Dual roles of F123 in protein homodimerization and inhibitor binding to biotin protein ligase from Staphylococcus aureus

Authors


Summary

Protein biotinylation is catalysed by biotin protein ligase (BPL). The most characterized BPL is from Escherichia coli where it functions as both a biotin ligase and a homodimeric transcriptional repressor. Here we investigated another bifunctional BPL from the clinically important Staphylococcus aureus (SaBPL). Unliganded SaBPL (apo) exists in a dimer-monomer equilibrium at low micromolar concentrations – a stark contrast to E. coli BPL (EcBPL) that is monomeric under the same conditions. EMSA and SAXS analysis demonstrated that dimeric apo SaBPL adopted a conformation that was competent to bind DNA and necessary for it to function as a transcription factor. The SaBPL dimer-monomer dissociation constant was 5.8-fold tighter when binding the inhibitor biotin acetylene, but unchanged with biotin. F123, located in the dimer interface, was critical for homodimerization. Inhibition studies together with surface plasmon resonance analyses revealed a strong correlation between inhibitor potency and slow dissociation kinetics. A 24-fold difference in Ki values for these two enzymes was explained by differences in enzyme:inhibitor dissociation rates. Substitution of F123 in SaBPL and its equivalent in EcBPL altered both inhibitor potency and dissociation. Hence, F123 in SaBPL has novel roles in both protein dimerization and ligand-binding that have not been reported in EcBPL.

Introduction

Biotin protein ligase (BPL) is an important metabolic enzyme that catalyses the ubiquitous and essential process of attaching biotin onto biotin-dependent enzymes; a family of enzymes that play key roles in various metabolic pathways (Polyak et al., 2012). Protein biotinylation proceeds through a conserved, two-step ordered reaction mechanism. In the first partial reaction, the intermediate biotinyl-5′-AMP is produced from biotin and ATP. Subsequently, the biotinyl moiety is transferred with the release of AMP onto the ε-amino group of a specific lysine residue present in the biotin domain of all biotin-dependent enzymes. BPLs can be divided into three structural classes (reviewed Pendini et al., 2008a). All three classes share a highly conserved catalytic module that is central to enzymatic biotinylation. Class I enzymes consist solely of this catalytic module, which is a fusion of an SH2-like domain with an SH3-like domain (Wilson et al., 1992; Bagautdinov et al., 2005). X-ray crystal structures have been reported for the Class I enzymes from Pyrococcus horikoshii (Bagautdinov et al., 2005), Aquifex aeolicus (Tron et al., 2009) and Mycobacterium tuberculosis (Gupta et al., 2010; Duckworth et al., 2011). It is known that Class II BPLs contain the conserved catalytic module with an additional winged helix–turn–helix domain at the N-terminus that facilitates DNA binding and allows them to also function as transcriptional repressors. Thus, the Class II enzymes are truly bifunctional proteins. The most studied Class II isozyme is from the prototypical bacteria Escherichia coli (Wilson et al., 1992; Weaver et al., 2001b; Wood et al., 2006), with the X-ray crystal structures of a second example from Staphylococcus aureus recently being reported (Soares da Costa et al., 2012a; Pendini et al., 2013). In contrast, Class III BPLs from mammals and fungi are more complex with a larger N-terminal extension that bears no homology to the DNA binding domains present in Class II enzymes, nor any other protein yet identified in the protein database (Mayende et al., 2012). An X-ray crystal structure for a Class III enzyme has not been reported. However, primary structure alignments, in combination with the available structural data on the smaller BPLs, reveals a common biotin-binding loop that undergoes a disordered to ordered transition upon the binding of biotin (Wilson et al., 1992; Bagautdinov et al., 2005). Within this loop resides a GRGRX motif, where X is divergent between homologues. Amino acids in this motif play key roles in stabilizing the BPL:biotinyl-5′-AMP complex (Kwon and Beckett, 2000). However, the role of the divergent residue X has not been extensively investigated.

An interesting observation from crystallography, size exclusion chromatography analysis and analytical ultracentrifugation (AUC) experiments is the difference in oligomerization states among all BPLs. M. tuberculosis BPL exists predominately as a monomer in its apo, biotin and biotinyl-5′-AMP liganded forms (Purushothaman et al., 2008). This is in sharp contrast to P. horikoshii BPL that constitutively exists as a dimer, regardless of its liganded state (Bagautdinov et al., 2005). E. coli BPL (EcBPL) undergoes a more complex transition from essentially monomer in its apo form (KD 2-1 > 1 mM) to homodimer in the presence of biotinyl-5′-AMP (KD 2-1 10 μM) (Streaker et al., 2002; Zhao and Beckett, 2008; Zhao et al., 2009). In EcBPL, amino acids within the GRGRX motif are located in the dimer interface and contribute to the network of hydrogen bonds and ionic interactions that stabilize the protein-protein interaction (Weaver et al., 2001b; Wood et al., 2006). In this case, dimerization of EcBPL is a prerequisite for DNA binding (Streaker et al., 2002). Based upon X-ray crystal structures, this dimerization is known to orientate the two N-terminal domains such that they are receptive to binding specific DNA recognition sequences. Binding of the EcBPL dimer to a palindromic target sequence in the promoter of the biotin biosynthesis operon bioO results in the concomitant repression of biotin biosynthesis (Eisenstein and Beckett, 1999; Streaker et al., 2002). It has been proposed that EcBPL utilizes the same protein surface for homodimerization and heterodimerization with the apo biotin domain substrates (Weaver et al., 2001a). Hence, competing protein–protein interactions are proposed to regulate the switch between transcriptional repressor and biotin protein ligase activities (Weaver et al., 2001a; Streaker and Beckett, 2003). Recent small angle X-ray scattering experiments, using SaBPL and the biotin domain from pyruvate carboxylase, provide additional support for this model (Pendini et al., 2013). This switch mechanism allows Class II BPLs to serve as a cellular sensor of biotin demand and regulator of its de novo synthesis. What is unclear is whether the same intricate mechanisms occur in other bacteria, such as S. aureus. Studies on the oligomerization state(s) of other Class II BPLs are thus required. Interestingly, class III Homo sapiens and Saccharomyces cerevisiae BPLs have been shown to be monomeric in solution (Polyak et al., 1999; Ingaramo and Beckett, 2009) and not subject to the same dynamic processes as the Class II counterparts.

In a recent report, we described a series of biotin analogues as BPL inhibitors that possess antibacterial activity against clinical isolates of S. aureus (MIC50 2–4 μg ml−1) (Soares da Costa et al., 2012b). These compounds serve as valuable tools for chemical biology studies and leads for a new class of antibiotic (Tieu et al., 2013). We observed unexpected selectivity between the BPLs from S. aureus, E. coli and H. sapiens. The most selective compound, biotin acetylene [(3aS,4S,6aR)-4-hept-6-ynyl-1,3,3a,4,6,6a-hexahydrothieno[3,4-d]imidazol-2-one], displayed 24-fold greater affinity for S. aureus BPL (SaBPL) compared with the E. coli equivalent. The X-ray crystal structure of biotin acetylene bound to SaBPL shows that it induces the same conformational changes as does biotin, especially those in the biotin-binding loop (RMSD 0.30 Å overall between the two structures) (Soares da Costa et al., 2012b; Pendini et al., 2013). A key difference between the two structures is that the BPL:biotin acetylene complex is stabilized through a direct hydrophobic interaction between the acetylene moiety and W127 that also resides in the biotin-binding loop (Soares da Costa et al., 2012b). This is in contrast to the BPL:biotin interaction that is stabilized through hydrogen bonding between the hydroxyl group on biotin and an NH backbone amide of R122 (Pendini et al., 2013). W127 is invariant among all BPLs and as such preferential binding cannot simply be rationalized by interactions involving this amino acid alone. In the current study, an explanation for the higher affinity of biotin acetylene for SaBPL compared with EcBPL was pursued. The oligomeric state of SaBPL in the absence and presence of ligand was investigated using a combination of AUC, electrophoretic mobility shift assays (EMSA) and small angle X-ray scattering (SAXS). The mechanism of inhibitor binding was further pursued using surface plasmon resonance (SPR) and a series of mutant proteins. We report that a single amino acid in SaBPL, F123, plays dual roles in dimer formation as well as stabilizing inhibitor binding.

Results

Oligomeric state of SaBPL in the absence and in the presence of ligands

As an initial first step towards addressing the molecular basis of inhibition by biotin acetylene, absorbance-detected sedimentation velocity experiments in the analytical ultracentrifuge were pursued to determine the quaternary structure of SaBPL in solution, and address the effect of inhibitor binding on oligomerization. Hexahistidine tagged protein was first purified using a His-Trap column as described in the Experimental procedures. A protocol previously devised to produce the enzyme in its non-liganded (i.e. apo) form was employed (Pendini et al., 2013). We confirmed that biotinyl-5′-AMP had not co-purified with SaBPL by performing two alternative biotinyl-transferase assays (Fig. S1). The Ni-NTA purified material was fractionated on a Superdex 200 size exclusion column and protein corresponding to the monomeric species was collected for subsequent analysis. Initially, AUC were performed on the apo material at three different protein concentrations. Unexpectedly, the oligomerization state was concentration-dependent, where the apo enzyme existed in a reversible self-association at these low micromolar concentrations with standardized sedimentation coefficients (s20,w) of 2.5 S and 3.8 S at both 26 μM (Fig. 1A) and 11.7 μM (Fig. 1B). At 3.9 μM, only one peak was observed with a s20,w value of 2.5 S (Fig. 1C). The [c(M)] distribution yielded apparent molecular masses of around 38 and 76 kDa, which are consistent with the predicted molecular masses of SaBPL monomer and dimer respectively (Table 1). This effect was subsequently quantified using sedimentation equilibrium analysis of SaBPL. Samples containing 11.7 μM, 7.8 μM and 3.9 μM were centrifuged at 13 000 and 19 000 r.p.m. until sedimentation equilibrium was achieved at each speed. The data were fitted to a range of models including monomer-dimer, monomer-trimer and monomer-tetramer association. The monomer-dimer association model gave the best global non-linear least-squares fit (Reduced χ2 = 0.06) with a KD2-1 of 29 ± 2 μM in the absence of ligand. Self-association of the class II EcBPL in the apo state has not been observed previously at these low protein concentrations (Kwon and Beckett, 2000; Streaker et al., 2002; Zhao et al., 2009).

Figure 1.

Concentration dependent dimerization of S. aureus BPL. Analytical ultracentrifugation analyses were performed and c(s) plotted as a function of s20,w (Svedberg) for apo SaBPL at (A) 26 μM, (B) 11.7 μM and (C) 3.9 μM. Residuals resulting from the c(s) distribution best fits are shown as a function of radius from the axis of rotation.

Table 1. Summary of hydrodynamic properties for SaBPL and mutants
  Massa (kDa) s20,wb (S) Apparent massc (kDa)f/f0d KD 2-1e (μM) KD 2-1 plus biotine (μM) KD 2-1 plus biotin acetylenee (μM)
  1. The data show the mean ± SD from three experiments. N/A, Not applicable as these mutants were shown not to dimerize in the sedimentation velocity experiments.
  2. aMolecular mass calculated from amino acid sequence.
  3. bStandardized sedimentation coefficient taken from the ordinate maximum of the [c(s)] distribution.
  4. cMolecular mass calculated from [c(M)] distribution.
  5. dFrictional ratio calculated using the method from SEDNTERP (Laue et al., 1992).
  6. eDissociation constant for dimer to monomer.
Wild-type SaBPL38.32.5 ± 0.238 ± 31.3629 ± 230 ± 25± 0
76.63.8 ± 0.276 ± 51.42
SaBPL-F123G38.32.5 ± 0.137 ± 21.30N/AN/AN/A
SaBPL-F123R38.32.5 ± 0.238 ± 11.30N/AN/AN/A

The unforeseen finding that apo SaBPL can self associate prompted EMSA and SAXS investigation into whether or not the dimer assembles into a conformation that is competent to bind DNA, and therefore of biological relevance. These studies employed double stranded 44-mer oligonucleotides predicted to contain the bioO operator sequence from S. aureus (Rodionov et al., 2002). We have previously demonstrated that holo SaBPL does bind to this probe using both SAXS and EMSA analysis (Pendini et al., 2013). Accordingly, the apo protein was assessed alongside holo SaBPL. Holo SaBPL was produced by pre-incubating the enzyme with a 10-fold molar excess of biotin and MgATP to pre-form a complex with the co-repressor biotinyl-5′-AMP (i.e. holo SaBPL). Both forms of the enzyme retarded movement of the bioO probe in the polyacrylamide gel, indicative of DNA binding (Fig. S2A). Non-related DNA probes failed to bind either enzyme, demonstrating the specificity of the DNA:protein interaction (Fig. S2A). SaBPL also failed to bind to truncated probes 22 nucleotides in length containing a single binding site (Fig. S2B), supporting the observation that the dimer binds to two half sites in the recognition sequence. A corresponding KD of 108 ± 6 nM was measured for holo SaBPL by varying the concentration of enzyme in the reaction with bioO (Fig. S2C). The apo enzyme also bound the DNA probe in the EMSA albeit with 6.0-fold lower affinity than holo SaBPL (KD apo 649 ± 43 nM; P < 0.01 apo vs holo) (Fig. S2C). SAXS analysis further confirmed the DNA binding activity of the apo dimer. Figure S3 shows the scattering data and Pairwise distribution function for the apo-SaBPL/bioO complex. While the radius of gyration (Rg) and Dmax calculated for both the apo and holo SaBPL with bioO were similar, both values differed greatly from those calculated from the scattering of either protein or DNA alone (Table 2). This would be expected for the formation of a protein:DNA complex. Furthermore, the ab initio SAXS structure of apo SaBPL in complex with bioO was in excellent agreement with the model of the enzyme dimer in complex with DNA (Fig. 2), as previously reported for holo SaBPL (Pendini et al., 2013). The SAXS analysis, together with the EMSA data, confirm the dimerization and DNA binding activity of the apo protein, which represents a key difference between the S. aureus and E. coli BPLs.

Table 2. Structural parameters calculated from SAXS data
 RG (Å)a(Guinier)RG (Å)b (P(r))Dmax (Å)cχ2dNSDeRef
  1. aRG – radius of gyration given by Guinier approximation.
  2. bRG – estimated from Pair wise distribution function.
  3. cDmax – maximum molecular dimension from P(r) function.
  4. dGoodness of fit of theoretical end experimental scattering curves calculated using FOXS (Schneidman-Duhovny et al., 2010).
  5. eNormalized Spatial Discrepancy (NSD) for ab initio SAXS models calculated from DAMAVER (Volkov and Svergun, 2003).
bioO alone38.3 ± 0.338.71351.80.65This study
apo SaBPL alone22.0 ± 0.122.8900.560.51Pendini et al. (2013)
bioO + apo SaBPL44.5 ± 0.244.41450.60.81This study
bioO + holo SaBPL46.2 ± 0.446.31501.11.5Pendini et al. (2013)
Figure 2.

SAXS analysis of apo SaBPL in complex with bioO. Ab initio reconstruction of the apo-SaBPL/bioO complex derived from the scattering data, overlaid with a molecular model derived from crystallographic structural data and modelling.

Inhibitor binding induces dimerization

Previous reports using E. coli BPL have shown that inhibitors can induce protein dimerization (Brown et al., 2004). Accordingly, we next addressed the effect of ligand binding on the oligomerization of SaBPL using sedimentation velocity experiments. Interestingly, in the presence of 100 μM biotin (i.e. 100 × Km), the equilibrium did not shift towards the dimer even at the highest protein concentration of 26 μM (Fig. 3). However, 100 μM biotin acetylene did shift the equilibrium significantly in favour of the dimeric 3.8 S species (Fig. 3). This effect was subsequently quantified using sedimentation equilibrium analysis with the same conditions as reported above. Once again, data were fitted to a number of models, with the monomer-dimer model yielding the best global non-linear least squares fit (Reduced χ2 = 0.06). The dissociation constant was calculated to be 5 μM in the presence of biotin acetylene, which is 5.8-fold lower compared with the unliganded enzyme.

Figure 3.

Inhibitor induced dimerization of S. aureus BPL. Analytical ultracentrifugation analyses were performed and c(s) plotted as a function of s20,w (Svedberg) for 26 μM apo SaBPL in the presence of either no ligand (black curve), 100 μM biotin (blue curve) or biotin acetylene (red curve). A representative residuals plot resulting from one of the c(s) distribution best fits is shown as a function of radius from the axis of rotation.

Characterization of SaBPL-F123G, SaBPL-F123R and EcBPL-R119F

The X-ray crystal structure of the holo SaBPL dimer [PDB ID 3RKW (Pendini et al., 2013)] was examined in order to identify amino acids located in the protein:protein interface that might potentially be required for dimerization. The interaction surface covers 1020 Å2 of solvent accessible surface area and contains an extensive network of hydrogen and ionic bonds (Fig. S4). The C-terminal and catalytic domains orientate themselves to make contact with appropriate amino acid residues on the opposing subunit. Importantly, two intersubunit bonding interactions at the dimer interface were observed that are not present in the EcBPL dimer. Notably R122 and F123 both contact the side-chain of D200 on the partner subunit: R122 through hydrogen bonding and F123 through a hydrophobic interaction with the γ-carbon (Fig. 4). This suggests a potential role in protein dimerization for the phenylalanine in the GRGRF123 motif. Previous studies on EcBPL have shown that the introduction of an aromatic amino acid at the equivalent position (R119W) does not effect ligand binding or enzyme activity (Kwon and Beckett, 2000), but it does significantly impede protein dimerization (Barker and Campbell, 1981; Zhao et al., 2009). Given the findings on the EcBPL-R119W mutant, it was intriguing that the equivalent position in native SaBPL should also be an aromatic amino acid. A series of mutant proteins was thus generated in order to define a possible function for F123 in protein oligomerization and inhibitor binding. In particular, the side-chain of F123 was substituted to glycine (F123G), and an arginine substitution (F123R) was also separately generated to match the equivalent residue in the E. coli enzyme. The reverse mutant EcBPL-R119F was also generated and characterized. CD spectra confirmed that all mutant proteins fold similarly in solution compared with their wild-type counterparts (Fig. S5). Therefore, any changes in catalytic efficiency or substrate binding are not attributable to gross perturbations in the secondary structure upon mutation.

Figure 4.

Staphylococcus aureus BPL dimer. SaBPL in complex with biotinyl-5′-AMP (yellow stick) is shown in its dimeric form, with one monomer coloured tan and the other cyan. The biotin-binding loop is highlighted in magenta. Enlarged in box are key amino acids required for dimerization. Figure was prepared using USCF Chimera with PDB 3RKW (Pendini et al., 2013).

Phe123 involvement in dimerization

Analytical ultracentrifugation was again employed to determine the oligomeric state of the two SaBPL mutant proteins and test the hypothesis that F123 was required for dimerization. AUC sedimentation velocity experiments were performed using a protein concentration of 26 μM. Fitting the data to a continuous sedimentation coefficient [c(s)] distribution model revealed a single peak at 2.5 S for both proteins (Table 1 and Fig. S6), indicative of a monomer. Interestingly, neither biotin nor biotin acetylene at 100 μM induced the formation of a dimer for either mutant protein (see Fig. S6). Thus, substitution of F123 abolished dimerization of SaBPL, highlighting that an aromatic amino acid at position X in the GRGRX motif plays a key role in the dimerization of SaBPL.

Analysis of inhibitor binding by enzyme kinetics

A comparison of the kinetic parameters for the mutant enzymes with the corresponding wild-type enzymes is shown in Table S1. The kinetic constants for biotin and MgATP were determined by steady-state kinetics (Soares da Costa et al., 2012a) and found to be in good agreement with published values (Chapman-Smith et al., 2001; Soares da Costa et al., 2012a). As has been observed previously with EcBPL-R119W (Kwon and Beckett, 2000), the F123G and F123R substitutions in SaBPL had minimal effect upon the KM for MgATP and biotin. Also, the R119F substitution in EcBPL had no effect on the enzyme's turnover rates (kcat). In contrast, the kcat for biotin and MgATP were significantly compromised for both SaBPL mutant proteins to ∼ 10% that of wild-type enzyme. Inhibition studies were then performed to quantify the effect of biotin acetylene on the activities of mutant BPLs (Table 3). We have previously reported biotin acetylene to be a potent inhibitor of both S. aureus and E. coli BPLs, but with a 24-fold preference for SaBPL (Soares da Costa et al., 2012b). The F123R substitution resulted in Ki = 7.2 ± 0.9 μM. Interestingly, the inhibition constant observed for SaBPL-F123R was in excellent agreement with the value obtained for wild-type EcBPL (7.3 ± 1.0 μM, P < 0.01 EcBPL vs SaBPL-F123R; Fig. 5A). Thus, a single, non-conserved amino acid at position 123 in SaBPL is sufficient to account for the higher affinity of SaBPL for biotin acetylene. In support of this observation, substitution of the EcBPL R119 residue to a phenylalanine yielded a mutant that was equally sensitive to biotin acetylene as the wild-type S. aureus enzyme (Ki = 0.4 ± 0 μM, P < 0.01 EcBPL-R119F vs SaBPL; Fig. 5B). The mutation of F123 to a glycine resulted in a 11-fold increase in Ki value compared with the wild-type enzyme.

Table 3. In vitro inhibition constants and SPR analysis of biotin acetylene binding
BPLKi (μM)Ki relative to WTka (× 103 M−1 s−1)kd (× 10−3 s−1)kd relative to WTKD (μM)
  1. The data show the mean ± SD from at least three experiments. N/A, Not applicable as on and off rates were outside measurable constraints.
  2. aKD determined using a steady state affinity model.
  3. bKD determined using kinetic analysis.
WT SaBPL0.3 ± 0N/AN/A0.2 ± 0.1a
SaBPL-F123G3.5 ± 0.511.72.9 ± 0.24.4 ± 0.214.11.5 ± 0.2b
SaBPL-F123R7.2 ± 0.9241.7 ± 0.29.2 ± 0.729.66.2 ± 0.4b
WT EcBPL7.3 ± 1.01.4 ± 0.29.2 ± 0.86.8 ± 1.5b
EcBPL-R119F0.4 ± 00.05N/AN/A0.030.2 ± 0.1a
Figure 5.

Inhibition of BPL activity by biotin acetylene. The activity of BPLs from S. aureus (red curves and symbols) and E. coli (blue curves and symbols) were measured in the presence of varying concentrations of biotin acetylene. The inhibitory activities of the two wild-type BPLs were compared against (A) EcBPL-R119F mutant (black curves and symbols) and (B) the SaBPL-F123R mutant (orange curves and symbols). Inhibition constants from these analyses are shown in Table 3.

Analysis of inhibitor binding by SPR

Surface plasmon resonance was employed in order to further dissect the interaction between BPL and biotin acetylene. Apo BPLs were applied to activated surfaces of a CM5 sensorchip for protein immobilization at a concentration known to represent primarily monomeric protein (i.e. 3.9 μM, > 5-fold less than KD2-1). The wild-type SaBPL, SaBPL-F123G, SaBPL-F123R, wild-type EcBPL and EcBPL-R119F were all investigated, with one channel left blank in each experiment as a reference and to take into account changes in refractive index associated with the buffer. Varying concentrations of biotin acetylene were injected over the immobilized proteins for 120 s, followed by a 300 s dissociation phase (Fig. 6). Data were fitted to a 1:1 ligand-binding model and the association and dissociation rates (kd) determined where possible (Table 3). The association rates were less than 2-fold different for all enzymes. However, the greatest variance between the proteins was observed in the dissociation rates, with excellent correlation apparent between the relative kd values and the in vitro potency of the biotin acetylene inhibitor. For wild-type SaBPL and the EcBPL-R119F mutant the dissociation of the inhibitor was slow, thereby preventing accurate off rates from being quantified. The binding constants for these proteins were calculated using a steady state binding model, and these compared favourably with the Ki values determined in the inhibition assays (Table 3). Substitution of F123 with arginine or glycine greatly enhanced the inhibitor dissociation rates (SaBPL-F123G 4.4 ± 0.2 × 10−3 s−1; SaBPL-F123R 9.2 ± 0.7 × 10−3 s−1), with the kd of the arginine substitution being the same as wild-type EcBPL (9.2 ± 0.8 × 10−3 s−1, P > 0.05). For these proteins KD values were determined using a kinetic binding model and, again, their affinities were in excellent agreement with the in vitro inhibition assays (Table 3). Together our data demonstrate that a phenylalanine residue at position 123 in SaBPL plays dual roles in both protein dimerization and ligand binding. These represent key differences between SaBPL and the well-characterized EcBPL.

Figure 6.

SPR binding of biotin acetylene. SPR sensorgrams measuring biotin acetylene binding to either (A) wild-type SaBPL, (B) SaBPL-F123G, (C) SaBPL-F123R, (D) wild-type EcBPL and (E) EcBPL-R119F. All enzymes were immobilized onto a CM5 chip and data subtracted from a blank reference cell that contained no enzyme. Concentrations of biotin acetylene injected were 0 (red), 1.6 μM (green), 3.2 μM (blue) and 6.4 μM (pink).

Discussion

SaBPL and EcBPL belong to a common class of BPLs that function as both biotin ligases and transcriptional repressors. EcBPL is the best-characterized example of these, and we reasoned that the large volume of genetic, structural, kinetic and mutagenic data about this enzyme would provide valuable information to assist in the design of inhibitors against SaBPL. Surprisingly, we discovered differences between the two enzymes despite the high degree of conservation in enzyme mechanism, as well as primary and tertiary structures. Importantly, our studies have revealed novel roles for an aromatic residue at position X in the GRGRX motif in both protein dimerization and selective binding of the inhibitor biotin acetylene. This is potentially important since other clinically important bacteria (e.g. Streptococci, Enterococci and Legionella sp) also contain BPLs with a phenylalanine at this position and these are likely to be subject to the mechanisms described here for S. aureus. Our study revealed the unexpected discovery that apo SaBPL exists in an apparent equilibrium between two oligomeric states, monomer and dimer, at low micromolar concentrations (KD2-1 29 ± 2 μM). This is in sharp contrast to the EcBPL that exists as a monomer in solution under similar conditions (KD 2-1 > 1 mM) (Streaker et al., 2002). Previous studies have demonstrated that the affinity of EcBPL for bioO is dependent on the stability of the protein homodimer (Streaker et al., 2002), so it was not surprising that both apo and holo SaBPL bound bioO with only a 6-fold difference in affinity, as measured by EMSA.

Sedimentation velocity experiments showed that biotin acetylene significantly shifted the equilibrium in favour of the SaBPL dimer formation in solution, and this is likely to contribute to the mechanism of inhibition. Interestingly, biotin itself did not alter the monomer – dimer equilibrium of SaBPL under the same conditions. One possible explanation for this is that biotin does not induce the conformational changes to the protein necessary for homodimerization. However, a comparison of the X-ray structures of SaBPL in complex with biotin, biotin acetylene or biotinyl-5′-AMP can be superimposed with RMSD of only 0.4 Å between the three structures (Soares da Costa et al., 2012a; Pendini et al., 2013). Furthermore, all three ligands induce the same conformational changes in the crystallized enzyme. An alternative explanation is provided by the binding kinetics observed in our SPR experiments (Soares da Costa et al., 2012a). The on and off-rates for SaBPL binding biotin are so fast that the KD could only be estimated using a steady-state affinity model (Soares da Costa et al., 2012b), whereas biotin acetylene establishes an enduring partnership with SaBPL characterized by a slow dissociation rate and it is this ligand bound conformation that facilitates dimerization. A slow off-rate is similarly observed between the reaction intermediate biotinyl-5′-AMP for both SaBPL (Soares da Costa et al., 2012a) and EcBPL (Xu and Beckett, 1994; Kwon et al., 2002). In EcBPL, the dissociation of the enzyme-biotin complex is 1000-fold faster than dissociation of the enzyme-intermediate complex (Xu and Beckett, 1994). Biotin binding has been linked to a considerably greater loss in entropy than binding of the reaction intermediate (Xu et al., 1996). The enthalpy cost associated with intermediate binding is somewhat ‘off set’ by the enthalpy that would have otherwise been realized from the formation of non-covalent bonds between the ligand and the repressor monomer (Brown et al., 2004). A stable BPL:biotinyl-5′-AMP complex is of importance since the adenylate serves both as the activated intermediate in transfer of biotin to the biotin domain (i.e. ligase activity of monomeric BPL) and as the co-repressor in site-specific DNA binding (transcriptional repressor activity of dimeric BPL) (Prakash and Eisenberg, 1979; Solbiati and Cronan, 2010; Chakravartty and Cronan, 2012). A BPL inhibitor that occupies the active site needs to effectively compete against the stable interaction involving the native reaction intermediate.

The X-ray crystal structures of SaBPL [PDB ID 3RKW (Pendini et al., 2013)] and EcBPL [PDB ID 2EWN (Wood et al., 2006)] reveal key differences between the dimer interfaces of the two proteins. We reasoned this analysis would provide molecular clues to explain the difference in the stability of the two apo BPL dimers and further understand the mechanism of DNA binding and subsequent regulation of gene expression. Of particular importance were two residues in the biotin-binding loop of SaBPL, namely R122 and F123. Both amino acid residues form intersubunit contacts with D200 on the partnering subunit. Our AUC analyses clearly demonstrate the importance of the F123 in stabilizing the dimer as substitution with either glycine or arginine abolished self-association. This is different to EcBPL where R118 (the equivalent of R122 in SaBPL) forms contacts with D176 and the backbone oxygen of I187 in the same polypeptide chain. R118 also hydrogen bonds with the phosphate group of biotinyl-5′-AMP through its side-chain and a backbone amine (Wood et al., 2006). Conversely, the guanidinium side-chain of R119 forms hydrogen-bonding interactions with the side-chain of E140 and backbone oxygen of G196 on the partner subunit (Wood et al., 2006). This potentially explains why a tryptophan substitution at R119 (the equivalent of F123 in SaBPL) had a dramatic effect upon EcBPL dimerization (Barker and Campbell, 1981; Kwon and Beckett, 2000).

Our data also demonstrate that F123 in SaBPL has a previously unknown role in stabilizing inhibitor binding. To eliminate the possibility that the differences in binding affinity (Ki) observed with biotin acetylene was a direct cause of dimerization, we designed SPR experiments such that the BPLs were immobilized onto the sensorchip at a concentration where the proteins were predominantly monomeric in solution. We also included the F123R and F123G SaBPL mutants that were shown to be monomeric in solution by AUC studies. Hence, we reasoned that these experiments would dissect the bimolecular interaction between monomeric SaBPL and biotin acetylene. However, the possibility of dimer formation on the sensorchip surface due to effects such as effective protein concentration increases at the membrane surface cannot be ruled out. Here we observed a strong inverse correlation between the dissociation rates and inhibitor potency, as measured by the inhibition constants Ki. Those BPLs that formed a long-lived binding complex with biotin acetylene, characterized by slow dissociation kinetics, were also most susceptible to inhibition. Reduced affinity for ligands biotin and biotinyl-5′-AMP, caused by mutations in the biotin-binding loop, have likewise been attributed to enhanced dissociation rates in the E. coli BPL (Kwon et al., 2002). Unexpectedly, the inhibition and SPR analysis performed here on SaBPL identified a role for F123 in the selective binding mechanism. X-ray crystal structures of dimeric SaBPL have shown that the side-chain of F123 does not bind directly to the inhibitor (Soares da Costa et al., 2012b). We suggest that the role of this amino acid is to help stabilize the biotin-binding loop, thereby preventing the dissociation of the biotin analogue from its binding site. Understanding the disordered-to-ordered conformational change performed by the biotin-binding loop, and the role of the ligand to stabilize this transition, will be essential considerations for the future design of BPL inhibitors with promise in antibacterial discovery.

Experimental procedures

Materials

Chemicals were purchased (unless otherwise stated) from Sigma-Aldrich Pty (Castle Hill, Australia). Protein concentration was measured using the Bradford assay (Bradford, 1976). All enzymes were stored in 50 mM Tris HCl pH 8.0, 1 mM DTT, 5 mM EDTA and 5% (v/v) glycerol at −80°C.

Nucleic acid manipulations

Cloning of the S. aureus gene with a C-terminal His6-tag, and construction of the recombinant expression vector, are described in Pendini et al. (2008b). A hexahistidine tag was similarly engineered onto the C-terminus of EcBPL to facilitate purification. Site-directed mutagenesis was performed using a QuikChange® Mutagenesis Kit (Stratagene). All oligonucleotides used in this study were purchased from Geneworks (Adelaide, Australia) and are shown in Table S2. Mutant plasmids were confirmed through DNA sequencing (SA Pathology, Adelaide, Australia).

Protein methods

The production of hexahistidine-tagged SaBPL and EcBPL enzymes is described in Supporting Experimental Procedures. Samples were further purified by size exclusion chromatography on a Superdex 200 column, eluting with a buffer containing 50 mM Tris HCl pH 8.0, 1 mM DTT, 5 mM EDTA and 5% (v/v) glycerol. Confirmation of the purification of SaBPL in its non-liganded state (i.e. apo) was confirmed using two alternative biotinyl transferase assays described in Supporting Experimental Procedures. BPL activity was measured using an in vitro biotinylation assay (Polyak et al., 1999), and inhibition by biotin acetylene was performed as described (Soares da Costa et al., 2012a). Circular dichroism (CD) spectroscopy (Soares da Costa et al., 2012a) and SPR (Soares da Costa et al., 2012b) were also performed as previously reported. SAXS was performed as described in Supporting Experimental Procedures and in (Pendini et al., 2013). The SPR data for biotin acetylene binding to BPL was analysed using one of two methods. Data were fitted to a 1:1 binding model and when association and dissociation rates could be determined the KD was calculated using kinetic analysis (KD = kd/ka). When an accurate off rate could not be measured steady state analysis was employed to determine the KD. The EMSA is described in Supporting Experimental Procedures.

Analytical ultracentrifugation

Sedimentation velocity and sedimentation equilibrium studies were performed using an XL-I analytical ultracentrifuge (Beckman-Coulter) with an 8-hole An-50-Ti rotor. Double sector centrifuge cells with quartz windows were loaded with 380 μl of sample and 400 μl of TBS reference buffer (20 mM Tris, 150 mM NaCl, pH 8.0) for sedimentation velocity experiments or 100 μl sample and 120 μl reference for sedimentation equilibrium experiments. Initial scans were carried out at 3000 r.p.m. to determine the optimum wavelength and radial range for the experiments. For sedimentation velocity analyses, absorbance versus radial profiles were measured in continuous mode using a step size of 0.003 cm at a single wavelength (280 nm) every 6 min employing a rotor speed of 40 000 r.p.m. at a temperature of 20°C for enzyme concentrations of 26.0, 11.7 and 3.9 μM. Sedimentation equilibrium scans were also performed at 20°C but in step mode with a step size of 0.001 cm with 10 averages using rotor speeds of 13 000 and 19 000 r.p.m. for enzyme concentrations of 11.7, 7.8 and 3.9 μM. The partial specific volume of SaBPL and mutants (0.7332 ml g−1), buffer density (1.00499 g ml−1) and viscosity (1.0214 cp) were calculated using the program SEDNTERP (Laue et al., 1992). Sedimentation velocity data were fitted to a continuous sedimentation coefficient [c(s)] distribution model using the program SEDFIT (Perugini et al., 2000; 2002; Schuck, 2000; Schuck et al., 2002). Sedimentation equilibrium data were fitted to various self-associating models using the program SEDPHAT (Vistica et al., 2004; Burgess et al., 2008; Voss et al., 2010).

Acknowledgements

This work was supported by the National Health and Medical Research Council of Australia (application 1011806), and Adelaide Research and Innovation's Commercial Accelerator Scheme. MAP would like to acknowledge the Australian Research Council for Future Fellowship support and MCJW is an Australian NHMRC Senior Research Fellow. We also acknowledge the computer resources of the Victorian Partnership for Advanced Computing and the Adelaide Protein Characterization Facility for access to the BIAcore and the CD spectrometer. The authors acknowledge that they have no conflict of interest to declare.

Ancillary