Crystal structures of RbsD leading to the identification of cytoplasmic sugar-binding proteins with a novel folding architecture.

RbsD is the only protein whose biochemical function is unknown among the six gene products of the rbs operon involved in the active transport of ribose. FucU, a paralogue of RbsD conserved from bacteria to human, is also the only protein whose function is unknown among the seven gene products of the l-fucose regulon. Here we report the crystal structures of Bacillus subtilis RbsD, which reveals a novel decameric toroidal assembly of the protein. Nuclear magnetic resonance and other studies on RbsD reveal that the intersubunit cleft of the protein binds specific forms of d-ribose, but it does not have an enzyme activity toward the sugar. Likewise, FucU binds l-fucose but lacks an enzyme activity toward this sugar. We conclude that RbsD and FucU are cytoplasmic sugar-binding proteins, a novel class of proteins whose functional role may lie in helping influx of the sugar substrates.

known. The only clue about the function of RbsD is that it enhances the ribose uptake and growth of a mutant E. coli strain lacking the rbs operon but containing plasmid-encoded RbsK and mutated glucose transporter ptsG on its genome, which allows transport of D-ribose at low affinity (5). MCD-Ribose in solution exists as multiple forms: ␤-pyranose (58.9%), ␣-pyranose (20.3%), ␤-furanose (13.3%), ␣-furanose (7.4%), and open chain aldehyde form (0.13%) (6). In contrast, D-ribose in biomolecules is exclusively in the furanose forms, the five-membered ring forms. The cyclic forms of the sugar interconvert spontaneously via the least populated acyclic aldehyde. Surprisingly, ribokinase binds and phosphorylates the ␣-furanose form of D-ribose, the least populated ring form (7). For an efficient supply of the ␣-furanose form of D-ribose to ribokinase, an enzymatic activity of increasing the interconversion between the different forms of the sugar may be required, or the spontaneous conversion of the sugar at physiological conditions is fast enough so that such an enzyme activity is unnecessary.
A sequence alignment shows that RbsD is found in a variety of bacteria and that an evolutionarily conserved protein FucU (Fucose, Unknown) is a close paralogue of RbsD, with any compared pair of RbsD and FucU homologues sharing about 20% amino acid sequence identity with each other. Prokaryotic FucU exists as a component of fucose regulon (8 -10), whose gene products are involved in the uptake and utilization of L-fucose (6-deoxy-L-galactose) (11). The fucose regulon encodes the following seven components: fucose permease (L-fucose H ϩ symporter), fucose isomerase, fuculose kinase, fuculose-1-phosphate aldolase, 1,2-propanediol oxidoreductase, regulatory protein, and FucU (9). Of these, FucU is the only protein whose function is unknown. L-Fucose is the major component in various oligo-and polysaccharides and glycosides in mammals as well as bacteria and plant (12). The isomerization of L-fucose into L-fuculose, which is subsequently phosphorylated into fuculose 1-phosphate, is the first step in the degradation of the sugar (13). Or L-fucose is converted into L-fucose 1-phosphate by fucose kinase as the first step in the synthesis of fucosecontaining oligo-and polysaccharides (11). The conservation of FucU in higher organisms including human highlights the functional significance of the protein. Both FucU and RbsD, containing no signal sequence, must be cytoplasmic proteins.
We sought to elucidate the biochemical function of RbsD and FucU on the basis of the three-dimensional structure of RbsD or FucU to provide a complete picture of the energy-driven transport of ribose and fucose. We determined the structure of Bacillus subtilis RbsD, which reveals a novel protein fold that associates into a homodecameric assembly. Crystallographic and other physicobiochemical studies led to the conclusion that RbsD binds specifically the ␤-anomeric forms of D-ribose and FucU binds L-fucose, revealing the existence of cytoplasmic sugar-binding proteins. The name is coined reminiscent of the periplasmic sugar-binding proteins that are one component of the proteins involved in the active transport of various neutral sugars. A potential role of these proteins may lie in helping translocation of sugar substrates into cells.

MATERIALS AND METHODS
Overexpression, Purification, and Crystallization of RbsD-E. coli RbsD and B. subtilis RbsDs were cloned from respective genomes and expressed in E. coli BL21(DE3) (Novagen). The purification and crystallization of E. coli RbsD were reported (14). The cells were grown in Luria-Bertani media containing 0.1 mg ml Ϫ1 ampicillin. The expression of the protein was induced by 1 mM isopropyl-␤-D-thiogalactopyranoside (IPTG) at an optical density of 0.6 at 37°C for 7 h. Both RbsDs were produced without a tag, which was the right choice since the both N and C termini are involved in the oligomerization of the proteins. B. subtilis RbsD was purified by ammonium sulfate precipitation (50 -60% fractionation) and column chromatographic separation employing a Hitrap Q (Amersham Biosciences) and a Hiload 26/60 Superdex 200 sizeexclusion column (Amersham Biosciences). The purified protein was concentrated to 10 mg ml Ϫ1 in 20 mM Tris-HCl buffer (pH 7.4) containing 0.2 M NaCl and 2 mM dithiothreitol. The crystals of B. subtilis RbsD were grown by hanging-drop vapor diffusion method against reservoir solution containing 20% polyethylene glycol 4000, 19% isopropyl alcohol, and 100 mM sodium cacodylate (pH 6.5).
Overexpression and Purification of E. coli FucU-The E. coli FucU was amplified by the PCR technique from E. coli cell lysate. The PCR products were ligated into pET21a vector (Novagen). The resulting vector was introduced into E. coli BL21(DE3) strain. The expression of recombinant FucU protein was induced by 1 mM IPTG at an optical density of 0.4 at 37°C for 7 h. Bacterial lysate was prepared by sonication in a buffer solution containing 20 mM Tris-HCl (pH 7.4), 0.1 M NaCl, and 1 mM dithiothreitol. The protein was purified by ammonium sulfate precipitation (30% fractionation) and using a Q-Sepharose fastflow column (Amersham Biosciences). The purified FucU was dialyzed against 20 mM sodium phosphate buffer (pH 7.6) containing 1 mM ␤-mercaptoethanol.
Overexpression and Purification of E. coli Ribokinase-The E. coli RbsK gene was amplified by the PCR method from E. coli cell lysate. The PCR products were ligated into pPROEX HTa vector (Invitrogen) and introduced into E. coli BL21(DE3) strain. The protein was expressed as a fusion protein containing a His 6 tag at the N terminus. The cells were grown in Luria-Bertani media containing 0.1 mg ml Ϫ1 of ampicillin. The expression of the protein was induced by 1 mM IPTG at an optical density of 0.6 at 18°C for 7 h. Bacterial lysate was prepared by sonication in buffer A containing 20 mM Tris-HCl buffer (pH 7.4) and 0.1 M NaCl. The fusion protein bound to a nickel-nitrilotriacetic acid column (Qiagen) was eluted with buffer A containing 200 mM imidazole after washing the column with buffer A containing 20 mM imidazole. The eluted fractions containing RbsK were dialyzed against buffer containing 30 mM Tris-HCl buffer (pH 8.0).
Structure Determination of RbsD in Complex with Glycerol-Initially, we tried to solve the structure of E. coli RbsD. Although we obtained the crystals of the protein (14), difficulty with cryocooling of the crystals hampered the determination of the structure. Subsequently, we switched to B. subtilis RbsD, whose crystals were easier to handle. The crystals belong to the space group C2 with unit cell dimensions of a ϭ 123.52, b ϭ 108.66, c ϭ 83.31 Å, and ␤ ϭ 128.68°. The asymmetric unit of the crystal contained a pentameric ring of RbsD molecules. A data set at three different wavelengths was collected with a selenomethionyl RbsD crystal at 100 K on the beamline 6B of the Pohang Accelerator Laboratory at Pohang, Korea, and were processed using the HKL software (15). Before cryocooling, crystals were briefly immersed in the same precipitant solution containing 7% glycerol. Ten selenium sites in the asymmetric unit of the crystal were located and used for phase determination at 2.8 Å with the program SOLVE (16), and phases were subsequently improved by density modification with the program RESOLVE (17). The electron density was of excellent quality showing nearly all features of protein side chains. The program O (18) was used for model building, and refinement was performed against 2.3-Å native data with the CNS package (19). From the beginning of the chain tracing, binding of glycerol to the protein was apparent.
Determination the Structures of RbsD and the Protein in Complex with Ribose or Ribose 5-Phosphate-Diffraction data were collected at room temperature with crystals mounted on a capillary tubes using a Rigaku Raxis IV 2ϩ area detector system on a rotating anode generator. For the structure determination of the complexes, a small amount of D-ribose or D-ribose 5-phosphate was introduced directly into the crystallization drops using a hair, which then sat at least overnight before data collection. The RbsD structure and the structures of the complexes were solved by direct refinement of the structure of the RbsD-glycerol complex against each of the two data sets. The resulting models fell within the limits of all the quality criteria of the program PROCHECK (20). Table I shows the crystallographic and refinement statistics of the RbsD structures.
NMR Spectroscopy-13 C NMR spectra of D-[1-13 C]ribose and 1 H NMR spectra of L-fucose were collected with Bruker DRX500 spectrometer at 25°C with proton decoupling. For 13 C NMR spectra, a total of 16 transients (ribose only) or 256 transients (ribose ϩ RbsD) in 64 K data points, relaxation time of 2.0 s, and spectral width of 10,000 Hz were used. For 1 H NMR spectra, a total of 64 transients in 32 K data points, relaxation time of 2.0 s, and spectral width of 7,000 Hz were used. No line broadening was employed in processing free induction decays with XWIN-NMR software (Bruker Instruments). The 13 C chemical shift was referenced from the reported value for C-1 of the ␣-pyranose form of D-ribose (6), and the 1 H chemical shift was referenced from that of the water peak set to 4.6 ppm.
Isothermal Titration Calorimetry-All measurements were carried out at 25°C on a MicroCalorimetry System (Microcal). RbsD and FucU were dialyzed against a buffer solution containing 20 mM sodium phosphate (pH 7.6) and 1 mM ␤-mercaptoethanol. D-Ribose, D-ribose 5-phosphate, and L-fucose were dissolved in the same buffer. The samples were degassed for 20 min and centrifuged to remove any residuals prior to the measurements. Dilution enthalpies were determined in separate experiments (titrant into buffer) and subtracted from the enthalpies of the binding between the proteins and the sugars. Data were analyzed using the Origin software (OriginLab Corp.).

RESULTS AND DISCUSSION
Description of the Structure of RbsD-Unless stated otherwise, RbsD refers to B. subtilis RbsD throughout the text. The structure of RbsD was determined with multiple-wavelength anomalous dispersion phasing by using one selenomethionine derivative crystal of the protein (Table I). The final model, refined against data to 2.3-Å resolution, consists of all the residues, 5 glycerol molecules, 2 chloride ions, and 369 water molecules. The electron densities for all the backbone atoms are well defined. The 5 RbsD molecules in the asymmetric unit of the crystal are associated in a head-to-tail fashion to form a pentameric ring structure, which tightly interacts with the adjacent pentameric ring that is related by the crystallographic 2-fold axis (Fig. 1a). The inner and the outer diameter of the ring is ϳ20 and 80 Å, respectively, whereas the thickness of the decameric assembly is ϳ60 Å. Extremely tight interactions between the two pentameric toroids obscure physical distinction of the two, whereas the 5 molecules on a pentameric toroid are discernible. We reported previously that E. coli RbsD is an octamer in solution on the basis of an analytical centrifugation analysis (14). However, the structure of E. coli RbsD, determined by the molecular replacement method, shows that it is also a decamer (data not shown). The incorrect conclusion is ascribed to the non-globular shape of the quaternary structure of the protein. The subunit structure can be described as an ␣/␤ fold with the central 6-stranded ␤-sheet sandwiched by three ␣-helices on one side and one ␣-helix plus one ␤-strand-like loop on the other side (Fig. 1b). The N-terminal loop (residues 1-9) of one molecule is extended from the main body of the structure and tightly interacts with three adjacent molecules on the decameric assembly. A strong and round electron density was found surrounded by the amino group of Lys-2 and water molecules at the interface of the pentameric rings (Fig. 1c). This cage, which is completely buried inside the protein, is most likely to arise from binding of a chloride ion. The same feature is found in the structure of E. coli RbsD, and Lys-2 is conserved in all but two RbsD homologues in the sequence data bases, suggesting that this ion cage is important for the assem-bly and stability of the decameric structure. A search for analogous folding architectures in the PDB data base using the program Dali (21) revealed that the structure of RbsD is distinct from any other known protein folds. The structure of a hypothetical trimeric protein YjgF (Protein Data Bank code 1QU9; Z-score 3.5), listed as the closest match, is obviously different from that of RbsD. Unlike the structure of RbsD, one side of the central ␤-sheet of YjgF has no peptide segment and serves as the trimeric interface of the protein.
We found that the clefts between the RbsD subunits bind glycerol (data not shown) when the crystals were immersed in the cryoprotectant containing 7% glycerol before data collection. It has been commonly observed that enzymes with a toroidal structure have the active site at the interface between adjacent subunits (for examples, 2-Cys peroxiredoxin, ribulosebisphosphate carboxylase/oxygenase, cyanase, muconolactone isomerase, and glutamine synthetase), indicating that the oligomer formation is vital for the enzyme activity. The interaction between RbsD and glycerol is weak when assessed by isothermal titration calorimetry (ITC). It was not possible to measure the dissociation constant K D by this method. The high concentration of glycerol present in the cryoprotectant presumably allowed the observation of the bound glycerol. Glycerol may be one of the polyol compounds that are able to bind fortuitously to the binding pocket of RbsD that is designed to interact most strongly with D-ribose (discussed below). The interaction of RbsD with glycerol is presumably physiologically irrelevant.
RbsD Lacks an Enzyme Activity toward Ribose-We recorded 13 C NMR spectra of D-[1-13 C]ribose in the presence and absence of RbsD. The addition of RbsD did not give rise to new peaks, demonstrating that the protein lacks an enzyme activity to convert D-ribose into other compounds. Instead, it caused significant broadening of the resonances arising from the 13 C-1 atoms of the ␤-pyranose and ␤-furanose forms but not those of the ␣-furanose and ␣-pyranose forms (Fig. 2). These spectroscopic data indicate that RbsD binds both the ␤-anomeric forms of D-ribose but not the ␣-anomeric forms of the sugar. Because the NMR spectrum does not indicate binding of the ␣-anomeric D-ribose, RbsD is not an anomerase, which catalyzes the interconversion between the ␣and ␤-anomeric forms of the sugar. We considered the possibility of RbsD being an enzyme that could increase the interconversion between the ␤-anomeric forms and the open chain form of D-ribose. If RbsD has this enzyme activity, the line broadening of the two NMR peaks would be due to chemical exchange process between the ␤-furanose and ␤-pyranose forms of D-ribose with the open chain aldehyde. Such an enzyme activity could enhance the availability of the least populated ␣-furanose form of D-ribose for ribokinase. For example, when the ␣-furanose form is depleted (a non-equilibrium situation), rapid conversion of ␤-anomeric forms into the open chain form should increase the appearance of the ␣-furanose form through the spontaneous conversion of the open chain form into the ␣-furanose form. We cloned and overexpressed E. coli ribokinase and measured the enzyme activity of phosphorylating ribose in the presence and absence of RbsD according to the method reported in the literature (22). RbsD did not enhance the enzyme activity of ribokinase (data not shown), indicating that the protein does not have this hypothetical enzyme activity. We also compared the rates of the change in the optical rotation of freshly dissolved solid D-ribose, which is exclusively in the ␤-furanose form, in the presence and absence of RbsD. Consistent with the enzymatic assay of ribokinase, the protein did not enhance the rate.
Binding of the ␤-Pyranose Form of D-Ribose-The exposure of the intersubunit clefts to the bulk solvent in the RbsD crystals facilitated the determination of the structures of RbsD in complex with D-ribose or D-ribose 5-phosphate using the crystals soaked with each of the two compounds. In order to avoid competition of glycerol for binding to the clefts, data were collected at room temperature. With a D-ribose-soaked crystal, less featured "fat" electron density was found at each of the 10 intersubunit clefts (Fig. 3). The poorly defined electron density, despite the high resolution of the data (1.95 Å), suggested the binding of ribose molecules in different configurations. Given the NMR data demonstrating the binding affinity of RbsD for the both ␤-pyranose and ␤-furanose form of D-ribose, we first interpreted the density with the most populated ␤-pyranose form of D-ribose. The modeled ␤-pyranose in a standard chair conformation mostly accounted for the electron density (Fig.  3a). The sugar-binding mode reveals that one of the two adjacent RbsD molecules provides a predominant contribution over the other. Asp-28, His-98, Lys-102 (via a water molecule), Tyr-120, and Asn-122 of one subunit and only His-20 of an adjacent subunit are in contact with the bound ribose. The aromatic ring of Tyr-120 is packed against the hydrophobic part of the bound sugar, a typical pattern frequently observed in the proteinsugar interactions. The bound ribose molecule fits tightly into the intersubunit cleft and leaves almost no room (Fig. 3b).
Binding of the ␤-Furanose Form of D-Ribose-Interpretation of the fat electron density with the ␤-furanose form of D-ribose was impossible. The electron density did not show the feature of the -CH 2 OH group of the furanose form. We determined the binding mode of the ␤-furanose form indirectly by elucidating the 2.05-Å resolution structure of RbsD in complex with Dribose 5-phosphate, whose ring form is exclusively the furanose form. As expected, the electron density for the five-membered ring was defined well enough to allow the unambiguous fitting of the sugar ring (Fig. 3c). The phosphorylated sugar that we used was the mixture of the ␣and ␤-anomers. The electron density indicated that RbsD binds selectively the ␤-anomer of D-ribose 5-phosphate, which is consistent with the NMR data. The result also supports the conclusion that RbsD is not an anomerase. The atomic temperature factors of the -CH 2 OPO 3 2Ϫ group are over 80 Å, and the electron density for this group is missing (methylene part) or diffused (phosphate group). As the phosphate group of the bound sugar barely interacts with the protein, the ␤-furanose form of D-ribose would bind to the protein in the same mode observed for D-ribose 5-phosphate. Notably, Asn-122 does not interact with the bound ribose 5-phosphate. In order to confirm the correctness of the crystallographically deduced sugar-binding modes, we substituted His-98 with alanine, whose imidazole ring provides a hydrogen bond to the both ␤-anomeric forms of the sugar (Fig. 3). The NMR spectrum of D-[1-13 C]ribose in the presence of the mutant RbsD exhibits significantly less broadening of the two peaks corresponding to the ␤-anomers of D-ribose (Fig. 2) than in the presence of the wild-type RbsD, indicating that the sugarbinding affinity of the mutant is reduced. This is consistent with the deduced sugar-binding modes.
Sugar-binding Residues Are Virtually Invariant-A sequence alignment reveals that Asp-28, His-98, Asn-122, and His-20 are invariant, whereas Lys-102 and Tyr-120 are 100% homologously conserved, among the 40 deposited sequences of RbsD homologues except for three entries. The conservation of the sugar-binding residues is significant in that Pro-32 and Gly-105 are the only two residues that are more than 97% conserved among the RbsD homologues. One of the three exceptions is NP_326431 (annotated as Mycoplasma pulmonis RbsD, GenBank™), which contains two substitutions of His-20 with threonine and His-98 with asparagine (Fig. 4). The other two entries in GenBank™ are NP_407075 and ZP_00060076, which contain a substitution of Lys-102 with histidine and a substitution of Tyr-120 with asparagine, respectively. Interestingly, two of the three genes are not a component of a canonical rbs operon but are associated with genes coding for putative phosphotransferase enzyme II, A and B (NP_326431, Gen-Bank™), or araC family transcriptional regulatory protein (NP_407075, GenBank™), which does not share sequence homology with RbsR. The two genes are likely to be RbsD paralogues whose gene products have binding specificity for sugars different from ribose. The observation raises a possibility that the energy-driven transport of neutral sugars other than ribose and fucose may also require biochemical activity similar to that of RbsD in some organisms.
FucU Lacks an Enzyme Activity toward L-Fucose-FucU, a close relative of RbsD, is a highly conserved protein; E. coli FucU shares 50% sequence identity with human FucU (Fig. 4). We cloned and purified E. coli FucU, which shares 21% sequence identity with RbsD. The elution times of the two proteins from a size-exclusion Superdex 200 column were the same, indicating that FucU also adopts a decameric quaternary structure similar to that of RbsD. In FucU, the ribose-binding residues His-20, Asp-28, Tyr-120, and Asn-122 in RbsD are also 100% conserved, whereas His-98 and Lys-102 are substituted with and strictly conserved as arginine and tyrosine, respectively (Fig. 4). L-Fucose is D-ribose containing the 6-methyl group and having a different configuration at the C-2 position. Most likely, these substitutions provide FucU with specificity and affinity for binding L-fucose. We recorded the 1 H NMR spectrum of L-fucose in the presence and absence of FucU (Fig.  5). The 1 H NMR peaks of L-fucose did not change after a 2-h incubation of the mixture at room temperature, demonstrating that FucU lacks an enzyme activity of converting L-fucose to a product.
Weak Sugar-binding Affinity of RbsD and FucU-We analyzed the interactions of RbsD and FucU with sugar molecules by ITC. The isothermal titration curve for ribose 5-phosphate readily indicated that the binding affinity of RbsD for this phosphorylated sugar is quite weak and prevented us from obtaining the dissociation constant by this method (Fig. 6a). However, it was possible to analyze the titration curve obtained for the interaction between RbsD and D-ribose (Fig. 6b). The analysis led to the K D value of 0.93 mM. The data demonstrate that the ␤-pyranose form binds to RbsD more strongly than the ␤-furanose form of ribose. In the data analysis, we assumed that the ␤-pyranose form of D-ribose contributes solely to the heat release in the titration, and we used the equilibrium concentration of this form of the sugar in solution. Considering some contribution by the ␤-furanose form to the heat release during the titration, the actual K D values for the interaction between the two should be slightly higher than 0.93 mM. FucU binds L-fucose also weakly with the K D value of 1.61 mM (Fig.  6c). In analyzing the data, we assumed that FucU binds ␣and ␤-forms of L-fucose with the same affinity, because the anomeric specificity of FucU is unknown.
The binding affinity between RbsD and D-ribose is quite low. The differential gain of free energy in the transfer of ribose from the solvent to the binding cleft of RbsD seems small due to the excellent hydrogen-bonding capacity of the sugar molecule in water. The mode of interaction between RbsD and ribose is uniquely distinguished from that between periplasmic RBP and ribose. Although the ribose of binding to RbsD does not induce a noticeable conformational change of the backbone or a side chain of the protein (data not shown), the binding of ribose to periplasmic RBP triggers a large domain movement that leads to the shielding of the bound ribose from the bulk solvent and the formation of new interactions between protein atoms (23). Consequently, the binding of ribose to periplasmic RBP is substantially tight, exhibiting the K D value of 0.13 M (24).
What Would Be the Biological Role(s) of RbsD and FucU?-We demonstrated that the biochemical function of RbsD and FucU is to bind specific forms of D-ribose and fucose, respectively. The residues at the sugar-binding site in the both proteins are remarkably conserved. Therefore, the functional role(s) of the proteins must be related to this sugar binding activity. We ruled out an effector function, because the ribosebound structures of RbsD exhibit no noticeable conformational change of the protein compared with the structure of the uncomplexed RbsD, and the oligomerization state of RbsD is not the result of the ribose binding. RbsD does not appear to be an essential enzyme at the downstream of RbsK, because the E. coli ptsG mutant cells lacking RbsD but containing RbsK can grow on ribose as the only carbon source (5). A 4-fold enhancement of the ribose uptake by expressing RbsD in this mutant (5) is presumably because of the action of the protein . The mixture was incubated for 2 h before the data collection. Residual water peak was suppressed by the pulsed field gradient method as described (27). upstream of RbsK. Noticeably, ribokinase specifically phosphorylates the ␣-furanose form of D-ribose, the least populated ring form of the sugar (7), whereas RbsD selectively binds the ␤-anomeric forms of D-ribose that account for about 72% equilibrium population of the sugar. Therefore, without directly competing with ribokinase, RbsD would be able to play a role that requires binding of D-ribose. In playing the role, the low affinity of RbsD for ribose may be required for minimal interference with the availability to ribokinase of the ␣-furanose form of D-ribose, which is generated through the nonenzymatic conversion of the free ␤-anomers. We do not know how RbsD would exert its cellular function by binding to the ␤-anomers of D-ribose. The functional role may lie in facilitating the influx of the sugar substrate, the event upstream of RbsK.
In prokaryotic cells, fucose permease appears primarily responsible for the energy-driven uptake of L-fucose (25). FucU may play a similar role as RbsD by binding L-fucose. We do not know yet the anomeric specificity of FucU in binding L-fucose due to unavailability of commercial C-13-labeled L-fucose and difficulties in obtaining suitable crystals of the protein. It remains to be determined whether, like RbsD, FucU may bind specifically an anomeric form of L-fucose that is not used by fucose isomerase or fucose kinase, the first enzymes in the utilization the sugar.
The membrane-bound fucose permease and ribose permease (RbsC) are unrelated proteins. The mutated ptsG is also unrelated to RbsC, but the overexpression of RbsD enhances the ribose uptake through this transporter (5). These observations suggest that RbsD and FucU do not interact directly with their respective permease component to exert function.
Concluding Remarks-The existence of periplasmic ligandbinding proteins has long been known. The biochemical functions and action modes of these proteins are well characterized. The novel protein fold of RbsD and the sugar binding activity of RbsD and FucU characterized in this study reveal the existence of the cytoplasmic sugar-binding proteins. Although RbsD genes are found only in bacteria, FucU genes are also found in mouse and human genomes, which underscores the functional importance of the protein. The two genes must have been derived from the same ancestral gene and evolved to have different sugar-binding affinities. It remains to be determined whether there are unidentified cytoplasmic sugar-binding proteins with different ligand specificity that act on the transport of other neutral sugars. The two RbsD homologues described above are potential candidates. It is possible that remote homologues of RbsD may play roles in the influx of other sugars. The biochemical function of RbsD and FucU identified in this study is to bind D-ribose and L-fucose, respectively. The sugar binding activity of the proteins should be tightly linked to the utilization of these sugars, because the gene expressions of RbsD and FucU are regulated as a part of the rbs operon or the fucose regulon in prokaryotes. In depth in vivo study should be necessary to elucidate the biological roles of RbsD and FucU. The presented structures and the common biochemical function of the two proteins delineated in this study are a footstep toward gaining a complete picture for the energy-driven transport of the neutral sugars.