Essential Amino Acid Residues of BioY Reveal That Dimers Are the Functional S Unit of the Rhodobacter capsulatus Biotin Transporter

Abstract

Energy-coupling factor transporters are a large group of importers for trace nutrients in prokaryotes. The in vivo oligomeric state of their substrate-specific transmembrane proteins (S units) is a matter of debate. Here we focus on the S unit BioY of Rhodobacter capsulatus, which functions as a low-affinity biotin transporter in its solitary state. To analyze whether oligomerization is a requirement for function, a tail-to-head-linked BioY dimer was constructed. Monomeric and dimeric BioY conferred comparable biotin uptake activities on recombinant Escherichia coli. Fluorophore-tagged variants of the dimer were shown by fluorescence anisotropy analysis to oligomerize in vivo. Quantitative mass spectrometry identified biotin in the purified proteins at a stoichiometry of 1:2 for the BioY monomer and 1:4 (referring to single BioY domains) for the dimer. Replacement of the conserved Asp164 (by Asn) and Lys167 (by Arg or Gln) in the monomer and in both halves of the dimer inactivated the proteins. The presence of those mutations in one half of the dimers only slightly affected biotin binding but reduced transport activity to 25% (Asp164Asn and Lys167Arg) or 75% (Lys167Gln). Our data (i) suggest that intermolecular interactions of domains from different dimers provide functionality, (ii) confirm an oligomeric architecture of BioY in living cells, and (iii) demonstrate an essential role of the last transmembrane helix in biotin recognition.

INTRODUCTION

Energy-coupling factor (ECF) transporters are a class of micronutrient importers in prokaryotes that contain features of ATP-binding cassette (ABC) transporters but differ from canonical ABC importers in several respects (4, 16). Significantly, ECF transporters do not rely on extracytoplasmic soluble substrate-binding proteins. Substrates of ECF transporters include the complete set of B vitamins and their metabolic precursors and degradation products, the transition metal ions Ni²⁺ and Co²⁺, and certain intermediates of salvage pathways. The capability to mediate vitamin uptake makes ECF transporters an essential inventory in human pathogens with restricted biosynthetic capacities, including members of the genera Streptococcus (23) and Mycoplasma (10), and in the food-borne pathogen Listeria monocytogenes (19).

Basically, all ECF importers are composed of a substrate-specific transmembrane protein (S unit), a moderately conserved transmembrane protein (T unit), and pairs of ABC ATPase domains (A units). The module of A and T units is called “energy-coupling factor” (ECF) for historical reasons (4, 16). ECFs can be dedicated to a specific S unit (subclass I ECF transporters) or can interact with many different S units in one organism (subclass II). The latter scenario is mainly found in Gram-positive bacteria and in a few archaea. In these cases, the ECF is encoded by linked genes, whereas the individual S unit genes are scattered around the genome. Recent biochemical and X-ray crystallographic analyses gave important insight into the structures of the riboflavin- and thiamine-specific S units RibU (25) and ThiT (5, 6) and provided clues to modularity of subclass II systems (5, 22).

ECF-type biotin transporters comprise a large group of importers with special properties (reviewed in reference 4). About one-third of those systems belong to subclass I and are encoded by operons. The order of the bioY (S unit), bioM (A unit), and bioN (T unit) genes varies in different organisms. The remaining two-thirds encompass about one half each subgroup II systems (BioY plus a shared ECF) and solitary BioY proteins which are predicted to function in the absence of any A and T units. The latter prediction is based on the observation that neither bioMN genes nor any gene for a T unit is recognizable in those genomes. The hypothesis that solitary S units may have this potential is corroborated by the fact that distinct S units of subclass I biotin and transition metal transporters show significant basal substrate uptake activity in the absence of their ECF (8, 11, 17, 20).

The BioMNY system of Rhodobacter capsulatus is the prototype of subclass I ECF transporters. It imports biotin molecules slowly but with very high affinity, and it was the first system for which the tripartite composition of S, T, and A units was shown biochemically (11). In vivo Förster resonance energy transfer (FRET) experiments (8), site-directed mutagenesis combined with functional assays (13), and in vitro cross-linking studies with isolated membranes (14) were applied to gain insight into the oligomeric composition and to localize critical interaction sites among the subunits. The BioMNY holotransporter is characterized as a high-affinity system (k_T ≈ 5 nM biotin) in vivo, whereas the solitary BioY has an approximately 50-fold-lower affinity (11).

Due to the small size of BioY (≈20 kDa) and the low number of transmembrane helices (only six), questions of whether this protein can function as a monomer or needs to dimerize arose. In vitro analyses of the oligomeric state of R. capsulatus BioY and a few BioY proteins from other organisms gave inconclusive results, because the proteins occurred in at least two (putative monomeric and dimeric) states in detergent solution (see Fig. S1 in the supplemental material). Our previous in vivo studies had identified FRET between BioY copies tagged with fluorophores (8). These data pointed to a dimeric state of BioY in vivo but could not answer the question whether dimerization is a prerequisite for function of this minimal transporter unit.

In the present study, we identified two highly conserved charged residues (D164 and K167) in the sixth transmembrane helix as essential for function of the R. capsulatus BioY. This discovery provided an instructive tool to address the question whether function of BioY requires its dimerization/oligomerization. We forced BioY into a dimeric state by fusing two copies of bioY in tandem, and we analyzed the consequences of D164 and K167 replacements in the BioY monomer and in the N-terminal or C-terminal half or in both halves of the covalently fused dimer. Similar approaches have been applied by others to prove the dimeric (e.g., of the LacS lactose transporter [9] or the EmrE multidrug exporter [21]) or monomeric (e.g., of the LacY lactose permease [18]) state of membrane transporters. Our results of in vivo biotin transport assays as well as in vitro data that quantify the stoichiometry of biotin binding argue for a dimer as the functional unit of BioY.

MATERIALS AND METHODS

Bacterial strains.

Escherichia coli XL1-Blue (Stratagene) was used for gene cloning and E. coli UT5600 (ompT) (New England BioLabs) for recombinant production of BioY variants as described previously (11). For biotin uptake assays, bioY alleles were expressed in the intrinsically biotin transport-deficient E. coli S1039 (12, 24). Static fluorescence anisotropy of fluorophore-tagged BioY variants was analyzed in suspensions of recombinant E. coli BL21(pLacI-Rare2) (Novagen).

Construction of dimeric, mutant, and monomeric yellow fluorescent protein (mYFP)-tagged bioY variants.

Plasmid pRcBioY, encoding an N-terminally His₁₀-tagged and C-terminally FLAG-tagged BioY (11), was the basis for all constructions. In this plasmid an NcoI restriction site overlaps with the bioY Met-1 codon immediately downstream of the His tag-encoding sequence, and a BglII site is located between the last codon of bioY and the FLAG-encoding sequence. A tandem bioY-bioY fusion was generated as follows. bioY was amplified with primers adding BglII recognition sites immediately upstream of the initiation codon and downstream of the last codon. The BglII-treated amplicon was inserted into the BglII site of pRcBioY. Transformants containing the correctly oriented bioY-bioY tandem fusion were selected upon restriction analysis. The encoded protein dimer contains two full-length copies of BioY (Fig. 1) separated by an Arg residue and a Ser residue that result from the BglII site in the DNA. The dimer contains a His₁₀ tag at the N terminus of the first half and a FLAG tag at the C terminus of the second half.

Fig 1.

Topological model of R. capsulatus BioY. The secondary-structure prediction is based on 3D modeling with the SWISS-MODEL server. The most strongly conserved residues among members of the BioY protein family are shown in solid circles. In the dimeric BioY, the two monomers are linked by two additional (Arg and Ser) residues as depicted in the lower right part.

The D164N, K167Q, and K167R mutations in bioY were created by two rounds of PCR. In the first round, products were generated with mutagenic forward primers and a reverse primer that overlaps with the 3′ end of bioY and inserts a BglII site. The products of this first round were used as primers in a second round of PCR together with a forward primer overlapping with the bioY initiation codon and inserting an NcoI site. These amplicons were treated with NcoI and BglII and used to replace the corresponding wild-type allele of pRcBioY. Plasmids encoding BioY dimers with individual mutations in one half were generated by insertion of the wild-type allele with NcoI sites at both ends into the NcoI site of plasmids encoding mutant monomers (to give BioY_WT-BioY_Mutant variants) or by insertion of the wild-type allele with two BglII ends into those plasmids (to give the BioY_Mutant-BioY_WT variants). Plasmids containing identical mutations in both halves were constructed by amplifying mutant alleles with the primers that add BglII sites to the ends and subsequent insertion of the BglII-treated amplicons into the corresponding template plasmids.

Plasmids encoding fusions of BioY variants to monomeric yellow fluorescent protein (mYFP) were the following. A plasmid in which the mYFP gene, flanked by NcoI sites, is fused to the 5′ end of bioY has been described recently (8). The mYFP cassette was excised from this plasmid with NcoI and inserted in frame into the NcoI site of the plasmid encoding dimeric BioY_WT-WT. The resulting plasmid codes for BioY_WT-WT with mYFP fused to the N terminus.

All in vitro-generated DNAs were verified by nucleotide sequencing. The sequences of the primers used for PCR are listed in Table S1 in the supplemental material.

Biotin uptake assay.

Biotin uptake by recombinant E. coli S1039 was assayed as described previously (11) with the following modifications. Cells were grown overnight in lysogeny broth (LB) supplemented with 1 mM isopropyl-β-d-galactopyranoside (IPTG) and 100 μg ml⁻¹ ampicillin,washed, and resuspended in glucose-free uptake buffer (35 mM sodium/potassium phosphate, pH 7.0). Cell suspensions were diluted in the same buffer to give an optical density at 578 nm (OD₅₇₈) of ≈0.2. Upon addition of [³H]biotin to a final concentration of 4 nM, the suspensions were incubated for 3.5 h at 37°C with shaking. Samples (0.5 ml) were filtered through nitrocellulose membranes, the filters were washed, and filter-bound radioactivity was quantified in a Packard TriCarb 2900 TR liquid scintillation counter.

Purification of BioY variants.

BioY variants were purified basically as described previously (11). Briefly, recombinant E. coli UT5600 cells (2-liter culture volume in LB) were harvested, resuspended in 35 mM sodium/potassium phosphate buffer (pH 7.0) containing EDTA-free protease inhibitor cocktail (Roche), and lysed by one passage through a Basic-Z pressure cell (Constant Systems Ltd.) at 1.7 × 10⁸ Pa. Membranes were pelleted by ultracentrifugation in a Sorvall T647.5 rotor at 36,000 revolutions per minute for 45 min at 4°C and solubilized in 6 ml 50 mM Tris-HCl (pH 8.0)–300 mM NaCl–20 mM imidazole–5% (vol/vol) glycerol–2% (wt/vol) dodecyl-β-d-maltoside (DDM) for 2 h. Nonsolubilized material was removed by a second ultracentrifugation step. Solubilized membrane protein was mixed with 0.5 ml Ni²⁺-nitrilotriacetate matrix (Invitrogen) and incubated overnight at 4°C with gentle rotating. After transfer into an empty column, the samples were washed twice with 100 ml 50 mM Tris-HCl (pH 7.5)–300 mM NaCl–5% (vol/vol) glycerol–0.05% (wt/vol) DDM containing 20 mM or 100 mM imidazole, respectively. Bound protein was eluted with this buffer containing 300 mM imidazole. The samples were desalted on PD-10 columns (GE Healthcare) and subsequently concentrated using Amicon filtration units (Millipore) with a 10-kDa (for monomeric BioY) or 30-kDa (for dimeric BioY) cutoff.

Quantification of biotin by MS.

Purified wild-type and mutant monomeric and covalently fused dimeric BioY proteins (250-μl samples; 10 μM in 50 mM Tris-HCl [pH 7.5]–300 mM NaCl–5% glycerol–0.05% DDM) were denatured at 95°C for 5 min. Denatured protein was pelleted by centrifugation (10 min at 15,000 rpm and room temperature). Fifty microliters of the supernatant (or of heat-treated biotin standard solutions, for calibration purposes) was subjected to Agilent 1200 high-performance liquid chromatography using a Zorbax 300SB C₁₈ column (300-Å stable bond, narrow bore, 2.1 mm by 150 mm, 5-μm particle size). The samples were eluted isocratically over 10 min with a flow rate of 0.1 ml/min using 25% acetonitrile–0.1% formate in water as the mobile phase. Biotin typically eluted at between 5 and 6.5 min. The high-performance liquid chromatograph was coupled to an Agilent MSD electrospray ionization time-of-flight (ESI-TOF) mass spectrometer (MS). Nitrogen was used as the sheath gas at a flow rate of 12 liters/min and a temperature of 350°C. The cone voltage was 250 V, the fragmentor voltage was set to 150 V, and the capillary voltage was 3,000 V. Spectral data were collected over an m/z range from 205 to 400 in positive-ion mode. Masses 245.0954 and 267.0774, representing [biotin + H]⁺ and [biotin + Na]⁺, respectively, were extracted from the total ion chromatogram using the Agilent MASSHUNTER software. Extracted chromatograms were integrated, and the biotin content was calculated from peak areas by comparison with the peak areas resulting from biotin solutions of known concentration. For each experimental setup the system was calibrated with freshly diluted biotin standard. As indicated in Fig. S3 in the supplemental material, between 200 pmol and 1 nmol of biotin could be quantified by this protocol. The protein concentration in the samples was determined prior to heat treatment by Peterson's modification of the Lowry protocol (15) using a commercial kit (Sigma TP0300) and bovine serum albumin as the standard. To validate the colorimetric assay, the concentration of purified wild-type BioY estimated by this assay was compared to the value obtained by photometric quantitation using the molecular mass of the N-terminally His₁₀- and C-terminally FLAG-tagged BioY (22.134 kDa) and a theoretical molar absorption coefficient (ε₂₈₀ = 26,470 M⁻¹ cm⁻¹, calculated with the ProtParam tool at http://web.expasy.org/protparam/). Those values differed only by approximately 17%, indicating the reliability of the Peterson protocol. Our attempts to determine the BioY concentration by quantitative amino acid analysis upon peptide bond hydrolysis failed, because BioY was recalcitrant to total hydrolysis even after prolonged incubation (120 h) in 6 M HCl at 110°C.

Analysis of fluorescence anisotropy.

Recombinant E. coli BL21 cells were inoculated in 10 ml of LB to an initial OD₅₇₈ of 0.1 and grown for 2 h at 37°C with shaking. Expression of bioY alleles under the control of the lac promoter was induced upon addition of 1 mM IPTG, and incubation was continued for 1 h. Cells were harvested by centrifugation (2,800 × g for 10 min at 10°C), washed, and resuspended in 35 mM sodium/potassium phosphate buffer, pH 7.0, to give an OD₅₇₈ of 5. The samples were diluted 1:5 in the same buffer prior to the assays, and 200-μl aliquots were subjected to fluorescence analysis. Static fluorescence anisotropy measurements were performed at 23°C using a Horriba FluoroMax-4 spectrofluorometer. The samples were excited at 470 nm, and mYFP-based fluorescence intensities were recorded at 520 nm, parallel (I_vv) and perpendicular (I_vh) to the vertical polarization of the exciting light beam. Fluorescence anisotropy was calculated according to the equation r = (I_vv − g · I_vh)/(I_vv + 2 · g · I_vh).

Fluorescence intensities I_hv (horizontally polarized excitation and perpendicular emission detection) and I_hh (horizontally polarized excitation and parallel emission detection) of a dilute solution of Alexa Fluor 488 in 50% (vol/vol) ethanol were measured at 520 nm (excitation at 470 nm) to calculate the g factor by the equation g = I_hv/I_hh.

Three aliquots of each sample were measured, and the fluorescence anisotropy values were averaged. Those values of biological replicates were averaged, and the means are shown ± the standard errors of the means in Fig. 8.

Fig 8.

Fluorescence anisotropy analysis of N-terminally mYFP-tagged (indicated by asterisks) BioY variants. Suspensions of washed cells were subjected to spectrometric analysis. Each suspension was tested in triplicate. The means ± standard errors of the means from three biological replicates are shown. *WT+WT indicates that those cells produced both a tagged (*WT) and an untagged (WT) monomeric BioY peptide.

RESULTS

Topological model of R. capsulatus BioY and conserved residues among the BioY protein family.

In order to inactivate BioY, we searched for essential amino acid residues and their location. In silico predictions on the membrane topology of S units by secondary-structure analysis tools give ambiguous results. This is mainly due to an unusual fold as was observed by crystal structure analyses of the S units RibU (25) and ThiT (5). The two proteins have a similar six-transmembrane-helix architecture with a very short transmembrane helix II in common, although the amino acid sequence identity is negligible (<15%). Using the SWISS-MODEL server (http://www.swissmodel.expasy.org) (2), we were able to model a three-dimensional (3D) structure for the R. capsulatus BioY which resembles those of RibU and ThiT. The secondary structure derived from this model is shown in Fig. 1. A six-transmembrane-helix architecture is compatible with hydropathy profile analysis and in silico analyses using the TOPCONS server (http://topcons.cbr.su.se). The most strongly conserved amino acid residues among the approximately 200 BioY sequences contained in the SEED database (http://www.theseed.org) are highlighted in Fig. 1. Besides a few conserved residues, mainly glycine and proline, the F₁₆₀xxxD₁₆₄xxK₁₆₇ signature (x is any residue) in transmembrane helix VI is a striking feature. The high degree of conservation of D₁₆₄ and K₁₆₇ pointed to an important role in biotin recognition, and we chose these two residues for mutagenesis.

Asp164 and Lys167 are critical for biotin transport.

In order to analyze the roles of the two strongly conserved D164 and K167 residues in BioY proteins, D164N, K167R, and K167Q mutations were generated and inserted into the R. capsulatus BioY monomer. A charge-conserving D164E variant was not constructed because a glutamate residue is found in some BioY sequences at this position, suggesting that this replacement may be tolerated. All BioY variants were detectable by immunoblotting (Fig. 2B), indicating that the replacements did not interfere with stability. Moreover, the mutant proteins could be purified from isolated membranes of the recombinant cells (not shown). Analysis of biotin uptake by E. coli cells producing the mutant BioYs gave clear-cut results: the three mutant proteins were devoid of any transport activity, demonstrating that D164 and K167 are essential residues (Fig. 2A).

Fig 2.

Replacement of D164 and K167 in monomeric BioY. (A) Biotin uptake activity of recombinant E. coli producing the variants. Buffered cell suspensions were incubated in the presence of [³H]biotin, the cells were filtered, and filter-bound radioactivity was quantified. Mean values from triple determinations ± standard deviations are shown. (B) Western blot with anti-FLAG antibodies (detecting the C-terminal FLAG epitope of the BioY variants). Cell suspensions with a final OD₅₇₈ of 6 were treated with SDS-containing buffer, the samples were centrifuged, and 30-μl aliquots of the supernatants were subjected to SDS-PAGE and immunoblotting.

Construction and properties of a fusion protein with tandemly repeated copies of BioY.

Our attempts to characterize the oligomeric state of R. capsulatus BioY in vitro were hampered by the fact that the protein could not be isolated in a monodisperse form in detergent solution. Rather, gel filtration analysis reproducibly identified at least two species in various amounts, a slow-eluting (putative monomeric) form and a fast-eluting form which likely represents the dimer. Aiming at a BioY variant in a monodisperse state, we cloned and expressed in E. coli nine bioY genes from other proteobacteria. As shown in Fig. S1 and S2 in the supplemental material, all BioY species conferred biotin uptake activity on recombinant E. coli, and the majority of the purified proteins eluted in at least two fractions during gel filtration. Thus, the formation of monomeric and oligomeric states in detergent solution seems to be a general property of members of the BioY family rather than a specific feature of the R. capsulatus BioY.

In order to characterize the properties of dimeric R. capsulatus BioY, we forced the protein into a dimeric state by constructing a tandem fusion of two BioY copies with a tail-to-head orientation. As illustrated in Fig. 3, wild-type monomeric (BioY_WT) and dimeric (BioY_WT-WT) forms conferred comparable biotin uptake activity on the recombinants. There was no indication of proteolytic cleavage of the dimer that would result in monomeric forms (Fig. 3B).

Fig 3.

Properties of dimeric BioY. (A) Biotin uptake of recombinant E. coli producing monomeric (BioY_WT) or dimeric (BioY_WT-WT) wild-type BioY. Mean values from triple determinations ± standard deviations are shown. (B) Coomassie blue-stained gel after SDS-PAGE of affinity-purified monomeric and dimeric BioY. Small amounts of a faster-migrating form (a putative product of proteolytic degradation) are visible in the case of the dimer, but monomeric BioY is not visible.

Since the well-characterized S units RibU (3, 25) and ThiT (6) had been shown to contain bound riboflavin and thiamine, respectively, upon purification, we analyzed the biotin contents of purified BioY_WT and BioY_WT-WT. The two proteins were isolated from membranes of cells grown in yeast extract-containing (and thus biotin-rich) LB medium. Purified proteins were denatured, the denatured protein was pelleted, and the biotin content in the supernatants was calculated upon quantitative ESI-TOF mass spectrometry. The results are summarized in Fig. 4. BioY_WT and BioY_WT-WT contained 0.43 and 0.23 mol biotin per mol of a single BioY domain, respectively, correlating roughly with 1:2 and 1:4 (bound biotin per single BioY domain) stoichiometries.

Fig 4.

Copurification of biotin with monomeric and dimeric wild-type BioY. (A) Protein was purified from independent cultures grown in complex medium. The samples were heat denatured, and the biotin content in the supernatants was quantified by ESI-TOF MS. For quantification of the stoichiometry of biotin binding to BioY_WT and BioY_WT-WT, the molecular masses of the monomer and the dimer, respectively, were used. The protein concentration was determined prior to heat treatment by Peterson's modification of the Lowry protocol (15) using a commercial kit (Sigma TP0300). The mean biotin contents (x̄) ± standard deviations from eight (BioY_WT) and seven (BioY_WT-WT) individual determinations with independent biological replicates are indicated. (B) Coomassie blue-stained gel strips after SDS-PAGE of independently purified monomeric and dimeric BioY, indicating that dimer preparations did not contain monomeric BioY.

Properties of dimeric BioY variants with replacements of Asp164 or Lys167 in one or both halves.

In the next series of experiments, D164 or K167 was replaced in one or both halves of the dimeric BioY to give the BioY_D164N-D164N, BioY_K167R-K167R, BioY_K167Q-K167Q, BioY_WT-D164N, BioY_D164N-WT, BioY_WT-K167R, BioY_K167R-WT, BioY_WT-K167Q, and BioY_K167Q-WT variants. The results of biotin uptake assays with recombinant E. coli producing wild-type or mutant BioY dimers are shown in Fig. 5. As expected on the basis of the results described above for the monomer mutants, the dimers with double replacements (D164N-D164N, K167R-K167R, and K167Q-K167Q) were inactive. To discriminate between the two possibilities that the individual domains of BioY dimers are autonomous transporters or need to interact, dimers with only one half inactivated were analyzed. If the two BioY domains in the dimers function independently as biotin transporters, mutations introduced in only one half would be anticipated to reduce the activity to 50%. This result, however, was not observed. Rather, the D164N and K167R exchanges, if present in only one half of the dimer, led to approximately 25% activity, whereas variants with a single K167Q exchange retained approximately 75% activity (Fig. 5). These values are significantly different (P < 0.01 in two-tailed paired Student's t tests) from 50% wild-type activity. The transport activities of mutant dimers with a single D164N or K167R replacement differ significantly (P < 0.001) from those of the corresponding variants with one K167Q exchange. Interestingly, single replacements in all three cases had a more prominent and statistically significant (P < 0.001) effect in the N-terminal BioY domain than in the C-terminal domain. This result suggests that the C-terminal half of the covalently fused dimeric BioY is slightly less active than the N-terminal half, a finding which may be due to restricted flexibility of the first transmembrane helix of the C-terminal domain.

Fig 5.

Biotin uptake by recombinant E. coli producing dimeric BioY with individual or double replacements of D164 and K167. Mean values from triple determinations ± standard deviations are shown.

The ability of mutant BioY dimers to keep biotin tightly bound during purification was assayed by quantitative mass spectrometry. Figure 6B indicates that wild-type and mutant dimers could be purified in a stable form. As illustrated in Fig. 6A, double replacements abolished biotin binding. Single replacements in the N-terminal domain left the binding capacity almost unaffected. Statistically, the biotin contents of the three single-mutant dimers were significantly different (P < 0.02) from 50% of the wild-type dimer. Again, this result is not compatible with a model in which the two halves of the BioY dimers act as independent entities.

Fig 6.

Mass spectrometric quantification of biotin copurified with mutant dimeric BioY. (A) Purified protein (from cells grown in complex medium) was heat denatured and pelleted, and biotin in the supernatants was quantified by ESI-TOF MS. Mean values from multiple biological replicates ± standard deviations are shown. The protein concentration was determined prior to heat treatment by Peterson's modification of the Lowry protocol (15) using a commercial kit (Sigma TP0300). (B) Coomassie blue-stained gel strips after SDS-PAGE of purified proteins (a representative purification for each dimer) identified stable dimers for all constructs.

The data argue in favor of a model in which two BioY domains of different dimers interact to give the functional unit. As depicted in the left panel of Fig. 7, interactions among BioY dimers, each containing two wild-type domains (WT-WT, solid circles) result in functional species that mediate biotin transport. The three variants with double mutations (only open circles) are inactive. The model proposes four types of dimer assemblies for variants with a single replacement. Dimerization of BioY dimers with an exchange in only one domain leads to WT/mutant, mutant/WT, mutant/mutant, and WT/WT domain interactions. Interpreting the biotin transport data (Fig. 5), we assume that WT/mutant and mutant/WT interactions result in active transporters in the case of the K167Q mutation and in inactive variants in the case of the D164N and K167R mutations, explaining the uptake activities of approximately 75% and 25%, respectively (Fig. 5 and 7). Analysis of the capacity to bind biotin gave a slightly different picture (right panel in Fig. 7). The wild-type BioY dimer (WT-WT) bound biotin at a stoichiometry of 1:4 (mol biotin/mol single BioY domain), suggesting that two BioY domains of different dimers are required for tight binding of substrate in the living cell. Biotin did not copurify with the D164N-D164N, K167R-K167R, and K167Q-K167K double mutants, indicating a strongly reduced or abolished substrate affinity. All three mutant-WT dimers apparently retained full biotin-binding capacity (Fig. 6), although a somewhat reduced value (3:16 versus 1:4 compared to the WT-WT dimer [Fig. 7, right panel]) would be expected. This deviation is within the error range of the mass spectrometric assay. The findings indicate that interactions of BioY_D164N and BioY_K167R domains with a BioY_WT domain result in species that cannot transport the vitamin but have the ability to bind the substrate. In the case of the BioY_K167Q/BioY_WT domain pair, both binding capacity and transport activity are retained.

Fig 7.

Summary and interpretation of biotin transport by and biotin binding to dimeric BioY variants. Dumbbells, dimeric BioY; solid circles, wild-type BioY domains; open circles, mutant BioY domains. Random interactions of four wild-type domains in two dimers result in conformations capable of transporting biotin and binding biotin at a stoichiometry of one biotin per four BioY domains. The latter finding suggested that intermolecular interaction between single BioY domains from two different dimers results in the active form. Dimers with two mutant domains neither bind nor transport biotin. Results with dimeric BioY containing a replacement in only one domain were more complex. Approximately 25% transport activity was found in the case of the D164N and K167R replacements, and approximately 75% transport activity was found in the case of the K167Q replacement. These findings are in line with a scenario according to which (D164N or K167R)/WT and K167Q/WT domain interactions give inactive or active transporters, respectively. Biotin binding analyses identified comparable stoichiometries for wild-type dimers and all mutant dimers with a replacement in one half. Assuming that dimerization of a wild-type domain and a mutant domain gives an intact binding site and that all domains can interact randomly, a binding stoichiometry of 3:16 is expected, which is close to the 1:4 stoichiometry for the wild-type dimer.

Fluorescence anisotropy analyses uncover oligomerization of BioY dimers in vivo.

Based on the transport and binding properties of wild-type and mutant BioY dimers, we hypothesized that those dimers oligomerize in vivo to form double dimers with a total of four BioY domains. This assumption was experimentally tested by fluorescence anisotropy measurements upon production of BioY dimers N-terminally linked to monomeric yellow fluorescent protein (mYFP). The rationale of the technique is based on the fact that energy transfer among neighboring identical fluorophores (homo-FRET) leads to depolarization and thus decreases the anisotropy of the fluorescence emission (see reference 1 for a review). The fluorescence anisotropies of mYFP-BioY_WT and mYFP-BioY_WT-WT were compared in the first experiment. In a previous study, hetero-FRET was observed in vivo when BioY_WT variants N-terminally or C-terminally tagged with the donor mCerulean and the acceptor mYFP were coproduced in E. coli (8), pointing to an oligomeric state of BioY. Oligomerization of tagged BioY_WT, and hence interaction of the mYFP tags, is the basis of homo-FRET, which most likely accounts for the low fluorescence anisotropy (0.2115) detected in these samples (Fig. 8). This hypothesis was tested by three series of experiments. First, parallel production of untagged and tagged BioY_WT should decrease homo-FRET and hence increase anisotropy, since apart from interactions of tagged monomers, interaction of untagged and tagged monomers occurs. For the latter, homo-FRET is absent. Therefore, fluorescence anisotropy of a recombinant expressing an untagged bioY gene from one plasmid (conferring streptomycin resistance) plus the mYFP-bioY fusion from the ampicillin resistance-conferring plasmid was analyzed. The mean anisotropy of those cells containing mYFP-BioY_WT plus untagged BioY_WT was significantly higher (0.2315; P < 0.01) than the anisotropy of cells harboring only the mYFP-BioY_WT construct (Fig. 8). Second, we predicted that oligomerization of mYFP-BioY_WT-WT should also result in lower overall homo-FRET (and therefore in increased anisotropy compared to that for mYFP-BioY_WT), because mYFP-BioY/untagged BioY domain interactions will occur in addition to mYFP-BioY/mYFP-BioY (and undetectable BioY/BioY) domain interactions. Indeed, the experimental data correlated well with this assumption, because a significantly higher mean anisotropy (0.2324; P < 0.001) was detected for the mYFP-BioY_WT-WT construct (Fig. 8). Third, coproduction of mYFP-BioY_WT-WT and BioY_WT-WT should further increase fluorescence anisotropy compared to that for the mYFP-BioY_WT-WT construct since the probability of tagged/untagged domain interactions is increased. Again, the experimental values for this combination (0.2524; P < 0.01) strongly corroborate the prediction. We therefore concluded that interactions of untagged BioY domains with fluorophore-tagged BioY domains reduce homo-FRET and hence increase fluorescence anisotropy of the latter. Our data clearly indicate that covalently linked BioY dimers oligomerize in vivo.

DISCUSSION

Shortly after their description as a novel type of ABC transporters (11, 17) and the discovery of their wide distribution among prokaryotes (4, 16), several physiological, biochemical, and mechanistic properties of ECF transporters were analyzed. Shared use of the energy-coupling factor (i.e., the A1, A2, and T components) by various S units in the case of subclass II ECF transporters has been shown both in vivo (13, 16) and in vitro with purified proteins in detergent solution or upon reconstitution in proteoliposomes. Based on the in vitro data, a 1:1:1:1 (A1/A2/T/S) quaternary structure for subclass II systems has been proposed (22). Various S units (e.g., the riboflavin-specific RibU, the folate-specific FolT, and the thiamine-specific ThiT) have been analyzed in detail and were found to bind their substrates in vitro with affinity constants in the nanomolar or subnanomolar range (3, 6, 7). Coinciding results from different laboratories demonstrated that transport activity of subclass II S units strictly depends on the energy-coupling factor (5, 13, 16, 22, 25). Crystal structure analyses of RibU (25) and ThiT (5) gave insight into the molecular basis of high-affinity substrate binding and, in the case of ThiT, provided clues for the mode of S unit/T unit interactions. In both cases, two molecules of the S unit were found in the asymmetric unit of the crystals. A biological relevance of this type of dimer, however, was considered unlikely, because the membrane plane at the dimer interface would have to be rotated by a large angle in the case of ThiT, and likewise, the membrane-spanning distance would be only 2 nm and highly charged surface areas would be exposed to the lipid phase in the case of RibU.

Residues in transmembrane helix VI of both RibU and ThiT are involved in substrate recognition through hydrogen bonding (5, 25). Tyr146 and Asn151 of ThiT are in contact with the pyrimidine moiety of thiamine. In RibU, Asn164 and Lys167 form hydrogen bonds with carbonyls (both residues) and the hydrogen atom of a ring nitrogen (Asn164) of the isoalloxazine ring of riboflavin. The observation in our present study that Asp164 and Lys167, which are strongly conserved among BioY proteins, are essential for the function of BioY suggests that these two residues are involved in biotin recognition. Hydrogen bonds may be formed with the carboxylate of the valeryl side chain or the ureido ring of the biotin molecule.

Work on the R. capsulatus BioMNY biotin transporter, the prototype of subclass I ECF systems (i.e., those with a dedicated energy-coupling factor), had uncovered unforeseen traits. The solitary BioY was shown by kinetic analyses of biotin transport into recombinant E. coli cells producing this protein to mediate low-affinity, high-capacity substrate uptake (11). In the presence of the A (BioM) and T (BioN) components, the system was converted into a high-affinity, low-capacity transporter. The function as a transporter of the R. capsulatus solitary BioY corresponds to the fact that about one-third of bioY-containing genomes encode neither BioMN nor a shared A1A2T energy-coupling factor (4). Moreover, the results in the present study suggest that activity in the solitary state is a typical property of at least many BioY proteins (see Fig. S2 in the supplemental material).

The small size of BioY, which might be insufficient for substrate transport if the protein were a monomer, and the property of BioY family members to occur in two oligomeric states in detergent solution (see Fig. S1 in the supplemental material) prompted us to analyze the in vivo oligomeric composition of the R. capsulatus BioY. In a recent hetero-FRET study using BioY variants individually tagged with the FRET donor mCerulean and the FRET acceptor mYFP, we identified physical interaction among BioY monomers in living bacteria independent of the presence of BioMN (8), but the question of whether oligomerization correlates with transport activity was left open. The results of the present study allow the conclusion that oligomerization of BioY could permit substrate release into the cytoplasm upon tight binding and thus is a requirement for transporter function in vivo. This conclusion is based on several lines of evidence. (i) Monomeric BioY and a covalently linked dimeric BioY isolated from cells grown in (biotin-containing) complex medium were found to contain biotin at a stoichiometry of 1:2 and 1:4 per single BioY domain, respectively. This finding suggests that in the living cell two BioY peptides are involved in tight binding of one biotin molecule and that in the case of the covalently linked dimer, two BioY domains from different dimers form the biotin-binding unit. Restricted flexibility of the two domains of one covalently linked dimer may prevent proper assembly of those two domains. (ii) Asp164 and Lys167 in transmembrane helix six of BioY proteins are strongly conserved and, based on the results of the present study, play a central role in substrate binding and the transport process. Replacement of those residues inactivated the monomeric BioY but affected the covalently linked dimer differentially. The behavior of the K167R and K167Q variants was particularly instructive. Dimeric BioY variants containing the replacements in only one domain were only slightly affected in biotin binding but showed significantly (K167Q, 25% reduction) or strongly (K167R, 75% reduction) diminished biotin transport activities. These results are not compatible with the assumption that the two BioY domains in the covalently linked dimers act as independent units. Rather, they again point to domain interactions between different dimers to give the functional (BioY)₂ assembly. (iii) The predicted oligomerization of monomeric BioY and covalently linked BioY dimers in vivo was confirmed by fluorescence anisotropy analyses of cell suspensions containing fluorophore-tagged variants.

Questions of whether the two domains of the same covalently linked dimer can interact to form an active unit need additional experimentation. As mentioned above, it is possible that the linker with only two additional amino acid residues between the two BioY domains in the present construct is not flexible enough to allow productive domain-domain interactions within the same dimer.

Recent analysis of BioM-BioN interactions in the R. capsulatus biotin transporter mapped major contact sites to a cytoplasmic loop in the BioN membrane protein and to a short stretch of amino acid residues adjacent of the Q loop in the BioM ABC ATPase (14). In addition, dimerization of BioN in a part of the BioMNY pool was observed in that study. The findings of the present study, together with those results, point to permanently or transiently occurring supramolecular complexes as the functional units of the biotin transporter.