Crystal structure of the apo forms of psi 55 tRNA pseudouridine synthase from Mycobacterium tuberculosis: a hinge at the base of the catalytic cleft.

The three-dimensional structure of the RNA-modifying enzyme, psi55 tRNA pseudouridine synthase from Mycobacterium tuberculosis, is reported. The 1.9-A resolution crystal structure reveals the enzyme, free of substrate, in two distinct conformations. The structure depicts an interesting mode of protein flexibility involving a hinged bending in the central beta-sheet of the catalytic module. Key parts of the active site cleft are also found to be disordered in the substrate-free form of the enzyme. The hinge bending appears to act as a clamp to position the substrate. Our structural data furthers the previously proposed mechanism of tRNA recognition. The present crystal structure emphasizes the significant role that protein dynamics must play in tRNA recognition, base flipping, and modification.

Post-transcriptional modifications of various forms of RNA are essential cellular processes (1). Naturally occurring RNA molecules (tRNA, rRNA, and small nuclear RNA) contain a variety of nonstandard ribonucleosides such as ribothymidine and pseudouridine. Pseudouridine (⌿), the C-glycoside isomer of uridine, is the most abundant of the modified bases found in RNA (2).
Several studies have shed light on the biological implications of pseudouridine bases. Conserved ⌿ residues are believed to play important roles in RNA stability, codon recognition, spliceosomal assembly, and other functions (2,(5)(6)(7). The importance of ⌿ residues is underscored by their tendency to occur in RNA molecules near functionally important regions such as near the peptidyl transfer site in rRNA (8). It has been demonstrated that the deletion of PUS genes results in impaired growth (RluD) and competitive disadvantage (TruB) in Escherichia coli (9 -11). The human homolog of TruB (dyskerin) is implicated in an X-linked hematopoietic disorder called dyskeratosis congenita (12). Suggestions have been put forward that, apart from their usual catalytic role, certain PUS enzymes (e.g. TruB) may also act as chaperones for RNA folding (11,13). Crystal structures have been determined for TruA (14), TruB (15,16), RsuA (17), and RluD (18,19) from E. coli and TruB from Thermatoga maritima (16). A comparison of those structures indicates a common ancestral linkage and a shared enzymatic mechanism involving a catalytic aspartate residue (17,20). The structures of TruB from E. coli (EC-TRUB) and T. maritima (TM-TRUB) in complex with an RNA segment mimicking the T stem-loop of tRNA demonstrates that TruB accesses the target U55 of its tRNA substrate by a mechanism involving a base-flipping conformational change (15,16). A combination of rigid docking followed by induced fit binding was proposed for tRNA recognition based on the significant conformational change observed between the structures of RNA-bound and -free forms of EC-TRUB (16).
Here we present the structure of ⌿55 tRNA pseudouridine synthase from Mycobacterium tuberculosis (MTB-TRUB) at 1.9-Å resolution in two substrate-free forms. The structures depict an interesting mode of ␤-sheet bending movement affecting the shape of the catalytic cleft as well as a major disordering of the active site in the absence of the RNA substrate. Implications of the above observations are discussed in relation to protein function. The structure of MTB-TRUB was determined as part of the M. tuberculosis structural genomics initiative (21).

EXPERIMENTAL PROCEDURES
Cloning-The Rv2793c gene was amplified by PCR using M. tuberculosis H37Rv genomic DNA as the template, with a forward primer (5Ј-CCATATGGCTAGCGCAACCGGCCCCGGAATCGTGGTTATCGA-C-3Ј) that introduced an NdeI site (underlined) and a reverse primer (5Ј-CGGCGACGATGCACCCCGGGGGTGTACCACGAGGTAAGCTT-3Ј) that introduced a HindIII site (underlined). The forward primer also inserted an alanine codon (GCT) immediately following the start codon to enhance protein expression (22), while the reverse primer introduced a thrombin recognition sequence to the C terminus. The PCR product was cloned into pCR-Blunt II-TOPO (Invitrogen). Following sequence confirmation, the gene was subcloned into pET22b (Stratagene), which added a hexahistidine tag to the expressed protein, trailing the thrombin recognition sequence. The resulting C-terminal extension had the amino acid sequence of GVPRGKLAAALEHHHHHH immediately following the natural protein sequence.
Each gram of cell pellet was resuspended and lysed in 5 ml of lysis buffer (20 mM Tris, pH 8, 0.3 M NaCl, 10 mM imidazole, 2 mM ␤-mercaptoethanol, 2 g/ml DNase I, 0.2 mg/ml lysozyme, and 1:100 protease inhibitor mixture (Sigma)). The lysate was clarified by centrifugation at 27,000 ϫ g for 30 min. The soluble, recombinant protein was initially purified using nickel-nitrilotriacetic acid Superflow resin (Qiagen). Because of the sensitivity of the protein to high concentrations of imidazole, the peak elution fractions were pooled and dialyzed against 20 mM Tris, pH 8, and 0.3 M NaCl soon after purification. The dialyzed sample was concentrated using Centricon YM-10 concentrators (Millipore) and further purified on a Superdex 75 column (Amersham Biosciences) equilibrated with 20 mM Tris, pH 8, and 0.3 M NaCl. The peak fractions were pooled and concentrated to 20 mg/ml. The purified protein was Ն99% pure as estimated by SDS-PAGE. Liquid chromatography-mass spectrometry confirmed the molecular weight of the recombinant protein and the purity of the sample.
Crystallization-The protein was initially crystallized as rod clusters after 4 weeks at room temperature by sitting drop vapor diffusion in which 1 l of the protein concentrate was mixed with 1 l of crystallization reagent (0.2 M (NH 4 ) 2 SO 4 , 0.1 M Tris, pH 8.5, and 25% (w/v) PEG 3350 (Hampton index screen formulation 69)) at room temperature. Diffraction quality crystals were grown with microseeding. In the optimization trial, pre-equilibrated drops of 10 l of protein mixed with 10 l of reservoir reagent gave larger and better diffracting crystals. The final crystallization condition was 110 mM (NH 4 ) 2 SO 4 , 0.1 M Tris, pH 7.5, and 17% (w/v) PEG 3350. Crystals grew to ϳ700 ϫ 400 ϫ 80 m.
Data Collection-Prior to shock-freezing the crystals in the nitrogen stream for data collection, each crystal was soaked for 5 min in a series of reagents for cryoprotection. The reagents consisted of the same components as the crystallization reservoir solution with the addition of PEG 3350 at a concentration of 25%, followed by 30 and 35%. It was noted that PEG concentrations Ͼ35% caused crystals to fracture. All data were collected at the Advanced Light Source beamline 5.0.1 equipped with an ADSC Q4 CCD detector ( ϭ 1.0332 Å). DENZO and SCALEPACK (23) programs were used for data processing and scaling. Data statistics are provided in Table I, Part A.
Phasing-The molecular replacement method was used to solve the structure of MTB-TRUB using the AMoRE program (24). The protein component of the EC-TRUB structure (Protein Data Bank code 1K8W, Ref. 15) was used as a search model. The search model was converted to a polyserine model, and an atomic displacement parameter (B factor ) of 20 Å 2 was applied to all atoms. An unambiguous molecular replacement solution was obtained for which the initial correlation coefficient was 0.43 (R factor 0.48) for two molecules in the asymmetric unit for 12-4 Å data.
Refinement-Rigid body refinement was performed in REFMAC5 (25) prior to the atomic refinement step using data to 3 Å. An initial attempt to build an atomic model using the ARP (26) automated rebuilding protocol with data to 1.9 Å failed. A conformational difference between the two chains was obvious from an inspection of the electron density map. Both chains were manually rebuilt in the core region, and partial models were refined in REFMAC5. The ARP automated model building protocol, interspersed with refinement using REFMAC5, was subsequently used to build the rest of the model. Manual rebuilding and checking were done using the graphics package O (27). A small portion of the data set was used to calculate R free for cross-validation (28). Moderate restraints were used initially for non-crystallographic symmetry-related molecules at all refinement stages. Individual isotropic B factor values were refined. ARP was used to add and scrutinize waters. Translation-libration-screw (TLS) refinement (29) was performed at the final stages of the refinement (with a fixed B factor of 20 Å 2 ) without any non-crystallographic symmetry restraints applied.
In an attempt to understand the molecular motion in terms of rigid body vibrations, we performed three parallel TLS refinements in REF-MAC5 under the following conditions: (i) case I, an entire protein molecule as a rigid body (two TLS groups); (ii) case II, the catalytic and the PUA (pseudouridine and archeosine) domain in each molecule as rigid bodies (four TLS groups); and (iii) case III, each molecule constituting three rigid bodies (segments I and II and the PUA domain; six TLS groups). The TLSANL program (30) was used to analyze the resultant TLS matrices. However, the difference in R factor and R free was only marginally better in case III than in case II or case I. Evidently the hinge-based flexibility of the protein is not captured by the minor rigid body vibrations manifested by the proteins in their crystalline environment.
Model Quality-The quality of the final model was quite satisfactory (Table I,  Surface Area Computation-The solvent-excluded surface area of the molecule was calculated using the MSMS program (35) with a probe radius of 1.4 Å. All hydrogen atoms were generated using the CNS program (36) prior to the surface area computation. The DINO graphics package (www.dino3d.org) was used to visualize and inspect the surface.
Software Used for Structure Analysis and Illustration-The BRUTE option of LSQMAN (34), the FIT program (bioinfo1.mbfys.lu.se/ ϳguoguang/fit.html), the LSQ option of O, and several programs from the CCP4 suite (37) were used for structural superposition and analysis. DSSP (38) was used for secondary structure assignment. The DINO and MOLSCRIPT (39) programs were used for making the illustrations.

RESULTS
Overall Structure of MTB-TRUB-The crystal structure of MTB-TRUB was determined with two molecules in the asymmetric unit. Here, we refer to the two molecules as MTB-TRUB-A and MTB-TRUB-B (A and B correspond to the chain names of the two polypeptides in the Protein Data Bank coordinate set). The final structure contains residues 3-292 and 1-294 from 298 residues in the A and B chains, respectively. Residues 115-142 in MTB-TRUB-A and residues 112-139 in MTB-TRUB-B are missing in the electron density map (henceforth referred to as the disordered region) and are not modeled. The structure is composed of two domains (Fig. 1), namely the large N-terminal catalytic domain (residues 1-226) and the short C-terminal RNA-binding domain (residues 227-298). Both domains are of the (␣ ϩ ␤) type. The catalytic domain contains 12 ␤-strands (11 ␤-strands in the B chain), four ␣-helices, and two 3 10 helices (Fig. 1). A mostly anti-parallel, mixed, multiply bifurcated ␤-sheet (henceforth, the central ␤-sheet) forms the core of the catalytic domain. The C-terminal domain contains a four-stranded mixed ␤-sheet flanked by one ␣-helix on each side.
Hinge Bending-A key observation from the crystal structure is the presence of a significant conformational difference, occurring in the central ␤-sheet, between the two copies of the enzyme visualized in the crystal asymmetric unit (Fig. 1B). As a result, a portion of the catalytic domain (residues 72-199; segment I) moves essentially as a rigid body with respect to the remainder of the catalytic domain (residues 1-71 and 200 -226; segment II) between the two chains (Table II and Fig. 2). The observed rigid body motion can be described by an ϳ13°rota- tion around an effective hinge axis (Fig. 1B) with a translational (screw) component less than 0.5 Å. The movement is created by a few small torsional rotations localized to the hinge region centered around residues 71-72 and 199 -200 (maximal ⌬/⌬ of ϳ8°). As a result, a segment of the central ␤-sheet formed by strands ␤10-␤4-␤9-␤8-␤7 moves to a new local conformational energy minimum, whereas the hydrogen-bonding pattern with the rest of the protein remains intact. The effective hinge axis runs close to the C␣ atom of Cys-173 (␤9), the carbonyl group of Leu-199 (␤10), and the C␥ atom of Tyr-70 (␤4) at the base of the catalytic cleft (Fig. 1B). Thus, the base of the catalytic cleft can function as a mechanical hinge. Interestingly, all three of these residues take part in active site formation.
The rigid body approximation to the conformational difference between the two molecules breaks down for two short stretches of segment I, i.e. in the ␤9-␣3 loop (Fig. 3, A and B) and in all the visible residues between residue 111 and residue 149. Residues 112-114 (␤7) in the MTB-TRUB-A chain form an anti-parallel ␤-sheet interaction with the neighboring ␤8 strand. These residues are missing in the MTB-TRUB-B chain.
The overall structure of MTB-TRUB-A is generally similar to the previously published structures of TruB from E. coli (EC-TRUB, Protein Data Bank code 1K8W; 2.2 Å r.m.s.d. between 255 pairs of C␣ atoms (15)) and T. maritima (TM-TRUB; Protein Data Bank code 1R3E; 1.5 Å r.m.s.d. between 222 pairs of C␣ atoms; Ref. 16). On the other hand, structural superposition using the segment II residues reveals noticeable displacements of the segment I residues between the different TruB structures with maximum displacements of corresponding C␣ atoms ranging from 2 to 5 Å (Fig. 4, A and B).
The Active Site-The observed conformational variability leads to certain differences in the active sites of the two MTB-TRUB molecules. The active site cleft is occupied by a few water molecules in both the A and B protein molecules (Fig. 3,  A and B). The wall of the cleft is lined mainly by hydrophobic residues (Tyr-70, Leu-199, Tyr-178, Arg-180, Asp-42, Lys-68, Cys-173, Ile-179, Thr-177, Arg-146, and Thr-40; the underscored residues are conserved in the TruB family). Those residues from the shifted segment I that contribute to the catalytic cleft are repositioned in the B chain with respect to the A chain. The Tyr-70 side chain orientation and the Asp-42/Arg-180 salt link, as observed in the structures of other TruB homologs, is maintained in both molecules of MTB-TRUB. It has been suggested that this salt link serves to keep the nucleophilic aspartate in the deprotonated charged form in an otherwise low dielectric environment (15). Based on the different orientations of Asp-42 and Arg-180 in the RNA-bound and substrate-free forms of EC-TRUB, a substrate-induced "conformational switch" has been proposed to align the Asp-42 side chain for catalysis (16). However, our structure suggests that such a conformational change may not necessarily be substrate-induced, because a comparable side chain reorientation of Asp-42 and Arg-180 is observed between the A and B molecules of substrate-free MTB-TRUB (Fig. 3). The observed structures of MTB-TRUB suggest that the two Arg-180 and Asp-42 side chain orientations represent alternate conformations with similar energies that are capable of interconverting by normal protein "breathing." Significant differences between the two conformations of MTB-TRUB are also observed for the ␤9-␣3 loop housing the conserved CXXGXY motif (Cys-173/Gly-176/Tyr-178, henceforth called the CGY loop). The alternate conformation of the Lys-68 and Tyr-178 side chains, the CGY loop movement, and the above mentioned bending motion combine to make the uridine substrate (U55) binding pocket in MTB-TRUB-B less accessible (Fig. 3). The pocket is more open to the bulk solvent in the MTB-TRUB-A conformation. Movement of the Tyr-179 residue (which corresponds to Tyr-178 in MTB-TRUB) was also observed in the substrate-free EC-TRUB, resulting in a more accessible U55 pocket (16).
The Disordered Thumb Region-One of the key observations from the present structure is that a significant portion of that part of the enzyme that grips the tRNA molecule in the major groove (the "thumb") is fully disordered in both molecules of MTB-TRUB (Fig. 4, A and B). Residues 121-152 in EC-TRUB form a protrusion or so-called thumb that constitutes one side of the binding cleft and accommodates the flipped bases of the substrate (15). The highly conserved thumb region makes numerous interactions with the flipped bases and the cognate sugar/phosphate backbone in the major groove of bound RNA. This protruding part is mostly disordered in both molecules of the MTB-TRUB as well as the apo form of EC-TRUB (16). The disordered thumb in the MTB-TRUB structure reinforces the previous suggestion (16) that this region is likely unstructured in solution and apparently undergoes a disorder-to-order transition upon binding RNA.  (72-111, 149-174, 181-199 The C-terminal Domain-A recurring theme in RNA modification enzymes is the acquisition of small RNA binding domains to form an extended RNA binding surface (40). The C-terminal domain of the MTB-TRUB is a PUA (40) domain, which is commonly found fused to RNA modification enzymes (PUS, RNA methylases, and archaeosine tRNA-guanine transglycosylase (ATG); Ref. 41), but occasionally as a single domain protein. This short domain contains several exposed basic residues. In the search for structures similar to the C-terminal domain of MTB-TRUB using DALI (42), apart from EC-TRUB (Z_score ϭ 7.8, 1.9 Å r.m.s.d. for 64 aligned C␣ atoms) the most similar structure identified was the RNA-modifying enzyme archaeosine tRNA-guanine transglycosylase (Protein Data Bank code 1J2B, Z_score ϭ 6, 2.9 Å r.m.s.d. for 63 aligned C␣ atoms).
The apo form of EC-TRUB also shows a hinge-bending motion with respect to the RNA-bound EC-TRUB involving a significant movement of the C-terminal domain (Fig. 4) (16). The overall orientation of the C-terminal domain with respect to the rest of the protein is marginally different for the two MTB-TRUB molecules (ϳ1 Å r.m.s.d.; Table II; Fig. 1B). This domain is envisaged as another flexible element of the TruB structure that assists in tRNA recognition (16).
The RNA Binding Surface-A plot of the acidic and basic residues in the solvent-excluded substrate binding surface of MTB-TRUB shows the parts of the surface that would form the tRNA binding surface (Fig. 1D). The RNA binding surface is quite similar in all of the reported structures of TruB (EC-TRUB, TM-TRUB, and MTB-TRUB). TruB enzymes do not necessarily require an entire tRNA molecule for recognition and catalysis (43,44). However, the extensive tRNA binding surface in TruB apparently serves in vivo to stabilize parts of the tRNA molecule beyond the recognition element. DISCUSSION The crystal structure of MTB-TRUB at atomic resolution provides two views of the substrate-free enzyme. The two quite different conformations observed in the present structures apparently have similar energies with the distinct conformations being preferred in the two different packing environments in the crystal asymmetric unit. Both molecules make about the same number of intermolecular contacts (contact distance of Յ4 Å, computed using CONTACT; Ref. 37) with the surrounding protein chains in the crystal, whereas the B form makes a few additional intra-molecular interactions at the catalytic cleft. The two alternate conformational substates that are stabilized by weak crystallographic packing interactions almost certainly reflect a range of conformations to which the enzyme has access in vivo. Interestingly, the different conformers reveal the presence of a hinge region at the bottom of the active site cleft. In support of this, a preliminary structure of MTB-TRUB has also been determined at 3-Å resolution from a monoclinic crystal form with four molecules in the asymmetric unit (data not shown). Although those structures are not described in detail here because of the relatively poor resolution, the same kind of protein flexibility is exhibited in that crystal form as well. A few small and concerted main chain torsional rotations at the base of this binding pocket create significant structural changes at the distal end of the protein; the observed shifts are as high as 6 Å between corresponding C␣ atoms. The hinge bending motion apparently serves to modulate the shape and volume of the U55 binding pocket.
How might the aforementioned hinge motion and the thumb reordering assist the enzyme action in vivo? The answer may lie in the energetic obstacles that TruB and other enzymes must encounter as they attempt to flip the bases out of their polynucleotide substrates prior to catalysis. In vitro kinetic studies suggest that the spontaneous base-flipping rates in duplex DNA and RNA (45) are ϳ4 or 5 orders of magnitude slower than the base-flipping rates observed in the presence of various enzymes (46,47). Not only do these enzymes provide a complementary active site for the flipped out substrate conformation, but they also apparently reduce the kinetic barriers faced by the substrate as it undergoes major conformational changes. In the reaction carried out by TruB, it is the T-loop of the tRNA molecule that flips out (15). In the canonical Lshaped tRNA, the T-loop makes tertiary interaction with the D-loop. These interactions are lost in the base-flipped form of the tRNA. Interestingly, it has been suggested that TruB can act on the alternative -form of tRNA (44), which is a conformation in which the canonical T-D interactions are already lost.
Although we do not yet have a detailed understanding of how TruB and related enzymes help to provide a kinetically accessible pathway for major substrate rearrangements, it seems logical that protein movements would play a critical role. The ordering of the swinging thumb region during the tRNA docking could facilitate base flipping (16). The subsequent placing of the flipped bases in the catalytic cleft and the tight gripping of the uracil ring (like a "molecular clamp") for an in-line nucleophilic attack by Asp-42 would be assisted by the observed mode of the hinge-bending movement. The chemistry requires the repositioning of the uracil ring inside the catalytic pocket to bring the C5 atom of the uracil base close to the C1Ј atom of the ribose group for covalent bond formation. The hinge bending could facilitate the above process during the chemical reaction. Thus, the crystal structures of TruB strengthen the view that protein flexibility and dynamics play key roles in tRNA recognition and modifying reactions involving base-flipping mechanisms.