The FtsJ/RrmJ Heat Shock Protein of Escherichia coli Is a 23 S Ribosomal RNA Methyltransferase*

Ribosomal RNAs undergo several nucleotide modifi-cations including methylation. We identify FtsJ, the first encoded protein of the ftsJ-hflB heat shock operon, as an Escherichia coli methyltransferase of the 23 S rRNA. The methylation reaction requires S -adenosylme-thionine as donor of methyl groups, purified FtsJ or a S 150 supernatant from an FtsJ-producing strain, and ri- bosomes from an FtsJ-deficient strain. In vitro , FtsJ does not efficiently methylate ribosomes purified from a strain producing FtsJ, suggesting that these ribosomes are already methylated in vivo by FtsJ. FtsJ is active on ribosomes and on the 50 S ribosomal subunit, but is inactive on free rRNA, suggesting that its natural substrate is ribosomes or a pre-ribosomal ribonucleoprotein particle. We identified the methylated nucleotide as 2 * - O -methyluridine 2552, by reverse phase high performance liquid chromatography analysis, boronate Purification— The RrmJ/FtsJ protein was purified using the IMPACT y -CN (intein mediated purification with an affinity chitin-binding tag) kit from New England Biolabs, Inc. according to manufac- turer’s instructions. The recombinant RrmJ/FtsJ protein fused to a chitin-binding domain (CBD) was purified by affinity chromatography on a chitin affinity column (20, 21). The CBD of the recombinant protein was spaced from the RrmJ/FtsJ protein by an intein that undergoes self-cleavage by addition of 1,4-dithiothreitol, thus releasing the predicted wild type RrmJ/FtsJ protein. The eluted RrmJ/FtsJ protein was pure, as judged by SDS-polyacrylamide gel electrophoresis (22). It was stored at 2 20 °C in Column Buffer supplemented with 30% glycerol. Nucleoside Analysis— Identification of the methylated nucleotide was performed using a reverse-phase HPLC protocol previously de- scribed for the analysis of modified nucleosides in tRNA (23, 24). Purified RrmJ/FtsJ and [ methyl - 3 H]AdoMet were used to 3 H-methylate the 50 S ribosomal subunits from the FtsJ-deficient strain, as described above. The 3 H-methylated 23 S rRNA was phenol-extracted, ethanol- precipitated, and dissolved in distilled water. This [ 3 H]rRNA (200 m g in 50 m l of distilled water) was digested overnight at 37 °C by 40 m l (0.3 unit/ m l) of nuclease P1 (Roche Molecular Biochemicals) in 30 m M ace-tate buffer, pH 5.3, and 30 m l of 10 m M ZnCl 2 . The solution was further digested for 2 h at 37 °C by 5 m l (0.01 unit/ m l) of snake venom phosphodiesterase I (Sigma), 10 m l (0.1 unit/ m l) of bacterial alkaline phospha- tase (Sigma), ribosomal H-methylated 23 ethanol-precipitated, and in distilled This [ 3 H]rRNA (0.7 pmol of methyl- 3 H/ pmol) was digested with nuclease P1, snake venom phosphodiesterase I, and bacterial alkaline phosphatase The resulting nucleoside mixture was analyzed by reverse-phase HPLC, and 5-min elution were collected and counted for 3 H radioactivity. H-methylated nucleo- side thus located in corresponding to elution between H-methylated nucleoside, boronate

The ribosome is a complex ribonucleoprotein particle that is responsible for translation of messenger RNAs into proteins. In Escherichia coli, it is composed of 23, 16, and 5 S ribosomal RNAs and of about 52 proteins. Twenty-one of them assemble with the 16 S rRNA to form the 30 S ribosomal subunit, while the 31 others assemble with the 23 and 5 S rRNA to form the 50 S ribosomal subunit (1). Ribosomal proteins and rRNAs cooperate both in the assembly and in the activity of the ribosome (1). The functional domains of the ribosome include a GTPase center, a peptidyl transferase center, and A-, P-, and E-tRNA binding sites; they involve specific regions of the rRNAs and one or several ribosomal proteins (1). The 16 and 23 S ribosomal RNAs and several ribosomal proteins are methylated at specific sites. The mature E. coli 16 and 23 S rRNAs have 10 and 14 methylated nucleotides, respectively (2,3). The methyl groups are clustered at the functional domains, e.g. the A-and P-tRNA binding sites for 16 S rRNA, and the peptidyl-transferase center for 23 S rRNA (4). Most of the modified nucleotides are conserved (5); however, their functions are poorly understood. It has been suggested that methylation could modulate rRNA maturation, affect stability of rRNA structures, or alter translation rates. The nucleotide methylation of rRNAs can modify the susceptibility of ribosomes to antibiotics that target them (6); it is reasonable to expect that the binding of other ligands can be affected as well. In E. coli, three 16 S RNA methyltransferases have been identified, RsmA (also known as KsgA) (7), RsmB (2,8) and RsmC (9). In contrast, little is known about genes involved in 23 S rRNA modifications. The 23 S rRNA displays 23 nucleotide modifications, of which 14 correspond to methylation. Recently, the rrmA gene encoding a 23 S rRNA methyltransferase that forms m 1 G 745 was identified (10). In the present study, we show that the heat shock protein FtsJ methylates the 50 S ribosomal subunit on its 23 S rRNA in vitro. ftsJ is the upstream gene of a bi-cistronic operon that includes hflB (also known as ftsH) (11,12). Whereas the HflB heat shock protease has been extensively studied, the function of FtsJ had not yet been characterized. In view of its newly established function, FtsJ has been renamed RrmJ and its encoding gene, rrmJ.
Preparation of Ribosomal 50 and 30 S Subunits-Ribosomal subunits were separated by sedimentation through 5-20% sucrose gradients prepared in buffer 2 (buffer 1 supplemented with 0.4 M NaCl) (16). The pelleted subunits were resuspended in buffer 1 and stored at Ϫ70°C.
Preparation of Ribosomal RNA-rRNAs were prepared from ribosomes by phenol extraction and ethanol precipitation. They were dissolved in 50 mM Tris, pH 7.4, 10 mM MgCl 2 and stored at Ϫ20°C (17).
RrmJ/FtsJ Purification-The RrmJ/FtsJ protein was purified using the IMPACT-CN (intein mediated purification with an affinity chitinbinding tag) kit from New England Biolabs, Inc. according to manufacturer's instructions. The recombinant RrmJ/FtsJ protein fused to a chitin-binding domain (CBD) was purified by affinity chromatography on a chitin affinity column (20,21). The CBD of the recombinant protein was spaced from the RrmJ/FtsJ protein by an intein that undergoes self-cleavage by addition of 1,4-dithiothreitol, thus releasing the predicted wild type RrmJ/FtsJ protein. The eluted RrmJ/FtsJ protein was pure, as judged by SDS-polyacrylamide gel electrophoresis (22). It was stored at Ϫ20°C in Column Buffer supplemented with 30% glycerol.
Nucleoside Analysis-Identification of the methylated nucleotide was performed using a reverse-phase HPLC protocol previously described for the analysis of modified nucleosides in tRNA (23,24). Purified RrmJ/FtsJ and [methyl-3 H]AdoMet were used to 3 H-methylate the 50 S ribosomal subunits from the FtsJ-deficient strain, as described above. The 3 H-methylated 23 S rRNA was phenol-extracted, ethanolprecipitated, and dissolved in distilled water. This [ 3 H]rRNA (200 g in 50 l of distilled water) was digested overnight at 37°C by 40 l (0.3 unit/l) of nuclease P1 (Roche Molecular Biochemicals) in 30 mM acetate buffer, pH 5.3, and 30 l of 10 mM ZnCl 2 . The solution was further digested for 2 h at 37°C by 5 l (0.01 unit/l) of snake venom phosphodiesterase I (Sigma), 10 l (0.1 unit/l) of bacterial alkaline phosphatase (Sigma), 30 l of 10 mM magnesium acetate, and 30 l of 0.5 M Tris buffer, pH 8.3. The resulting hydrolysate was analyzed by HPLC using a Spectra-Physics liquid chromatograph equipped with a Supelcosil LC 18S column (250 mm ϫ 4.6 mm, Supelco France) maintained at 26°C in a controlled-temperature oven, a ternary solvent delivery system at a flow rate of 1 ml/min, and a forward Spectra Focus scanning detector. For location and isolation of the labeled nucleoside, 5-min fractions were collected, and counted for 3 H radioactivity. Each nucleoside component detected in the only radioactive fraction was characterized by comparing its retention time and UV spectrum with those of ribonucleoside reference standards. The 3 H-methyl nucleoside present in the radioactive fraction was finally identified as 2Ј-O-methyluridine (Um) by chromatography of that fraction on a boronate affinity column (Bio-Rad, Affi-Gel 601) following the manufacturer's recommended protocol.
Hybridization-Protection Studies-3 H-Methyl-labeled 23 S RNA (0.6 pmol) was hybridized with a 200-fold excess of deoxyoligonucleotide in a 50-l reaction mixture that also contained 40 mM MES, pH 6.4, 400 mM NaCl, 5 mM EDTA, and freshly deionized 80% (v/v) formamide, as described in Ref. 18. The samples were heat-denatured at 90°C for 10 min and then cooled at room temperature for 15 min. They were diluted with nine volumes of ice-cold RNase buffer (10 mM Tris, pH 7.5, 300 mM NaCl, 5 mM EDTA) and RNase T1 (1-10 Sankyo units/pmol of 23 S RNA) was then added. The samples were digested for 30 min at 42°C, precipitated with 10% trichloroacetic acid, filtered on HAWP Millipore filters, and counted for radioactivity. Miscellaneous-Products were reagent grade from Sigma. [methyl- 3 H]S-Adenosylmethionine was purchased from Amersham Pharmacia Biotech.

FtsJ/RrmJ Is a Putative Methyltransferase-
The ftsJ gene encodes a putative cytoplasmic alkaline protein of 209 amino acids. The gene is not essential under laboratory conditions (11), and it is transcribed from two promoters, one of which is a heat shock promoter (25). Putative methyltransferases can be predicted (26) by the presence of three motifs that are presumably involved in AdoMet binding (27,28). The FtsJ sequences VVTQIGGKGRIIACDLL, PAVDIPR, and VLAPGGSFVV (amino acid positions 69 -84, 136 -142, and 158 -167, respectively) are similar to the consensus motifs I, II, and III (27,28). Putative FtsJ AdoMet binding motifs were previously reported (29); however, the authors did not find motif II, and proposed a different motif I (which is also different from that reported in Ref. 27). FtsJ displays strong similarities with several protein isoaspartate methyltransferases, with the PrmA ribosomal protein methyltransferase and with a putative S-adenosylmethioninedependent enzyme from yeast (29 -31). Since our results show that FtsJ methylates the 23 S ribosomal RNA, and since the ftsJ-deficient strain AR1147 did not show significant filamentation at 30, 37 and 42°C (data not shown), we refer to the protein and to its coding gene as RrmJ (for ribosomal RNA methylation J) and rrmJ, respectively.
Ribosomal RNA Derived from a FtsJ/RrmJ-deficient Strain Is Methylated by Extracts from a FtsJ/RrmJ-producing Strain-We first checked whether RrmJ catalyzes protein methylation, by incubating supernatants and particulate fractions from a RrmJ-deficient strain and from the RrmJ-overproducing strain in the presence of [methyl-3 H]AdoMet. Under appropriate conditions, we observed a 3 H-labeled band of molecular mass higher than 200 kDa on SDS-polyacrylamide gels. This band was resistant to prior protease treatment and disappeared if samples were treated with RNase A before electrophoresis, suggesting that RrmJ could be involved in rRNA methylation. We therefore assayed the activity of different fractions on ribosomes. Crude extracts from the RrmJ-overproducing strain (J) and from the RrmJ-deficient strain (⌬) were prepared by mechanical lysis of bacteria. These extracts were separated into supernatant fractions (S J and S ⌬ , respectively) and ribosomal fractions (R J and R ⌬ , respectively), by ultracentrifugation. Incorporation of methyl groups from [methyl-3 H]AdoMet into phenol-extracted material was analyzed by polyacrylamide gel electrophoresis (Fig. 1). Incubation of ribosomes from the strain depleted for RrmJ (R ⌬ ) with the S 150 supernatant from the RrmJ-overproducing strain (S J ) leads to incorporation of labeled methyl groups in a band that migrates around 16 S RNA. This band (lane 7) could correspond either to 16 S rRNA or, as confirmed below, to a degradation product of 23 S rRNA (1). This radioactive band disappears if the sample is treated for 10 min with RNaseA before loading the gel (data not shown). No significant incorporation of methyl groups into RNA was observed upon incubation of S ⌬ with R ⌬ (lane 4), S ⌬ with R J (lane 5), S J with R J (lane 6), S J or S ⌬ alone (lanes 1 and 2, respectively), S J plus S ⌬ (lane 3), or R ⌬ or R J alone (lanes 8 and 9, respectively). The above experiments were also performed with the wild type strain C600 and gave results similar to those obtained with the RrmJ-overproducing strain (data not shown). These results show that S J but not S ⌬ methylates R ⌬ . An rRNA methyltransferase is thus present in the supernatant of RrmJ-producing strains but not in the RrmJ-deficient strain. This methyltransferase is likely to methylate ribosomes in vivo, since ribosomes from RrmJ-producing strains do not ac-cept labeled methyl groups in vitro.
RrmJ/FtsJ Purification-We purified RrmJ to confirm its proposed activity in vitro. A chimera protein of RrmJ fused to a chitin-binding domain was purified by affinity chromatography on a chitin affinity column. The CBD is separated from the RrmJ protein by an intein that undergoes self-cleavage by addition of 1,4-dithiothreitol, thus releasing the predicted wild type RrmJ protein from the chitin-bound intein tag (20, 21) (Fig. 2). AD494 pINTRRNJ induced by IPTG expressed a soluble protein of 78 kDa corresponding to the chimera CBD-RrmJ protein (Fig. 2, lanes 2 and 3). The protein released is of 23 kDa, corresponding to the expected molecular mass of RrmJ (lane 6).
RrmJ/FtsJ-dependent Methylation of the 23 S Ribosomal RNA-Ribosomes extracted from the RrmJ-deficient strain were incubated with [methyl-3 H]AdoMet, in the absence or presence of purified RrmJ. Incorporation of methyl groups into rRNA was analyzed by trichloroacetic acid precipitation and autoradiography of polyacrylamide gels. RrmJ is able to incorporate 0.8 pmol of methyl groups/pmol of ribosomes in trichloroacetic acid-precipitable material in 20 min, compared with 0.03 pmol of methyl groups/pmol of ribosomes incorporated in the absence of RrmJ (Fig. 3A). After the plateau has been attained, addition of more rRNA (22 pmol) leads to an additional methyl incorporation (18 pmol), whereas addition of more enzyme does not produce any effect (Fig. 3A). Analysis of the methylation products on a polyacrylamide gel reveals a major band at the position of 23 S RNA, and a weaker band around 16 S (Fig. 3B, lanes 2-5). When the S 150 supernatant from the RrmJ-overproducing strain is used instead of purified RrmJ, two bands migrating at similar positions are also observed (lane 8). However, the 23 S band is weak, and the 16 S band is strong; it is likely that the 16 S band reflects the action of RNases on 23 S RNA. The in vitro specificity of the methylation catalyzed by RrmJ is demonstrated by the inability of ribosomes from the RrmJ-overproducing strain and from the wild type strain to accept labeled methyl groups from [methyl-3 H]AdoMet (lanes 6 and 7, respectively). These results suggest that the in vitro methylation catalyzed by RrmJ is representative of a physiological methylation occurring in vivo. Interestingly, the weak incorporation of labeled methyl groups by ribosomes from the wild type strain (lane 7) suggests that RrmJ methylates these ribosomes in vivo at 37°C, even in the absence of heat shock (RrmJ is a heat shock protein, which is also expressed from a non-heat shock promoter; Ref. 25).
Several rRNA methyltransferases, such as RsmB (2), function equally well in EDTA as in the presence of magnesium ions whereas others, like RsmC, require magnesium ions for an efficient methylation reaction (9). The RrmJ methyltransferase shows a stringent dependence on magnesium concentration, with a peak centered around 2 mM magnesium, and its activity is severalfold lower in the presence of 10 mM EDTA or in the presence of 10 mM magnesium (data not shown). This result is consistent with the fact that RrmJ is more active on ribosomes rather than free rRNAs as the methylation substrate (see below).
RrmJ/FtsJ Is Active on Ribosomes but Not on Free rRNAs-Some rRNA methyltransferases, such as RsmB (2), methylate rRNA rather than ribosomes, while others, such as RsmC (9), are active on ribosomes but not on free rRNA. We compared the RrmJ-dependent transfer of methyl groups from [methyl-3 H]AdoMet to ribosomes and ribosomal RNAs from the RrmJdeficient strain. Analysis of the methylation reaction on a polyacrylamide gel (Fig. 4) shows that RrmJ actively incorporates methyl groups in the 23 S rRNA of ribosomal particles ( lanes  1-4), whereas there is no significant incorporation in the same amount of free rRNA (lanes [5][6][7][8]. Similar results were obtained by measuring the incorporation of radioactivity from [methyl-3 H]AdoMet in trichloroacetic acid-insoluble material (data not shown). Thus, the RrmJ methyltransferase strongly prefers ribosomes to rRNAs as the methyl group acceptor.
RrmJ/FtsJ Methylates the 23 S rRNA of 50 S Ribosomal Subunits-Purified RrmJ catalyzes efficient incorporation of methyl groups into trichloroacetic acid-precipitable material with 50 S ribosomal particle as substrate, whereas no incorporation occurs with 30 S ribosomal particles (Fig. 5). Similarly, the S 150 supernatant from the RrmJ-overproducing strain cat- alyzes incorporation of methyl groups into 50 S ribosomal particles (analyzed by autoradiography of a polyacrylamide gel), but not into 30 S ribosomal particles (not shown).
Identification of the Methylated Nucleotide by HPLC-50 S ribosomal particles were methylated in vitro by purified FtsJ/ RrmJ and [methyl-3 H]AdoMet, as described above. The 3 H-methylated 23 S rRNA was phenol-extracted, ethanol-precipitated, and digested with nuclease P1, snake venom phosphodiesterase I, and bacterial alkaline phosphatase. The resulting nucleoside mixture was analyzed by reverse-phase HPLC, and 5-min elution fractions were collected and counted for 3 H radioactivity, as described under "Experimental Procedures." As shown in Fig. 6, most of the 3 H radioactive material was located in fraction 6 corresponding to HPLC retention times between 25 and 30 min. This fraction was rechromatographied on the same HPLC column (1-min elution fractions were collected), and the radioactive peak eluted in the 27-28-min fraction (data not shown). Its elution time is similar to that of 2Ј-O-methyluri- H-methylated 23 S rRNA was phenol-extracted, ethanol-precipitated, and dissolved in distilled water. This [ 3 H]rRNA (0.7 pmol of methyl-3 H/ pmol) was digested with nuclease P1, snake venom phosphodiesterase I, and bacterial alkaline phosphatase The resulting nucleoside mixture was analyzed by reverse-phase HPLC, and 5-min elution fractions were collected and counted for 3 H radioactivity. The 3 H-methylated nucleoside was thus located in fraction 6 corresponding to the HPLC elution between 25 and 30 min retention time. The arrow shows the elution position of the 3 H-methylated nucleoside, which was identified as [ 3 H]Um, given that its retention time, UV absorption spectrum, and boronate affinity gel behavior are similar to those of the reference standard Um. dine, Um (27.3 min; data not shown). Furthermore, the radioactive fraction exhibited the same UV absorption spectrum as Um. Since 3-methyluridine (m 3 U) co-elutes with Um in our HPLC conditions, and also displays an UV spectrum similar to that of Um (data not shown), we loaded the radioactive fraction onto a boronate affinity column, which retains nucleosides containing an unmodified ribose, like m 3 U, but not nucleosides containing 2Ј-O-methylated ribose, such as Um. The radioactive nucleoside was not retained on the boronate column (data not shown), and could thus be identified as [ 3 H]Um.
Localization of the Site of Methylation-E. coli 23 S RNA contains three methylated uridines, located at positions 1915, 2449, and 2552 (3). Uridine 2552 has been characterized as Um, whereas the nature of the methylated uridines at positions 1915 and 2449 has not been determined (3). To determine which uridine is methylated by FtsJ/RrmJ, hybridization-protection studies were conducted using deoxyoligomers complementary to the RNA sequences spanning each of the methylated uridines. As shown in Fig. 7, oligomer 2552 (which spans the 2543-2562 RNA region) strongly protects the site of the 3 H-methyl group from RNase T1 digestion, whereas oligomer 1915 (which spans the 1906 -1925 RNA region) and oligomer 2249 (which spans the 2240 -2259 region) are totally ineffective. These results show that RrmJ/FtsJ is a 23 S RNA methyltransferase, which methylates uridine 2552 at the 2Ј OH of the ribose. DISCUSSION In the present study, we show that RrmJ methylates the 23 S RNA of the 50 S ribosomal subunit in vitro. The methylation reaction requires S-adenosylmethionine as donor of methyl groups, ribosomes from an rrmJ-deficient strain, and purified RrmJ or a S 150 supernatant from a RrmJ-producing strain. The requirement for ribosomes from a strain deficient in RrmJ as methyl acceptor suggests that the in vitro methylation described in this work reflects in vivo methylation of 23 S ribosomal RNA by RrmJ. The inability of ribosomes from wild type strains to act as substrate for in vitro methylation reactions has been reported by others (15). The fact that ribosomes from the wild type strain grown in normal conditions at 37°C are poor acceptors of methyl groups in vitro suggests that RrmJ methylates these ribosomes in vivo, even in the absence of heat shock. However, a residual ability to accept labeled methyl groups, compared with an inability of ribosomes from RrmJ-overproducing strains to do so, suggests that they are not completely methylated. This opens the possibility for a stress-dependent methylation of ribosomal RNA in vivo by the heat shock protein RrmJ. 23 S ribosomal RNA is clearly identified as the substrate of RrmJ methyltransferase; purified RrmJ incorporates methyl groups into a band which migrates as 23 S RNA, and 50 S ribosomal particles but not 30 S ribosomal particles can act as methyl acceptors. Consequently, the radioactive band around 16 S RNA observed in the presence of crude extracts probably results from a degradation of 23 S RNA into 16 S RNA. Such degradation of 23 S RNA has frequently been observed (1).
RrmJ methyltransferase is active on ribosomes or ribosomal particles, rather than on free RNA. This property is shared by RsmA and RsmC enzymes (9,32), and is true of most rRNAmodifying enzymes, with the exception of RsmB (2). RrmJmediated methylation requires magnesium; this requirement has also been reported for activity of other rRNA methyltransferases, and appears to be characteristic of reactions involving ribosomes or ribosomal particles (9).
Little is known about the physiological roles of ribosomal RNA methylation, since most experiments have been made in vitro, and since only few rRNA methyltransferases have been identified. Ribosomes reconstituted in vitro with completely unmodified rRNA are active, suggesting a minor role of RNA methylation in ribosome function. In contrast, mutations of 23 S ribosomal RNAs at several methyl-modified regions have dramatic effects on cell viability, reflecting the importance of these sites for proper ribosome function in vivo. FtsJ catalyzes a 2Ј-O-ribose methylation of Um-2552. This is a universally conserved nucleotide within the peptidyltransferase center of domain V of the large subunit rRNA (33). Um-2552 is adjacent to bases protected by tRNA bound in the A site (33), and mutations in one of these bases, U-2555, affects translational accuracy (34). The growth properties, translation properties, polysomes profile and antibiotic sensitivity of the ftsJ-deficient strain are under investigation.
In E. coli, the rluA, rluC, and rluD (35) genes encode pseudouridine synthase, which modifies 23 S rRNA. Curiously, an allele of rluD is a suppressor of a thermosensitive mutation in hflB, the downstream gene of the rrmJ-hflB operon. hflB encodes an ATP-dependent protease involved in the degradation of s32 and other substrates. These observations raise the possibility that the rrmJ-hflB operon might encode two proteins involved in ribosomal regulation.
RrmJ homologues are found in eubacteria, archea, and eukaryotes. However, among 24 completely sequenced microorganisms, only 9 of them encode putative RrmJ homologues. If the rRNA methylation mediated by RrmJ is conserved, it could be performed by methylases lacking obvious sequence similarities with RrmJ. Surprisingly, the JM 23 human protein encoded by chromosome X has 35% identity (52% conserved) over 192 amino acids with RrmJ, suggesting that, in at least certain higher organisms, rRNA methylation could be performed by an enzyme closely related to the E. coli RrmJ.