et al., 2015; Zeng et al., 2020; Woong Park et al., 2011). Biotin synthesis is an energy-consuming process, requiring 20 equivalents of ATP in the E. coli pathway and at least 5 enzymes are required to assemble a molecule of biotin (Feng et al., 2013). Hence tight regulation is required to avoid costly overproduction.

Bacterial biotin synthetic genes are generally found clustered on the genome and organized in operons. A bioWAFDBI operon is found in Bacillus subtilis (Perkins et al., 1996) whereas in Escherichia. coli, the biotin synthetic genes bioBFCD and bioA are divergently transcribed under control of the bioO operator that controls both promoters (Sirithanakorn and Cronan, 2021). These operon organizations facilitate negative regulation by the bifunctional biotin protein ligase, BirA. When biotin is in excess it is converted by BirA to biotinoyl-5´-AMP. The BirA-biotinoyl-5´-AMP complex dimerizes and binds to bioO to repress transcription of the operon genes (Sirithanakorn and Cronan, 2021).

Dethiobiotin synthetase (DTBS) catalyzes the penultimate step of the biotin biosynthesis: the formation of the ureido ring of dethiobiotin (DTB) from (7R, 8S)-7,8-diaminopelargonic acid (DAPA, formal name 7,8-diaminononanoate), CO₂ and ATP (Fig. 1). Three steps are involved in the DTBS reaction: First, CO₂ is consumed in formation of N7-carbamate of DAPA, the carbamate moiety is then activated by ATP to give carbamic-phosphoric acid anhydride and finally attack of the anhydride by the DAPA N8 amino group results in closure of the ureido ring with release of inorganic phosphate and ADP (Gibson et al. 1997). The mechanism is essentially that proposed by Krell and Eisenberg in 1970 (Krell and Eisenberg, 1970) and has been conclusively established by kinetic crystallography (Käck et al., 1998).

Fig 1.

The reaction scheme of the BioD dethiobiotin synthetase (Krell and Eisenberg, 1970) (Käck et al., 1998). It remains unclear if BioD catalyzes formation of the N7 carbamate or selects the N7 carbamate from the mixture of N7 carbamate, N8 carbamate and the dicarbamate that spontaneously forms in solution (Gibson et al., 1995).

Several laboratories have reported that transcription of E. coli ynfK, a gene encoding a putative BioD paralog, (BioD and YnfK are 51% identical, Fig. 2) is expressed only in anerobic cultures and its expression requires the FNR transcription factor (Constantinidou et al., 2006; Myers et al., 2013). Upon shift to anaerobiosis transcription of ynfK increased 5- to 9-fold depending on the method used to measure transcription (microarray versus RNA-Seq) and the growth media used (Constantinidou et al., 2006; Myers et al., 2013; Kang et al., 2005). In Salmonella enterica serovar Typhimurium (hereafter S. enterica) a similar FNR-dependent increase in ynfK transcription (8.4-fold) was reported upon shift to anaerobic conditions in a microarray study (Fink et al., 2007) and YnfK protein levels were reported to decrease by 10-fold in anaerobic cultures upon deletion of fnr (Wang et al., 2019). These data are supported by chromatin immunoprecipitation experiments that detected binding of FNR to a region upstream of the E. coli ynfK coding sequence (Grainger et al., 2007). These data plus their high amino acid sequence identity raised the possibility that YnfK might be an anaerobically expressed counterpart of BioD. Unlike bioD, the genome location of ynfK is far removed from the biotin operon. The riddle is why E. coli and even rather distant γ-proteobacterial relatives have retained ynfK. The sole E. coli biotinylated protein, AccB, is required for fatty acid (hence membrane) synthesis (Cronan, 2021) and E. coli does not require biotin when grown anaerobically on minimal media indicating the pathway is functional. The BioD reaction requires only CO₂ (readily generated from the carbon source) plus DAPA and ATP. Is YnfK a dethiobiotin synthase?

Fig. 2. Sequence alignment BioD and YnfK of E. coli.

Unweighted sequence alignments was performed using T-Coffee with the default settings and displayed using Jalview. Positions having 30% or greater are highlighted. Residues where residue substitution were made are indicated by arrows. The sites denoted are those of BioD.

Results

YnfK is a functional dethiobiotin synthetase

We first tested anaerobic growth of a ΔbioD strain on minimal medium and observed very weak growth (Fig. 3A). However, the growth observed was dependent on YnfK since a ΔbioD ΔynfK strain failed to grow anaerobically unless supplemented with biotin. A further test of the annotations of YnfK as having DTBS activity was to express the protein from plasmids derived from the arabinose inducible pBAD322 vector in a ΔbioD Δynfk double deletion strain. The strain carrying the ynfK encoding plasmid, however, showed only weak growth under anaerobic conditions after 3 days of incubation at 37°C. In attempts to improve the expression of ynfK, we expressed the ynfK gene using three different vector promoters the pQE2 vector carrying the powerful T5 promoter, vector pBAD33 carrying the 300 bp chromosomal fragment upstream of bioB (pBAD33-pbioB300) and the pBAD33 vector carrying a 300 bp chromosomal fragment upstream of ynfK (pBAD33-Pynfk300). The various promoter plasmids encoding ynfK were transformed into the E. coli MG1655 ΔynfK ΔbioD strain and the resulting strains tested for growth aerobically and anaerobically after incubation for 3 days at 37°C. However, these strains still showed only weak growth (Fig. S1). We tested YnfK expression levels in vivo relative to BioD using hexahistidine-tagged versions of YnfK and BioD in vector pQE2 and found that YnfK levels were 4-fold lower than BioD levels (Fig. S2) in agreement with the ribosome profiling data to be discussed below.

Fig. 3. Anaerobic growth of E. coli mutant strains.

A. The strains were streaked on M9 minimal medium containing 2 nM biotin, starved by avidin addition (avidin is a very robust protein that binds biotin very tightly with a Kd ~ 10⁻¹⁵M) and then restreaked on M9 biotin-free minimal medium and incubated at 37°C under anaerobic conditions. Panels A and B show anaerobic growth of E. coli DTBS null mutant strains and anaerobic growth of E. coli mutant strains with DAPA supplementation, respectively. The black hazy color is cleaved S-Gal (3,4-cyclohexenoesculetin β-D-galactopyranoside, a substrate for β-galactosidase, that unlike the commonly used X-Gal (5-bromo-4-chloro-indolyl-β-D-galactopyranoside), is chromogenic under anerobic conditions. S-Gal was added to facilitate detection of faint growth.

The low YnfK expression level could explain at least in part the low activity observed in complementation assays. However, it remained possible that YnfK was also a poorly active DTBS. The DTBS activity of YnfK was measured using a spectrophotometric assay. For kinetic studies with purified YnfK and BioD, DTBS activity was measured by monitoring ADP formation, using a coupled pyruvate kinase/lactate dehydrogenase assay (Gibson et al., 1995; Yang et al., 1997). The disappearance of NADH was continuously monitored at 340 nm. At saturating levels of CO₂, initial velocity kinetic techniques were used to measure K_M values of the enzymes for both ATP and DAPA. Compared to BioD, YnfK has at least 2-fold higher K_M for DAPA and a 2-fold higher K_M for ATP (Table 1), indicating that YnfK is less enzymatically competent than BioD, although it has a respectable in vitro activity for a biotin synthetic enzyme. The higher K_M of YnfK for DAPA may explain the very slow anaerobic growth of ΔbioD strains since DAPA synthesis limits YnfK activity in vivo (as shown below).

Table. 1.

Kinetic Parameters for BioD, YnfK and the YnfK derivative proteins (pH 7.5, 25°C) measured by the spectrophotometric assay.

DTBS Activity	BioD	YnfK	S13E	T114A	I184Y	S13E/I184Y	T114A+I184Y
KM (DAPA) (μM)	0.25	0.50	0.20	0.15	0.16	0.20	0.21
Vmax (DAPA) (μM·min−1)	11.25	11.74	17.44	27.97	20.98	17.20	17.85
kcat (DAPA) (min−1)	2.16	1.96	4.85	5.38	3.75	3.91	4.06
kcat/KM (DAPA) (min−1·μM−1)	8.66	3.91	24.23	35.86	23.42	19.55	19.31
KM (ATP) (μM)	0.17	0.34	0.10	0.15	0.13	0.18	0.22
Vmax (ATP) ((μM·min−1)	9.09	8.63	11.98	11.33	11.09	8.20	12.06
kcat (ATP) (min−1)	1.75	1.44	3.33	2.18	1.98	1.86	2.74
kcat/ KM (ATP) (min−1·−μM1)	10.29	4.23	32.62	14.15	15.36	10.35	12.63

The estimated errors are ± 15%.

Substitution of YnfK with BioD active site residues improves YnfK enzymatic activity

Sequence alignments of DTBS from five different bacterial species (E. coli, Serratia marcescens, Bacillus sphaericus, Brevibacterium flavum and Mycobacterium leprae) showed 27 conserved residues, all of which are involved in substrate and metal ion binding (Schneider and Lindqvist, 1997). In crystallographic studies of E. coli DTBS-substrate complexes five active site residues (T11, E12, K15, K37 and S41) were reported to function in catalysis and play important roles in substrate binding (Huang et al., 1995). Three of the five residues, T11, K15 and K37 are conserved in all BioD DTBS sequences reported to date and in YnfK. The positively charged side chains of the two lysine residues and the hydroxyl side chains of T11 and S41were proposed to play important roles in substrate binding and stabilization of reaction intermediates. S41 is conservatively replaced with threonine in some sequences (Yang et al., 1997) but is replaced with lysine in YnfK. E12 was assigned as a catalytic acid in the phosphate transfer step and/or a general base in the ring closure steps (Yang et al., 1997). Unlike the other four residues, E12 is less conserved among DTBS sequences. In a previous study, E12A showed a 4-fold decrease in the K_M for DAPA but retained a normal K_M for ATP whereas E12D had a K_M for DAPA similar to that of the YnfK wild type enzyme, but a 6-fold decrease in the K_M for ATP (Yang et al., 1997). Mutagenesis also indicated a possible role for E12 in synergistic binding between DAPA and ATP (Yang et al., 1997). In YnfK E12 is replaced with serine.

Considering the important role of E12 and the distinct chemical properties of Glu and Ser, we constructed the S13E mutant protein in hopes of increasing the enzymatic activity of YnfK. The functionality of the S13E protein was tested in the ΔbioD ΔynfK strain complementation assay. Vectors expressing the S13E protein were transformed into the E. coli MG1655 Δynfk ΔbioD strain and the plates were incubated at 37°C either aerobically or anaerobically for 20 h. Strains expressing the S13E enzyme gave better anaerobic growth than strains carrying wild type YnfK (Fig. 4). We then purified the S13E enzyme and performed kinetic assays to obtain the K_M values for ATP and DAPA. The S13E YnfK protein showed enhanced substrate binding towards both ATP and DAPA, reaching that of BioD (Table 1).

Fig. 4. Complementation of the ΔynfK ΔbioD strain with plasmids encoding YnfK derivatives having BioD active site residues.

The proteins were expressed from the PbioB300 promoter in vector pBAD33. The biotin-free plates were incubated under anaerobic conditions for 20 h and contained S-Gal. A more detailed analysis of growth rates is given in Fig. S3.

E. coli DTBS functions only as a homodimer because DAPA binds at the interfaces between the two subunits and forms hydrogen bonds with residues from both homodimer subunits (Yang et al., 1997). In E. coli BioD, Y187 played an important role in binding DAPA. One of the DAPA carboxyl oxygens forms a hydrogen bond to the hydroxy group of Y187. We constructed a YnfK I184Y enzyme to test if the enzymatic activity of YnfK was improved. Binding of DAPA induces local conformational changes, mainly confined to loop regions (Huang et al., 1995).

In BioD, residues G117-A118-Gl19 participate in DAPA binding and interaction with the phosphate-binding loop (Huang et al., 1995) Yang et al., 1997). In YnfK, however, the non-polar A118 is replaced by T118 and hence, we constructed the T114A enzyme. The first test of the function of the mutant I184Y and T114A enzymes was the complementation assay. Strains carrying I184Y or T114A gave better growth (colonies appeared more rapidly) than the strain carrying wild type YnfK (Figs. 2B and S3). We then purified the two altered proteins and performed kinetic assays. Both I184Y and T114A proteins showed enhanced substrate binding towards both ATP and DAPA, reaching that of BioD (Table. 1).

Double substitutions of active site residues were constructed by combining the single site substitutions constructed earlier and complementation assays were performed. Two of the three double substitution proteins, S13E/I184Y and T114A/I184Y, gave better growth of the ΔynfK ΔbioD strain than wild type YnfK whereas the S13E/T114A protein gave similar but somewhat slower growth than the wild type YnfK protein (Fig. 4). The two double substitution (S13/I184Y and T114/I184Y) were purified and steady state kinetic assays were carried out. Reduced K_M values for DAPA were observed in the double substitution proteins, although not as low as that of wild type BioD. The T114/I184Y protein showed a slightly lower K_M for ATP than wild type YnfK whereas the S13E/I184Y protein showed a higher K_M than the wild type YnfK (Table 1). Combined with the complementation assay results, all three single point mutations showed enhanced enzymatic performance compared with wild type YnfK whereas the two double substitution proteins showed only modestly improved enzymatic performance.

The oligomerization states of BioD, YnfK and the YnfK active site substitution proteins

As noted above E. coli DTBS functions as a homodimer. We tested the ability of BioD, YnfK and the purified mutant proteins to dimerize using chemical cross-linking with ethylene glycol bis(succinimidyl succinate (EGS). Nearly all BioD molecules became crosslinked in the presence of 2 mM EGS whereas only 41 % of YnfK molecules became crosslinked under the same conditions (Fig. 5A), indicating that YnfK dimerization was weaker than that of BioD. Moreover, the YnfK dimer bands seemed broader and more diffuse than those of BioD perhaps suggesting “slippage” during the crosslinking reaction due to a weaker interface. Markedly better crosslinking was seen in the presence of DAPA (Fig. 5A) (DAPA was added to YnfK and incubated 5 min prior to addition of EGS because otherwise EGS would react with the amino groups of DAPA and inactivate the substrate). BioD dimerization is required to bind DAPA and thus poor crosslinking is consistent with the low activity of YnfK.

Fig. 5. EGS cross-linking analysis of the oligomerization state of BioD, YnfK, and the YnfK residue substitution proteins.

M: Protein standard markers. The dimer and monomer (mono) are indicated by arrows. A. EGS cross-linking analysis of BioD and YnfK. Increased YnfK crosslinking in the presence of DAPA is also shown. B. EGS cross-linking analysis of YnfK and the YnfK active site substitution proteins. The protein standards (in KDa) are 95, 72, 55 (heavy band), 43, 34, 26, and 17. Note that the YnfK crosslinking of gel A in the absence of DAPA could not be quantitated by scanning due to overloading. We therefore repeated the YnfK crosslinking with 2 mM EGS and loaded increasing amounts on a single gel (Fig S4) and scanned several lanes in the linear range with a GE Typhoon FLA7000 scanner in the fluorescent shadow mode and found that only 41% of YnfK was crosslinked.

EGS crosslinking of all the YnfK-derived mutant proteins (Fig. 5B) was also tested. The purified proteins were incubated with indicated concentrations of EGS and separated on 4–20% gradient SDS-PAGE gel. The I184Y, T114A/I184Y and S13E/T114A proteins showed higher levels of dimerization compared with YnfK at 2 mM EGS and the dimer bands were sharper than those of YnfK. However, the S13E and T114A proteins showed only modestly improved crosslinking with 2 mM EGS (Fig. 5B). Note that the I184Y, T114A/I184Y and S13E/I184Y proteins showed less diffuse dimer bands than YnfK consistent with their greater activity.

Addition of DAPA to the growth medium markedly improved YnfK function

The enzyme that synthesizes DAPA, BioA (S-adenosylmethionine-8-amino-7-oxononanoate transaminase) is a remarkably poor enzyme. The literature gives two turnover values that differ by 10-fold (Eliot et al., 2002; Stoner and Eisenberg, 1975). However, the higher value is only 17 molecules/min/molecule of BioA (k_cat of 0.13 s⁻¹) (Stoner and Eisenberg, 1975). Moreover, ribosome profiling shows that BioA is expressed at very low levels (259 copies/cell in a wild type strain) (Li et al., 2014). The level should increase about five-fold under biotin starvation conditions, but this would remain 2 to 3-fold lower than BioD, a more efficient enzyme. Although bioA is part of the classical E. coli bio operon under control by the same operator as bioABFCD, it is divergently transcribed. The overlapping face-to-face promoters show similar amplitudes of regulation although the bioA promoter seems 2- to 3-fold weaker than the bioBFCD promoter (Barker and Campbell, 1980; Cronan, 1988). This seems ill-advised since a stronger bioA promoter would seem advantageous. We tested our prediction that YnfK activity is limited by the supply of DAPA. DAPA is known to enter E. coli where it bypasses loss of BioA (Rolfe and Eisenberg, 1968) (Fig. 3B) and thus we added DAPA to the biotin-free medium and observed greatly improved anaerobic growth of a ΔbioD strain (Fig. 3B). Hence, YnfK is starved for DAPA, yet another limitation of its ability to make DTB.

FNR directly activates transcription of ynfK

FNR is a global regulator of bacterial anaerobiosis and the ynfK FNR binding site was found about 90 bp upstream of the coding sequence. Wild type FNR is inactive under aerobic conditions, thus the aerobically stable and active FNR D154A was constructed (Shan et al., 2012). Prior size exclusion chromatography data (Moore et al., 2006) (Lazazzera, Bet, al., 1993) reported that the FNR D154A protein behaved as a dimer in size exclusion chromatography. We repeated this experiment and obtained the same result, a value of about 64.9 kDa, slightly larger than the value expected from two monomers of 27.9 kDa (data not shown) that can probably be attributed to asymmetric shape of the FNR dimer (Volbeda et al., 2015).

Expression of the FNR D154A protein resulted in increased transcription of ynfK as measured using a β-galactosidase reporter. A DNA fragment containing the 196 bp upstream of ynfK was ligated into pBAD33 carrying the lacZY genes and the plasmid transformed into E. coli MG1655 ΔlacZY derivative either lacking FNR (Δfnr) or with FNR overexpression from a plasmid. A nearly five-fold increase was observed in strains over-expressing FNR D154A whereas a three-fold increase was observed in strains over-expressing wild type FNR (Fig. 6A). A nearly 1.5-fold decrease in signal was observed in the strain lacking FNR. These observations confirm that FNR is involved in the activation of ynfK transcription.

Fig. 6. FNR directly activates YnfK transcription.

A. Anaerobic expression of ynfK-lacZY fusions strain grown anaerobically. The strains were grown in defined medium supplemented with 500 μM IPTG for 6 at 37°C. The values are given above the bars and the results are the average of three independent experiments (the error bars denote standard error of the mean). The strain background was a ΔlacZY derivative of E. coli MG1655 and the fusion was ynfK196-lacZY carried by plasmid pBAD33. The wild type (wt) and Δfnr strains also carried the empty pQE2 vector whereas the FNR and FNRD154A proteins were expressed from the pQE2 vector.

B. Electrophoretic mobility shift assays of DNA binding by FNR.

EMSA showed that FNR binds to a site upstream of gene ynfK. A C. acetobutylicum bioY fragment was used as negative control (Song et al., 2021). Gel shifts are indicated by the arrows.

C. Sequence alignment of E. coli FNR binding site. Unweighted sequence alignments were performed using T-Coffee with the default settings and displayed using Jalview. Positions having 50% or greater conservation with the consensus sequence are highlighted. E. coli FNR, FNR consensus binding site; E. coli ynfK FNR, FNR binding site of E. coli ynfK. The 3’ end of the ynfK binding site is 85 bp upstream of the coding sequence.

To assess the binding affinity of FNR towards the deduced binding site in vitro, we performed electrophoretic mobility shift assays (EMSAs). The indicated concentrations of purified FNR D154A was incubated with a 125 bp DNA fragment containing the FNR binding site centered 90 bp upstream of the ynfK gene. A gel shift was observed with 100 nM FNR D154A protein and a nearly complete gel shift was observed when FNR D154A 1 mM was added (Fig. 6B). The BirÁ binding site of Clostridium acetobutylicum bioY (Song et al., 2021) served as negative control in the assay and no shift was observed.

Discussion

DTBS catalyzes formation of the ureido ring of dethiobiotin (Fig. 1). YnfK is a functional DTBS but is more poorly expressed and enzymatically slightly weaker than BioD. The measured K_M of YnfK for DAPA is twice that of BioD. Our substitutions of BioD active site residues for those of YnfK were based on the alignment of Fig. 2 which argued that the two proteins may have similar structures. This is borne out by the model of YnfK (entry P0A6E9) in the very recently released AlphaFold Protein Structure Database (alphafold.ebi.ac.uk). AlphaFold is a dramatically improved state of the art ab initio protein structure prediction algorithm (Jumper et al., 2021). For the majority of proteins tested, the models produced by AlphaFold from their amino acid sequences accurately match determined crystal structures with Cα root-mean-square-deviation at 95% residue coverage with a resolution of 0.96Å. In the present case the AlphaFold model of YnfK can be readily superimposed on the determined 0.95Å structure (PDB 1BY1) of BioD (Fig. 7). Indeed. the root mean square deviations (RMSD) for Cα carbons of the AlphaFold structure predicted for YnfK and the BioD crystal structure are 0.629Å and 0.734Å in two independent analyses. Hence, our use of BioD to probe YnfK activity is strongly justified by the AlphaFold model and the residue substitutions which increased YnfK activity. Therefore, YnfK and BioD can be considered as isoforms. Direct regulation of ynfK transcription seems primarily the property of FNR. However, transcripts of the upstream gene mlc (also called dgsA), which encodes a global regulator of carbohydrate metabolism seem likely to read through ynfK since there is no known or predicted transcription terminator within the intragenic region of 125 bp. Transcription of the mlc gene is highly complex (three promoters, activation and repression by Crp plus repression by Mlc) whereas the mlc-ynfK intragenic region contains only three recognizable sequences: two highly predicted promoters and the FNR binding site.

Fig. 7. The AlphaFold model of the YnfK structure superimposed on the 0.95Å structure of BioD using ChimeraX.

The YnfK model is shown in blue and the BioD structure(Sandalova et al., 1999) is shown in yellow. The single residue mutations are shown in pink. The root mean square deviation (RMSD) between 201 pruned atom pairs is 0.734Å (across all 219 pairs:1.610). The magnesium ion is shown in green. ATP: adenosine-5-triphosphate. DNN: 7,8-diamino-nonanoic acid. Based on the modeling, Ser13 and T114 are directly involved in substrate binding, whereas I184Y is located in the dimerization interface (note that AlphaFold predictions are currently restricted to monomers). A parallel, independent analysis by Dr. Yuanyuan Hu of this laboratory gave a Cα carbons RMSD of 0.629Å over 80% of the shorter BioD residues. The RMSD for all carbon atoms was 0.646Å over 1079 C atoms indicating that not only the Cα carbons but a significant number of the sidechains could be superimposed. The few discrepancies in the superimpositions are segments of the model that were predicted with low or very low confidence (e.g. the carboxy terminus).

The in vivo activity of YnfK is almost certainly much lower than what we measure in vitro because DAPA improves YnfK dimer formation (Fig. 5A). Since YnfK is starved for DAPA, dimerization in vivo is likely to be poor resulting in the low activity seen. Another factor in the low activity of YnfK relative to BioD is that it is more poorly expressed (Fig. S2). The expression levels of both BioD and YnfK have been measured by ribosome profiling. When grown in minimal medium there are 675 BioD molecules/cell versus 30 YnfK molecules/cell (Li et al., 2014). However, the levels of both proteins must be corrected for the culture conditions of the profiling assay. The BioD value is for a strain that has a wild type bio operon so the cells have sufficient biotin and biotin synthesis would be down-regulated 4 to 5-fold relative to the completely depressed levels seen upon biotin starvation (Cronan, 1988). In the case of YnfK ribosome profiling was done under aerobic conditions where FNR is inactive. Anaerobiosis to activate FNR would give a roughly 9-fold increase in YnfK expression. Therefore, even under anaerobic conditions in wild type cells YnfK protein levels would be about 40% those of BioD. Moreover, in the absence of DAPA only about 40% of YnfK monomers form dimers suggesting that relative to BioD perhaps only about 16% of YnfK would be able to synthesize DTB. These factors together with scarce supply of DAPA seem very likely to be the major contributors to the poor ability of YnfK to support anaerobic growth.

Expression of E. coli YnfK allowed very slow anaerobic growth of a bioD deletion strain due to activation of ynfK transcription by FNR. S. enterica serovar Typhimurium also encodes two DTBS, BioD in the bio operon and YnfK. Sequence alignments show that E. coli BioD shares 74% identity with S. enterica BioD whereas only 48% identity was seen between the S. enterica BioD and YnfK, a result similar to that seen in E. coli. The similarity between the E. coli and S. enterica BioDs is somewhat lower than generally seen between orthologous proteins of the two bacterial species (~90%).

Similar to the E. coli case, S. enterica BioD shares 48% identity with its YnfK. However, the YnfK proteins of E. coli and S. enterica have striking identity (93%) indicating that the two proteins are orthologs. The YnfK sequences and genome neighborhoods are almost completely conserved in all sequenced E. coli and S. enterica strains (Fig. 7). In both S. enterica and E. coli FNR is a global regulator that adapts cells to anaerobic growth by upregulating genes for anaerobic metabolism and shutting down genes of aerobic metabolism. The transition to anaerobic growth is crucial for pathogenesis in S. enterica and this plays a strong role in pathogenesis. Strains lacking FNR (Δfnr) are non-motile and fail to survive in macrophages, indicating a role for FNR in virulence (Fink et al., 2007). Indeed, a fnr deletion was included in a vaccine strain designed to protect against S. enterica (Zhao et al., 2021). Moreover, biotin synthesis has been shown to be essential for pathogenesis by Francisella novicida (Feng et al., 2014; Feng et al., 2015), Moraxella catarrhalis (Zeng et al., 2020) and Mycobacterium tuberculosis (Woong Park et al., 2011). Note that mammals cannot synthesize biotin and blood/serum levels are very low (~1 nM) (Bhagavan and Coursin, 1967) so the pathogen must synthesize this essential vitamin.

In conclusion ynfK encodes a weakly expressed enzyme that poorly dimerizes and is starved for its DAPA substrate. Given these handicaps of YnfK activity why, as shown in the DTBS phylogeny (Fig. 8), has the encoding gene not only been retained but conserved in diverse γ-proteobacteria? Was YnfK an early form of what became BioD? If so, why has it been retained and placed under anaerobic control? This puzzle aside, YnfK is a functional dethiobiotin synthase (DTBS) and thus we suggest that the ynfK gene designation be replaced with bidA for biod anaerobic. Note that YnfK has been denoted BioD2 in some protein databases.

Fig. 8. Phylogeny of DTBS enzymes.

The phylogenetic tree was constructed using the default setting of Neighbor Joining, Blosum62 in Jalview. Each bacterial species is given followed by the UniProt code.

Experimental Procedures

Strains, chemicals and culture media

The strains and plasmid used in the work are given in Tables S1 and S2, respectively Reagents and chemicals were obtained from Research Products International unless otherwise noted. New England BioLabs supplied restriction enzymes and T4 DNA ligase. Oligonucleotides were purchased from Integrated DNA Technologies and are listed in Table S3. PCR amplification was performed using Q5 high fidelity DNA polymerase (New England BioLabs) according to manufacturer protocols. DNA constructs were sequenced by ACGT, Inc. The rich medium for growth of E. coli was LB broth. The defined medium for E. coli was M9 salts supplemented with 0.5% glycerol and 0.1% vitamin-free Casamino Acids (Difco). The defined medium for protein purification in E. coli (protein purification medium) was M9-XS1 salts supplemented with 2 mM MgSO₄, 0.5x Trace Metals, 0.5% vitamin-free Casamino Acids, 0.4% glycerol and 1% glucose. The M9-XS1 salts contains (in g L⁻¹) Na₂HPO₄, 7.1; K₂HPO₄, 3.4; NaCl, 0.5; NH₄Cl, 2.675 adjusted to pH 8.0–8.2. One hundred mL of 1000X Trace Metals, contained 50 mL 0.1 M FeCl₃, 2 mL 1 M CaCl₂, 1 mL 1M MnCl₂, 1 mL 1M ZnSO₄, 1 ml 0.2 M CoCl₂, 2 ml 0.1 M CuCl₂, 2 ml 0.2 M NiCl₂, 2 mL 0.1 M Na₂MoO₄, 2 ml 0.1 M H₃BO₃ and 37 ml MilliQ H₂O. Antibiotics were used at the following concentrations (μg ml⁻¹): kanamycin sulfate, 50; chloramphenicol, 20; and ampicillin 100. The 6x-His Tag Monoclonal Antibody (4E3D10H2/E3) was purchased from Thermo Fisher Scientific, catalog # MA1–135, RRID AB_2536841. Goat anti-Mouse IgG (H+L) Secondary Antibody, HRP was purchased from Thermo Fisher Scientific, catalog # 31430, RRID AB_228307. Gel filtration standards was purchased from BIO-RAD, catalog # 1511901. Ni-NTA agarose, lot No. 139290134 was purchased from QIAGEN. The protein standard was the Blue Protein Standard, NEB# P7718S from New England BioLabs.

Plasmids and plasmid constructions

Complementation plasmids

All PCR amplified E. coli genome segments were from strain MG1655 genomic DNA. The ynfK gene was PCR amplified with primers SXJ394 and SXJ395 and ligated into EcoRI and SalI digested pBAD322C to give plasmid pXS32.The bioD gene was PCR amplified with primers SXJ392 and SXJ393 and ligated into EcoRI and SalI digested pBAD322C to give plasmid pXS33. DNA fragments containing ynfK and 300 bp upstream of ynfK were PCR amplified from strain MG1655 genomic DNA using primers SXJ505 and SXJ506, and pBAD33 was PCR amplified using primers SXJ507 and SXJ508. These two fragments were assembled using a Gibson Assembly kit (NEB #E5510S from New England BioLabs) to construct pXS34. DNA fragment containing bioD was PCR amplified using primers SXJ509 and SXJ510, and pBAD33 carrying 300 bp upstream of ynfK was PCR amplified from pXS34 using primers SXJ511 and SXJ511. The two fragments were assembled using Gibson Assembly to create pXS35. DNA fragment containing 300 bp upstream of bioB was PCR amplified from strain MG1655 genomic DNA using primers SXJ517 and SXJ520, and pBAD33 carrying ynfK was PCR amplified from pXS34 using primers SXJ518 and SXJ519. These two fragments were assembled using Gibson Assembly to create pXS36. DNA fragment containing 300 bp upstream of bioB was PCR amplified from strain MG1655 genomic DNA using primers SXJ521 and SXJ523, and pBAD33 carrying bioD was PCR amplified from pXS35 using primers SXJ522 and SXJ524. These two fragments were assembled using Gibson Assembly to create pXS37. The ynfK gene was PCR amplified using primers SXJ535 and SXJ536 and ligated into KpnI and HindIII digested pQE2 to give plasmid pXS38. The bioD gene was PCR amplified using primers SXJ533 and SXJ534 and ligated into KpnI and HindIII digested pQE2 to give plasmid pXS39.

Site directed mutagenesis

The Q5 Site-Directed Mutagenesis Kit from New England Biolabs (catalog # E0554S) was used in constructing site directed mutants following manufacturer’s instructions. Plasmid pXS43 that encodes the YnfK S13E mutation, was PCR amplified from pXS36 using primers SXJ585 and SXJ586. Plasmid pXS44 that encodes YnfK T114A was PCR amplified from pXS36 using primers SXJ579 and SXJ580. Plasmid pXS45 that encodes the YnfK I184Y was PCR amplified from pXS36 using primers SXJ581and SXJ582. Plasmid pXS46 that encodes YnfK S113E/T114A was PCR amplified from pXS43 using primers SXJ579 and SXJ580. Plasmid pXS47 which encodes YnfK S113E/I184Y was PCR amplified from pXS43 using primers SXJ581 and SXJ582. Plasmid pXS48 that encodes YnfK T114A/I184Y, was PCR amplified from pXS45 using primers SXJ579 and SXJ580. Plasmid pXS55 that encodes FNR D154A was PCR amplified from pXS55 using primers SXJ597 and SXJ598.

Protein expression and purification plasmids

The ynfK and bioD genes were PCR amplified using primers SXJ305 and SXJ306, SXJ307 and SXJ308, respectively. The 5´ primers contained NcoI sites whereas the 3´primers contained HindIII sites and were ligated into pET28b cut with NcoI and HindIII to construct plasmids pXS30 and pXS31, respectively. The gene that encodes the YnfK S13E was PCR amplified from pXS43 using primers SXJ305 and SXJ306, and ligated into NcoI and HindIII digested pET28b to give plasmid pXS49. The gene ynfK T114Awas PCR amplified from pXS44 using primers SXJ305 and SXJ306, and ligated into NcoI and HindIII digested pET28b to give plasmid pXS50. The gene ynfK I184Ywas PCR amplified from pXS45 using primers SXJ305 and SXJ306, and ligated into NcoI and HindIII digested pET28b to give plasmid pXS51. The gene ynfk S13E/I184Y was PCR from pXS47 using primers SXJ305 and SXJ306, and ligated into NcoI and HindIII digested pET28b to give plasmid pXS52. The gene ynfk T114A/I184Y was PCR amplified from pXS48 using primers SXJ305 and SXJ306, and ligated into NcoI and HindIII digested pET28b to give plasmid pXS53. The gene fnr D154A was PCR amplified from pXS55 using primers SXJ609 and SXJ610, and ligated into NcoI and HindIII digested pET28b to give plasmid pXS56.

Plasmids for western blotting

The hexahistidine tagged ynfK gene was PCR amplified using primers SXJ560 and SXJ395 and ligated into EcoRI and SalI digested pBAD322CM to give plasmid pXS40. The DNA fragment encoding hexahistidine tagged YnfK was PCR amplified using primers SXJ558 and SXJ556 and pBAD33 carrying 300 bp upstream of bioB was PCR amplified from pXS36 using primers SXJ557 and SXJ559. These two fragments were assembled using Gibson Assembly to create pXS41. DNA fragment containing hexahistidine tagged ynfK was PCR amplified using primers SXJ554 and SXJ556, and pBAD33 carrying 300 bp upstream of ynfK was PCR amplified from pXS34 using primers SXJ555 and SXJ557. These two fragments were assembled using Gibson Assembly to create pXS42.

LacZY fusion plasmids

The wild type fnr gene was PCR amplified using primers SXJ485 and SXJ486 and ligated into KpnI and HindIII digested pQE2 to give pXS54. The DNA fragment carrying 196 bp upstream of ynfk was PCR amplified using primers SXJ493 and SXJ495, pBAD33 carrying lacZY genes was PCR amplified from a pXS58 using primers SXJ494 and SXJ496. These two fragments were assembled using Gibson Assembly to create pXS57.

All plasmids were verified via sequencing and are listed in Table S1. DNA fragments cloned into pBAD322C were verified using primers SXJ62 and SXJ63. DNA fragments cloned into pBAD33 were verified using primers SXJ431 and SXJ513. Plasmid pXS57 was verified using primers SXJ431 and SXJ432. DNA fragments cloned into pET28b were verified using primers SXJ192 and SXJ193. DNA fragments cloned into pQE2 were verified using primers SXJ365 and SXJ366.

Complementation analysis under biotin starvation conditions

The E. coli strain XS395 was transformed with pBAD322C, pBAD33, pQE2, pXS32, pXS33, pXS34, pXS35, pXS36, pXS37, pXS38, pXS39, pXS43, pXS44, pXS45, pXS46, pXS47 or pXS48, respectively. Transformants were selected on LB plates supplemented with chloramphenicol or ampicillin at 37°C. Single colonies were picked and streaked on defined medium plates supplemented with chloramphenicol or ampicillin. After four cycles of restreaking the colonies were then sub-cultured into 1 ml of minimal media lacking biotin and containing 0.05 units of avidin. These cultures were incubated at 37°C for 5 h, centrifuged at 13,000xg for 2 min to pellet the cells and the pellets were washed 5 times with 1 ml of M9 medium and resuspended in 200 μl of M9 medium. Cells were then streaked on biotin-free M9 minimal medium supplemented with chloramphenicol or ampicillin. Arabinose was added to 0.02% in the medium in cells transformed with pBAD322C, pXS32 and pXS33. IPTG (500 μM) was added to the medium in cells transformed with pQE2, pXS38 and pXS39. For cell growth under anaerobic conditions, 300 mg L⁻¹ S-gal (3,4-cyclohexenoesculetin β-D-galactopyranoside, Sigma-Aldrich catalog # S7313), 10 μM IPTG, 500 mg L⁻¹ ferric ammonium citrate, 25 mM NaNO₃ and 5 mM adenosine 3´, 5´-cyclic monophosphate sodium salt monohydrate (Sigma-Aldrich, catalog # A6885) were added into medium for visualization of strain growth.

Purification of the BioD, YnfK, YnfK substitution proteins and FNR

E.coli BL21 Star (DE3) was transformed with pXS30 for YnfK purification as were BioD and the YnfK active site substitution derivatives. The strains was grown at 37°C in protein purification medium with kanamycin to an OD₆₀₀ of 0.8 and expression was induced by addition of IPTG to 1 mM. Following growth for an additional 16 h at room temperature (24°C), the cells were recovered by centrifugation and suspended in lysis buffer (50 mM HEPES, 10 mM imidazole, 0.5 mM TCEP, 250 mM NaCl, 5% glycerol, pH 7.5). The cells were lysed by passage through a French pressure cell, the lysates were centrifuged to remove unbroken cells and cellular debris and the supernatant was added to Ni NTA beads (Qiagen) and incubated for 1 h before the beads were added to a disposable 10 ml polypropylene column. The column was washed with 30 ml of wash buffer (lysis buffer containing 60 mM imidazole). Proteins was eluted in 1 ml fractions with elution buffer (lysis buffer containing 250 mM imidazole). The fractions were analyzed by SDS-PAGE to determine purity. Pure fractions were combined and dialyzed against storage buffer (lysis buffer lacking imidazole). Aliquots were flash frozen and stored at −80°C. The purity of each of the DTBS proteins is shown by the no crosslinker added lanes of Fig. 5 whereas that of FNR D154 is given in Fig. S3.

β-Galactosidase assays

E. coli strains were grown overnight in defined media supplemented with chloramphenicol and ampicillin. Cultures were diluted to OD₅₉₅ of 0.1 in defined media containing 500 μM IPTG and grown at 37°C for 4 h. β-Galactosidase activity was assayed following permeabilization of the cells with lysozyme.

Electrophoretic mobility shift assay of DNA binding

The predicted C. acetobutylicum DNA binding site upstream of bioY2 was PCR amplified from C. acetobutylicum genomic DNA with primers SXJ129 and SXJ130 and used as negative control. The predicted FNR binding site upstream of ynfK was PCR amplified with primers SXJ607 and SXJ608. All DNA fragments were 125 bp in length. The PCR products were assayed on a 2.0% agarose gel and purified using a QIAquick PCR Purification Kit (Qiagen). DNA concentrations were determined at OD₂₆₀ by using a NanoDrop 2000c. The DNA binding reaction contained 50 mM HEPES (pH 7.5), 50 mM NaCl, 10% glycerol, 20 nM DNA, 1 mM MgCl₂ and indicated concentrations of FNR D154A. The binding reactions were incubated at 37°C for 30 min and then loaded into a 6% DNA retardation gel (Invitrogen). The gels were run in 0.5X TBE at 100 V for 100 min. The gel was stained with SYBR Green I nucleic acid gel stain (Invitrogen) and visualized using Bio-Rad Chemidoc XRS and Quantity One software.

Western Blotting

Protein samples were separated on a 12% SDS-polyacrylamide gel and transferred by electrophoresis to Immobilon-P membranes (Millipore) for 60 min at 90 V. The membranes were pre-blocked with TBS buffer (20 mM Tris base and 150 mM NaCl, pH 7.5) containing 0.1% (v/v) Tween-20 and 5% non-fat dry milk. For His₆-tagged protein visualization, the membranes were probed for 1 h with 6x-His Tag monoclonal antibody diluted 1:2000 in the above buffer and washed 4 times with TBS buffer. Following incubation with goat anti-mouse IgG (H+L) secondary antibody, the labeled proteins were stained with SuperSignal West Femto Maximum Sensitivity Substrate (Thermo Scientific, Lot # SC246945, ProD # 34095) and detected using Quantity One software.

Size exclusion chromatography

The Superdex200 Increase 10/300 GL column was purchased from GE Life Sciences and run following the manufacturer instructions. The running buffer was 0.05M phosphate buffer containing 0.15M NaCl, and 0.5 mM TCEP (pH 7.4) with a flow rate of 0.35mL/min. Gel filtration standard was purchased from BIO-RAD (catalog # 1511901) and used following manufacturer instructions and protein samples were loaded as the manufacturer directed.

Chemical cross-linking of proteins

Protein samples was incubated with or without indicated concentration of ethylene glycol bis[succinimidylsuccinate] (EGS) at room temperature for 30 min. SDS-loading dye was added and samples were heated to 99°C for 5 min. Samples were loaded on 4–20% gradient SDS polyacrylamide gels (BioRad) and run at 110 V for 75 min.

Sequence alignments

Sequence alignments were processed by Jalview 2 to generate the final figure (Waterhouse et al. 2009).

P1 phage transduction

Mutations in ynfK and fnr were constructed by P1 transduction from mutants that were obtained from the E. coli Genetic Stock Center or from the Dr. Cari Vanderpool lab. Strain XS44 was transduced with P1 phage harvested from strain XS273 to obtain strain XS444. The FRT-flanked antibiotic resistant markers were eliminated by the FLP activity of pCP20, and the strains were then cured of pCP20 by incubation at the nonpermissive temperature (42°C) to obtain strain XS446. Strain STL114 was transduced with a P1 phage stock grown on strain XS254 to obtain strain XS395. Gene mutations in ynfK was confirmed by PCR analysis using primers SXJ311 and SXJ312. The mutations in fnr waere confirmed by PCR analysis using primers SXJ377 and SXJ378. Mutations in bioD was confirmed by PCR analysis using primers SXJ317 and SXJ318. All mutations were confirmed by sequencing of PCR products.

DTBS coupled spectrophotometric assay

The kinetic parameters K_M and k_cat of ATP and DAPA with each enzyme were determined from the initial velocity data measured by limiting one substrate and saturating with the others. DTBS activity could be measured spectrophotometrically by monitoring either ADP or P¡ formation. ADP production was assayed using the pyruvate kinase/lactate dehydrogenase couple (Cleland, 1979a). The disappearance of NADH was continuously monitored at 340 nm (6.22 mM-1‚cm-1 at 340 nm) at 25°C using a DU800 spectrophotometer (Beckman Coulter) to calculate the initial velocity. In a total volume of 500 μl, a typical assay contained 100 mM potassium-HEPES, pH 7.5, 20 mM NaHCO_3, 10 mM MgCl₂, 5 mM KCl, 5 mM ATP, 5 mM PEP, 0.2 mM NADH, 10 μl of pyruvate kinase/lactic dehydrogenase enzymes from rabbit muscle (Sigma Aldrich, SKH# P0294), 20 μl of DTBS, and varying amounts of DAPA (Toronto Research Chemicals, Cat # D416550). For measuring K_M of ATP, 5 mM ATP was replaced with 5 mM DAPA, and varying amounts of ATP was added into the reaction.

Data availability statement

All data are available on request from the authors