Solution NMR Structure of Selenium-binding Protein from Methanococcus vannielii

Abstract

Selenium is an important nutrient. The lack of selenium will suppress expression of various enzymes that will lead to cell abnormality and diseases. However, high concentrations of free selenium are toxic to the cell because it adversely affects numerous cell metabolic pathways. In Methanonoccus vannielii, selenium transport in the cell is established by the selenium-binding protein, SeBP. SeBP sequesters selenium during transport, thus regulating the level of free selenium in the cell, and delivers it specifically to the selenophosphate synthase enzyme. In solution, SeBP is an oligomer of 8.8-kDa subunits. It is a symmetric pentamer. The solution structure of SeBP was determined by NMR spectroscopy. Each subunit of SeBP is composed of an α-helix on top of a 4-stranded twisted β-sheet. The stability of the five subunits stems mainly from hydrophobic interactions and supplemented by hydrogen bond interactions. The loop containing Cys⁵⁹ has been shown to be important for selenium binding, is flexible, and adopts multiple conformations. However, the cysteine accessibility is restricted in the structure, reducing the possibility of the binding of free selenium readily. Therefore, a different selenium precursor or other factors might be needed to facilitate opening of this loop to expose Cys⁵⁹ for selenium binding.

Early studies examining selenium metabolism focused on the toxic potential of this unique element. Symptoms of selenium poisoning were observed in animals that ingested plants containing high levels of selenium (1). The interest in studying the biochemical nature of this trace element has increased dramatically, because it was discovered that low levels of selenium are essential (2). Defects such as liver necrosis, white muscle disease, and certain cardiac and skeletal muscle degenerations are linked to deficiency in selenium in mammals (for review please see Ref. 3). Recently, selenium has been shown to play a central role in redox processes in the cell (4, 5). The amount of selenium that a cell intakes that is necessary for survival is in a very narrow range (6). Therefore, a sophisticated regulating mechanism is needed to control selenium levels in the cells. Despite its importance, the molecular details of selenium metabolism and transport in living cells are still not completely understood.

One pathway of selenium incorporation is associated with the biosynthesis of proteins containing selenocysteine. This pathway is best characterized in Escherichia coli. The insertion of the selenocysteine amino acid in a specific location in a protein is directed by the in-frame UGA codon and requires the products of four genes selA, selB, selC, and selD in E. coli (7–9). The activated form of selenium in this pathway is selenium phosphate and is produced by the selD gene product, selenophosphate synthetase (8, 10).

Considering the toxicity of selenium, a transport carrier is needed to convert inorganic selenium to the appropriate precursor and deliver it to the proper biosynthetic pathways. The transport mechanism must be highly specific to selenium (11). Furthermore, due to the highly reactive and unstable nature of selenium, the transport mechanism has to preserve the physical state of the selenium during the transport.

Recently a novel new family of proteins that lacks selenocysteine but can bind selenium has been identified in the anaerobic methane-producing organisms. In Methanococcus vannielii, the gene corresponding to a family member of these proteins encodes an 8.8-kDa polypeptide. This protein was purified from the organism as a 42-kDa protein and referred to as the selenium-binding protein (SeBP)3 (12, 13). The recombinant protein was shown to bind selenium when the trace element was added to the growth media of the E. coli. However, the selenium incorporation into the protein was extremely low (13). Further characterization of SeBP shows that its oligomeric state is preserved under extreme conditions of pH, temperature, and denaturant (14). Under denaturing SDS-PAGE conditions, several protein bands corresponding to various oligomers such as dimer, trimer, and tetramer could still be found. This observation was attributed to the high content of hydrophobic amino acids in the SeBP sequence (13).

SeBP contains a single cysteine residue (Cys⁵⁹) per subunit that is important for selenium binding. Alkylation of this cysteine abolishes selenium incorporation (14). A number of studies on E. coli system have shown that selenodiglutathione (GS-Se-SG) can be used as the precursor from which selenium substrate can be derived (14). Incubation of glyceraldehyde-3-phosphate dehydrogenase (GAPDH) from E. coli with this precursor in the presence of selenophosphate synthetase has resulted in selenium incorporation into GAPDH (15). Biochemical studies show that a significant incorporation of selenium can only be observed when SeBP was completely reduced and incubated with the selenodiglutathione as donor in the presence of mild denaturant (1 m guanidine-HCl). The cysteine residue in SeBP appears to be inaccessible. A total of 0.25 equivalents of selenium per SeBP subunit was observed (14). This stoichiometry suggests a 1:4 coordination ratio of the selenium by the cysteines. Despite all of these biochemical data, it is still not clear how SeBP can bind selenium with high selectivity and transport it to the proper biosynthetic pathways. Our study attempts to address, at the atomic level, the high stability of the SeBP oligomer, its mechanism for selenium binding, the accessibility of the cysteine residue, and possible transport of the selenium to the proper enzyme.

First, we determined that SeBP exists as a pentamer in solution. Subsequently, we solved the solution structure of SeBP by Nuclear Magnetic Resonance spectroscopy. The five subunits form a barrel-shaped assembly, where the reactive cysteine is located at the loop close to the core. Solving the SeBP structure is the first step in obtaining a full picture of how this potentially toxic element can be safely managed as an important nutrient for a healthy organism.

MATERIALS AND METHODS

Recombinant Protein—The protein was expressed using E. coli BL21(DE3) harboring pSBP2 (13). The homogeneity of the protein was accomplished using a 4-step purification procedure: anion exchange chromatography (DEAE-Sepharose), hydrophobic interaction chromatography (phenyl-Sepharose), affinity chromatography (anti-SeBP antibody beads), followed by size exclusion chromatography (TSK SW2000). Particularly, anti-SeBP antibody used here was obtained from rabbit by immunizing recombinant SeBP that was purified using the method described previously (13). The anti-SeBP was conjugated to Affi-Gel 15 (Bio-Rad) to prepare anti-SeBP antibody beads. The protein sample from previous steps was mixed with the beads, then the bound proteins were eluted using ImmunoPure Gentle Ag/Ab elution buffer (Pierce).

Uniformly ¹⁵N-labeled protein was obtained from the culture using minimal medium containing ₁₅NH₄Cl and [U-¹²C]glucose. Uniformly ¹⁵N-, ¹³C-labeled protein was from medium containing ¹⁵NH₄Cl and [U-¹³C]glucose. ¹⁵N-, ¹³ C-₂H-enriched protein was obtained from D₂O-based culture with ¹⁵NH₄Cl and [U-¹³C]glucose. Partially ¹³C-labeled protein was obtained using medium containing 10% [U-¹³C]glucose and 90% [U-¹²C]glucose (16). A mixture of ¹³C-labeled subunit and unlabeled subunit was prepared as follows: the uniformly ¹⁵N-, ¹³C-labeled protein and unlabeled protein were mixed at a 1:1 ratio, denatured using urea, then refolded by removing urea through dialysis. All NMR samples contained 0.1–1.0 mm protein in 20 mm Tris/HCl, pH 7.1, 90% H₂O/10% D₂O, or in 20 mm deuterated-Tris/DCl, pD 7.1, 100% D₂O.

Multi-angle Light Scattering—A size exclusion chromatography column (BioSep-SEC-S 2000, Phenomenex) was coupled to a multi-angle detector (DAWN EOS, Wyatt Technology) equipped with a 30 milliwatt linearly polarized GaAs laser for light scattering photometry. The average molecular weight was determined by the Debye fitting method in the ASTRA software package (Wyatt Technology). A refractive index of buffer of 1.330 and a refractive index increment (dn/dc) of 0.184 were used. The samples studied were 0.5 mm SeBP in a volume of 200 μl. The buffer was 20 mm Tris/HCl, pH 7.1, and the flow rate used was 0.5 ml/min.

Mass Spectrometry—The mass of the intact form of Sebp was determined by direct infusion at 1 μl/min into an electrospray mass spectrometer (Agilent G1969A mass spectrometer with time of flight detector). The mass range from 145 to 200 amu was scanned at 10,000 transients/scan. Voltages were: capillary, 5000 V; fragmentor, 235 V; skimmer, 60 V; and octopole RF, 250 V. The ion source gas temperature was 150 °C, drying gas flow was 10 ml/min, and nebulizer gas pressure was 10 psig. It is important to decrease the gas temperature from the usual 350 °C to preserve the quaternary structure.

NMR Spectroscopy—Spectra were acquired at 45 °C on Bruker 600 or 800 MHz NMR spectrometers, processed using NMRPipe (17), and analyzed with PIPP (18). For the sequential assignment, HN(CO)CACB (19) and HNCACB (20) experiments were carried out using the ¹⁵N-,¹³C-,²H-enriched protein. The assignments of ¹H and ¹³C resonances were done based on four-dimensional ¹⁵N/¹³C-edited NOESY (21) and four-dimensional ¹³C/¹³C-edited NOESY (22) experiments using the ¹⁵N-, ¹³C-labeled protein. Stereospecific assignments for the methyls of leucines and valines were achieved from ¹H-¹³C CT-HSQC experiments (23) using the partially ¹³C-labeled protein (16). The spin-lattice relaxation time constants in the rotating frame (¹⁵N T_1ρ) for backbone amides were calculated from the peak intensities measured using a conventional pulse program (24) with the ¹⁵N-labeled protein.

Proton homonuclear NOEs were obtained from three-dimensional ¹⁵N-edited NOESY (25) with a 300-ms mixing time, four-dimensional ¹⁵N/¹³C-edited NOESY with a 100-ms mixing time, and four-dimensional ¹³C/¹³C-edited NOESY with a 90-ms mixing time. The three-dimensional ¹⁵N-edited NOESY spectra were recorded using 1024 × 256 × 128 complex data points in F3, F2 (¹H), and F1 (¹⁵N) dimensions. The spectral widths were 8399.262, 6024.096, and 1582.279 Hz in the F3, F2, and F1 dimensions, respectively. The four-dimensional ¹⁵N/¹³C-edited NOESY spectra were recorded using 768 × 60 × 120 × 32 complex data points in F4, F3 (¹⁵N), F2 (¹H), and F1(¹³C) dimensions. The spectral widths were 9124, 1563, 50,000, 2587 Hz in the F4, F3, F2, and F1 dimensions, respectively. The four-dimensional ¹³C/¹³C-edited NOESY spectra were recorded using 1024 × 128 × 30 × 30 complex data points in F4, F3 (¹H), F2 (¹³C), and F1(¹³C) dimensions. The spectral widths were 9615, 10,000, 3449, 344.9 Hz in the F4, F3, F2, and F1 dimensions, respectively.

Scalar couplings for one-bond N-H in the presence and absence of Pf1 phage (12 mg/ml) were obtained from the modified version of CT-HNCO experiments using the ¹⁵N-, ¹³C-, ²H-enriched protein. Scalar couplings for one-bond C-H were obtained from BRCT-3DJ (26) experiments using the ¹⁵N-, ¹³C-labeled protein with or without Pf1 phage. The ^3hJ_NC′ scalar couplings across hydrogen bonds were detected using HNCO TROSY experiments with a dephasing delay time of 66.6 ms at the field strength of 18.8 Tesla (27). The three-dimensional ¹³C-edited, ¹²C-filtered NOESY experiment (28) was recorded to confirm contacts between the subunits.

Structure Calculation—Peak intensities from NOESY experiments were translated into proton-proton distance restraints. TALOS (29) predicted the backbone dihedral angles (φ, ψ) from ¹³C^α, ¹³C^β, ¹³C′, ¹H^α, and ¹⁵N^H chemical shifts. Statistically significant angles were used as structural restraints with at least 10° margins. Generic hydrogen bond distance restraints were employed for α-helical regions that were determined based on secondary ¹³C^α and ¹³C^β chemical shifts and medium range NOE patterns. Hydrogen bond distance restraints were also employed for β-strand regions based on the observation of the ^3hJ_NC′ connectivity and NOE patterns in addition to the prediction from secondary ¹³C^α and ¹³C^β chemical shifts. Residual dipolar couplings for N-H and C-H were calculated from the difference in corresponding scalar couplings measured in the presence and absence of Pf1 phage. Residual dipolar coupling restraints were separated into mobile and rigid regions determined based on the backbone dynamic data.

Structures of SeBP were calculated by distance geometry and simulated annealing protocol (30) with the incorporation of dipolar coupling restraints (31) using the XPLOR-NIH (32). Structure calculations employed 811 proton-proton distance restraints, 40 hydrogen bond restraints, 36 φ and 36 ψ angle restraints, 63 N-H, 42 C^α-H^α, and 54 side chain C-H dipolar couplings per subunit. Among 811 proton-proton distance restraints, 250 were intraresidue, 163 were sequential, 99 were medium-range (2 < i < 4), and 299 were long-range (i > 5). The structure of SeBP is defined by 43 intersubunit proton-proton distance restraints. The final set of 10 structures was chosen from 100 calculated structures on the basis of their overall energies. No distance and dihedral restraints are consistently violated by more than 0.4 Å and 5°, respectively. All molecular illustrations were produced using the program MOLMOL (33).

Denaturation Study of SeBP—The mutation of Ile-9 to Ala was made using the QuikChangeII site-directed mutagenesis kit (Stratagene), and the mutant protein I9A was produced and purified as a wild-type protein. The wild-type and mutant SeBP were incubated in 50 mm sodium phosphate, pH 7.0, 1 mm EDTA, 100 mm NaCl, 10 mm DTT for 30 min at room temperature in the anaerobic chamber. Reduced proteins were dialyzed against the same buffer without DTT. Differential scanning calorimetry (DSC) was carried out with a N-DSC II (Calorimetry Sciences) at a heating rate of 1 °C/min. Sample proteins at 0.8 ∼ 1 mg/ml were used for each scanning. The data analysis was performed using CpCalc V.2.1 (Calorimetry Sciences).

RESULTS

Preliminary NMR Studies—A total of 64 resonances corresponding to backbone H-N could be observed in the ¹H^N-¹⁵N HSQC spectrum of the 81-residue polypeptide (Fig. 1A). Resonances of the N-terminal two residues could not be observed, along with the peaks corresponding to Asp¹³–Ile¹⁵ and Phe⁵⁵–Leu⁶⁵. The fact that we could not observe these resonances suggests that some parts of SeBP are undergoing conformational exchange and contain some structural variations. We eliminated possible solvent exchange contribution to the missing resonances by pH titration experiments (pH 6.3–7.5) that showed no significant changes in the NMR spectrum as well as resonance line widths of SeBP (data not shown). The freshly prepared NMR sample did not contain any disulfide bonds. The sample that was left for several weeks after preparation showed the same HSQC spectrum as the freshly prepared sample. The sample in which DTT was added also showed the same HSQC spectrum. Together with the previous finding that Cys⁵⁹ is not easily modified (14), we can eliminate the possibility of missing resonances due to the oxidative state of Cys⁵⁹. Upon closer inspection of the HSQC spectrum, multiple resonances with decreased intensity can be observed for residues preceding and trailing the region corresponding to Phe⁵⁵–Gly⁶⁵ (Fig. 1, B and C). This indicates multiple conformations that are sampled at different rates for different residues. It should be noted, however, that we were able to assign some side-chain atoms for Ile¹⁵, Pro¹⁶, Phe⁵⁵, Ile⁵⁷, and Leu⁶⁵, based on the analysis of NOESY spectra.

FIGURE 1.

¹H-¹⁵N HSQC spectra of SeBP. A, ¹H-¹⁵N HSQC spectra shows that the number of cross peaks for backbone amides is roughly 75% of the number of residues of SeBP, indicating that the protein is composed of a symmetric assembly of the 81-resiude monomer. B, peak intensity of the ¹H-¹⁵N HSQC cross peaks is plotted as a function of the residue number. The residues from Ala⁵⁴ to Leu⁶⁵ were not observed, and nearby residues exhibited low intensities. C, enlarged view of ¹H-¹⁵N cross peaks corresponding to residues Gly⁴⁸–Ile⁵³ and Gly⁶⁶–Lys⁷². The cross peaks for residues next to the non-observable region, such as Leu⁵², Ile⁵³, Gly⁶⁶, and Tyr⁶⁷, show a doublet or a triplet, indicating that this region undergoes a conformational exchange.

Oligomeric State of SeBP—Based on earlier studies, SeBP was proposed to be a homotetramer (13), (14). The 4:1 stoichiometry of SeBP to selenium would suggest that a tetramer is indeed plausible. While the molecular mass of the subunit, as derived from MALDI-TOF, was 8,804 Da, the estimated size of the protein obtained from native PAGE as well as size exclusion chromatography was 42 kDa. Interestingly the denaturing SDS-PAGE showed the highest SeBP band migrated as a 33-kDa protein. Therefore it is natural to deduce from the available molecular size data that SeBP was tetrameric. However, for determining structures by NMR, it is crucial for the exact state of the oligomers to be determined.

The size of the SeBP oligomer will be directly manifested in its NMR relaxation. The ¹⁵N transverse (T₂) relaxation time is a strong measure of the overall correlation time or the overall tumbling rate of the molecule, which in turn reflects its size. The ¹⁵N T₂ values measured at 42 °C and 800 MHz ¹H frequency are shown in Fig. 2. The average ¹⁵N T₂ for the nonflexible portions of the protein is 47 ms. Based on this average value and using the Debye-Stokes-Einstein relationship, assuming a spherical oligomer, the size of SeBP is estimated to be 45 kDa. This is in the range of the previously estimated size of the protein. In addition, a light scattering experiment was performed on the SeBP. The result is shown in Fig. 3. A monodispersed distribution of molecular weights clustered around 43 kDa was found. This is consistent with the NMR relaxation data. What is unequivocally convincing data of the oligomeric size of SeBP was obtained from the mass spectrometry measurement. Under mildly ionizing conditions, the oligomeric state of SeBP could be preserved during the ion spray stage of the mass spectrometer, albeit as a fraction of the population of the full sample. The measured mass of SeBP under this condition is 44,116.9 Da. All of these data taken together indicate that SeBP is pentameric.

FIGURE 2.

Measured value of ¹⁵N T₂ plotted against residue number. ¹⁵N T₂ was acquired at 45 °C and at 800 MHz ¹H frequency.

FIGURE 3.

Multi-angle light scattering data of SeBP. The continuous and broken lines represent the trace of the scattered light and signal from the refractive index detector in volts, respectively. Filled circles, across the eluted peak at ∼19.5 min after injection, indicate the calculated averaged molecular mass. The size of the protein calculated from the hydrodynamic radius is 43 kDa.

Structure of a Subunit of SeBP—NMR data were collected to provide a total of 811 NOEs per subunit for the determination of the SeBP structure. Dihedral restraints were obtained based on TALOS analysis of the chemical shifts of SeBP. The dihedral and NOE data indicated that SeBP contains a single α-helix and four β-strands (Fig. 4A). In addition, dipolar couplings were measured for ¹H-¹⁵N and ¹H-¹³C for the protein aligned in Pf1 phage. An NMR experiment to measure scalar coupling through the hydrogen bond was performed on SeBP to establish the hydrogen bond network present in the protein. Because of the size of SeBP; thus short T₂ values and limited signal to noise, only 4 hydrogen bonds could be directly observed. This however provides important information on how theβ-strands are arranged. A twisted β-sheet is formed by two strands, β2 and β4, in an antiparallel fashion as the through-hydrogen bond ^3hJ_NC′ connectivities were observed between Tyr²¹–H^N and Lys⁷²–C′, between Leu²³–H^N and Ile⁷⁰–C′, between Ala²⁶–H^N and Gly⁶⁸–C′, as well as between Ile⁷⁰–H^N and Gly²⁴–C′, in addition to the NOE connectivities. In contrast, the β1- and β3-strands form a short, parallel β-sheet that is defined mainly by NOEs.

FIGURE 4.

The structure of one subunit of SeBP. A, secondary structure composition of SeBP. The protein consists of one α-helix. B, a stereoview of the backbone (N, C^α, C′) superposition of 20 calculated structures of SeBP subunit. The regions with secondary structures are color-coded as in A. The atomic r.m.s.d. about the mean of the coordinate for the regions with the secondary structures was 0.40 ± 0.06 Å for backbone heavy atoms and 0.95 ± 0.12 Å for all heavy atoms. C, a ribbon representation of the subunit structure. The view on the left has the same orientation as B, while the right view has been rotated 90° about the vertical axis.

The structure of a single subunit of SeBP is illustrated in Fig. 4, B and C. Each subunit of SeBP is a twisted β-sheet containing four strands with the single α-helix capping one side of the sheet. Strand β1 consists of residues Ile⁸–Thr¹⁰ and is followed by a loop connecting it to strand β2 (Leu²⁰– Ser²⁷). Some resonances of the residues (Asp¹³ and Glu¹⁴) in the loop connecting two strands, β1 and β2, were not observed. Residues Thr²⁸–Asn³² form the loop connecting strand β2 to the single helix in SeBP. The helix is composed of residues Val³³–Ala⁴⁶. This helix is followed by a short loop (Lys⁴⁷–Met⁵⁰) that connects it to strand β3 (Gly⁵¹–Ile⁵³) in the structure. A long loop (Ala⁵⁴–Gly⁶⁶) containing the single cysteine residue (Cys⁵⁹) in SeBP connects strand β3 to strand β4 (Tyr⁶⁷–Lys⁷²). Almost all ¹H^N-¹⁵N resonances (Ala⁵⁴–Leu⁶⁵) belonging to residues in this long loop were not observed. As indicated by the increased ¹⁵N transverse (T₂) relaxation times at both termini (Fig. 2), the N-terminal (residues Met¹–Phe⁷) and C-terminal loops (residues Glu⁷⁵–Ala⁸¹) are flexible. The other regions showed well-defined structures except the β3-β4 loop that showed structural dispersion. This is in agreement with the HSQC spectra that suggested possible conformational variations.

Overall Structure of the SeBP—The SeBP structures were calculated from 5 sets of a completely extended starting structure to a pentamer of a barrel-shaped assembly. Several of the long range NOEs obtained from four-dimensional ¹³C/¹³C-edited NOESY experiment using uniformly ¹⁵N-, ¹³C-labeled protein sample, such as: Ile⁸–Ser²², Ile⁹–Ile²⁵, Ile²⁵–Phe⁵⁵, and Ile⁵⁷–Leu⁶⁵, were also detected in three-dimensional ¹³C-edited, ¹²C-filtered NOESY experiments using the mixture of ¹³C-labeled subunit and unlabeled subunit, and therefore included in the calculation as intersubunit contacts. The structures converged to a single conformation (Fig. 5) with the lowest overall energies and acceptable precision as shown in Table 1. The N and C termini of each subunit are located at one side of the pentamer (designated as top), while the loop containing Cys⁵⁹ is at the opposite side (designated as bottom). The top of the SeBP seems to have a slight opening, suggesting a possible solvent entry, while the bottom is closed. The center of SeBP consists primarily of residues of which NMR resonances were not observed, such as Asp¹³–Glu¹⁴ and Phe⁵⁵–Cys⁵⁹.

FIGURE 5.

The NMR structure of SeBP. A, an ensemble of 10 calculated structures. The view on the bottom has been rotated 90° about the horizontal axis. B, ribbon representations of SeBP structure. Both of the top and bottom views have the same orientation as A. C, a close-up view of the contact between two subunits. Sticks indicate side chains of the designated residues. Red balls indicate backbone H^N atoms, and green balls indicate O atoms. Hydrogen bond formations between β1- and β3-strands are; from Ile⁸-O to Leu⁵²-H^N, and from Leu⁵²-O to Thr¹⁰-H^N.

TABLE 1.

Structure statistics for the NMR structure of SeBP

	R.m.s. deviationsa
	10 Lowest energy conformers	Lowest energy conformer
Restraints
NOE distance, Å (4055)	0.020 ± 0.002	0.023
Hydrogen bonds, Å (200)	0.024 ± 0.005	0.022
Dihedrals, (°) (360)	1.04 ± 0.13	1.15
Residual dipolar couplings, Hz (795)
DNH (315)	0.71 ± 0.06	0.67
DCH (480)	1.25 ± 0.01	1.24
Deviations from idealized covalent geometry
Bonds, Å (6278)	0.0042 ± 0.0002	0.0041
Angles (°) (11388)	0.77 ± 0.02	0.74
Impropers (°) (3130)	0.70 ± 0.05	0.58
Structure quality
Lennard-Jonesb potential energy (kcal mol-1)	-1373.6 ± 39.0	-1446.7
Ramachandran plot analysisc,d
Most favored regions	87.9% ± 1.3%	88.1%
Additionally allowed regions	8.5% ± 1.4%	7.6%
Coordinate precisiond
Backbone, Å	1.13 ± 0.08	1.12
All heavy atoms, Å	1.20 ± 0.08	1.20

^aThe statistics were calculated using the 10 lowest energy conformers.

^bThe Lennard-Jones van der Waals energy was calculated with the CHARMM (35) PARAM19/20 parameters and was not included in the structure calculation.

^cResults were calculated with Procheck-NMR (36).

^dOnly the structured regions of SeBP (Ile⁸-Thr¹¹, Gly¹⁷-Ile⁵³, and Gly⁶⁶-Ala⁷³) were analyzed.

NMR resonances from several residues in the loop containing the single cysteine in SeBP could not be observed. However, the conformation of this loop can still be defined, although not as well as the rest of the protein. Despite the β3-β4 loop being long (12 residues), it is mostly positioned inside the barrel, which has a scaffold composed by the α-helix and β-strands. This seems to restrict the loop enough from being completely disordered. In addition, several NOE contacts involving side chains of Phe⁵⁵, Ile⁵⁷, and Leu⁶⁵ to other parts of the protein assist in defining the possible conformations of this loop. It is still important to point out that this loop does not adopt a single conformation. This is reflected in the high r.m.s.d. values in the calculated NMR structures for this loop.

One important question is to understand the high stability of SeBP. The primary forces that stabilize the subunit complex in SeBP are hydrophobic interactions between the subunits. For example, Ile²⁵ showed intersubunit NOEs to Ile⁹ and Val³⁷ (Fig. 5C). NOEs between Ile⁸ and Ser²² were also observed. In the β1-strand, Ile⁸-O and Thr¹⁰-H^N form hydrogen bonds with the H^N and O of Leu⁵² from the β3-strand to form a parallel β-sheet, while backbone H^N and O atoms of Ile⁹ are free, and facing toward Ser²². There is a high likelihood of a hydrogen bond formation between the backbone H^N atom of Ile⁹ and the side chain O^γ atom of Ser²². Such a hydrogen bond would enhance the interactions between subunits and increases the stability of the SeBP.

The electrostatic surfaces of the SeBP are shown in Fig. 6. The outside of the barrel shows bands of positive and negative surfaces. The inner side from the top shows a concentrated negatively charged region. In the bottom of the SeBP, where the β3-β4 loop has undefined conformations, a mix of positive and negative charges can be found.

FIGURE 6.

The electrostatic potential surfaces of SeBP. Positively charged areas are colored blue, and negatively charged areas are in red.

Stability of SeBP—To support our NMR study that showed the stability of SeBP is due to hydrophobic interactions among subunits, we mutated a residue at the interface between the subunits, and investigated its thermal denaturation. We selected Ile⁹ to be mutated because it resides at the interface, and showed intersubunit NOEs to Gly²⁴ and Ile²⁵, being involved in hydrophobic interactions. The structural integrity of I9A mutant protein was confirmed to be the same as that of the wild-type protein by circular dichroism (data not shown).

The thermal stability of the I9A mutant protein, however, was found to be very different from that of the wild-type protein. As shown in Fig. 7, the wild-type SeBP shows a thermal transition at 88.0 °C while the I9A mutant shows a much earlier transition at 70.8 °C in DSC. The mutant protein maintains less interaction forces compared with the wild-type protein, and thus is more heat-sensitive.

FIGURE 7.

DSC of wild type and I9A mutant proteins. The solid line represents the DSC scan for wild-type SeBP protein under the conditions described under “Materials and Methods.” The dotted line represents the DSC scan for I9A mutant protein under the same condition.

Structural Similarity to Other Proteins—The combination of four strands and a helix in a protein structure obviously is not uncommon. However, their arrangement in SeBP is quite unique. The twisted β-sheet is formed by aligning the β-strands 1, 3, 4, and 2 in order. Strands β1, β2, and β3 are parallel and strand β4 is antiparallel with the rest. A search of fold similarity was performed using the DALI program (34) against a representative set of structures in the PDB data base. The DALI search revealed that the pentameric assembly of SeBP is similar to that of a pentameric protein found in other bacteria. The list from the DALI search includes 1Y2I from Shigella flexneri and 2GTC from Bacillus cereus. These two proteins appear to be homologues of each other with 52% sequence identity, but do not show significant similarity to SeBP. The crystal structures 1Y2I and 2GTC were determined as part of a structural genomics effort, and their function is yet unknown. Their structural comparisons are illustrated in Fig. 7. When compared with SeBP, two-thirds of the 1Y2I and 2GTC structures exhibit similarity to the SeBP. The 1Y2I protein has one cysteine in its sequence, but it is located in the middle of strand β2 (indicated by a ball in Fig. 8). There is no cysteine in the sequence of 2GTC. The β3-strands of 1Y2I and 2GTC are elongated and contain a β-sheet along with β2- and β4-strands in an antiparallel fashion. Instead, the corresponding region of SeBP is a loop, of which the backbone amide resonances are unobservable, and expected not to adopt a fixed β-sheet conformation. The overall scaffold of SeBP is similar to those of 1Y2I and 2GTC, suggesting perhaps a general architecture for highly stable proteins where the bottom part of it serves as a platform to define its unique biological function. The coordinates for the 10 lowest energy structures of SeBP have been deposited in the PDB with the accession number 2JZ7.

FIGURE 8.

Comparison between SeBP (A) and structurally similar proteins. The coordinates were obtained from the PDB data base with accession numbers 1Y2I (B) and 2GTC (C). The comparable regions are colored similarly. The middle panels show two of the five subunits to clearly compare the core packing of these proteins.

DISCUSSION

Selenium-binding protein of M. vannielii exists as a pentamer in solution. The interaction between the subunits is stabilized by hydrophobic interaction as well as hydrogen bonds. The strong subunit associations could be related to the protein function to transport selenium in the cell. A stable pentamer provides an excellent framework for safely sequestering the selenium from other factors in various cellular environments. Breakdown in the SeBP during the selenium transport will have adverse consequences to the organism because of the high toxicity of this element. The low number of dynamic and exposed loops connecting various secondary structure elements in SeBP contributes to its stability as well. This feature in conjunction with the amino acid compositions of the loops can reduce the potential of degradation of SeBP by proteolysis.

The single cysteine (Cys⁵⁹) located in the β3-β4 loop is the key residue for selenium reactivity. The loop does not adopt a defined conformation. NMR data indicated that the loop undergoes conformational exchange and conformational variations. However, the SeBP structure revealed that the motion of the loop is limited within the barrel. At the outset, it was anticipated that the Cys residues would be accessible to selenium derivatives, but it was found in the previous study (14) that the Cys residues were not completely exposed without a denaturant. Our structure supports the fact that not all Cys residues are freely accessible. This partial accessibility may be related to the regulated reaction of SeBP with the selenium derivatives. The high specificity for selenium must come from the local environment around the Cys⁵⁹. Several charged (Asp³¹, Asp⁶¹, Lys⁶³) as well as some hydrophobic residues (Thr⁵⁸, Ala⁶⁰, Phe⁶⁴, Leu⁶⁵) are surrounding the Cys⁵⁹. Even though the conformation of this loop is not well determined, therefore the relative placement of these residues are not fixed with respect to Cys⁵⁹, they do offer a possibility to provide specificity for the substrate. In fact, the multiple conformations adopted by this loop may be necessary to provide selectivity in selenium binding.

Access to the Cys⁵⁹ can be achieved from either side of the SeBP core. Entrance from the top of the core, where a small opening is already present, has to involve interaction of the substrate with residues lining the core, which are mostly hydrophobic. Access from the bottom of the core will have to involve the opening of the Cys⁵⁹ loop to expose the cysteine. Either mechanisms will affect the on and off rate of substrate binding. This will also add to the stability of the bound selenium. The need for different selenium precursor or other factors to induce the opening of the Cys⁵⁹ loop to expose the cysteine to bind selenium cannot be ruled out. On the other hand, it is likely that only interaction with the selenophosphate synthase can induce a conformational change in the SeBP that exposes the bound selenium for the enzyme. This provides the selectivity in delivery of the selenium for this particular enzyme.

Selenium can be found in a variety of compounds, where the most common oxidation numbers of selenium are 6, 4, 2, -1, and -2, including selenophosphate (H₂SePO^–₃) and selenodiglutathione (GS-Se-SG). The possible coordination states of selenium as well as the measured stoichiometry between selenium and SeBP (14) do not seem to correspond to the five cysteines found in the pentameric SeBP. It is still not clear what important role one additional cysteine plays in the binding of selenium. Naturally a structure of SeBP with bound selenium will address this question and complete the picture of molecular interactions that define the function of this protein. As shown previously, the binding of selenium to SeBP in vitro is not trivial (14). This has to be carried out under partially denaturing and anaerobic conditions with the SeBP fully reduced, and it involves selenodiglutathione as the precursor of the selenium. So far the yield of bound selenium to SeBP under our sample conditions is only a few percent, eliminating the possibility of structurally studying this complex. It is clear that future structural studies are needed to address some of the questions regarding the selenium biosynthetic pathways in this organism. Nevertheless, our current SeBP structure affords the first glimpse of how this protein can manage the potentially toxic selenium in the cell and deliver it as an important nutrient for the organism.