Crystal structure of a fibrillarin homologue from Methanococcus jannaschii, a hyperthermophile, at 1.6 Å resolution

Overall structure and topology

Mj0697 has a globular two-domain structure with overall dimensions of ∼50 × 35 × 35 Å (Figure 1A). Composed of a total of seven α-helices and 12 β-strands, it is divided into a smaller N-terminal domain (residues 1–54) and a larger C-terminal domain (residues 55–230). The N-terminal domain represents a novel fold, forming a crescent-shaped arrangement consisting of five β-strands. The C-terminal domain, which is connected to the N-terminal domain by a short helix (residues 57–64), has a mixed α/β structure with alternating α-helices and β-strands. The core of this C-terminal domain is a seven-stranded β-sheet (β₁–β₇), flanked by three α-helices on one side (α₀, α₁ and α₂) and four α-helices on the other (α₃, α₄, α₅ and α₆) (Figure 1A). The strands of the β-sheet are arranged in the order 3–2–1–4–5–7–6, and all strands, except for the seventh strand, are in parallel orientation (Figure 1C).

In the crystal, two N-terminal domains join to form a homodimer related by a crystallographic 2-fold axis (Figure 1B and C). The dimer interface is a β-barrel of 10 β-strands, with a hollow centre ∼5 Å in diameter, and composed of β-strands from each monomer (Figure 1B). The buried surface area at the dimer interface is ∼1700 Ų. Native gel electrophoresis indicated that Mj0697 dimerizes in solution (data not shown), suggesting that the interdomain sheet observed in the crystal may be a functional molecular interface.

Despite the absence of any significant sequence similarities, the overall fold of the C-terminal domain is similar to the catalytic domain common to many S-adenosyl-l-methionine (AdoMet)-dependent methyltransferases, such as several DNA methyltransferases, including: HhaI (Cheng et al., 1993; Klimasauskas et al., 1994), TaqI (Labahn et al., 1994), HaeIII (Reinisch et al., 1995) and PvuII (Gong et al., 1997); a few small molecule methyltransferases, such as catechol O-methyltransferase (COMT) (Vidgren et al., 1994) and GNMT (Fu et al., 1996); two RNA methyltransferases, VP39 (Hodel et al., 1996) and ErmC′ (Bussiere et al., 1998); and one protein methyltransferase, CheR (Djordjevic and Stock, 1997). The consensus topology for the methyltransferase catalytic domain is a seven-stranded β-sheet flanked by three α–helices on each side of the sheet (Figure 2A). The C–terminal domain of Mj0697 differs from the methyltransferase consensus topology only by the addition of a minihelix (α₅). Similarly to methyltransferases, a Rossmann fold is formed in Mj0697 by the first three β–strands (β₁, β₂ and β₃) and the two helices that lie above the surface of the β-sheet closest to the N-terminal domain (designated α₁ and α₂) (Eklund et al., 1981).

Comparison of the C-terminal domain with methyltransferases

Alignment of the C_α positions of the C-terminal domain of Mj0697 with the catalytic domain of the methyltransferases shows that the backbone configurations are almost identical. The root-mean-squared deviation (r.m.s.d.) of the C_α positions of all the corresponding residues in the conserved seven-stranded β-sheet region (49 residues) between Mj0697 and these methyltransferases varies from 1.3 to 2.1 Å. Among these, Mj0697 is most similar to COMT (PDB accession number 1VID) with an r.m.s.d. of 1.3 Å. Three-dimensional superposition of Mj0697 and 1VID is shown as a stereo view of the C_α trace in Figure 2B. A structure-based sequence alignment of Mj0697 and 1VID is shown in Figure 2C.

Sequence motifs characteristic of AdoMet-dependent methyltransferases can also be identified in this domain of Mj0697. These motifs are present in all fibrillarin homologues (Figure 3). In Mj0697, a majority of them are located around a small pocket at the C-terminal ends of the parallel β-strands (β₁–β₄) as seen in other methyltransferases. These include a loop between α₁ and β₁ corresponding to GASAG (82–86), an acidic residue at the end of β₂ (E105) and a small residue juxtaposed with this acidic residue at the end of β₁ (G82). In all methyltransferases, these motifs are important for AdoMet interactions.

Expression and purification

Mj0697 was overexpressed at room temperature in Escherichia coli strain BL21(DE3) containing the expression vector pET21a (Wang et al., 1999) with the Mj0697 gene and an additional plasmid encoding rare E.coli tRNAs. Soluble Mj0697 protein was purified by heating E.coli cell lysate at 80°C for 30 min, followed by a single-column elution (Pharmacia HiTrap Q). Selenomethionine (Se-Met)-substituted protein was produced by overexpression in E.coli B834, a methionine auxotroph, and cultured in minimal media supplemented with 45 μg/ml selenomethionine.

Crystallization and data collection

The crystals of Mj0697 were grown by the vapour diffusion method from a solution containing 25 mg/ml protein, 10% PEG 4000, 10% isopropanol and 0.05 M sodium citrate pH 5.6, equilibrated against a solution containing 20% PEG 4000, 20% isopropanol and 0.1 M sodium citrate pH 5.6. The crystals are in the C₂ space group with unit cell dimensions of a = 121.4 Å, b = 43.2 Å, c = 55.3 Å and β = 96.9°. Purification and crystallization of the Se-Met-substituted protein were performed under essentially the same conditions as for the native protein (Wang et al., 1999). Diffraction data for the heavy-atom derivative crystals were collected on a Rigaku R-axis IIC imaging plate system, using CuK_α radiation from a Rigaku RU200 HB rotating anode operated at 40 kV and 100 mA. The data for the multiwavelength anomalous diffraction of the Se-Met crystals were collected at the National Synchrotron Light Source (NSLS), Brookhaven beamline X4A corresponding to wavelengths 0.96373, 0.97894 and 0.97921 Å.

Structure determination

X-ray diffraction data sets were processed using program packages DENZO and SCALEPACK (Otwinowski, 1993). Crystallographic statistics for X-ray diffraction data, phase calculation and model refinement are summarized in Table II. The crystals were first phased using the multiple isomorphous replacement (MIR) method. Heavy-atom derivatives were initially analysed using XTALVIEW (McRee, 1992). The CCP4 package (Collaborative Computational Project, 1994) was used for heavy-atom refinement and phase calculation. Two derivatives were used, a mercury heavy-atom derivative (mersalyl acid) and a Se-Met-substituted crystal. Each asymmetric unit was found to contain a single mercury site and five Se-Met sites employing both Patterson and cross-Fourier analysis. After solvent flattening and density modification, the resulting phases were used to calculate an electron density map, which revealed a seven-stranded β-sheet of the larger C-terminal domain. However, the quality of the map was not good enough to trace the backbone completely.

Subsequently, a second Se-Met data set was collected at the three wavelengths close to the selenium absorption edge at NSLS beamline X4A. This multiwavelength anomalous diffraction data set was treated as a special case of MIR (Ramakrishnan and Biou, 1997). Using HEAVY (Terwilliger et al., 1987), positions were verified for the five selenium atom sites, resulting in close agreement with those previously obtained by MIR. MLPHARE (CCP4) was used for heavy-atom position refinement and phase calculation. The phase was calculated utilizing data collected from the wavelength corresponding to the maximum anomalous signal as the heavy-atom derivative data set, and the data at a remote wavelength from the selenium absorption edge as the native data set. After solvent flattening and density modification, an experimental electron density map was calculated with the CCP4 package using data between 15 and 2.0 Å. The resulting map was of sufficient quality to identify most regions of the molecule, with the exception of several loop regions (Figure 4).

Model building and refinement

The initial chain tracing and all subsequent model building were performed using O (Jones et al., 1991). Crystallographic refinement was performed using X-PLOR (Brünger, 1992a) with cycles of bulk solvent correction, energy minimization and simulated annealing. The model was refined against 90% of the data set with the maximum anomalous signal, while 10% was used to calculate the free R-factor (Brünger, 1992b). At later stages of refinement, the X-PLOR-refined model was also refined using the program CNS (Brünger et al., 1998). After the free R-factor was reduced to 26% and the R-factor was reduced to 23%, 185 water molecules were identified from the refined electron densities in the F_o – F_c map. The final model was refined to 1.6 Å resolution with an R-value of 22% and a free R-value of 24%, and includes all the residues of Mj0697 (Table II). None of the amino acid residues lie within the disallowed regions of a Ramachandran plot (Laskowski et al., 1993). The atomic coordinates have been deposited in the Protein Data Bank with accession code 1FBN.