Ligand-induced conformational changes in a thermophilic ribose-binding protein

Background Members of the periplasmic binding protein (PBP) superfamily are involved in transport and signaling processes in both prokaryotes and eukaryotes. Biological responses are typically mediated by ligand-induced conformational changes in which the binding event is coupled to a hinge-bending motion that brings together two domains in a closed form. In all PBP-mediated biological processes, downstream partners recognize the closed form of the protein. This motion has also been exploited in protein engineering experiments to construct biosensors that transduce ligand binding to a variety of physical signals. Understanding the mechanistic details of PBP conformational changes, both global (hinge bending, twisting, shear movements) and local (rotamer changes, backbone motion), therefore is not only important for understanding their biological function but also for protein engineering experiments. Results Here we present biochemical characterization and crystal structure determination of the periplasmic ribose-binding protein (RBP) from the hyperthermophile Thermotoga maritima in its ribose-bound and unliganded state. The T. maritima RBP (tmRBP) has 39% sequence identity and is considerably more resistant to thermal denaturation (appTm value is 108°C) than the mesophilic Escherichia coli homolog (ecRBP) (appTm value is 56°C). Polar ligand interactions and ligand-induced global conformational changes are conserved among ecRBP and tmRBP; however local structural rearrangements involving side-chain motions in the ligand-binding site are not conserved. Conclusion Although the large-scale ligand-induced changes are mediated through similar regions, and are produced by similar backbone movements in tmRBP and ecRBP, the small-scale ligand-induced structural rearrangements differentiate the mesophile and thermophile. This suggests there are mechanistic differences in the manner by which these two proteins bind their ligands and are an example of how two structurally similar proteins utilize different mechanisms to form a ligand-bound state.


Background
Bacterial periplasmic binding proteins (PBP) are receptors for extracellular solutes in metabolite uptake [1], chemotaxis [2], and intercellular communication [3] processes. The PBPs collectively constitute a structural protein super-family characterized by two pseudo-symmetric domains that are linked by a hinge formed by two or three βstrands connecting the domains; a ligand-binding site is situated at the interface between the two domains [4]. Each domain adopts a three-layered α/β/α sandwich fold and is classified into one of three structural sub-categories (group I/ribose-binding protein fold, group II/maltosebinding protein fold, and group III/Vitamin B12-binding protein fold) [5] according to β-strand topology.
Ligand-free PBPs adopt an open conformation in which the inter-domain interface is exposed to solvent. Solute binding induces a conformational change to form a closed state in which the ligand is bound at the domain interface and buried by the surrounding protein [6][7][8]. This closed form typically binds to other molecular components to trigger downstream cellular processes such as chemotaxis [9], quorum sensing [3], and transmembrane ligand transport [10]. Eukaryotic receptors that contain the PBP fold as part of multi-domain proteins are also regulated by ligand-induced conformational coupling mechanisms [11].
A collection of PBP structures determined in both apo and ligand-bound states (Table 1 and references therein) has provided a wealth of information on the ligand-induced domain motions of PBPs. Analysis of the ligand-induced conformational changes in PBPs has to differentiate between different types of motions: large-scale (interdomain) movements, loop movements, relative intradomain movements of secondary structure elements, and amino acid side-chain reorganization. Large-scale changes in PBPs can be described as a rigid body motion of the two domains, characterized by bending/twisting motions around two axes [12]. The magnitude of this hinge-bending motion ranges from 62° in a mutant E. coli ribosebinding protein [8] to as little as 14° in the leucine-binding protein [13]. PBPs such as the E. coli ribose-binding protein (RBP) [8] and allose-binding protein [7] have been shown to adopt a series of intermediate values in their apo state suggesting that the observed states represent snapshots of a continuum between two extremes: the defined closed form, and a less precisely defined fully open conformation.
In E. coli RBP (ecRBP) small-scale backbone movements are restricted to the hinge region, whereas the secondary structure elements in the two domains and the amino acids in the binding pocket adopt essentially the same conformations in both the apo and ribose-bound forms [8]. However, in E. coli leucine-binding protein, not only the hinge region, but also loops and amino acid sidechains in the binding pocket show ligand-induced changes [13], many of which are restricted to one domain. This difference in conformational changes between the domains has been postulated to imply ordered interactions between the protein and ligand [8,13,14].
The ligand-induced conformational changes have not been described previously in a thermophilic PBP. We have characterized the stability, determined the ligand-binding properties, and solved the X-ray crystal structures of the apo and ligand-bound forms of a thermophilic periplasmic ribose-binding protein from the hyperthermophile Thermotoga maritima (tmRBP), the mesophilic homolog, ecRBP, of which has been studied in detail [8,15,16]. The ecRBP and tmRBP proteins share 39% amino acid sequence identity, but differ by 52°C in apparent thermal stability. We find that the interdomain motions, although not of the same magnitude, exhibit similar movements. The amino acids in the tmRBP sugar-binding pocket undergo ligand-induced conformational changes, whereas their conformations in apo ecRBP are essentially pre-formed for ligand binding.

Expression
The RBP gene was identified in the T. maritima genome sequence [17] as open reading frame (ORF) tm0958, based on sequence similarity to the E. coli RBP, and genetic linkage of this ORF within a putative operon that contains sequences for ABC transporters characteristic of a ribose transport system [18]. ORF tm0958 was amplified from T. maritima genomic DNA using the polymerase chain reaction. The resulting DNA fragment was cloned into a pET21a vector with a C-terminal hexa-histidine tag preceded by a glycine-serine linker. The nucleotide sequence of the recombinant was confirmed by DNA sequencing. Overexpression of this ORF in E. coli produced ~50 mg of pure protein per liter of growth medium, which was purified by immobilized metal affinity chromatography [19] followed by gel filtration chromatography.
The gel filtration elution profile of tmRBP consists of two peaks, one of which is consistent with a monomeric tmRBP (34 kDa), the other consistent with a ~55 kDa protein ( Figure 1). SDS-PAGE of the resulting fractions revealed that both peaks contain tmRBP. The fractions corresponding to the 55 kDa protein also contain significant amount of a ~20 kDa species (Figure 1). Tryptic digestion of this 20 kDa protein, followed by MALDI mass spectrometry peptide mapping [20], revealed that it corresponds to a truncated form of the full-length tmRBP (Figure 2). The 55 kDa protein is therefore a heterodimer consisting of one full-length and one truncated copy of tmRBP. Neither full-length, nor truncated homodimers were observed. Analysis of the tm0958 DNA sequence suggests that this truncation may result from translation initiation at methionine 142 (numbering according to NCBI NP 228766), which is preceded by a ribosome binding site ( Figure 2). This interpretation is further supported by the M142A mutant tmRBP in which the 20 kDa truncation is absent (data not shown).

Thermal Stability
The apparent thermal stability ( app T m ) of full-length monomeric wild-type tmRBP was determined by thermal denaturation using circular dichroism (CD) [21]. In the Expression and purification of the tm0958 ORF absence of denaturant, no significant change in the CD signal could be observed as a function of temperature (data not shown). All measurements were therefore carried out in the presence of the chemical denaturant guanidine hydrochloride (GdCl) to bring thermal denaturation into a measurable range. Melting curves were found to fit a two-state model [21,22]. An app T m in the absence of GdCl was determined by linear extrapolation of a series of melting point determinations carried out at different GdCl concentrations [23] ( Figure 3) and was found to be 108°C. tmRBP is significantly more stable than the mesophilic ecRBP ( app T m value is 56°C (Figure 3)). Addition of the 20 kDa truncation has no effect on the app T m value of the full-length wild-type monomeric protein (data not shown).

Ligand Binding
Ribose binding was observed as a ligand-mediated change in the app T m of full-length wild-type monomeric tmRBP in the presence of 5.5 M GdCl. Under these conditions the app T m is 71°C in the absence of sugar and 97°C in the presence of 1 mM ribose, indicating that tmRBP is a ribosebinding protein, as predicted from sequence homology ( Figure 3). For the ligand-bound form (1 mM ribose), an Peptide mapping of the tm0958 ORF gene products

Structure Determination
Crystals of ribose-complexed tmRBP were grown using a full-length wild-type construct (residues 30-323) that lacks the periplasmic signal sequence (residues 1-29). The apo-protein was crystallized using a construct that consisted of residues 30-310 (numbering according to NCBI NP 228766), containing a M142A mutation to prevent expression of the in-frame ORF. We were unable to obtain crystals of the heterodimeric form. The apo-protein and ribose-complex diffract to 1.4 Å and 2.15 Å resolution and were refined to R cryst /R free values of 18.0/20.3 and 19.3/ 22.3 respectively. The X-ray crystal structure of ribosebound tmRBP was solved by molecular replacement using ecRBP as the search model [24]. The apo-form of tmRBP was solved by separately searching with the amino-and carboxy-terminal domains of the ribose-bound form of tmRBP. Data collection, refinement, and stereochemistry statistics are summarized in Table 2.

Overall Structure and Comparison of the E. coli and T. maritima apo proteins
The apo forms of ecRBP [8] and tmRBP adopt the same overall fold. However, the relative inter-domain angles [25] differ significantly (43° for ecRBP; 28° and 20° for the two molecules in the tmRBP unit) ( Figure 4). The hinge in ecRBP is very flexible as evidenced by the number of crystal forms that differ in the inter-domain closure angle [8]. The two molecules found in the tmRBP asymmetric unit differ in the inter-domain closure angle by 10°, analogous to the conformational heterogeneity observed in ecRBP [8].
The construct used to crystallize the apo-form of tmRBP was a C-terminally truncated form of the protein (13 amino acids). It is possible that the absence of this region could in some way influence the observed conformation of apo form of tmRBP. However, superimposition of the tmRBP ribose complex C-terminal domain onto the C-terminal region of the apo protein suggests that these this region does not form interdomain interactions in the absence of ligand.

Overall Structure and Comparison of the E. coli and T. maritima ribose complexes
The structure of the tmRBP ribose-complex is similar to the ribose complexes observed in ecRBP [24] and a thermophilic RBP obtained from Thermoanaerobacter tengcongensis [26] (tteRBP). Both structures superimpose on tmRBP with a 1.2 Å RMSD calculated over C α atoms (Figure 4). The structures tteRBP and ecRBP are almost identical [26]; comparisons are described therefore only for ecRBP. The largest differences between ecRBP and tmRBP are at the C-termini, where tmRBP is extended by an addi-Thermal stability of tmRBP  [27,28]. We postulate that these C-terminal extensions form inter-domain interactions that may be important for modulating the intrinsic free energy difference between the apo and closed forms in the absence of ligand (Miklos, Cuneo and Hellinga; in preparation).
Although ribose is commonly found as a furanose carbohydrate in biological molecules (e.g. nucleic acids), all periplasmic RBPs, including tmRBP, bind the β-anomer of D-pyranose ribose [24,26] (as initially postulated by Koshland [29]). β-D-pyranose ribose the most prevalent form in solution under ambient conditions (59%) [30]. The ligand-binding site of tmRBP is composed of a network of polar amino acids which is identical in sequence and hydrogen-bonding pattern to the E. coli protein [24] ( Figure 5). Seven polar amino acids make a total of eleven hydrogen-bonds with the ribose. One residue in ecRBP (Q235) has been postulated to be important for both ligand-binding and hinge-bending; in the closed form it forms hydrogen-bonds with the ligand and amino acids from both domains [8,15]. The equivalent residue (Q244) and the amino acids which it interacts with are conserved in tmRBP. This pattern of conservation suggests that similar mechanisms couple ligand-binding to conformational changes in both proteins [8,14].
The ribose is wedged between three aromatic amino acids (W15, F16 and F172) which make extensive van der Waals interactions with the sugar ring. In ecRBP the equivalent aromatic binding pocket residues are all phenylalanines. Alignment of tmRBP and ecRBP structures indicates that the six-membered ring of W15 in tmRBP is equivalent to F15 in ecRBP ( Figure 5).

Open to Closed Transition: Global Changes
The addition of ribose to tmRBP induces a 28° hingebending motion [25] mediated about residues 102-105, 244-249, and 271-275. The hinge-bending motion of tmRBP is smaller than the 43° change observed in ecRBP [8]. In both ecRBP and tmRBP, the effects of these motions on the backbone are confined largely to the hinge region ( Figure 6).
The two molecules in the tmRBP asymmetric unit have slightly different degrees of closure, indicative of an intrinsic flexibility of the hinge, as observed in ecRBP [8] and E. coli allose-binding protein [7]. Molecule B is related to molecule A by a 10° closing about the hinge. This movement is limited to one of the two strands (residues 101-106) which connect the two domains ( Figure 6). The magnitude of C α torsion changes transitioning between the open and closed states is significantly greater for molecule B than molecule A ( Figure 6); the average B-factors of the two molecules are the same.

Open to Closed Transition: Local Changes
In ecRBP and tmRBP local ligand-induced changes are restricted largely to the hinge region, the N-terminal amino acids that interact with ribose, and the hinge amino acid (Q235 and Q244 in ecRBP and tmRBP respectively) that interacts with the ribose (Figure 7 and Table  3). The amino acid side-chains in the C-terminal domain of tmRBP remain fixed in the same rotameric state in both apo and ligand-bound forms (Figure 7 and Table 3). By contrast, the side-chain torsional changes in the N-termi-nal domain of tmRBP are significant; in particular, W15 and F16 undergo torsional movements about χ 1 and χ 2 ( Figure 7 and Table 3). This ligand-induced binding pocket rearrangement of the N-terminal domain is also observed in ecRBP, but of smaller magnitude than in tmRBP, and is restricted to three polar amino acids (N13, D89, and R90) ( Table 3).

Solvent Interactions in tmRBP and ecRBP
Water molecules play an important role in the hinges of PBPs [7]. Analysis of the conservation pattern of bound water molecules among various PBPs identifies critical water molecules that participate in inter-strand hydrogen bonding in the hinge, in place of amino acid side-chains [7,8]. The positions of four bound water molecules are conserved (separated by less than 1.5Å in the aligned structures) in the open forms of ecRBP and tmRBP. One of these water molecules (HOH5 in tmRBP, W1 in ecRBP) is conserved in both the open and closed forms of group I PBPs [7]. This water molecule remains fixed in position in both the apo and ligand-bound forms. It is postulated to act as a "ball bearing" by serving as a fixed intra-hinge rotation point for the two domains [7]. It also mediates indirect interstrand hydrogen bonding. Another water molecule conserved among other group I PBPs, W2 [7], is absent from tmRBP. When present, this water mediates cule, replacing the water-mediated hydrogen bonds with inter-strand hydrogen bonds [8]. W2 is absent from the ribose-bound tmRBP structure, as it is in E. coli arabinosebinding protein [7]. In both instances, the hinge conformation allows for inter-strand hydrogen-bonding to satisfy the water-mediated hydrogen-bonds that would be formed [7,8,24].

Conclusion
We have characterized the ligand-binding properties of a putative ribose-binding protein identified in the genomic sequence of the extremophilic bacterium T. maritima and solved its X-ray crystal structure in the absence and presence of ribose. The structure reveals that tmRBP has high structural similarity to its mesophilic homolog ecRBP. Polar ligand interactions and ligand-induced global conformational changes are conserved [8,24]. Local structural whereas the C-terminal domain remains fixed. Based on hydrogen-bonding pattern (6 and 5 hydrogen-bonds with the N-and C-terminal domains respectively) and buried surface area (55Å 2 and 35Å 2 with the N-and C-terminal domains respectively) it has been postulated that ordered binding occurs and ribose initially interacts with N-termi-  Binding pocket organization of the apo and ribose-bound tmRBP nal domain of ecRBP [8]. If an order of interaction can be established from analysis of structure, it is likely to proceed with ribose initially interacting with the C-terminal domain of the apo tmRBP, as the entropic costs of fixing the side-chains for ligand binding should be reduced for a pre-ordered binding site.
Water molecules have been suggested to play an important mechanistic role in the evolution and adaptation of the PBP hinge [7]. In particular, two water molecules, (W1 and W2), are closely associated with the hinges of group I PBPs [7]. In tmRBP, W1, which is postulated to act as a "ball bearing" in the ligand-mediated conformational change, is conserved in both the apo-and ribose-bound forms. On the other hand, W2, which is involved in mediating important inter-hinge contacts in apo-and ligandbound group I PBPs, is absent in both forms of tmRBP. In tmRBP the inter-strand hydrogen bonds form directly in the hinge. These differences in water interactions in the hinges of PBPs suggest local structural differences can supplant the need for W2, whereas the role of W1 cannot be accommodated through differences in main-chain geometry or side-chain identity.
Ligand-induced hinge bending motion is a key characteristic of the periplasmic binding protein superfamily. Analysis of PBP structures has provided a detailed description of this class of conformational change [7,[12][13][14]. The detailed comparative analysis of the open to closed transition of the thermophilic tmRBP and mesophilic ecRBP presented here illustrates the subtle differences in the mechanism and magnitude of the ligand-induced conformational changes, and the interplay between global and local conformational changes in this protein superfamily.

Cloning Over-expression and Purification
The tm0958 gene was amplified from T. maritima genomic DNA (American Type Culture Collection) by the stickyend PCR method [31] using the following primers to make the full-length tmRBP (residues 30-323) and the construct used to crystallize the apo form of tmRBP (residues 30-310) (numbering according to NCBI Protein Database NP 228766: PO 4--TATGAAAGGAA AGAT-GGCTATTGTGATCTCC and for the 5'-TGAAAGGAA AGATGGCTAT TGTGATCTCC end of the genes; PO 4--AATTCTA ATGGTGATGGTGATGGTGACTGCCTTCT-TCTTTTCTGCCGTAAGCAGTG and CTAATGGTGATGGTGATGGTGACTGCCTTCTTCTTT-TCTGCCGTAAGCAGTG for the 3'end of the full-length tmRBP gene, PO 4 -AATTCTAATGGTGATGGTGATGGT-GACTGCCTTCTCTTGTCACCAGCTCAACAGTGAC and CTAATGGTGATGGTG ATGGTGACTGCCTTCTCTTGT-CACCAGCTCAACAGTGA C for the 3' end of the tmRBPapo gene [31]. The 30-323 construct which was used to crystallize the apo-form additionally contains an M142A mutation to prevent translation of the truncated form of tmRBP. The resulting fragments were cloned into the NdeI/EcoRI sites of a pET21a (Novagen) plasmid for overexpression in E. coli. This ORF lacks the periplasmic signal sequence. The coding sequence starting at lysine 30 was cloned in-frame with an ATG start codon. A hexa-histidine affinity tag, preceded by a glycine-serine linker, was fused in-frame at the carboxy terminus to facilitate purification by immobilized metal affinity chromatography (IMAC). Protein concentration was determined spectrophotometrically (ε 280 = 41,000 M -1 cm -1 ) [32]. The resulting gene product was expressed and purified by IMAC and gel filtration as described [23]. Pooled IMAC fractions were concentrated to 12 ml and were loaded onto a Superdex 26/60 S75 (Amersham) gel filtration column that was previously that was previously calibrated with blue dextran, bovine serum albumin, chicken serum albumin, chymotrypsin and lysozyme.

Tryptic Digest and Mass Spectrometry
Proteins were excised from a 12% Tris-HCl SDS-PAGE gel and were digested in-gel using the Pierce In-gel Tryptic Digest Kit. Mass spectra were acquired on an Applied Biosystems Voyager DE MALDI-TOF mass spectrometer using an α-cyano-4-hydroxycinnamic acid matrix with a 300 ns delay time.

Circular Dichroism
Circular dichroism (CD) measurements were carried out on an Aviv Model 202 CD spectrophotometer. Thermal denaturations were determined by measuring the CD signal at 222 nm (1 cm path length) as a function of temperature, using 1.0 μM of full-length wild-type monomeric tmRBP (10 mM Tris-HCl pH 7.8, 150 mM NaCl) in the presence or absence of 1 mM ribose at several GdCl concentrations extrapolated to 0 M GdCl [23]. Protein samples were incubated for 15 minutes prior to collecting data. Each measurement includes a 3-second averaging time for data collection and a 60 second equilibration period at each temperature. Data were fit to a two-state model [22].

Crystallization and Data Collection
Crystals of full-length wild type ribose-complexed tmRBP were grown using 3:1 stoichiometric ribose:protein ratio by micro-batch under paraffin oil in drops that contained 2 μl of the protein solution (15 mg/ml in 10 mM Tris pH 7.8, 20 mM NaCl, 1.5 mM ribose) mixed with 2 μl of 0.1 M MES pH 6.0, 20% (w/v) PEG 8000 and 0.1 M RbCl. Crystals of the C-terminally truncated M142A apoprotein were grown in micro-batch drops containing 2 μl of the protein solution (15 mg/ml in 10 mM Tris pH 7.8, 20 mM NaCl) mixed with 2 μl of 0.1 M Bis-Tris pH 5.9, 25% (w/ v) PEG 3350, 0.2 M NaCl. Diffraction quality crystals typ-ically grew within two weeks at 17.0°C. The ribose-complexed crystals diffract to 2.15 Å resolution and belong to the I222 space group (a = 72.1 Å, b = 98.2 Å, c = 131.1 Å) ( Table 2). The apo tmRBP crystals diffract to 1.4 Å resolution and belong to the F222 space group (a = 120.9 Å, b = 136.8 Å, c = 144.5 Å) ( Table 2). Crystals were transferred stepwise to a cryoprotectant solution consisting of the original precipitant solution with an additional 15% ethylene glycol or glycerol, after which they were mounted in a nylon loop and flash cooled in liquid nitrogen. All data were collected at 100 K on the SER-CAT 22ID beam line at the Advanced Photon Source. Diffraction data were scaled and integrated using HKL2000 [33].

Structure Determination Methods, Model Building and Refinement
The structure of ribose-complexed tmRBP was determined by molecular replacement utilizing the AMore program, where the ligand-bound form of the E. coli ribose-binding protein was used as the search model [34]. The N-and Cterminal domains of ribose-complexed tmRBP were used as a search model in Phaser to solve the apoprotein structure [35]. In both cases, rotation, translation, and fitting functions revealed a single clear solution yielding higher correlation coefficients and a lower R factor than all the others. Manual model building was carried out in the programs O and COOT and refined using REFMAC5 [36][37][38].

Structural Analysis
The final model for ribose-complexed tmRBP includes one intact monomer (residues 30-323), one ribose molecule, and 142 water molecules. The final model for the apoprotein includes two intact monomers (residues 30-310) and 627 water molecules. The models exhibit good stereochemistry as determined by PROCHECK [39] and MolProbity [40]; final refinement statistics are listed in Table 2. PDB coordinates and structure factors of ribosecomplexed tmRBP and apoprotein have been deposited in the RCSB Protein Data Bank under the accession codes 2FN8 and 2FN9 respectively.
Large-scale hinge bending motions were analyzed with the DynDom web server [25]. Local C-alpha torsional changes were analyzed with LSQMAN [41].