A conserved regulatory mechanism in bifunctional biotin protein ligases

Abstract

Class II bifunctional biotin protein ligases (BirA), which catalyze post‐translational biotinylation and repress transcription initiation, are broadly distributed in eubacteria and archaea. However, it is unclear if these proteins all share the same molecular mechanism of transcription regulation. In Escherichia coli the corepressor biotinoyl‐5′‐AMP (bio‐5′‐AMP), which is also the intermediate in biotin transfer, promotes operator binding and resulting transcription repression by enhancing BirA dimerization. Like E. coli BirA (EcBirA), Staphylococcus aureus, and Bacillus subtilis BirA (Sa and BsBirA) repress transcription in vivo in a biotin‐dependent manner. In this work, sedimentation equilibrium measurements were performed to investigate the molecular basis of this biotin‐responsive transcription regulation. The results reveal that, as observed for EcBirA, Sa, and BsBirA dimerization reactions are significantly enhanced by bio‐5′‐AMP binding. Thus, the molecular mechanism of the Biotin Regulatory System is conserved in the biotin repressors from these three organisms.

Introduction

Cellular homeostasis relies on communication between metabolism and gene expression. The bifunctional biotin protein ligases (BirA) link metabolic demand for and production/uptake of biotin.1, 2, 3, 4 In the E. coli system protein : protein interactions are critical for effecting this linkage. The ligase binds to biotin followed by ATP to produce biotinyl‐5′‐adenylate (bio‐5′‐AMP)5 (Fig. 1). Adenylate binding to BirA enhances the equilibrium association constant and Gibbs free energy of dimerization of the protein by 1000‐fold and −4 kcal/mol, respectively.6, 7 Given that homodimerization is a kinetic prerequisite to biotin operator (bioO) binding, transcription repression is regulated by holoBirA dimer availability8 (Fig. 1). When metabolic demand for biotin is high, holoBirA dimerization does not occur because the monomer preferentially binds to the abundant acetyl‐CoA carboxylase, resulting in biotin transfer to the biotin carboxyl carrier protein (BCCP) subunit to activate the carboxylase for fatty acid biosynthesis (Fig. 1). At low biotin demand, when the apoBCCP concentration is decreased, holoBirA monomer accumulates, dimerizes and binds to bioO to repress transcription of the biotin biosynthetic operon, thereby limiting biotin production when the cellular demand is low.

Figure 1.

The Escherichia coli biotin regulatory system: BirA catalyzes synthesis of bio‐5′‐AMP from biotin and ATP and the resulting holoBirA either transfers biotin to apoBCCP or forms a homodimer that binds to the biotin operator to repress transcription.

Bifunctional ligases are present in a broad range of bacteria and archaea.4 While some of these ligases regulate only transcription of the biotin biosynthetic genes, others are predicted to regulate expression of biosynthetic genes and/or those that code for biotin transport into the cell.4 Few of these proteins have been subjected to biochemical studies. Studies of the Staphylococcus aureus and Bacillus subtilis enzymes provide a mixed picture of the functional energetics of the Biotin Regulatory System in these organisms. Both in vivo and in vitro measurements performed on the B. subtilis BirA (BsBirA) indicate that, like the E. coli enzyme (EcBirA), transcription regulation and biotin operator binding are very sensitive to biotin concentration.9, 10 However, measurements of S. aureus BirA (SaBirA) binding to its cognate operator site using electrophoretic mobility shift assays (EMSA) indicate a relatively small difference in the overall binding energetics for the bio‐5′‐AMP‐bound and ligand free forms of the protein, a result that predicts little sensitivity of transcription regulation to biotin concentration.11 By contrast, more recent fluorescence anisotropy measurements of the binding reveal a large difference in the bioO binding energetics of the two species.10

Repression complex assembly by the EcBirA occurs in two steps including dimerization followed by bioO binding.8 Moreover, in the E. coli system, the large difference in the overall two‐step repression complex assembly energetics in the absence and presence of bio‐5′‐AMP results from the enhanced dimerization of holoBirA relative to apoBirA.7, 12 Previous measurements suggest that this is not the case for SaBirA. Homodimerization measurements performed using combined sedimentation velocity and equilibrium methods indicate that apoSaBirA dimerizes in the micromolar range of protein concentration and with energetics identical to those of the biotin‐bound species.11 By contrast, apo‐ and biotin‐bound EcBirA dimerize in millimolar concentration range.7 No measurements of holoSaBirA dimerization have been published. Given that SaBirA, like EcBirA, represses transcription in vivo in a biotin‐dependent manner,10 the apparent relatively tight dimerization of the apo‐repressor is difficult to rationalize because it should enable DNA binding and repression in the absence of biotin.

In this work, we have used sedimentation equilibrium to measure coupling between small ligand binding and homodimerization of the bifunctional Sa and BsBirA proteins. Results of these measurements indicate that these two proteins behave similarly to the E. coli enzyme in coupling between ligand binding and self‐association. The one exception is that biotin alone, which has little effect on EcBirA dimerization, enhances SaBirA dimerization. The results support conservation of the mechanism of response to biotin concentration for the S. aureus, B. subtilis, and E. coli Biotin Regulatory Systems.

Results

Sa and BsBirA preparations are free of biotin or bio‐5′‐AMP contamination

Since biotin protein ligases exhibit coupling between small ligand binding and dimerization, unambiguous interpretation of sedimentation equilibrium data obtained on any of these proteins requires knowledge of the ligation state. Previous studies indicate that biotin or bio‐5′‐AMP may co‐purify with biotin protein ligases and a standard protocol developed in this laboratory,13 which involves exposure of the ligases to a molar excess of apoBCCP in the presence of ATP, can eliminate such contamination.9, 10, 13 The Sa and BsBirA preparations used in the current studies were subjected to this published protocol. Before performing sedimentation measurements the proteins were assayed for the presence of contaminating biotin or bio‐5′‐AMP.14 In the assay each purified ligase was incubated with ATP and apoBCCP in the absence and presence of biotin, and the resulting products were subjected to analysis by MALDI‐TOF mass spectrometry. The spectra reveal that in the absence of added biotin no biotinylated BCCP was detected for either Sa or BsBirA. By contrast, the expected mass shift of 225 ± 1 Da was observed in the control reaction to which biotin had been added (Table 1, Fig. 2). Thus, neither the Sa or BsBirA preparation used in the measurements reported in this work was contaminated with biotin or bio‐5′‐AMP.

Table 1.

Maldi‐ToF Analysis of BCCP Samples from Biotinylation Reactions

	−biotin
BirA species	Mass (daltons)	+biotin
SaBirA	9339 ± 1	9565 ± 1
BsBirA ‐Enzyme	9343 ± 1 9348 ± 1	9567 ± 1

Reactions were carried out in Standard Buffer at 20°C for 16 h and prepared for MALDI‐ToF MS as described in “Materials and Methods.” All reactions contained 500 μM ATP.

Figure 2.

Maldi‐ToF spectra of biotinylation products: reactions containing 20 μM E. coli apo‐BCCP, 500 μM ATP, and (A) 20 μM EcBirA, 40 μM biotin; (B) 20 μM SaBirA, 40 μM biotin; (C) 20 μM SaBirA; (D) 20 μM BsBirA, 40 μM biotin; (E) 20 μM BsBirA; and (F) 40 μM biotin were incubated at 20°C overnight for 16 h. The products were subjected to analysis using a SHIMADZU Axima‐CFR MALDI‐TOF instrument.

Dimerization measurements on SaBirA indicate coupling between small ligand binding and homodimerization

Coupling between small ligand binding and SaBirA homodimerization was investigated using sedimentation equilibrium measurements. For all species each measurement was carried out on protein solutions prepared at three concentrations that were subjected to centrifugation at three speeds. Moreover, the concentration versus radius profile data obtained for all species were of high quality [Fig. 3(A)].

Figure 3.

Absorbance versus radius profiles from sedimentation equilibrium measurements carried out on (A) 13, 10, 7 µM holoSaBirA or (B) 12, 9, 6 µM holoBsBirA at 17k (red), 20k (green), and 23k (blue) rpm. Solid lines are best‐fits to a monomer‐dimer model obtained from global analysis of nine datasets. The lower panels provide the residuals of the fit.

Results of the measurements performed on apoSaBirA indicate that it undergoes no detectable dimerization in the conditions used for the measurements. Analysis of data obtained at relatively low protein concentrations of approximately 10 μM indicates that oligomerization state of apoSaBirA is well‐described by a single species model, with an average molecular weight of 37 ± 1 kDa, consistent with the monomer molecular weight predicted from the sequence. In an effort to obtain an estimate of the apoSaBirA dimerization constant, measurements were performed at protein concentrations up to 100 μM. Analysis of the resulting data indicates that, even at these high concentrations, apoSaBirA is best described as a single species with a molecular weight consistent with that expected of the monomer [Table 2, Fig. S1(A)]. Attempts to analyze the data using a monomer‐dimer model indicate an unrealistically large equilibrium dissociation constant (K _D) and the fit to the model, as judged by the magnitude of the variance of the fit, that not as good as the fit to a single species model (Table 2). The data are consistent with a K _D for apoSaBirA dimerization in the millimolar or higher concentration range, which yields a lower limit for the Gibbs free energy of apoSaBirA dimerization of −4.0 kcal/moln (Table 3).

Table 2.

Testing Sedimentation Equilibrium Data for Sa and BsBirA Against Single Species and Monomer‐Dimer Association Models

			Single species		Monomer‐Dimer
Protein	Ligand	Initial concentrations (μM)	Weight average MW (kDa)	Variancec	Variancec
SaBirAa	–	100, 90, 80	37 ± 1	0.0057	0.0060
	biotin	80, 70, 60	54 ± 2	0.0073	0.0070
	bio‐5′‐AMP	12, 9, 6	70 ± 2	0.0072	0.0060
SaBirAb	–	17, 14, 11	37 ± 2	0.0042	0.0042
	bio‐5′‐AMP	12, 9, 6	60 ± 2	0.0062	0.0058
BsBirAa	–	13, 10, 7	43 ± 1	0.0059	0.0057
	biotin	13, 10, 7	43 ± 1	0.0046	0.0043
	bio‐5′‐AMP	13, 10, 7	68 ± 2	0.0058	0.0054

Measurements were carried out in (a) Standard Buffer (10 mM Tris–HCl, 200 mM KCl, 2.5 mM MgCl₂, pH = 7.50 ± 0.02 at 20°C) or (b) 20 mM Tris, 150 mM NaCl, pH = 8.0 at 20°C at three different rotor speeds. (c) Square root of the variance of the fit.

Table 3.

Thermodynamic Parameters Governing Sa and BsBirA Dimerization

Protein	Ligand	BirA loading concentration range (μM)	$K_{\mathrm{D}}^{\mathrm{c}}$ (μM)	ΔG d (kcal/mol)	Maximum fraction dimere (%)
SaBirAa	–	100–5	N/A	N/A	N/A
	biotin	100–40	150 ± 20	−5.1 ± 0.1	38
	bio‐5′‐AMP	19–5	3 ± 1	−7.6 ± 0.2	64
SaBirAb	–	17–4	N/A	N/A	N/A
	bio‐5′‐AMP	12–6	9 ± 3	−6.8 ± 0.2	49
BsBirAa	–	25–7	300 ± 200	−4.7 ± 0.3	7
	biotin	25–7	300 ± 100	−4.8 ± 0.2	7
	bio‐5′‐AMP	20–7	4 ± 1	−7.4 ± 0.2	56

Measurements were carried out in (a) Standard Buffer (10 mM Tris–HCl, 200 mM KCl, 2.5 mM MgCl₂, pH = 7.50 ± 0.02 at 20°C) or (b) in 20 mM Tris, 150 mM NaCl, pH = 8.0 at 20°C at three different rotor speeds. (c) The reported equilibrium dissociation constants and Gibbs free energies for dimerization are mean and standard deviation of at least three independent measurements. (d) Gibbs free energies were calculated using equation ΔG = RT ln K _D. (e) The fraction dimer at highest protein concentration in the concentration versus radius profile.

Biotin and bio‐5′‐AMP binding significantly enhances SaBirA dimerization. For sedimentation equilibrium measurements performed on small ligand‐protein complexes, protein solutions were prepared in Standard Buffer at 1.5:1 molar ratio of the appropriate ligand to protein. The data obtained for the biotin‐bound protein are well‐described by a monomer‐dimer model as judged by both the distribution of the residuals and the variance of the fit [Table 2, Fig. S1(B)], and the analysis yields an equilibrium dissociation constant for dimerization of 150 ± 20 µM (Table 3). Sedimentation equilibrium measurements performed on holoSaBirA indicate even tighter dimerization for this species. Again, the data are well described by the monomer‐dimer model [Table 2, Fig. 3(A)] and the analysis yields an equilibrium dissociation constant of 3 ± 1 µM. Based on the equilibrium constants, the calculated Gibbs free energies for dimerization of biotin‐bound and bio‐5′‐AMP‐bound proteins are −5.1 ± 0.1 kcal/mol and −7.6 ± 0.2 kcal/mol, respectively (Table 3).

Dimerization properties of SaBirA are independent of buffer composition

Published measurements of SaBirA dimerization yielded results that differ significantly from those reported in the previous section.11 However, these measurements were carried out in buffer conditions different from those used in the current studies. The differences include the use of NaCl versus KCl as the monovalent salt at 150 mM compared to 200 mM in the current study and a 0.5 unit higher pH. To determine if the discrepancy in measured dimerization properties is due to the buffer conditions, sedimentation equilibrium analysis of SaBirA was repeated in a buffer identical to that used in the previous studies. Measurements performed on apoSaBirA indicate that the protein is monomeric [Table 2, Fig. S1(C)]. By contrast, data obtained with the bio‐5′‐AMP bound protein are well‐described by a monomer‐dimer model (Table 2). The equilibrium dimerization constant of 9 ± 3 µM obtained from the measurements is similar in magnitude to that obtained in Standard Buffer [Table 3, Fig. S1(D)].

Small ligand effects on BsBirA dimerization mirror those observed for EcBirA

Results of in vivo measurements of the biotin concentration‐dependence of transcription repression suggest that BsBirA and EcBirA share the same regulatory mechanism.9 To determine the relationship of these in vivo results to the BsBirA self‐association properties, sedimentation equilibrium measurements were carried out on apo, biotin‐bound, and holoBsBirA. Due to the limited solubility of BsBirA in Standard Buffer, relatively low protein concentrations ranging from 7 to 25 μM were used for all measurements. Nevertheless, the data obtained for all samples were of high quality [Fig. 3(B)].

Sedimentation equilibrium data obtained for apoBsBirA indicate that it dimerizes with a modest Gibbs free energy. The variances of the fits indicate that a monomer‐dimer model provides a slightly better description of the data than does a single species model [Table 2, Fig. S1(E)]. Moreover, the average molecular weight obtained from the single species fit of 43 ± 1 kDa is greater than the 39.1 kDa expected of the monomer [Table 2, Fig. S1(F)]. The equilibrium dissociation constant obtained from the data analysis is 300 ± 100 µM (Table 3). The biotin‐bound BsBirA also dimerizes weakly with a constant identical to that of apoBsBirA. Thus, consistent with observations made on EcBirA, biotin has no effect on BsBirA dimerization.6 In contrast, addition of bio‐5′‐AMP to BsBirA greatly enhances dimerization, yielding an equilibrium dissociation constant of 4 ± 1 μM [Table 3, Fig. 3(B)].

Coupling between Sa and BsBirA dimerization and ligand binding

The results of sedimentation equilibrium measurements on the Sa and BsBirA species can be used to calculate the coupling free energy between small ligand binding and dimerization for the two proteins. This coupling free energy, ΔG°_c, is defined as the difference between the Gibbs free energies for dimerization of the ligand‐bound and ligand‐free proteins [Fig. 4(A)]. The value for coupling between biotin binding and dimerization for SaBirA is estimated to be at least −1.1 kcal/mole, based on the lower limit for the Gibbs free energy of apoSaBirA dimerization of −4.0 kcal/mol. By contrast, consistent with EcBirA, no coupling is observed between biotin binding to BsBirA and its dimerization. Both Sa and BsBirA dimerization are significantly enhanced upon bio‐5′‐AMP binding. Again, using the lower limit estimate for the Gibbs free energy of apoSaBirA dimerization of −4.0 kcal/mol, the estimated coupling free energy associated with bio‐5′‐AMP binding is at least −3.6 kcal/mol. The calculated value for BsBirA is −2.7 ± 0.3 kcal/mol [Fig. 4(B)].

Figure 4.

Coupling between small ligand binding and dimerization. (A) Thermodynamic cycle demonstrating coupling between bio‐5′‐AMP (blue) binding and homodimerization of BirA (red). The coupling free energy, ΔG°_c,dim, is the difference between Gibbs free energies of holoBirA, ΔG°_dim,holo, and apoBirA, ΔG°_dim,apo., dimerization (B) Coupling free energies between biotin (red) and bio‐5′‐AMP (blue) binding for Bs, Sa, and EcBirA. *Previously published data.32

Discussion

Both Sa and BsBirA exhibit large thermodynamic coupling between bio‐5′‐AMP binding and dimerization. Sedimentation equilibrium analyses were performed on the apo, biotin‐bound, and bio‐5′‐AMP‐bound species of the two proteins. All measurements were performed on protein preparations that were free of biotin or bio‐5′‐AMP contamination, and excellent agreement was observed from the results of nonlinear least squares analysis of all data. In the absence of ligand, no apoSaBirA dimerization was detected, which allows an estimate of a lower limit of 1 mM for the equilibrium dissociation constant governing dimerization of this species. By contrast, SaBirA dimerization is enhanced upon the addition of either biotin or bio‐5′‐AMP, yielding estimates of the upper limits for the coupling free energies associated with the two ligands of −1.1 and −3.6 kcal/mol, respectively. Dimerization free energies of the BsBirA apo and biotin‐bound species, both of which can be accurately measured, are identical. However, addition of bio‐5′‐AMP leads to a significant enhancement of dimerization, consistent with a coupling free energy of −2.7 ± 0.2 kcal/mol. Thus, as previously demonstrated for EcBirA, dimerization of both Sa and BsBirA is significantly enhanced concomitantly with bio‐5′‐AMP binding.

The dimerization properties obtained for SaBirA in this work differ significantly from those previously reported. The current work indicates no detectable dimerization for apoSaBirA and a K _D of 150 ± 20 µM for the biotin‐bound species. By contrast, in the previous study the authors reported K _D values for apo and biotin‐bound SaBirA dimerization of 29 ± 2 µM and 30 ± 2 µM, respectively.11 Since the proteins used in both of the studies have C‐terminal His₆ affinity tags, the discrepancy cannot be explained by the absence or presence of a purification tag. The buffer conditions used for the previous studies differ from those used for the current measurement. However, sedimentation equilibrium measurements of apoSaBirA self‐association performed in this work in that same buffer also indicate no detectable dimerization. Moreover, consistent with the results summarized above, measurement of holoSaBirA dimerization in that buffer indicates large coupling between bio‐5′‐AMP binding and dimerization (Table 3). No measurements of holoSaBirA dimerization were included in the previous study. A possible source of the discrepancy in the results is the presence of residual biotin or bio‐5′‐AMP contamination in SaBirA preparations used in previous measurements. The authors indicated that as part of their purification protocol SaBirA protein was exposed to biotin acceptor protein to remove such contamination. They then used solid phase and immuno‐blotting methods to detect residual biotin contamination in their preparations.11 In this work a more direct mass spectrometry‐based method of measuring the shift in the mass of apoBCCP to that of holoBCCP also indicates no residual contamination of either the Sa or BsBirA preparation with biotin or bio‐5′‐AMP. It is possible that the methods employed by the previous authors were not sufficiently sensitive to detect contamination. Finally, the large coupling between bio‐5′‐AMP binding and SaBirA dimerization reported in this work is consistent with both the in vivo biotin concentration‐dependence of transcription repression and the recently reported effects of small ligands on sequence specific DNA binding to the Sa biotin operator sequence.10

The large coupling free energies between bio‐5′‐AMP binding and dimerization measured for Ec, Bs, and SaBirA support a conserved molecular mechanism of allosteric regulation for bifunctional biotin protein ligases/biotin repressors. The bifunctional ligases are widely distributed in eubacteria and archae.4 The EcBirA transcription repression function is regulated by combined demand for and supply of biotin.2, 3 In conditions of high biotin demand the bio‐5′‐AMP‐BirA complex is rapidly consumed in biotin transfer to the BCCP subunit of acetyl‐CoA carboxylase.15, 16 A decrease in biotin demand, which accompanies a decreased growth rate, allows accumulation of both biotin and the holoBirA dimer, which can bind to the biotin operator to repress biotin biosynthetic operon transcription.17 Henke and Cronan have demonstrated in in vivo measurements that, like EcBirA, both Sa and BsBirA show biotin‐concentration dependent transcription repression at their respective regulatory sequences.9, 10 They have, moreover, shown that over‐expression of the biotin acceptor proteins results in derepression of the transcription. Finally, they demonstrated that for each of these Class II biotin ligases, the overall affinity for the cognate operator site is dramatically increased in the presence of bio‐5′‐AMP. Results presented in this work indicate that, as previously reported for EcBirA, the enhanced repression complex assembly by Sa and BsBirA reflects bio‐5′‐AMP‐promoted dimerization.

The sequences of the three‐biotin repressor proteins suggest that conservation is not required to achieve similar allosteric function. Alignment of the three sequences reveals that relative to EcBirA Sa and BsBirA show 22 and 27% conservation, respectively (Fig. S2). Despite this low sequence identity, the three‐dimensional structures of Ec and SaBirA are very similar [Fig. 5(B)]. In the EcBirA structure, loops on the dimerization and ligand binding surfaces are known to play critical roles in bio‐5′‐AMP binding, dimerization, and coupling between the two processes.18, 19, 20, 21 Moreover, the disorder‐to‐order transitions that these loops undergo concomitant with bio‐5′‐AMP binding are key to allosteric activation of EcBirA dimerization. Nevertheless, with the exception of the glycine‐rich segment of the biotin‐binding loop that is required for biotin binding,22 the sequences of these loop regions in Sa, Bs, and EcBirA show no conservation [Fig. 5(A)]. Thus, it appears that in the bifunctional biotin ligases, similar functional allostery can be achieved with high sequence divergence.

Figure 5.

Sequence and structural comparison of Ec, Sa, and BsBirA. (A) Sequence alignment of Bs, Sa, and EcBirA surface loops on the ligand binding and dimerization surfaces. The output of the alignment obtained in Clustal Omega31 was used to generate the figure in JalView.33 Residues are colored based on the percentage identity with numbering from the EcBirA sequence. (B) Disorder‐to‐order transitions in SaBirA (top) and EcBirA (bottom) upon bio‐5′‐AMP (black) binding. The models were generated in Pymol34 using PDB files 3V8J:apoSaBirA,35 3V8L:holoSaBirA,35 1BIA:apoEcBirA,24 and 2EWN:holoEcBirA.25 Color coding for surface loops: SaBirA, 143–149, red; 196–202, green; 119–129, orange; 214–228, cyan; EcBirA, red, 140–146; green, 193–199; 116–124, orange; 211–222, cyan. Color coding for hydrophobic clusters: SaBirA, L216‐pink, F219‐yellow, I223‐brown, A227‐green; EcBirA, P126‐pink, F124‐orange, M211‐cyan, V214‐yellow, V218‐brown, V219‐green, W223‐white.

Little is known about the relationship of sequence to allosteric function. However, protein disorder is now well‐established as important for the thermodynamics of allostery,23 perhaps because it renders a protein segment poised to respond to an allosteric signal. In the disorder‐to‐order transition accompanying allosteric activation of EcBirA, hydrophobic side chains in the ligand binding surface loop condense around the bio‐5′‐AMP ligand to form a cluster24, 25 [Fig. 5(B)]. Perturbation of this cluster through alanine substitution compromises both bio‐5′‐AMP binding and its coupling to dimerization.19, 20 A similar clustering of hydrophobic residues in the ligand binding loop around bio‐5′‐AMP is observed in SaBirA, albeit with a very different sequence [Fig. 5(A)]. On the EcBirA dimerization surface the disorder‐to‐order transition accompanying allosteric activation includes extension of an α‐helix, and packing of the two neighboring loops. Although the sequences of the analogous loop segments diverge completely from those of EcBirA, the same helical extension and interloop packing are observed in comparison of the structures of apo and holoSaBirA [Fig. 5(B)]. Thus, in Ec and SaBirA even the detailed structural features of allosteric activation are conserved in the absence of sequence conservation.

Materials and Methods

Chemicals and biochemical

All reagents used were at least ACS grade. The d‐biotin (Sigma‐Aldrich) solutions were prepared in Standard Buffer containing 10 mM Tris (pH = 7.50 ± 0.02 at 20°C), 200 mM KCl, 2.5 mM MgCl₂, and stored at −80°C in 1 mL aliquots.5 Due to the absence of an absorption signal for biotin, these solutions were prepared by first weighing out the biotin using a microbalance and bringing the final solutions to full volume in a volumetric flask. Bio‐5′‐AMP was synthesized and purified as previously described and stored desiccated at −80°C.26, 27 The bio‐5′‐AMP solutions were prepared by dissolving lyophilized powder into Milli‐Q H₂O, aliquoting the solution into 1 mL volumes and stored at −80°C. The bio‐5′‐AMP concentration was determined by absorption spectroscopy at 259 nm using molar extinction coefficient of 15,400 M⁻¹cm⁻¹ and the integrity of the compound was checked by thin layer chromatography.

Expression and purification of Bs and SaBirA

Both Bs and SaBirA were expressed and purified using methods modified from previously published protocols.9, 10 The pET19b plasmid containing the N‐terminally histidine tagged BsBirA coding sequence or the pET28b plasmid containing the C‐terminally histidine tagged SaBirA coding sequence, both obtained from J. E. Cronan's laboratory, was transformed into E. coli strain BL21 (λDE3). Cells were grown at 37°C in LB broth containing 100 µg/mL ampicillin (BsBirA) or 50 µg/mL kanamycin (SaBirA). Once OD₆₀₀ of the culture reached 0.8, protein expression was induced for 12 h at 30°C by addition of IPTG to a final concentration of 1 mM.

All purification steps were performed at 4°C. For SaBirA purification, the cell pellet collected by centrifugation was resuspended in lysis buffer containing 50 mM HEPES (pH = 7.8 at 4°C), 250 mM NaCl, 0.1 mM dithiothreitol (DTT), 10 mM imidazole, 5% glycerol and lysed by sonication. The crude cell lysate was centrifuged again at 8360g to remove cellular debris and the resulting supernatant was loaded onto HisPur™ Ni‐NTA resin (ThermoFisher scientific). The resin was washed with at least 10 column volumes (CV) of Ni‐NTA Wash Buffer (lysis buffer containing 60 mM imidazole), and SaBirA was eluted in the same buffer using a linear 60–250 mM imidazole gradient. Column fractions were analyzed by SDS‐PAGE and those containing SaBirA were pooled, dialyzed against SP sepharose starting buffer (50 mM Tris–HCl, 50 mM KCl, 5% glycerol, 0.1 mM DTT, pH = 7.5 at 4°C), and the resulting sample was loaded onto SP Sepharose Fast Flow resin (GE Healthcare). The resin was washed with at least 10 CV of starting buffer, and SaBirA was eluted using a linear 20–800 mM KCl gradient prepared in the same buffer. To remove residual biotin or bio‐5′‐AMP, fractions containing pure SaBirA were pooled and dialyzed against lysis buffer, then loaded onto the Ni‐NTA resin. At least 20 CV of a solution comprised of 1 mM ATP, 1 μM E. coli apo‐BCCP, and 0.5 mM MgCl₂ in lysis buffer was applied to the column and exposed to the resin‐bound SaBirA for at least 2 h to allow for complete biotin transfer. After the incubation, the resin was washed with at least 10 CV of Ni‐NTA wash buffer, after which SaBirA was eluted with Ni‐NTA elution buffer (lysis buffer containing 250 mM imidazole). Finally, fractions containing SaBirA were pooled, dialyzed against storage buffer containing 50 mM Tris‐HCl (pH = 7.5 at 4°C), 200 mM KCl, 5% glycerol, 0.1 mM DTT and stored at −80°C in 1mL aliquots.

The BsBirA was purified using a method similar to that used for SaBirA with a few modifications. The lysis buffer composition was 50 mM Tris–HCl, 250 mM NaCl, 0.5 mM tris(2‐carboxyethyl)phosphine (TCEP), 10 mM imidazole, 5% glycerol, pH = 8.7 at 4°C. The crude cell lysate was centrifuged at 48,400g for 30 min prior to column chromatography. In addition, due to the protein's low solubility in SP sepharose starting buffer, SP sepharose column chromatography was replaced by a second Ni‐NTA column chromatography step. The procedure for removing biotin or bio‐5′‐AMP contamination was identical to that used for SaBirA. Purified BsBirA was dialyzed against storage buffer containing 50 mM HEPES (pH = 7.8 at 4°C), 250 mM NaCl, 0.5 mM TCEP, 5% glycerol and stored at −80°C in 1mL aliquots.

Protein concentrations were determined by UV absorbance using molar extinction coefficients of 39,420 M⁻¹cm⁻¹ (BsBirA) and 45,380 M⁻¹cm⁻¹ (SaBirA) at 280 nm calculated from amino acid compositions.28 All proteins were at least 95% pure as assessed by Coomassie‐staining of samples subjected to SDS‐PAGE.

Biotinylation assay

A biotinylation assay, which is a modification of a previously described method,14 was performed to ensure that purified Bs and SaBirA were not contaminated with biotin or bio‐5′‐AMP. Solutions containing 20 μM Bs or SaBirA, 20 μM E. coli apo‐BCCP, and 500 μM ATP were prepared in Standard Buffer in the absence or presence of 40 μM biotin. As a negative control, a reaction containing all components except biotin protein ligase was also prepared. The resulting solutions were incubated overnight for 16 h at 20°C. To prepare for matrix‐assisted laser desorption/ionization time of flight (MALDI‐TOF) mass spectrometry analysis, protein solutions were desalted and exchanged into the matrix (50 mM α‐Cyano‐4‐hydroxycinnamic acid in 70% acetonitrile, 30% water) using C‐18 Ziptips (Millipore). Spectra were acquired using an Axima‐CFR MALDI‐TOF (Shimadzu) mass spectrometer in linear mode, with insulin (Sigma‐Aldrich) and cytochrome c from horse heart (Sigma‐Aldrich) used as calibration standards.

Sedimentation equilibrium

All proteins subjected to sedimentation equilibrium measurements were dialyzed exhaustively against the appropriate buffer (Standard Buffer or 20 mM Tris, 150 mM NaCl, pH = 8.0 at 20°C) prior to measurements. After filtering the protein solutions through 0.22 μm PES syringe filters (Simsii), concentrations were determined by UV spectroscopy. Solutions were prepared at three final protein concentrations by dilution into dialysis buffer. For solutions containing biotin or bio‐5′‐AMP, the ligand was added immediately before centrifugation at a 1.5:1 ligand : protein molar ratio.

Sedimentation equilibrium measurements were performed using a Beckman Optima XL‐I Analytical Ultracentrifuge (Beckman Coulter). A four‐hole An‐60 Ti rotor (Beckman Coulter) and 12‐mm six‐channel cells equipped with charcoal‐filled Epon centerpieces and quartz windows were used in all experiments. Centrifugation was carried out at three speeds of 17,000, 20,000, and 23,000 rpm for 8 h at each speed and absorbance scans were acquired in step mode with five averages and a spacing of 0.001 cm. Overlay of a second scan acquired after an additional 1 h of centrifugation at each speed with the 8 h scan indicated that equilibrium had been achieved. The wavelengths used for data acquisition were 280, 295, or 300 nm, depending on the particular samples analyzed. At least three independent sedimentation measurements were carried out for each species, apo, biotin, and bio‐5′‐AMP, of Sa and BsBirA.

Data analysis

Sedimentation equilibrium data were analyzed using Nonlin29 in the Heteroanalyisis version 1.1.60 package downloaded from Center for Open Research Resources and Equipment (COR²E) at the University of Connecticut (http://core.uconn.edu/resources/biophysics). Absorbance versus radius profiles were globally analyzed using a single ideal species model to obtain the reduced molecular weight $\sigma$ using the following equation:

$$c\left(r\right)=c\left(r_0\right)e^{\sigma \frac{\left(r^2-r_o^2\right)}{2}}+\delta$$ (1)

where $c\left(r\right)$ is the protein concentration at radius $r$, $c\left(r_0\right)$ is the protein concentration at an arbitrary reference radial position $r_0$, and $\delta$ is the baseline offset. The best‐fit reduced molecular weight was used to calculate the weight average molecular weight $M$ of the sample using the following expression30:

$$\sigma =\frac{M(1-\overline{v}\rho )}{RT}{\omega }^2$$ (2)

where $\overline{v}$ is the partial specific volume of the protein, ρ is the buffer density, ω is the angular velocity of the rotor, R is the gas constant, and T is the temperature. Partial specific volumes of 0.7366 cm³/g for BsBirA and 0.7335 cm³/g for SaBirA and buffer densities were calculated using Sednterp (http://sednterp.unh.edu/).

Data were also analyzed using a monomer‐dimer self‐association model to obtain the equilibrium dissociation constant K _D for homodimerization:

$$c\left(r\right)=c\left(r_o\right)e^{{\sigma }_m\frac{\left(r^2-r_o^2\right)}{2}}+\frac{1}{K_D}{[c\left(r_o\right)e^{{\sigma }_m\frac{\left(r^2-r_o^2\right)}{2}}]}^2+\delta$$ (3)

where ${\sigma }_m$ is the reduced molecular weight for Bs or SaBirA monomer calculated from Eq. (2), and $r$, $c\left(r\right)$, $c\left(r_0\right)$, and $\delta$ have the same significance as indicated for Eq. (1). The reduced molecular weight of the dimer is assumed to be twice that of the monomer.

Sequence alignment

Sequences of 34 class II bifunctional protein ligases were selected4 and alignment was carried out using Clustal Omega.31 The output file was analyzed using the Sequence Identity And Similarity (SIAS) tool (http://imed.med.ucm.es/Tools/sias.html) to obtain pairwise sequence identities, calculated using the mean length of sequences.

Supporting information

Supporting Information.