Archease from Pyrococcus abyssi Improves Substrate Specificity and Solubility of a tRNA m5C Methyltransferase*

Members of the archease superfamily of proteins are represented in all three domains of life. Archease genes are generally located adjacent to genes encoding proteins involved in DNA or RNA processing. Archease have therefore been predicted to play a modulator or chaperone role in selected steps of DNA or RNA metabolism, although the roles of archeases remain to be established experimentally. Here we report the function of one of these archeases from the hyperthermophile Pyrococcus abyssi. The corresponding gene (PAB1946) is located in a bicistronic operon immediately upstream from a second open reading frame (PAB1947), which is shown here to encode a tRNA m5C methyltransferase. In vitro, the purified recombinant methyltransferase catalyzes m5C formation at several cytosines within tRNAs with preference for C49. The specificity of the methyltransferase is increased by the archease. In solution, the archease exists as a monomer, trimer, and hexamer. Only the oligomeric states bind the methyltransferase and prevent its aggregation, in addition to hindering dimerization of the methyltransferase-tRNA complex. This P. abyssi system possibly reflects the general function of archeases in preventing protein aggregation and modulating the function of their accompanying proteins.

the precise nature of this function has remained undetermined. The structures of two archeases, MTH1598 from the Archaea Methanobacter thermoautotrophicum (2) and TM1083 from the thermophile bacterium Thermotoga maritima (3), have been solved. The two structures superimpose quite well (root mean square deviation (r.m.s.d.) 5 ϭ 2.6 Å) despite the relatively low identity (14.8%) in their amino acid sequences (1). Intriguingly, the two archease structures are also similar to that of the heat shock protein Hsp33 (r.m.s.d. ϭ 3.1 Å) (4,5) and the gyrase inhibitory protein GyrI (r.m.s.d. ϭ 3.1 Å) (6). Protein Hsp33 is activated by oxidative stress and protects unfolded proteins against aggregation, whereas GyrI interacts specifically with DNA gyrase to inhibit its supercoiling activity. The structural resemblance to these proteins suggests that archeases might be engaged in similar functions.
Archeases contain two SHS2 modules (also called ␤␣␤2 folds) that are found in other proteins, including the ATPase FtsA, the RNA polymerase subunit Rpb7p, and GyrI (7). The two SHS2 modules in GyrI are arranged in a tandem and opposite orientation, whereas in the MTH1598 and TM1083 archeases the one SHS2 module is inserted within the other. Hsp33 has no SHS2 fold, although it possesses two slightly different S2HS2 modules (␤2␣␤2) arranged in the same manner as in the MTH1598 and TM1083 archeases. The S2HS2 modules probably mediate protein-protein interactions, as has been postulated for SHS2 domains (7). Further clues as to the function of archeases come from genome analyses, which reveal that all archease genes are adjacent to an ORF encoding a protein involved in nucleic acid processing, such as a DNA gyrase, a polymerase, or an RNA helicase (1,7). The structure and the genomic location of archeases suggest that they might interact directly with these proteins. In support of this hypothesis, we additionally note that the C-terminal end of a putative Thiobacillus denitrificans tRNA nucleotidyltransferase is an archease sequence (EMBL data base number Q3SGD7).
In the genome of the extreme thermophile Pyrococcus abyssi, an archease gene (PAB1946) is located immediately upstream from PAB1947 forming a putative bicistronic operon (8,9). PAB1947 contains the Sun/NOL1/NOP2 family signature (Prosite profile PS01153) suggesting that it encodes an RNA m 5 C methyltransferase. To date, the specific RNA modifica-tions sites of four RNA m 5 C methyltransferases have been identified. Two of these enzymes, the Escherichia coli rRNA m 5 C methyltransferases RsmB (10,11) and RsmF (originally annotated as YebU) (12), are site-specific and modify 16 S rRNA at positions C967 and C1407, respectively. The two other characterized m 5 C methyltransferases modify eukaryal tRNAs. The Trm4 enzyme from Saccharomyces cerevisiae targets multiple cytidines within different tRNA, modifying positions 34 or 40 in an intron-dependent manner and positions 48 and/or 49 in an intron-independent manner (13). The orthologous Trm4 enzyme in humans appears to target only cytidine 34 within the intron-containing tRNA Leu (14). The m 5 C modification is common in archaeal and eukaryal tRNAs (15) (summarized in Fig.  1) and is frequently found at nucleotides 48 and 49, where it improves base stacking and enhances tRNA stability (16,17). Of the presently characterized m 5 C methyltransferases, P. abyssi PAB1947 shows greatest sequence similarity to the Trm4 enzyme (13,18,19), suggesting a similar substrate specificity.
In this study, we determine the functions of the gene products encoded by the P. abyssi PAB1946 (the archease) and PAB1947 (the putative methyltransferase). We establish that PAB1947 does indeed encode a tRNA m 5 C methyltransferase, and we localize its tRNA methylation sites by a combination of biochemical and mass spectrometric techniques. Furthermore, we establish how the substrate specificity of the tRNA m 5 C methyltransferase and its tendency to aggregate are influenced over a range of temperatures by the P. abyssi archease. The study provides the first experimental evidence for how an archease and its co-expressed enzyme operate as a functional couple.

EXPERIMENTAL PROCEDURES
Cloning of the P. abyssi Open Reading Frames PAB1946 and PAB1947-Recombinant versions of the archease (PAB1946) and putative methyltransferase (PAB1947) proteins were constructed with a His 6 tag at their N-terminal ends. The genes were amplified from P. abyssi genomic DNA using the following PCR primers: AAAACCATGGGTCATCATCATCATCA-TCACAAGAGATGGGAGCACTATG (5Ј-primer) and AAA-ACTCGAGTCATATGTCGGGGACAAG (3Ј-primer) for PAB1946, and AAAACCATGGGTCATCATCATCATCAT-CACATGGACTACAAGGAAGAAT (5Ј-primer) and AAAA-CTCGAGTCACCTCGGCTTCCTTATCTTAG (3Ј-primer) for PAB1947. The 5Ј-primers created the ATG start codon in the NcoI restriction site followed by six histidine codons. The 3Ј-primers introduced an XhoI restriction site immediately downstream from the TGA stop codon. The amplified fragments were cut with NcoI and XhoI restriction enzymes, and were cloned into the same sites in the expression vectors pET28b and pET15b (Novagen) for PAB1946 and PAB1947, respectively.
Overexpression and Purification of Recombinant Archease and PAB1947 Proteins-Archease and the PAB1947 enzyme were overexpressed in separate clones of E. coli BL21 (DE3) CodonPlus RIL (Stratagene). Transformed cells were grown at 37°C in LB medium containing chloramphenicol and were supplemented with kanamycin for the archease plasmid and ampicillin for the PAB1947 plasmid. Recombinant protein expression was induced by 1 mM of isopropyl thiogalactopyranoside when the culture absorbance reached 0.7 at 600 nm. After a further 3 h at 28°C, cells were harvested and resuspended in 5 volumes of buffer A (50 mM sodium phosphate, pH 7.5, 300 mM NaCl, 10 mM imidazole, 10% glycerol, 5 mM ␤-mercaptoethanol) supplemented with 1% (v/v) Protease Inhibitor Mixture (Sigma). For the PAB1947 enzyme, NaCl was increased to 1 M to prevent co-purification of nucleic acids. Cells were lysed by sonication and centrifuged at 15,000 ϫ g for 30 min at 4°C.
The archease supernatant was added to nickel-nitrilotriacetic acid Superflow TM resin (Qiagen) and mixed during 30 min at 4°C. Special care was required to avoid aggregation of the PAB1947 enzyme, which was loaded in batches onto the resin, and purification was carried out at room temperature. Resin columns were washed with 20 mM imidazole in buffer A, and proteins were eluted with buffer A containing 250 mM imidazole for archease and 150 mM imidazole for the PAB1947 enzyme. The final respective yields were 45 and 35 mg of recombinant protein per liter of culture. After purification, archease was dialyzed against 25 mM Tris-Cl, pH 7.5, 100 mM NaCl, 10% glycerol, 1 mM DTT; for the PAB1947 enzyme, the buffer contained 300 mM NaCl. Both proteins were stored in aliquots at Ϫ80°C.
Cloning of the P. abyssi tRNA Asp Gene-The P. abyssi tRNA Asp gene was cloned for transcription by T7 RNA polymerase. Overlapping DNA oligonucleotides were annealed, elongated, and then amplified using Platinum TM Pfx DNA polymerase (Invitrogen). The oligodeoxynucleotides sequences were as follows: (ϩ) primer, CCCAAGCTTAATACGACTCA-CTATAGCCCGGGTGGTGTAGCCCGGCCTATCATGCG-FIGURE 1. Occurrence of m 5 C in archaeal tRNA sequences. The data are a compilation of the 54 halophilic archaeal tRNAs and the five structures from thermophilic and hyperthermophilic archaea that are presently available (15). The conventional nucleotide numbers are encircled. In the boxes, the first number gives the frequency of m 5 C, and the second is the number of archaeal tRNAs with cytosine at this position. Nucleotide m 5 C is most common at positions 48 and 49 and is found at both sites in nine of the tRNAs. No more than two m 5 C are present per archaeal tRNA.
GGACTGTCAC; and (Ϫ) primer, CGCGGATCCTGGCGCC-CGGGCCGGGATTTGAACCCGGGTCGCGGGAGTGAC-AGTCCCGCATGATAGGCCG. The (ϩ) primer contained a HindIII site (underlined) upstream from the T7 promoter (italic type), and the (Ϫ) primer contained BamHI (underlined) and MvaI sites (boldface type), positioned to cleave the template for in vitro transcription. The PCR product was digested with HindIII and BamHI and was inserted into the same sites in plasmid pUC18.
T7 in Vitro Transcriptions of RNA-Transcription of the tRNA genes with radiolabeled [␣-32 P]CTP were transcribed in vitro from the T7 promoter as described previously (20). The plasmid templates were digested with MvaI for P. abyssi tRNA Asp and E. coli tRNA 2 Tyr and with BamHI for P. abyssi tRNA 1 Leu prior to transcription. Nonradioactive tRNAs were synthesized by using the RiboMAX TM large scale RNA production system-T7 (Promega). All tRNA transcripts were purified on 6% polyacrylamide gels; yields were around 100 g of transcript from 10 g of plasmid template. The rRNA fragments radiolabeled with [␣-32 P]CTP were T7-transcribed in vitro from single-stranded DNA templates (21).
Methyltransferase Assays-Methylation reactions were carried out in 50 l of 25 mM Tris-Cl, pH 7.5, 50 mM KCl, 2 mM DTT, 80 M AdoMet (i.e. in large excess), 1 mg/ml RNase-free bovine serum albumin (Worthington), with 1 nM 32 P-radiolabeled RNA and 200 nM PAB1947 enzyme. In some experiments, radiolabeled tRNA (1 nM) and unlabeled tRNA (200 nM) were mixed to increase the substrate concentration. Reactions were quenched by phenol/chloroform extraction; tRNA were digested with nuclease P1 (Roche Applied Science); and modified nucleotides were identified by two-dimensional thin layer chromatography (20). Radiolabeled mono-nucleotide spots were quantified using a Storm TM system (GE Healthcare).
Methyltransferase assays using nonradioactive RNA and 0.67 M tritiated AdoMet (GE Healthcare) were carried out at 50°C to avoid thermal degradation of the cofactor (13). Other components in the reaction were unchanged, except that 0.3 g of tRNA transcript or poly(C) RNA (Roche Applied Science) was used. In some samples, 1.2 or 4 M archease was added, corresponding to a 6-or 20-fold molar excess over the PAB1947 enzyme. MALDI Mass Spectrometry Analysis-Nonradioactive tRNA transcripts at 200 nM were methylated as above with the exception that KCl was replaced with ammonium acetate, pH 7.5. After 10 min of incubation at 80°C, reactions were stopped by phenol/chloroform extraction, and tRNA was recovered by ethanol precipitation with 300 mM ammonium acetate, pH 5.5. The tRNA was digested with 10 -20 units of RNase T1 (U. S. Biochemical Corp.) in 0.5 l of 0.5 M 3-hydroxypicolinic acid at 37°C for 2 h. Cyclic phosphates were hydrolyzed with HCl, and the RNA oligonucleotides were dried and redissolved in H 2 O (12).
Mass spectra were recorded in reflector and positive ion mode on a Voyager-DE STR mass spectrometer (PerSeptive Biosystems); spectra were smoothed and calibrated using "m/z" software (Proteometrics Inc). Tandem mass spectra were recorded in positive ion mode on a MicroMass MALDI Q-Time-of-Flight Ultima mass spectrometer (22). The colli-sion energy used for tandem mass spectrometry was varied between 30 and 110 eV. All tandem mass spectra were smoothed using the MassLynx software supplied by the manufacturer.
Determination of m 5 C Level in T1 Oligonucleotides of P. abyssi tRNA Asp -[␣-32 P]CTP-labeled P. abyssi tRNA Asp was methylated as described above with or without a 10-fold molar excess of archease. After 10 min at 80°C, the tRNA was extracted with phenol and recovered by ethanol precipitation. The tRNA was then digested by overnight incubation in 25 mM Tris-Cl, pH 7.5, at 37°C with 10 units of RNase T1 and 10 g of unlabeled carrier tRNA (yeast total tRNA; Roche Applied Science). The individual RNA oligonucleotides were purified on denaturing 20% polyacrylamide gels, and after extraction as described (23), they were digested with nuclease P1. The resulting mononucleotides were analyzed by two-dimensional thin layer chromatography (20).
Gel Filtration and Stokes Radius Determination-The interaction of PAB1947 enzyme with archease was studied by gel filtration on a Superdex TM 75 HR 10/30 column (GE Healthcare) equilibrated with 25 mM Tris-Cl, pH 7.5, 100 mM NaCl, 10% glycerol, 5 mM EDTA. Samples, in 200 l of the same buffer containing 150 g (20 M) PAB1947 enzyme and/or 450 g (130 M) archease, were incubated for 30 min at ambient temperature (18 -20°C) before loading onto the column and eluting at a flow rate of 0.5 ml/min. For Stokes radius (R S ) determination, the column was calibrated with bovine ␥-globulin (158,000 Da; R S ϭ 45 Å), chicken ovalbumin (44,000 Da; R S ϭ 27.5 Å), equine myoglobin (17,000 Da; R S ϭ 20 Å) (Bio-Rad), and bovine brain tubulin (100,000 Da; R S ϭ 41.5 Å (24)). The void volume of the column was determined from the elution volume of blue dextran. The Stokes radii of archease, PAB1947 enzyme, and their complexes were determined using Equation 1, according to Ref. (25), Binding of tRNA to PAB1947 or the archease-PAB1947 complex was followed by Superdex TM 200 HR 10/300 GL gel filtration (GE Healthcare) as described above. The PAB1947-tRNA complex was formed between 16 M PAB1947 enzyme and 2.8 M (15 g) natural E. coli tRNA 2 Tyr (Sigma); to form the tripartite complex, 100 M (360 g) archease was added before the E. coli tRNA 2 Tyr .

JOURNAL OF BIOLOGICAL CHEMISTRY 18713
Sedimentation Velocity and Molecular Mass Calculation-Sedimentation velocity experiments were carried out with a Beckman Optima XL-A analytical ultracentrifuge equipped with a 60 Ti four-hole rotor and cells with two-channel 12-mm path length centerpieces. Measurements were made at 50,000 rpm and at 18°C with PAB1947 enzyme at 0.3 mg/ml and archease at 0.9 mg/ml in 25 mM Tris-Cl, pH 7.5, 100 mM NaCl, and 10% glycerol. The solvent density was 1.032 g⅐cm Ϫ3 , and partial specific volumes of PAB1947 enzyme (0.7441 cm 3 ⅐g Ϫ1 ) and archease (0.7405 cm 3 ⅐g Ϫ1 ) were calculated using SEDNTERP software (26). The apparent distributions of sedimentation coefficients were calculated using Svedberg (26) and Sedfit (27)(28)(29) programs.
The apparent molecular masses of the proteins were calculated from the Stokes radii and sedimentation coefficients using the modified Svedberg Equation 2 (25), where M is molecular mass; is viscosity of the medium (10 Ϫ2 g⅐cm Ϫ1 ⅐s Ϫ1 ); N is Avogadro's number; R S is Stokes radius, s is sedimentation coefficient, v is partial specific volume, and is density of the medium. Sedimentation Equilibrium-Sedimentation equilibrium experiments were performed at 10,000 rpm and at 18°C. Radial scans of the absorbance at 280 nm were taken at 3-h intervals, and equilibrium was reached after 24 h of centrifugation. The base line was recorded at 60,000 rpm, at the end of the experiment. The data were analyzed with XLAEQ and EQASSOC programs (Beckman) to calculate molecular weights.
Aggregation Analyses-The PAB1947 enzyme was adjusted to a final concentration of 0.06 mg/ml in 20 l of 25 mM Tris-Cl, pH 7.5, 50 mM KCl, 2 mM DTT, with or without 0.18 mg/ml of archease or BSA. Samples were incubated for 10 min at 4, 60, 65, 70, or 80°C and were then centrifuged at 12,000 ϫ g for 15 min. The pellets were resuspended in 20 l of the same buffer and were analyzed by SDS-PAGE together with the supernatants.

RESULTS
The P. abyssi ORF PAB1947 Encodes a tRNA m 5 C Methyltransferase-BLAST analysis (30) of the P. abyssi genome identified five potential RNA m 5 C methyltransferases. Of these, the sequence encoded by PAB1947 is most similar to the tRNA m 5 C methyltransferase from S. cerevisiae, Trm4p (13,18,19).
The P. abyssi PAB1947 gene was cloned in E. coli and expressed with an N-terminal His tag. The activity of the purified recombinant PAB1947 protein was tested at 50 and 80°C against tRNA transcripts containing radiolabeled cytidine. The incubation temperatures in vitro were lower than that for optimal growth of P. abyssi (103°C) to reduce breakdown of the individual components of the reaction. After methylation, nucleotide monophosphates were derived from the transcripts by nuclease P1 digestion and were analyzed by thin layer chromatography. Analyses of P. abyssi tRNA Asp (Fig. 2A), P. abyssi tRNA 1 Leu , and E. coli tRNA 2 Tyr transcripts (not shown) demonstrated that the recombinant PAB1947 protein is indeed an m 5 C methyltransferase and uses AdoMet as the methyl donor.
Time courses of the methylation reactions were performed at 80°C with P. abyssi tRNA Asp and tRNA 1 Leu (Fig. 2B). After 10 min, the PAB1947 enzyme had methylated 1.6 cytosines per tRNA Asp and 2.6 cytosines in tRNA 1 Leu . The tRNA substrates at 200-fold higher concentrations were also methylated, and this resulted in a lower final level of m 5 C incorporation (Fig. 2B).°°°F The mesophilic E. coli tRNA 2 Tyr was tested only at 50°C to avoid degradation at the higher temperature. The PAB1947 enzyme is active on all the tRNA substrates at 50°C (Fig. 2C), although methylation is slower than at 80°C. Surprisingly, small fragments of ribosomal RNA (Fig. 2D) were also methylated by the PAB1947 enzyme, despite having no obvious structural similarities to tRNAs.
The Unassisted PAB1947 Enzyme Methylates Multiple Cytosines in P. abyssi tRNA-Pyrococcal tRNAs have been shown previously to contain m 5 C (31), although the positions of these modifications were not determined. Here we have mapped the methylation sites in P. abyssi tRNA Asp (Fig. 3A) after incubation at 80°C under conditions where 0.3 mol of m 5 C was incorporated per mol of tRNA. The methylated tRNA was digested with the guanosine-specific RNase T1, and the oligonucleotides were analyzed by MALDI mass spectrometry. In the mass spectra (Fig. 3B), proportions of six oligonucleotides were 14 Da larger than would be expected from their predicted masses (Fig. 3C), indicating that the recombinant PAB1947 enzyme had methylated at least six different nucleotides in the tRNA Asp . The number of methylation sites is potentially higher, as some oligonucleotides (such as in the oligonucleotide CCCG) contain several cytosines and could be substoichiometrically modified at more than one cytosine under the conditions used here.
From the compilation of known methylation sites (Fig. 1), only the fragments 35-43 and 47-51 were expected to contain targets for methylation. When tRNA was incubated with the PAB1947 enzyme at 50°C instead of 80°C, methylation became more specific for the 47-51 sequence, indicating that this oli-gonucleotide contained the preferred target. This methylated oligonucleotide (m/z 1622.3) was analyzed by tandem mass spectrometry, and the primary site of PAB1947 methylation was unambiguously identified as the tRNA cytidine 49 (Fig. 4). The methylation site was localized more precisely to the cytosine base of nucleotide 49 from the combination of ions, including y 3 that had lost methylcytosine (m/z 863.1), the free methylcytosine ion (m/z 126.1), and the lack of any methylribose ion (at m/z 111.1).
Nonspecific Methylation in Vitro of Poly(C) RNA by PAB1947-Our observation that a large variety of RNAs could serve as substrates for the PAB1947 enzyme in vitro was tested further using a poly(C) RNA. The unstructured poly(C) RNA was effectively methylated at 50°C, with most of the tritiated label from the AdoMet methyl donor becoming incorporated into the RNA (Fig. 5B). However, when the recombinant PAB1947 enzyme was substituted with a P. abyssi cell extract, no methylation of the poly(C) RNA occurred. Under comparable conditions, the cell extract efficiently catalyzed m 5 C formation in P. abyssi tRNA Asp transcripts, and retained this activity despite extensive dialysis (data not shown). Clearly, the naturally occurring tRNA m 5 C methyltransferase activity in P. abyssi cells has greater substrate specificity than the isolated recombinant PAB1947 enzyme, and this indicated that at least one essential component was missing from our in vitro assays.
A clue as to what this component might be was given by its resilience to extensive dialysis. The high specificity of methylation exhibited by the P. abyssi cell extract was retained after dialysis through membranes with a cutoff at 14,000 Da. This ruled out the hypothetical component being a small molecule, such as magnesium or a polyamine, which are known to stabilize RNA structures (reviewed in Refs. 32 and 33).
The Specificity of the PAB1947 Enzyme Is Dependent on Archease-An obvious candidate for the missing component was the P. abyssi archease, the gene for which (PAB1946) is located immediately upstream in the same operon as the PAB1947 methyltransferase. Because MALDI mass spectrometry is essentially a nonquantitative method, we used an alternative approach to determine the degree to which archease influences the methylation of different tRNA regions. [␣-32 P]CTP-labeled tRNA samples that had been modified by the PAB1947 methyltransferase in the absence or presence of archease were digested to completion with RNase T1. Oligonucleotides were purified on gels and subjected to further digestion with nuclease P1 followed by thin layer chromatography to estimate the amounts m 5 C, as described above. Addition of archease to the methylation reactions with P. abyssi tRNA Asp at 80°C increased modification within the oligonucleotide ACCCG that contains cytidine 49, while reducing modification at unspecific sites (Fig.  5A). This fits well with the progressive decline in the methylation of the nonspecific poly(C) RNA substrate after addition of archease (Fig. 5B). This effect is specific for the archease, because addition of comparable amounts of bovine serum albumin did not affect the methylation reaction (Fig. 5C).
To investigate further how archease increases the methyltransferase specificity, we determined whether the relative affinities of the methyltransferase for nonspecific RNAs compared with its tRNA substrate are affected by archease. Gel retardation assays for the P. abyssi tRNA Asp , together with Scatchard plots to quantify free and bound tRNA, are shown in Fig. 6. The apparent dissociation constant (K d ) of the methyltransferase for RNA is ϳ1 M ( Table 1). Upon addition of archease, the K d of the methyltransferase for the authentic substrate P. abyssi tRNA Asp increased about 3-fold, whereas the K d for the two nonspecific RNA substrates increased 10-fold.
Archease on its own is unable to bind tRNA (Fig. 6, lane 12) or small rRNA fragments (data not shown), indicating that its effect in reducing the affinity of the PAB1947-RNA interaction is not merely because of substrate sequestration. A more plausible explanation is that archease contributes to specificity by reducing methyltransferase binding and possibly catalytic efficiency at unspecific sites. The interaction between the archease and the methyltransferase was explored further using gel filtration and sedimentation velocity approaches.  Tripartite Complex of Archease, PAB1947 Enzyme, and tRNA-Gel filtration analyses of the isolated proteins showed that archease eluted in three peaks (Fig. 7A, fractions 10 -14), whereas the PAB1947 enzyme came out in one peak (fractions [13][14][15][16]. After the two proteins were allowed to interact, a complex was formed shifting the PAB1947 enzyme from fractions 13-16 to fractions 10 -11 (Fig. 7, A-C). In the absence of archease, the methyltransferase interacted with E. coli tRNA 2 Tyr to form a complex PAB1947-tRNA that eluted in fractions 13 and 14 ahead of the free tRNA in fraction 16 (Fig. 8A). Upon addition of archease to PAB1947 enzyme and the tRNA, a tripartite complex was formed, which eluted unexpectedly in the intermediate fractions 14 -16 and thus had a smaller apparent size than that of the PAB1947-tRNA complex (Fig. 8B). A small fraction of PAB1947-tRNA complex (fraction 13) remained unbound to archease.
Sedimentation velocity experiments (supplemental Fig. S1) on the isolated components revealed one species of the PAB1947 enzyme at 4.5 S and three species of the archease at 1.9 S, 3.8 S, and 6.5 S, with respective proportion of 50, 25, and 25% (Table 2). After interaction of the PAB1947 enzyme and archease components, three species were observed sedimenting at 1.9 S, 5.5 S and 7.7 S (50, 40, and 10%, respectively). These results are consistent with the gel filtration results (Fig. 7) and suggest that only the two archease species of 3.8 S and 6.5 S bind to the PAB1947 enzyme to form the 5.5 S and 7.7 S species, respectively. Apparent molecular masses of proteins and the PAB1947archease complexes were calculated from their Stokes radius determined by gel filtration and from their sedimentation coefficients (Table 2). These data are compatible with a monomer form of the PAB1947 enzyme (M), and monomer, trimer, and hexamer forms of the archease (A ϩ A 3 ϩ A 6 ). PAB1947 enzyme binds only the oligomeric forms of the archease. The proportion of unreactive, monomeric archease remains constant at around 50% (Table 2), and thus we tentatively propose the following scenario for complex formation: Sedimentation equilibrium measurements of the PAB1947-tRNA complex gave an apparent molecular mass of 144 Ϯ 10 kDa (data not shown), fitting with a stoichiometry of 2:2 and indicating that PAB1947-tRNA dimerizes (2 M ϩ 2T 7 (MT) 2 , where T is tRNA). The smaller size of the tripartite complex archease-PAB1947-tRNA measured by gel filtration thus indicates that archease prevents dimerization of PAB1947-tRNA.
Archease Protects PAB1947 Enzyme against Aggregation-The sedimentation velocity data indicate the solubility state of the proteins. The start and the end of the spectral curves from the isolated PAB1947 enzyme exhibit dispersed points (supplemental Fig. S1A) that are characteristic for a partially aggregated protein.
These findings are consistent with the difficulties experienced in keeping the recombinant PAB1947 enzyme soluble during purifi-

TABLE 1 Effects of archease on the affinity of PAB1947 enzyme for specific and nonspecific RNA substrates
The dissociation constants for PAB1947 enzyme on different RNAs in the presence or absence of archease were determined by gel retardation. The ribosomal RNA fragments are represented in Fig. 2D; the domain IV fragment is 72 nucleotides in length; the domain II fragment is 55 nucleotides. cation (see "Experimental Procedures"). The equivalent points of the curve are homogeneous for the archease/PAB1947 mixture (supplemental Fig. S1E), showing that the methyltransferase remains soluble under these conditions. These results suggest that the PAB1947 enzyme tends to aggregate, and this tendency is counteracted by archease.
To confirm this observation, the ability of archease to counteract thermally induced aggregation of the PAB1947 enzyme was analyzed using a sedimentation assay coupled with SDS-PAGE analysis (Fig. 9). In the absence of archease, the unprotected PAB1947 enzyme is almost completely aggregated after 10 min at 70°C (Fig. 9A). However, addition of a 6-fold molar excess of archease enables about half of the PAB1947 enzyme to remain soluble under the same conditions (Fig. 9B). Addition of bovine serum albumin (Fig. 9C) or the thermostable protein stathmin (24) (data not shown) offered no protection at 70°C. Archease also prevents aggregation of PAB1947 enzyme at 4°C, whereas BSA does not (Fig. 9D). The archease is itself an extremely thermostable protein and remains soluble up to at least 80°C.

DISCUSSION
Members of the archease family have been proposed previously to modulate and chaperone the nucleic acid processing activities of the proteins that are encoded immediately downstream (1,7). Here we provide evidence for this idea after determining the functions of the PAB1946 (archease) and PAB1947 (RNA methyltransferase) genes that are adjacent in the genome of P. abyssi. When the purified recombinant PAB1947 methyltransferase is allowed to function unrestrained, it catalyzes AdoMet-dependent formation of m 5 C at many locations in tRNAs, rRNA fragments, and even unstructured poly(C) RNA. Archease makes the PAB1947 enzyme more selective by reducing its affinity for nonspecific RNA as well as for nonspecific sites within tRNA, raising the respective K d values by 3-and 10-fold. In gel filtration (Fig. 8B) and in agreement with our gel retardation assays (Fig. 6), we note that a proportion of the PAB1947-tRNA complex remains unbound by archease, and we are therefore probably underestimating the improvement in specificity in the archease-bound fraction.
Thus, without the controlling influence of the archease, the PAB1947 methyltransferase interacts directly with numerous RNAs in addition to its preferred tRNA substrates. Similar cases of broad specificity among tRNA m 5 C methyltransferase have also been reported. For instance in yeast, depletion of the maturation enzyme tRNA guanylyltransferase leads to the methyltransferase Trm4p becoming more promiscuous by adding two extra methyl groups on cytosines at positions 48 and 50 in the tRNA His , where normally m 5 C is found only at position 49 (34). The human tRNA m 5 C methyltransferase shows greatest activity in vitro on tRNA substrates, although methylation occurs with other substrates such as E. coli rRNA and viral RNA (35). Similar in vitro activity was also observed for the putative murine tRNA m 5 C methyltransferase, Misu (36). Misu expression is upregulated in vivo by the proto-oncogene myc and is connected with myc-induced cell proliferation; arresting Misu expression leads to reduction of some tumors. The broad specificities of these methyltransferases become evident only under unusual conditions, and may be a remnant property of an ancestral RNA m 5 C methyltransferase, which has subsequently evolved along different paths to achieve higher specificity.
In vitro studies with the PAB1947 methyltransferase in the absence of the archease can undoubtedly be regarded as aberrant conditions. Such conditions promote the formation of a dimeric complex between the methyltransferase and its canon- ical tRNA substrate. The PAB1946 archease improves specificity by binding directly to the PAB1947 methyltransferase to form an archease-methyltransferase complex, which then interacts with the tRNA substrates. Under the direction of archease, the resultant methyltransferase-tRNA complex is in a monomeric rather than a dimeric state. Archease on its own is incapable of forming a stable complex with tRNA, and therefore functions by directly interacting with and modulating the methyltransferase, rather than by merely sequestering the reaction components. Two structural features of the archease appear to be important for this mode of action. First, it possesses patches of negative charge on its surface (theoretical pI ϭ 4.6), which could neutralize the predominantly positively charged regions of the PAB1947 methyltransferase and thereby prevent the methyltransferase from making nonspecific interactions with RNA phosphate backbones. Second, in addition to its monomeric state, the archease also forms homotrimers and homohexamers, and only these oligomeric forms bind the PAB1947 methyltransferase.
The structural and functional properties of archease are reminiscent of the molecular chaperone, heat shock protein 33 (Hsp33) (1, 7). Upon activation by oxidative stress, Hsp33 forms dimers, tetramers, and higher oligomers that are able to bind tightly to unfolded proteins and prevent their irreversible aggregation; the monomeric Hsp33 has no comparable function (37,38). The gel filtration profiles of archease ( Fig. 7) resemble those of Hsp33 obtained under oxidation conditions (38). The crystal structure of Hsp33 (4,5,39,40) suggests that a single domain interacts with its identical counterpart to form a dimer, in a process termed domain swapping (41). The crystal structure of the T. maritima archease TM1083 (Protein Data Bank code 1J5U) (3) exhibits domain swapping via the N-terminal ␤-strand (supplemental Fig. S2) with a contact area of 1300 Å 2 . This is in the upper range for nonphysiological crystal contacts (42), and it remains to be established whether the contacts are biological relevant and not merely crystal artifacts (1).
Based on theoretical considerations, it has been proposed that the TM1083 archease stabilizes the thermolabile DNA gyrase encoded by the adjacent downstream ORF, TM1084 (1). Such a mechanism is supported here by our experimental findings where the thermostable P. abyssi PAB1946 archease prevents aggregation of its downstream PAB1947 methyltransferase. In this respect, the archease proteins differ from the Hsp33 analogue, as the main archease function involves the  a Stokes radii were determined by gel filtration, see under "Experimental Procedures" for details. b Apparent molecular weights calculated from Stokes radii and sedimentation coefficients (see "Experimental Procedures"). c M is the PAB1947 methyltransferase; A is the archease (see text for discussion of putative compositions of complexes).
binding and protection of folded proteins rather than unfolded and aggregated proteins (43). However, there are exceptions among the heat shock proteins, and Hsp90 also binds native p53 to prevent this labile protein from irreversible thermal inactivation (44). It should be added that the full repertoire of pyrococcal archease functions remains to be investigated, and we cannot presently rule out that archeases might also act as classical molecular chaperones to bind and refold denatured proteins.
In conclusion, we show here that the ORF PAB1947 encodes an m 5 C tRNA methyltransferase which, under conditions elicited in vitro, shows extensive nonspecific activity. However, it is unlikely that such unrestrained activity would be evident under physiological conditions. The free PAB1947 methyltransferase has a marked tendency to aggregate, and its solubility and specificity require the presence of the PAB1946 archease, which is co-expressed in vivo. Similarly co-operative systems are likely to be functional in other thermophilic genera, where an archease gene is immediately upstream of an ORF encoding a nucleic acid processing enzyme. This order ensures that the archease gene is expressed ahead of its accompanying enzyme gene, so that the archease is already formed when its chaperone activity is required. The study present here is the first to experimentally test these ideas, and provides credence for the notion that members of the ubiquitous archease family serve to protect and direct the functions of adjacently encoded DNA/RNA processing proteins.