Biotin (vitamin H) is a key enzyme cofactor required in all three domains of life. Although this cofactor was discovered over 70 years ago and has long been recognized as an essential nutrient for animals, our knowledge of the strategies bacteria use to sense biotin demand is very limited. The paradigm mechanism is that of Escherichia coli in which BirA protein, the prototypical bi-functional biotin protein ligase, both covalently attaches biotin to the acceptor proteins of central metabolism and represses transcription of the biotin biosynthetic pathway in response to biotin demand. However, in other bacteria the biotin protein ligase lacks a DNA-binding domain which raises the question of how these bacteria regulate the synthesis of biotin, an energetically expensive molecule. A bioinformatic study by Rodionov and Gelfand identified a protein termed BioR in α-proteobacteria and predicted that BioR would have the biotin operon regulatory role that in most other bacteria is fulfilled by the BirA DNA-binding domain. We have now tested this prediction in the plant pathogen Agrobacterium tumefaciens. As predicted the A. tumefaciens biotin protein ligase is a fully functional ligase that has no role in regulation of biotin synthesis whereas BioR represses transcription of the biotin synthesis genes. Moreover, as determined by electrophoretic mobility shift assays, BioR binds the predicted operator site, which is located downstream of the mapped transcription start site. qPCR measurements indicated that deletion of BioR resulted in a c. 15-fold increase of bio operon transcription in the presence of high biotin levels. Effective repression of a plasmid-borne bioB-lacZ reporter was seen only upon the overproduction of BioR. In contrast to E. coli and Bacillus subtilis where biotin synthesis is tightly controlled, A. tumefaciens synthesizes much more biotin than needed for modification of the biotin-requiring enzymes. Protein-bound biotin constitutes only about 0.5% of the total biotin, most of which is found in the culture medium. To the best of our knowledge, A. tumefaciens represents the first example of profligate biotin synthesis by a wild type bacterium.
Biotin (vitamin H) is a universal cofactor essential for certain essential carboxylation, decarboxylation and transcarboxylation reactions in both prokaryotes and eukaryotes (Beckett, 2007; Hebbeln et al., 2007). Biotin is only active as a cofactor when covalently attached to a unique lysine residue of the enzyme where it acts as a ‘swinging arm’) that transfers intermediates from one active site to another. Although proteins modified by attachment of this coenzyme are rare (< 6 protein species/organism) they play key roles in such key metabolic processes as fatty acid biosynthesis, gluconeogenesis and amino acid degradation (Beckett, 2007). The attachment of biotin to its cognate proteins is catalysed by biotin protein ligase (BPL), a highly conserved and largely interchangeable enzyme essential in all three domains of life. Thus far, each organism has a single BPL that modifies all of the cognate proteins.
Plants, most fungi and most bacteria synthesize biotin whereas mammals and birds cannot. Biotin synthesis is generally highly regulated because the synthetic pathway is expensive. In the best understood pathway, that of Escherichia coli, the synthesis of one biotin molecule requires 20 ATP equivalents. Our knowledge of transcriptional regulation of biotin synthesis in bacteria is largely restricted to E. coli (Barker and Campbell, 1980; Cronan, 1989; Eisenstein and Beckett, 1999; Weaver et al., 2001) and Bacillus subtilis (Gloeckler et al., 1990; Speck et al., 1991; Bower et al., 1995; 1996; Lee et al., 2001). The E. coli biotin regulatory system is the better characterized and the central player is BirA, a dual function protein that is both a BPL and the repressor that specifically binds the biotin operator to repress transcription initiation of the biotin biosynthesis operon (Beckett, 2007) (Fig. S1). The fact that BirA acts as both an enzyme and a repressor imparts an unexpected regulatory sophistication in that expression of the bio operon is not only sensitive to intracellular biotin levels (Fig. S1), but also responds to the levels of cognate accepter proteins requiring biotinoylation for enzyme function (Fig. S1). The small molecule BirA regulatory ligand is biotinoyl-adenylate (biotinoyl-5′-AMP) rather than biotin itself (Fig. S1) (Weaver et al., 2001; Beckett, 2007; Chakravartty and Cronan, 2012). Biotinoyl-5′-AMP is the product of the first half-reaction of the BirA-catalysed BPL activity that proceeds in an ordered manner in which monomeric BirA protein first binds biotin and then ATP (Xu and Beckett, 1994). Tight binding of the adenylate triggers a monomer to dimer transition which is required for BirA DNA-binding activity (Streaker et al., 2002). Upon binding to the bio operator (bioO), a 40 bp inverted repeat that overlaps the promoters of the bifurcated bio operon (bioA and bioBFCD), the dimeric BirA + biotinoyl-5′-AMP complex blocks transcription (Fig. S1) (Otsuka and Abelson, 1978; Streaker and Beckett, 1998a,b). Current evidence indicates that the regulatory switch is the level of biotinoyl-5′-AMP. When biotin levels are low BirA is unable to make the adenylate and is thus unable to form the dimers required for operator binding and hence bio operon expression is derepressed. Even in the presence of sufficient biotin, biotinoyl-5′-AMP levels are low when an excess of acceptor protein is present. The enzyme-bound adenylate is attacked by the ε-amino group of the specific acceptor protein lysine residue resulting in cleavage of the adenylate and a biotinoylated protein (Fig. S1). Under either low biotin levels or high acceptor levels any biotinoyl-adenylate synthesized is rapidly consumed in biotinoylation of acceptor proteins and no significant level of the dimeric BirA-biotinoyl-5′-adenylate complex is present. Thus the bioO operator is seldom occupied and transcription is maximal (Fig. S1).
E. coli BirA is composed of three highly interacting domains. These are the N-terminal winged helix–turn–helix (HTH) DNA-binding domain, the central catalytic domain where biotin, ATP and biotinoyl-5′-AMP bind and the C-terminal dimerization domain (Wilson et al., 1992; Weaver et al., 2001). Analogous proteins are found in a broad range of bacteria and archaea (Rodionov et al., 2002). However, in some phylogenetic groups, such as α-proteobacteria and actinobacteria the BirA homologues lack N-terminal DNA-binding domains (Rodionov et al., 2002; Rodionov and Gelfand, 2006; Brune et al., 2012). Hence, it seems that in these bacteria, biotin synthesis is either regulated by another mechanism or is unregulated.
Agrobacterium tumefaciens, a well-studied member of the α-proteobacteria, is a plant pathogen often utilized in plant genetic engineering (Goodner et al., 2001; Wood et al., 2001). Like the other α-proteobacteria the A. tumefaciens birA gene encodes a protein that lacks a recognizable DNA-binding domain whereas the central and C-terminal domains readily align with those of E. coli (Fig. S2). We report that the A. tumefaciens BirA is a fully functional BPL. Computational predictions by Rodionov and Gelfand (2006) suggested that a GntR family transcription factor protein they named BioR, is the repressor of bioBFDAZ operon transcription. BioR was predicted to bind a bio operator, a 10 bp inverted repeat (TTATCTATAA) sequence. We demonstrate that the prediction of Rodionov and Gelfand is correct in that BioR binds the predicted operator and functions as a repressor of bioBFDAZ transcription. Unexpectedly, we were unable to demonstrate that DNA binding by this protein requires the presence of biotin or any detectable biotin derivative. Unlike the scenarios observed with E. coli and B. subtilis, A. tumefaciens produces orders of magnitude more biotin than that required for modification of its biotinylated proteins. This seems the first example of profligate biotin synthesis by a wild type bacterium.
Results and discussion
A. tumefaciens BirA is a bona fide Group I biotin protein ligase
Multiple sequence alignment analyses showed that A. tumefaciens BirA is similar to that of its close-relative Brucella melitensis, an animal pathogen (Fig. S2A). Both α-proteobacterial BirA proteins are significantly shorter than E. coli BirA and their smaller sizes seem due to lack of the E. coli N-terminal HTH domain (Fig. S2A). The multiple genome-scale phylogenetic analyses of Rodionov et al. (2002) indicated that bi-functional BPL enzymes (Group II) such as E. coli BirA are widespread in both Eubacteria and Archaea suggesting this type of BirA is the ancient form. Thus, it seemed very likely that the bacteria having BPLs that lack the N-terminal domain (Group I) have either evolved a different mode of biotin gene regulation or do not regulate synthesis of this cofactor.
To directly test the biochemical properties and physiological function of A. tumefaciens BirA we purified the protein to homogeneity following expression in E. coli (Fig. S3A). In contrast to the other characterized type I BirAs the A. tumefaciens protein seemed largely monomeric (Fig. S3B). Mass spectrometric analysis of tryptic peptides of the protein matched the expected sequence with 53% coverage (Fig. S3C). The A. tumefaciens BirA was shown to convert biotin and [α-32P]-ATP (Fig. 1) to the canonical biotinoyl-5′-AMP intermediate (Fig. 1C and D) and transferred the biotin moiety to the E. coli acceptor protein, AccB-87 as well as to an analogous B. subtilus AccB-86 biotin domain (Fig. 1A and B). Biotin attachment to the latter acceptor was verified by electrospray ionization mass spectrometry (Fig. 1E and F).
Function of the A. tumefaciens BirA was also tested in vivo by its ability to complement growth of E. coli strain BM4062 (birA85), a temperature sensitive strain that grows at 30°C, but not at 42°C due to loss of BirA function at the higher temperature (Fig. S4A). As expected from the in vitro results (Fig. 1), expression of the A. tumefaciens birA gene allowed growth at 42°C, whereas the mutant strain alone or carrying the empty vector pBAD24 not on either Cornell University Luria Broth (LB) agar plates (Fig. S4A) or in liquid media (Fig. S4B). Together, the combined in vitro and in vivo evidence demonstrate that A. tumefaciens birA encodes a bona fide Group I BPL that lacks a detectable regulatory domain (Fig. S2C and D).
Characterization of A. tumefaciens BioR
To maintain the biotin level required for central metabolism of A. tumefaciens without wasteful synthesis, it seemed likely that the bacterium would have evolved a transcription factor that would compensate for the lack of regulation by BirA. Recently, Rodionov and Gelfand (2006) predicted that a GntR family protein they named BioR would provide the missing regulation. Multiple sequence alignments of these BioR orthologues revealed that they all shared a relatively conserved HTH DNA-binding motif belonging to the GntR superfamily (Fig. S5A). Using the structure of (PSPTO_5454, PDB:3C7J), a Pseudomonas syringae GntR-like transcription factor of unknown function as template, we modelled the structure of BioR (Fig. S5B–D). However, the presence of a birA gene that encodes a putatively bi-functional 319 residue protein in the chromosome of the tomato pathogen P. syringae pv. tomato DC3000 (Buell et al., 2003) indicates that PSPTO_5454 is not involved in regulation of biotin synthesis. To test the function predicted for BioR it was expressed in E. coli and purified to homogeneity (Fig. S6A). Like the another GntR protein, FadR (Fig. S6B), chemical cross-linking assays suggested that the BioR protein forms dimers in solution (Fig. S6C). Liquid chromatography mass spectrometry of tryptic peptides validated the identification of the recombinant protein with 80% coverage to the sequence predicted from the DNA sequence (Fig. S6D).
Binding of BioR to the bioB promoter
Rodionov and Gelfand (2006) proposed that the BioR binding site was a conserved 10 bp sequence within a putative A. tumefaciens bioB promoter region (Figs 2A and 6E). Given the physical position of this BioR binding site BioR seemed likely to act as a repressor of bio operon expression in A. tumefaciens. We tested the ability of BioR to bind the predicted sequence using a 50 bp probe containing the Rodionov–Gelfand site and electrophoretic mobility shift (gel shift) assays performed with either crude cell extracts or purified BioR (Fig. 2B). Gel shift assays showed that both sources of BioR efficiently bound the bioB probe in a dose-dependent manner (Fig. 2B) in that nearly 90% of bioB probe was found in the DNA–protein complex in the presence of 1 pmol BioR (Fig. 2C and D). As previously seen at high FadR concentrations (Feng and Cronan, 2010; 2011b) BioR super-shifted bands were also seen at the higher protein concentrations (Fig. 2B and C). We favour the explanation that the super-shifted bands are due to a portion of the protein forming soluble aggregates (perhaps tetramers) in the mobility shift buffer as a consequence of the further concentration that occurs as the proteins enter the gel. Together, our in vitro data therefore indicate that the A. tumefaciens bioB palindromic site predicted by Rodionov and Gelfand (2006) is specifically recognized by BioR, but the physiological relevance of this interaction remained to be elucidated. As a test of relevance we extended this analysis to other α-proteobacteria (Fig. 2E). We prepared ten DNA probes derived from three different organisms and performed gel shift experiments. The A. tumefaciens BioR protein shifted three predicted Bradyrhizobium japonicum BioR binding sites, three putative Rhodobacter sphaeroides sites and four Brucella melitensis sites (Fig. 2E). Therefore it seemed likely that such BioR–operator interactions are widely conserved in α-proteobacteria.
BioR acts as a repressor
To test if BioR indeed functions as a repressor of the bioBFDAZ operon, we used qPCR and LacZ reporter assays to assay transcription of the operon. For the latter assays plasmid pRG-PbioB, a broad host range LacZ reporter plasmid in which the A. tumefaciens bioB promoter was fused to the lacZYA operon (Table S2) was constructed. The reporter plasmid was introduced into wild type strain NTL4 and three other strains: FYJ212 (ΔbioR::Km), FYJ283 (ΔbioBFDA) and FYJ284 (ΔbioBFDA ΔbioR::Km).
Analysis by qPCR showed that the bioBFDAZ operon expression in the ΔbioR strain was about 2 to 2.5-fold greater than that of the wild type strain in cultures grown on M9 minimal media lacking biotin supplementation (Fig. 3A). However, upon addition of biotin to a very high level (1 μM) expression of this bio operon in the ΔbioR strain was about 10 to 15-fold greater than that of the wild type strain (Fig. 3B). This was due to decreased expression in the wild type cultures because bio operon expression in the ΔbioR strain was insensitive to biotin (Fig. 4). The results obtained with the LacZ reporter system were in general agreement. In complex RB medium the LacZ activity of the ΔbioR strain was consistently about 1.5-fold higher than the wild type strain (Fig. 3C). Upon overexpression of BioR a 10-fold decrease in bioBFDAZ operon transcription was seen relative to the wild type strain (Fig. 3C). Similar results were observed when the strains were grown in M9 minimal media (not shown). To rule out the possible interference by endogenously produced biotin, we introduced the bioB-LacZ reporter and BioR-overexpression plasmids into a strain lacking the bioBFDA gene cluster [maintained on RB medium and/or M9 minimal medium with various biotin concentrations (1, 10, 100 or 1000 nM), Table S1]. In the strain lacking BioR, bioBFDAZ expression increased about twofold whereas overexpression of bioR in this strain led to six to eightfold repression of transcription (Fig. 3D). Hence, the prediction of Rodionov and coworkers that BioR functions as a repressor is correct. However, in the wild type NTL4 strain bioBFDAZ repression is very modest (Fig. 3C and D). We believe that the weak repression is due to a restricted supply of BioR because the protein binds the operator well in vitro and BioR overexpression gives appreciably greater repression (Fig. 3D). Hence, it seems that the bioB operator is not fully occupied in the wild type strain that carries the bioB-lacZ reporter plasmid [which is expected to have a copy number of 6–8 (Itoh et al., 1984)].
BioR senses biotin or a biotin metabolite
Figure 4 shows that addition of exogenous biotin to cultures of the wild type strain represses transcription of the biotin operon, although the repression is modest at concentrations expected to be physiologically relevant. However, cultures of the ΔbioR strain show no obvious change in expression of the bio operon genes even at 1 μM biotin (Fig. 4B). Therefore, BioR seems a biotin sensing protein, albeit a very inefficient one. BioR expression is insensitive to biotin supplementation (Fig. 4C and D) as expected from the absence of recognizable BioR binding site in the A. tumefaciens bioR promoter region (Rodionov and Gelfand, 2006). In contrast, bioR orthologues of other α-proteobacteria (e.g. Brucella and Paracoccus) have a putative BioR binding site upstream of bioR and thus may be auto-regulated, suggesting regulatory diversity in α-proteobacterial biotin metabolism.
Search for a possible BioR regulatory ligand
The most straightforward mechanism for BioR regulation of the biotin synthetic pathway would be for BioR to bind its operator and repress bio operon expression only upon binding biotin or a biotin derivative such as biotinoyl-5′-adenylate. However, the electrophoretic mobility shift assays were performed in the absence of added biotin (Fig. 2C and E), although it remained possible that BioR had acquired its ligand during expression in E. coli and that this interaction survived protein purification [this has been observed for E. coli BirA (Chakravartty and Cronan, 2012)]. To test this possibility we assayed two freshly purified BioR preparations of demonstrated DNA-binding activity for the presence of biotin or biotinoyl-5′-adenylate using a very sensitive bioassay (Fig. 5A). BioR samples (20 pmol) were placed in boiling water for 20 min to denature the protein and release any bound biotin or biotinoyl-5′-adenylate (the treatment would hydrolyse the very labile adenylate to free biotin). Bioassays showed that the BirA preparations contained no biotin under conditions where 2 pmol of biotin was readily detected (Fig. 5B). We also synthesized BioR in vitro using a cell-free coupled transcription–translation system in which individual components are well defined (Shimizu et al., 2001) and thus free of biotin. Gel shift assays showed that these BioR preparations bound the DNA probe (Fig. 5C) again indicating that BioR binding of its operator does not require biotin. Finally treatment of BioR preparations with streptavidin immobilized on agarose beads failed to block DNA binding in EMSA tests (not shown). Since, tetrameric streptavidin has an extraordinarily high affinity for biotin (dissociation constant of ∼ 10−14 M), it would effectively trap any free biotin or any biotin released from BioR. We also performed acid hydrolysis of BioR to test the covalently bound-biotin but failed to detect any traces of biotin or its derivatives (Fig. 5D).
We also tested the possibility that biotin acts to disassociate BioR from its cognate DNA palindromes. Recombinant BioR was mixed with various concentrations of biotin and assayed for DNA binding as above. Biotin addition had no effect on formation of BioR–DNA complexes nor did a variety of biotin related metabolites (acetyl-CoA, malonyl-CoA, malonyl-ACP, sodium pimelate, pimeloyl-ACP and dethiobiotin) (not shown). Since repression is similar in the wild type and bioBFDA strains intermediates in biotin biosynthesis are ruled out. Other possibilities are biotin-responsive covalent modification of BioR (e.g. phosphorylation) in its native host. Another possibility is a degradation product of biotin. Perhaps high biotin concentrations trigger a degradative pathway that degrades free biotin but not protein-bound biotin and one of the degradation products is the BioR regulatory ligand. Degradation of biotin by soil bacteria has been demonstrated (Li, 1964; Brady et al., 1965; Iwahara et al., 1969) although to our knowledge this has not been demonstrated in A. tumefaciens.
The A. tumefaciensbio gene cluster constitutes an operon
To this point we had assumed that the bioBFDAZ gene cluster is an operon transcribed from a promoter located upstream of bioB and that the BioR binding site was appropriately located to modulate transcription. We therefore mapped the promoter by an improved version of 5′-RACE (RLM-RACE) that showed the 5′-end of the A. tumefaciens bioB transcript is located 36 nucleotides upstream the BioB translation initiation codon (Fig. 6C and D). To our surprise we found that the BioR binding site appears to be located within the bioB coding sequence and the sequence is strictly conserved in all of the extant Agrobacterium sequences (although not in other α-proteobacteria). However, although the A. tumefaciens bioB translational initiation site has not been directly determined, the amino acid sequence encoded by the BioR binding site is conserved in all Brucella species and no alternate initiator methionine codons are present.
To test if the bioB transcript extends through the gene cluster as expected if the genes constitute an operon, PCR and RT-PCR assays were performed using multiple sets of primers (Table S2 and Fig. 6B). First, positive amplifications of five intragenic regions (1, 2, 3, 4 and 10) were obtained by RT-PCR analyses demonstrating (Table S2 and Fig. 6B) that the five genes are transcribed at similar levels as shown above (Fig. 3B). PCR was used to validate the primers used for the intragenic regions. Transcription of the intergenic regions was tested using other primer sets (Fig. 6B upper gels) and four intergenic amplicons (5, 6, 7 and 9) were observed. The same amplicons were observed in our RT-PCR analyses indicating that the five genes of the bio cluster comprise a transcription unit and hence an operon (Fig. 6A and B). It remains possible that there are internal promoters within the operon that were not detected in the 5′-RACE experiments (Fig. 6). We also performed RT-PCR analysis of bioR and found that it was expressed at levels comparable to the biosynthetic genes (Fig. 6A and B).
The production of biotin far exceeds the levels required for growth of A. tumefaciens
An energetically efficient regulatory system would produce biotin to match the amount needed to modify the proteins required for central metabolism. However, the very modest regulation seen at biotin levels sufficient to support growth of other bacteria suggested that biotin synthesis in A. tumefaciens may be essentially uncontrolled. This was tested by bioassay determination of the amount of biotin produced by the wild type strain and the fraction of protein-bound biotin. For comparison we also assayed these parameters in the E. coli K-12 strain MG1655 and B. subtilis 168. Total biotin produced by wild type A. tumefaciens was 7.67 × 105 molecules per viable cell, a value 20-fold greater than E. coli and almost 60-fold greater than B. subtilis (Table 1). The amount of biotin bound to the A. tumefaciens enzyme proteins was obtained by acid hydrolysis of purified cellular protein fractions and was found to constitute < 0.5% of the total biotin (Table 1). Hence, A. tumefaciens produces biotin in great excess over the levels needed to modify and activate the biotin-requiring enzymes. One explanation could be that that A. tumefaciens is unable to retain biotin. However, the cell-associated (presumably cytosolic) free biotin was ninefold higher than the protein-bound fraction. Finally, it was possible that A. tumefaciens degrades biotin and must overproduce the coenzyme to offset this loss. If so, we would expect that an A. tumefaciens biotin auxotroph would requirement more biotin to support growth than do biotin auxotrophs of E. coli K-12 and B. subtilis 168. This was tested by use of two ΔbioBFDA strains (FYJ283 and FYJ284) and two control strains, the wild type strain NTL4 and the ΔbioR::Km mutant (FYJ212) (Table S2). As expected the wild type and bioR mutant strains both grew well without biotin addition whereas the ΔbioBFDA strain failed to grow on minimal media lacking biotin. Upon addition of biotin 1 and 2 nM biotin gave significant growth and 5 nM biotin resulted in normal growth (Fig. 7). Since E. coli and B. subtilis biotin auxotrophs show normal growth with 4 nM biotin, the biotin requirement of A. tumefaciens is very comparable. The finding that A. tumefaciens grows at low nM biotin concentrations indicates the bacterium has an efficient uptake system for the vitamin. One might expect A. tumefaciens to have a somewhat greater biotin requirement than the other two bacteria because it has six putative biotin-modified proteins (Fig. S8) whereas E. coli has only a single biotinylated protein and B. subtilis has three. However, the A. tumefaciens proteins may not have to be fully biotinylated to support normal growth as was seen in E. coli (Cronan, 2001). Note that a system in which BioR interacts with BirA could provide a plausible regulatory system. However, our attempts to detect interaction by a bacterial two-hybrid system were unsuccessful (Fig. S9).
Table 1. Comparison of biotin levels in A. tumefaciens with those of E. coli and B. subtilis
Level (× 103 cfu−1)
The data are expressed in average ± standard deviation (SD).
cfu, colony-forming units.
No. of biotin molecules excreted
22.48 ± 9.65
0.72 ± 0.21
754.29 ± 373.53
Cytosolic biotin molecules
9.61 ± 4.57
9.73 ± 2.65
11.59 ± 2.18
No. of protein-bound biotin molecules
0.16 ± 0.11
1.66 ± 0.55
1.26 ± 0.36
The regulatory system used by A. tumefaciens to regulate biotin synthesis appears to be either at a rudimentary stage of development or has degenerated from a functional system. Although BioR binds its operator tightly in vitro, it is a weak repressor in vivo unless overexpressed. One reason for the very modest regulation by BioR is that the operator appears to lie within the bioB coding sequence. Although operators are found in similar positions in well-studied operons such as the E. coli lac (Oehler et al., 1990) and gal (Semsey et al., 2002) operons, repressor molecules bound to such sites generally exert their effects by interaction with other repressor molecules bound to a promoter-proximal site to give a high local concentration of repressor and DNA looping. This raises the possibility that A. tumefaciens may have not yet acquired (or has lost) an upstream site or that the site is so weak that BioR can only bind it when overproduced. However, there are examples of proteins (such as the Bacillus subtilis CodY repressor) that act as roadblocks of transcript elongation by binding at single site located well downstream of the transcription initiation site (Belitsky and Sonenshein, 2011).
A second shortcoming of BioR is that very high and non-physiological biotin concentrations are required to exert regulation. For these reasons, biotin operon regulation in A. tumefaciens seems a work in progress. Whether it is progressing towards tight regulation or towards complete loss of regulation is indeterminate. However, it remains possible that profligate synthesis of biotin provides some advantage to A. tumefaciens in its biological niche. Perhaps this provides biotin to other soil bacteria unable to make the cofactor and these recipients, in turn, somehow aid growth or survival of A. tumefaciens.
Bacterial strains and growth conditions
The E. coli strains used were all E. coli K-12 derivatives (Table S1) and the A. tumefaciens strains were all derived from NTL4, a genetically modified strain with sensitivity of tetracycline (Goodner et al., 2001; Luo et al., 2001; Wood et al., 2001). Generally, most of E. coli strains were grown at 37°C, whereas A. tumefaciens strains were maintained at 28–30°C. The temperature sensitive birA (Ts) strain BM4062 was grown at 30°C (Barker and Campbell, 1980). DH5α (λ-pir) was used for maintenance of derivatives of pWM91 having the oriR6K replication origin (Table S1) whereas DH10B (Novagen) was used for cloning and propagating large plasmids such as pRG970 and its derivatives (Van den Eede et al., 1992; van Dillewijn et al., 2001) (Table S1). The E. coli LB and RB media were described previously. The minimal medium was M9 medium supplemented with 0.4% glucose or other carbon source, 0.1% Vitamin-Free Casamino Acids and 0.001% thiamine. Although LB, RB and the minimal medium can support growth of A. tumefaciens, only mannitol-glutamate/Luria medium (5 g of mannitol, 1.16 g of monosodium glutamate, 12.5 g of LB broth powder, 0.1 g of MgSO4 and 1 mg biotin per litre, pH 7.0) was used for growth prior to preparation of competent cells (Cangelosi et al., 1991; van Asma, 1995). Antibiotics were added as required at the following concentrations (in mg l−1): sodium ampicillin, 100 for E. coli; sodium carbenicillin, 100 for A. tumefaciens; kanamycin sulphate, 25 for E. coli and 50 for A. tumefaciens; gentamicin sulphate, 50 for E. coli and 30 for A. tumefaciens and tetracycline HCl, 15 for E. coli and 3 for A. tumefaciens.
Plasmids and DNA manipulations
The pCR2.1-TOPO vector (Invitrogen) was used for PCR cloning and sequencing and the resulting plasmids were propagated in strain Top10 (Table S1). Two plasmids (28a-birA and 28a-bioR) were transformed into BL21 (DE3) strain for in vitro production of BirA and BioR proteins (Table S1). THe arabinose-inducible pBAD24 vector (Guzman et al., 1995) was used for functional complementation in E. coli, whereas the IPTG-inducible pSRKGm expression vector (Khan et al., 2008) was employed for functional analyses in A. tumefaciens. The birA coding sequence was amplified with the primers birA_at-F2 plus birA_at-R2 (Table S2) and the product was directly inserted into EcoRI and SalI sites of pBAD24 (Table S1). The bioR coding sequence was amplified with primers bioR_at-CF plus bioR_at-CR with (Table S2) and the product was inserted between the NdeI and NheI sites of pSRKGm (Table S1). The DNA fragment containing the bioB promoter region was amplified with primers PbioB_at-F2 plus PbioB_at-R2 (Table S2) and the product was ligated into the SmaI and BamHI sites of pRG970, a low-copy lacZ reporter vector (Van den Eede et al., 1992; van Dillewijn et al., 2001), giving pRG-PbioBat (Table S1).
Two different strategies were used for gene disruption in A. tumefaciens. To generate the ΔbioR mutant a gene disruption plasmid, pAW-LKR was constructed in which a kanamycin-resistance cassette is located between the two flanking bioR sequences (Table S1). For removal of the bioBFDA cluster, a marker-less knockout plasmid pAW-bio was assembled from DNA fragments located upstream and downstream of the targeted bioR (and/or bioBFDA) obtained by amplification with Platinum Pfx-DNA polymerase (Invitrogen). The kanamycin-resistance cassette of plasmid pKD4 was amplified with primers Km-F plus Km-R (Table S2). Overlapping PCR was carried out to make the fusion DNA fragments (LKR, and bio-up/do), each of which was directly cloned into the sucrose-sensitive suicide vector pWM91 via the two sites of XmaI and SpeI, giving pAW-LKR and pAW-bio respectively (Table S1). Note that 5% dimethylsulphoxide was added to all these PCR reaction systems to cope with the high GC percentage of A. tumefaciens genomes. All of the recombinant plasmids were verified by PCR detection, restriction enzyme digestion or direct DNA sequencing.
Expression, purification and identification of BirA and BioR proteins
The BirA and BioR hexahistidine-tagged proteins were produced in E. coli BL21 (DE3) carrying the appropriate expression plasmids (pET28a-birA and pET28a-bioR, Table S1). Induction of bacterial cultures was done with 0.3 mM IPTG at an OD600 of 0.6–0.8 at 30°C for 5 h gave partially soluble proteins (Feng and Cronan, 2009b; 2010; 2011b). The bacterial cells were pelleted by centrifugation (4200 g, 10 min), washed twice with ice cold PBS buffer (101.4 mM Na2HPO4, 1.8 mM KH2PO4, 137 mM NaCl, 2.7 mM KCl, 8% glycerol, pH 7.4), and dissolved in PBS buffer containing 20 mM imidazole was added. The clarified supernatant obtained following lysis in a French pressure cell and removal of bacterial debris by centrifugation (16 000 g for 30 min) was loaded onto a nickel chelate column (Qiagen). After washing with 20 column volumes of with PBS buffer containing 50 mM imidazole, the BirA (or BioR) protein was eluted with 150 mM imidazole. Appropriate protein fractions were pooled and dialysed against PBS buffer then concentrated by ultrafiltration (10 kDa cut-off, Amicon Ultra) (Feng and Cronan, 2010). The protein purity was judged by 12% SDS-PAGE, followed by staining with Coomassie brilliant blue R250 (Sigma). Finally, both BirA and BioR was further determined by liquid chromatography quadrupole time-of-flight mass spectrometry of tryptic peptides as described previously (Feng and Cronan, 2011a,b).
Chemical cross-linking assays
To reveal the solution structure of both BirA and BioR, chemical cross-linking assays with ethylene glycol bis-succinimidylsuccinate (Pierce) were conducted routinely as described (Feng and Cronan, 2010). In each chemical cross-linking reaction (15 μl in total), the purified BirA (or BioR) protein (∼ 5 mg ml−1) was incubated with cross-linker at various concentrations (from 1.0 to 20 μM) and kept 30 min at room temperature. Finally, the reaction products were subjected to SDS-PAGE separation.
Assays of in vitro BirA biotinoylation activity
The reaction of in vitro BirA catalysed biotinoylation was performed as previously reported (Chapman-Smith et al., 1994; Choi-Rhee et al., 2004; Chakravartty and Cronan, 2012) with minor modifications. The components in this assay (20 μl in total volume) were added as follows: Tris-HCl, 50 mM; KCl, 100 mM; MgCl2, 5.5 mM; ATP, 25 μM; DTT, 0.1 mM; biotin, 100 μM; a B. subtilis acceptor protein (BCCP86), 50 μM and BirA, 2.5 μM. For visualization of the biotinoyl-adenylate (bio-5′-AMP) intermediate, 3 μCi [α-32P]-ATP (American Radio-labeled Chemicals) was also added to the reaction mixture. Biotinoylation reactions plus or minus [α-32P]-ATP were incubated at room temperature. After about 1 h 1 μl of each reaction mixture was spotted on a thin-layer chromatographic plates (Analtech TLC Uniplates microcrystalline cellulose matrix) and the plates were developed in isobutyric acid-NH4OH-H2O (66:1:33). The dried TLC plates were exposed to ECL film (Amersham) at −80°C overnight for signal capture. To further differentiate the biotinoylated BCCP86 from its apo-form, the above reaction mixture without [α-32P]-ATP was dialysed against 2 mM ammonium acetate, dried under a stream of nitrogen and subjected to electrospray mass spectral analysis (Chapman-Smith et al., 1994).
In vitro production of BioR
To obtain functional BioR by a different method than in E. coli, we turned to the PURExpress in vitro protein synthesis kit (New England Biolabs), which is a cell-free transcription/translation system in which the components are well defined. As before the reaction system (total volume, 25 μl) was set up, in which 1 μg of purified pET28a-bioR plasmid (Table S1) was mixed with 12.5 μl of solution A and 5.0 μl of solution B (Feng and Cronan, 2009b; 2011a). The synthesized BioR protein was immediately used for gel shift assays.
Electrophoretic mobility shift assays
The assays of the interaction between the BioR and the promoters of the A. tumefaciens bio operon and those of other closely related α-proteobacteria were done essentially as previously reported (Feng and Cronan, 2009b; 2010; 2011b). In total, DNA probes containing the predicted BioR sites were synthesized by annealing two complementary primers (Table S2) by incubation in 10 mM Tris-HCl, 1 mM EDTA, 100 mM NaCl, pH 8.0 at 95°C for 15 min followed by slow cooling to 25°C and then digoxigenin (DIG) labelling by terminal transferase with DIG-ddUTP (Roche). The DIG-labelled DNA probes (0.2 pmol) were incubated with or without BioR protein in binding buffer (Roche) for 15 min at room temperature and then separated by native PAGE (8%) and visualized by exposure to ECL film (Amersham) as previously described (Feng and Cronan, 2010; 2011a; 2012).
To detect whether BioR expressed in E. coli contains bound biotin or a biotin derivative bioassays were conducted using the biotin auxotroph of E. coli, ER90 (ΔbioF bioC bioD) (Lin et al., 2010). The biotin assay plates were prepared as previously with minor improvements (Lin et al., 2010). Overnight cultures of strain ER90 (grown in 5 ml of defined M9 minimal medium with 1 nM biotin at 30°C) were collected by centrifugation, washed twice with 5 ml of M9 medium and grown in 100 ml of minimal medium lacking biotin for 9 h at 37°C to starve the cells for biotin. The bacteria were harvested by centrifugation, washed three times with M9 medium, suspended in 1 ml of the same medium and mixed into 150 ml of the defined M9 agar media supplemented with 0.1% (w/v) 2, 3, 5-triphenyl tetrazolium chloride, as a redox indicator. Finally, the mixture (5 ml per sector) was poured into Petri dishes sectored with plastic walls to avoid cross-feeding and a sterile paper disk (6 mm, BBL) was centred on the agar top of each sector.
To release the possible biotin bound by BioR protein, the tested sample were boiled in water for 20 min. In addition to the positive control biotin, 20 μl of protein samples were spotted onto the disk and kept at 30°C overnight. The growth of strain ER90 was visualized as a red deposit of formazan, indicating the presence of biotin on the disk. The diameter of the red deposit is proportional to the amount of biotin spotted on the disk.
To directly measure biotin pools produced by A. tumefaciens, E. coli and B. subtilis, overnight bacterial cultures grown in the defined minimal media (∼ 220 ml) were collected through centrifugation (4200 g, 30 min). The acquired supernatant cultures were further filtered with 0.22 μm filter (MILLEX-MP, Millipore) and kept for detection of biotin secreted outside of bacterial cells. Following three rounds of washing with 20 ml of minimal media, the bacterial pellet of each species was divided equally into two parts: one part was directly treated with 8 ml of 3M HCl and further hydrolysed in autoclave oven (121°C, 30 min), the other half dissolved in 15 ml of 100% TCA was subjected for centrifugation (18 000 g, 36 min) to harvest protein pellets. Similarly, the protein pellet dried in air was hydrolysed with 4 ml of 3M HCl in the autoclave for 30 min. Subsequently, the pH of each hydrolysis solution was adjusted to about 7.0, and concentrated appropriately. The former piece of hydrolysis solution represents the pool of biotin distributed in cytoplasm, whereas the latter part means to protein-bound biotin pool. Finally, the three samples were spotted on the bioassay discs as above. Different concentrations of biotin spotted on the same discs were used as a reference standard to estimate the level of biotin and its derivatives.
Transformation of A. tumefaciens
Electro-competent cells were prepared as previously described (Cangelosi et al., 1991) with minor changes. A single colony of A. tumefaciens was grown overnight in 2 ml of MG/L medium and then diluted 1:100 into 20 ml of MG/L medium in a 200 ml of flask with vigorous shaking at 30°C. When the OD600 reached 0.5–0.6, the cultures were chilled on ice for 30 min and then collected by centrifugation at 3600 g for 12 min at 4°C. The harvested pellets in each tube were suspended in 2 ml of iced-cold 1 mM HEPES (pH 7.0) and then centrifuged at 3800 g for 16 min at 4°C. Following two rounds of washing the cells with the same buffer 1 mM HEPES containing 10% (vol/vol) glycerol (pH 7.0) was used for the final wash. Finally, the bacterial pellets were suspended in 0.5 ml of ice-chilled HEPES plus 10% glycerol, divided into small aliquots which were frozen on dry ice and kept at −80°C until use.
Prior to electroporation, the frozen cell suspensions were thawed on ice, mixed with plasmid DNA (> 1 μg for gene disruption, 50–100 ng for delivery of plasmids) and then placed on ice for no more than 30 min prior to transfer into a chilled 2 cm gap cuvette (Bio-Rad). Electroporation parameters for the Gene Pulser apparatus (Bio-Rad) were 25 μF capacitance, 400 Ω pulse controller units and 2.5 kV electrical pulse to give a field strength of 12.5 kV cm−1 with an exponential decay constant of approximately 9–10 ms. The pulsed cells were quickly diluted with 600 μl of MG/L medium, grown at 30°C for 2 h and plated on MG/L agar plates containing appropriate antibiotics (The plates were incubated at 30°C for 36–48 h). Transformants containing reporter plasmids were verified by colony PCR and assay of β-galactosidase activity.
Disruption of bioR and bioBFDA and functional complementation
Three mutant strains of A. tumefaciens were constructed that included FYJ212 (ΔbioR::Km), FYJ283 (ΔbioBFDA), and a double mutant strain FYJ284 (ΔbioR::Km ΔbioBFDA) (Table S1 and Fig. S7). To generate the ΔbioR::Km strain the suicide plasmid pAW-LKR containing two antibiotic resistance cassettes (AmpR and KmR) plus a sacB gene was electroporated into competent cells of A. tumefaciens NTL4 and plated on MG/L plates with selection for resistance to carbenicillin (100 μg ml−1) and kanamycin (50 μg ml−1). In a second of screening, the candidate colonies were streaked on the same selective medium and onto MG/L plates containing 10% sucrose plus kanamycin (50 μg ml−1). Colonies that grew well on the latter plates, but not in the presence of carbenicillin were likely to be the desired ΔbioR::Km strains. Multiplex PCR assays were conducted using primers, bioR_check-F plus bioR_check-R and Km-F plus Km-R (Table S2) to confirm disruption of bioR. DNA sequencing of PCR products was also used to confirm the mutation.
In order to be able to manipulate biotin levels in A. tumefaciens two ΔbioBFDA biotin auxotroph strains were constructed (Table S1 and Fig. S7). The suicide plasmid pAW-bio was electroporated into the WT strain NTL4 or the ΔbioR::Km strain FYJ212 (Table S1) transformants resistant to carbenicillin or carbenicillin plus kanamycin respectively. In the procedure of second round screening for the double mutant strain FYJ284 (Δbio ΔbioR::Km), the disruptions were further screened and validated as above.
RNA isolation and RT-PCR
Total RNA preparations were isolated from the mid-log phase cells of A. tumefaciens NTL4 grown in either MG/L medium or defined M9 minimal medium using the RNeasy bacterial RNA isolation kit (Qiagen). After treatment with RNase-free DNase I (at room temperature for ∼ 1 h), the RNA samples were analysed by agarose gel electrophoresis to assess the quality of the rRNA preparations. To further rule out the possibility of trace DNA contamination of the RNA samples (Feng and Cronan, 2009b; 2010), a general PCR-based detection was also conducted using total RNA as template with primers 16S_at-F plus 16S_at-R (Table S2).
Reverse transcription-PCR experiments were performed on the validated RNA preparations as follows: 1 μg of RNA was mixed with 0.5 μg of random primers (11 μl in total), denatured (70°C for 5 min), and then chilled on ice (5 min). Reverse transcription (RT) was conducted with a reaction mixture (20 μl total volume) of denatured RNA template, 10 μl; random primers, 1 μl; ImProm-II 5× reaction buffer, 4 μl; MgCl2, 2.5 μl; deoxynucleoside triphosphate mix, 1 μl; recombinant RNasin RNase inhibitor, 0.5 μl; ImProm-II reverse transcriptase, 1 μl. The program of RT reaction is consisted of equilibration at 25°C for 5 min, extension at 42°C for 60 min, and inactivation of enzyme at 70°C for 15 min. The resulting cDNA (1 μl) served as template for PCR amplification of the biotin synthesis/sensing-related genes (e.g. bioR and bioBFDA) using specific primers (Table S2) on an Eppendorf thermal cycler.
Real-time quantitative RT-PCR
To address the altered expression of bioBFDA operon of A. tumefaciens in response to varied biotin concentrations of its growth environment, real-time quantitative PCR (qPCR) assays were conducted, based on the of SYBR Green dye method as previously reported (Feng and Cronan, 2009b; 2010). The qPCR reaction system (20 μl) contained the following components 12.5 μl of iQ™ SYBR Green Supermix, 1 μl each of the primers, 1 μl of the diluted cDNA sample, and 4.5 μl of sterile water. Data analyses were done in triplicate on a Mastercycler ep realplex (Eppendorf), instrument using a program of denaturing cycle at 95°C for 15 min, 45 cycles comprising 94°C for 20 s, 60°C for 20 s, and 72°C for 20 s, and a final step in which temperature is elevated in gradient from 60°C to 90°C for dissociating double stranded DNA products. The internal reference was the16S rRNA gene (Table S2) and water functioned as blank control to monitor cross-contamination of different cDNA samples. The relative transcription levels were measured by the method of Livak and Schmittgen (2001).
5′-RACE mapping of the bioB transcription start site
The tobacco acid pyrophosphatase-based 5′-RACE technique, RLM-RACE (Ambicon), was used to determine the transcription start site of A. tumefaciens bioB (Feng and Cronan, 2011a,b). The nested PCR reactions were established using the Outer Primer plus bioB-GSP and the Inner Primer plus bioB-nested primer was adopted (Table S2). The PCR program was a denaturing cycle at 95°C for 5 min followed by 35 cycles comprising 95°C for 30 s, 55°C for 30 s, and 72°C for 30 s. The purified PCR products were cloned into the pCR2.1 TOPO vector (Invitrogen) for direct DNA sequencing. The transcriptional start site was assigned to first nucleotide adjacent to the RLM-RACE adaptor (Feng and Cronan, 2009a,b).
Mid-log phase cultures in RB or defined M9 minimal media (with or without supplementation with thiamine or biotin), were collected by centrifugation, washed twice with Z Buffer (Miller, 1972) and assayed for β-galactosidase activity after lysis with sodium dodecyl sulphate-chloroform (Miller, 1972). The data were recorded in triplicate with more than three independent experiments.
The work was supported by National Institutes of Health (NIH) Grant AI15650 from National Institute of Allergy and Infectious Diseases (NIAID). We are grateful to Dr Peter Yau (Biotechnology Center, University of Illinois) for his technical assistance in Q-TOF. We thank the members of Dr S. Farrand's lab for sharing with us their expertise in genetic manipulation of A. tumefaciens and providing the NTL4 strain and relevant plasmids. The plasmid pAW19 and the bacterial two-hybrid system were kindly provided by the Metcalf and Slauch labs of the Department of Microbiology, University of Illinois at Urbana-Champaign.