Small methyltransferase RlmH assembles a composite active site to methylate a ribosomal pseudouridine

Eubacterial ribosomal large-subunit methyltransferase H (RlmH) methylates 23S ribosomal RNA pseudouridine 1915 (Ψ1915), which lies near the ribosomal decoding center. The smallest member of the SPOUT superfamily of methyltransferases, RlmH lacks the RNA recognition domain found in larger methyltransferases. The catalytic mechanism of RlmH enzyme is unknown. Here, we describe the structures of RlmH bound to S-adenosyl-methionine (SAM) and the methyltransferase inhibitor sinefungin. Our structural and biochemical studies reveal catalytically essential residues in the dimer-mediated asymmetrical active site. One monomer provides the SAM-binding site, whereas the conserved C-terminal tail of the second monomer provides residues essential for catalysis. Our findings elucidate the mechanism by which a small protein dimer assembles a functionally asymmetric architecture.

anticodon stem loop of tRNA, with K162 -positioned similarly to R154 of RlmH -stabilizing the backbone of adenosine 38, whose neighbor guanosine 37 flips out for modification.
The SAM-binding bights 1 and 3 from one monomer interact with the same bights of the second monomer via hydrophobic interactions (I74, L122, L125 and L127) ( Figure S1). This hydrophobic patch is conserved among RlmH orthologs (Fig. 1). The head-to-tail dimerization of RlmH contrasts the tail-to-tail dimerization of archaeal and eukaryotic counterparts responsible for m 1 Ψ or m 1 acp 3 Ψ methylation [48][49][50] , which resembles the bacterial SpoU sub-family of SPOUT methyltransferases ( Figures S5 and S6). Yet all SPOUT family members retain the over-hand-knot architecture comprising three bights forming the SAM binding pocket.
Cofactor binding rearranges the RlmH C-terminus. Both cofactor-bound (this work) and ligand-free (PDB ID 1NS5) RlmH form dimers. Dimerization buries ~13% (~1,100 Å 2 ) 51 of the total surface area of each monomer. In the RlmH · SAM dimer, the SAM-binding site of monomer A lies next to the C-terminus of monomer B, forming the putative helix 69-binding site. Comparing the structure of SAM-bound and ligand-free enzyme (Fig. 1) shows that cofactor binding rearranges the C-terminal loop into a 3 10 -helix in each subunit (Figs 1 and S3). Formation of the 3 10 -helix was not observed in the recent structures of SAM-bound OrfX, the Staphylococcus aureus homolog of RlmH 45 . The OrfX · SAM complex contains a disordered C-terminal loop and a phosphate ion bound next to the methyl group of SAM. The phosphate ion has been proposed to occupy the putative position of the phosphate backbone of the substrate rRNA nucleotide 45 , suggesting that the C-terminal tail may undergo an additional rearrangement when RlmH binds the ribosome. Thus, the C-terminal tail appears to play a role in the catalytic mechanism of RlmH, pre-arranging the C-terminal residues for RNA binding or positioning the SAM methyl group toward the RNA binding pocket, as observed for RlmH · SAM in this study (Fig. 1).
E138 and all C-terminal mutants are catalytically inactive. The catalytic residues of RlmH have not yet been identified 45 . Alignment of 5,000 non-redundant sequences of RlmH homologs reveals that the SAM-binding pocket residue H129, residues E138 and R142 near the methyl group of SAM and the C-terminal residues Y152, H153 and R154, as evinced in the crystal structure of RlmH · SAM, are widely conserved ( Figure S4). To test the functional importance of these amino acids, we generated mutant proteins and measured their catalytic activities. In particular, we sought to test whether active site pre-assembly requires Y152, H153 and R154 in the C-terminus and whether residues H129, E138 and R142 are critical for catalysis.
We purified seven single-amino-acid mutant RlmH proteins: H129A, E138A, E138Q, R142A, Y152F, H153F and R154A. The isosteric E138Q mutant was designed to test the importance of the carboxylate anion of E138 in catalyzing methyl transfer. Y152F tests the role of the hydroxyl group, which interacts with the carboxyl groups of E138 and the guanidinium of R142, in organizing the active site. H153F tests the role of the histidine hydrogen bond with S145, which connects the 3 10 -helix with α5 in the SAM-bound enzyme (Fig. 1).
None of the three C-terminal or E138 mutants (E138A, E138Q) mutants retained detectable catalytic activity. The failure of glutamine 138 to catalyze methyl transfer suggests that RlmH activity requires the acidic carboxyl group of the native glutamic acid.
Replacement of H129 or R142 with alanine reduced methyltransferase activity. H129A retained ~50% of the wild-type methylation activity, while R142A mutant protein had a K M for ribosomes ≥20 times greater and a k cat /K M ≥ 30-fold lower than wild-type RlmH (Table 1).

Wild-type and all mutant RlmH form dimers.
Loss of catalytic activity in the E138, Y152, H153 and R154 mutants might reflect (1) a failure to form dimers; (2) loss of substrate(s) binding; or (3) inactivation of the active site. We sought to distinguish among these explanations. First, we measured the oligomeric state of purified wild-type and mutant RlmH in solution using size-exclusion chromatography combined with multi-angle light scattering (SEC/MALS 53,54 ). The protein was dialyzed to remove endogenous ligand(s) that may co-purify with RlmH. SEC/MALS measures molecular mass irrespective of the shape of a protein or complex. At both ~20 and ~200 µM, each of the five proteins (wild-type, E138, Y152, H153, and R154) existed as a single peak of monodisperse particles with a molecular weight corresponding to a dimer ( Fig. 1 and Table S1). Conventional size exclusion chromatography suggests that the two other mutants, H129A and R142A, similarly retain the ability to dimerize, because the two mutants had elution profiles identical to that of wild-type RlmH (data not shown). We conclude that all the mutants analyzed here retain the dimeric state of wild-type RlmH.
RlmH mutants retain SAM binding. We next used isothermal titration calorimetry (ITC) to test whether each RlmH mutant retained wild-type affinity for binding SAM (Figs 3 and S6). For the mutant proteins, the apparent equilibrium dissociation constants for the first SAM binding site in the RlmH dimer (K D1 ) ranged from 3.0 ± 0.1 (for H129A) to 8.5 ± 0.7 µM, similar to that of wild-type RlmH (5.6 ± 0.3 µM; Table 2). The modestly increased SAM binding of the H129A mutant is likely due to non-polar environment in this region bolstered by the adjacent proline residues. The enhanced binding, however, comes at the cost of modestly decreased catalytic activity.
K D1 of wild-type and mutant RlmH are similar to those of other SPOUT methyltransferases: 4.5 µM for SpTrm10 55 ; 7.6 µM for ScTrm10 55 ; 20.0 µM for HiYibK/TrmL 56 ; 33.0 µM for SaTrm10 57 . The second binding site in the RlmH dimer remained unoccupied at SAM concentrations below 500 µM, consistent with the K D2 ~1 mM reported for OrfX 45 . A millimolar K D2 is consistent with our observation of two SAM molecules per dimer in the crystal structure, which was obtained using 2 mM SAM in the crystallization solution. None of the mutated residues directly contacts SAM in our structure (Fig. 1), consistent with the lack of effect of the mutations on SAM binding. We conclude that the loss of activity in mutant RlmH does not reflect a SAM-binding defect.

R154A mutant is deficient in ribosome binding.
To test whether the loss of catalytic activity in the RlmH mutants reflected a failure to bind ribosomes, we performed a competition assay in which the methylation efficiency of wild-type RlmH was measured in the presence of mutant proteins. If a mutant protein does not bind the ribosome, the efficiency of methylation by wild-type RlmH will be unaltered. In contrast, if a mutant protein retains near wild-type ribosome binding, competition of RlmH with the catalytically-inactive, mutant protein should reduce the efficiency of methylation. The competition assay was performed under single-turnover conditions, in which the concentration of wild-type RlmH was higher than the concentration of 70S ribosomes; varying concentrations of each catalytically inactive mutant were incubated with ribosomes before the addition of wild-type RlmH (Fig. 4).  Of the five inactive mutants, four inhibited wild-type RlmH. The two E138 mutants, Y152F, and H153F halved the extent of methylation at ~3 µM mutant concentration; with 20-40 µM mutant protein, no methylation by wild-type RlmH was detected. Thus, none of these four mutations impaired ribosome binding. At 5 µM, the R154A mutant had little effect on methylation by RlmH, and at 40 µM reduced methylation by just two-fold. Thus, mutation of R154 compromised ribosome binding. The presence of some ribosome binding at the higher concentration suggests that the R154A mutant is deficient in both ribosome binding and catalysis.

Discussion
Our data suggest that E. coli RlmH functions as a dimer, with the SAM binding pocket of one monomer and the C-terminal residues of its counterpart forming a composite active site. We propose that the C-terminal tail    Figure S7.
provides a residue that acts as a general base (see discussion below), while the other conserved residues form a network of interactions that organize the catalytic site. Like other SPOUT methyltransferases 58 RlmH forms a dimer at concentrations at which RlmH efficiently methylates the 70S ribosome (Figs 1 and 2). Despite differences in sequence, dimer topology and substrate specificity, the conserved overhand-knot architecture of SPOUT family members enables a dimer to position the SAM binding pocket of one monomer next to the catalytic residues of the second monomer. Such architecture of the composite active site appears to be a general feature of SPOUT methyltransferases.
Comparison of the SAM-bound RlmH structure with the crystal structure of apo-RlmH (PDB ID 1NS5) supports the earlier proposal 31, 45 that the RlmH catalytic pocket, comprising the SAM-binding and Ψ1915-binding sites, lies at the interface of two monomers (Fig. 5). Our SAM-bound structure reveals that SAM binding shifts the C-terminus of RlmH by as much as 14 Å, rearranging it into a 3 10 helix (Fig. 1, close-up). No other large rearrangements were observed. Kinetic analyses of RlmH mutants identify the conserved E138 and C-terminal residues Y152, H153 and R154 as essential for catalysis. The side chains of these amino acids lie ≥8 Å from the methyl group of SAM. Recent structures of a related tRNA-methyltransferase, TrmD, suggest that the proposed general base D169 can shift from ~7 Å to ~4.5 Å toward the methyl moiety of SAM upon tRNA binding 47 . Superposition of our RlmH structure with that of TrmD reveals that of the four catalytically critical residues in RlmH, Y152 is closest to D169 of TrmD (Figs 5 and S8). Loss of RlmH activity upon removal of the tyrosine hydroxyl group in the Y152F mutant is consistent with this residue serving as a general base, echoing the roles of catalytic tyrosines in some non-SPOUT methyltransferases 59,60 . We propose that the interaction of Y152 with R142, which itself forms a salt bridge with E138, serves to perturb the pK a of Y152, as would be required for this residue to act as a general base. Our mutational studies support a role for interactions within the E138-R142-Y152 triad in this process, as RlmH activity is substantially reduced in the R142 and E138 mutants. E138, R142 and Y152 are conserved widely among RlmH orthologs 28,31,37 .
H153 is also conserved, and an H153F mutant RlmH lacks detectable methyltransferase activity. The H153 side chain is rotated away from the methyl group of SAM and makes a hydrogen bond with the hydroxyl group of residue 145, which is always serine or threonine in RlmH orthologs. In apo-RlmH, the H153 side chain is oriented toward solvent and does not contact S145. Therefore, H153 likely stabilizes the formation of the C-terminal 3 10 -helix, which helps organize the RlmH active site. The penultimate residue of RlmH, R154, is replaced by lysine or histidine in some RlmH orthologs. In a small subset of RlmH orthologs, H153 is the terminal residue; these enzymes completely lack R154 and E155 (E. coli numbering). Our data suggest that R154 plays role in RNA binding or positioning: an R154A mutant was defective in ribosome binding and lacked catalytic activity even when its low affinity for ribosomes was compensated by using a high protein concentration (Fig. 4). In RlmH orthologs ending with H153 perhaps another residue adopts the role of R154.
The SPOUT methyltransferase superfamily contains proteins with diverse substrate specificities, primary structures, and lengths. RlmH comprises just 155 amino acids (~17 kDa), whereas Trm3p contains 1436 amino acids (~164 kDa). Yet all SPOUT methyltransferases form a common overhand knot structure. In addition to this overhand-knot architecture, all SPOUT methyltransferases use SAM as the methyl donor. It is therefore likely that the overhand knot architecture has been evolutionarily constrained by positive selection to retain the ability to bind SAM, particularly SAM in a bent conformation 58,61 . Structural alignment of bacterial and archaeal SPOUT methyltransferase structures containing SAM or SAM-like ligands reveals that the three SAM pocket-forming bights of RlmH are also present in other methyltransferases, albeit with variable lengths and topologies ( Figures S5 and S6). Bight 1, the least conserved among the three, contains a conserved threonine that interacts with the adenine moiety of SAM. Bight 2 contains a canonical glycine-rich motif ( Figure S9), which is also found in Rossman-fold enzymes 62 . Bight 3 lies adjacent to helix α5 (in RlmH), which participates in both dimerization and catalysis by displaying E138 and R142 on the same side of the helix. A similar arrangement of glutamic or aspartic acid followed by an arginine or lysine a few residues away ( Figures S5 and S6) exists in other bacterial SPOUT methyltransferases, consistent with the functional importance of this motif. The role of the general base, however, is likely played by different residues in SPOUT subfamilies. While larger methyltransferases employ an aspartate or glutamate, such as D169 in TrmD, RlmH lacks a similarly positioned conserved acidic side chain. Instead, a conserved tyrosine is positioned in the same area of the catalytic pocket ( Figure S8), supporting its proposed role in catalysis. Closely positioned R142 and H129 may be involved in catalysis. Their mutations to alanine, however, only modestly reduce k cat , suggesting that the direct role is unlikely unless a structural reorganization or a water molecule compensates for the loss of the side chain function. It is also possible that the less conserved E155 or the C-terminal α carboxyl group of RlmH acts as the general base. The general-base role of the terminal α carboxyl group has been shown for some unrelated enzymes, such as phospholipase 63,64 and N-myristoyltransferase 65 .
The RlmH dimer allows formation of the SAM binding pocket within one monomer, while the second monomer provides its C-terminal tail to complete the catalytic site (Fig. 5). Monomer B also stabilizes the SAM binding pocket of monomer A: SAM-binding bights 1 and 3 from one monomer interact with the same bights of the second monomer via hydrophobic interactions that are conserved among RlmH orthologs (Figs 1 and S2). The mechanism of RlmH provides new insights into the evolution and conservation of the overhand-knot SPOUT methyltransferases dimers and small-protein catalysts in general. Our structures suggest that SAM binding "pre-arranges" the C-terminal tail for binding of the RNA substrate and catalysis. Structures of ribosome-bound RlmH are required to provide additional insights into the mechanism of substrate recognition and catalysis.

Construction and purification of recombinant wild-type and mutant RlmH. rlmH from E. coli
JW0631 (ASKA Clone of the National Bio Resource Project, Keio University, Tokyo, Japan) was cloned into pET24b (Novagen, Darmstadt, Germany), using primers shown in Table S2. Gene sequencing confirmed proper subcloning of the rlmH gene, in which the first methionine is replaced by MHHHHHHV. The His 6 -tagged RlmH protein was expressed in BLR(DE3) cells (Novagen) at 37 °C. 1 mM isopropyl β-D-1-thiogalactopyranoside (IPTG) was added at OD 600 = 0.8. Following overnight growth at 16 °C, cells were harvested, suspended in buffer A (50 mM Tris-HCl, pH 7.5, 300 mM KCl, 10 mM imidazole, 5 mM β-mercaptoethanol), lysed using a microfluidizer at 18,000 psi, and clarified by centrifugation at 38,397 × g for 20 min. This clarified lysate was incubated with 1 ml of nickel resin for 1 h at 4 °C. Using batch/gravity flow immobilized metal ion affinity chromatography (IMAC), lysate with the affinity matrix resin were then packed into a pierce centrifuge column. Following binding of the His 6 -tagged protein, the column was washed with buffer A. RlmH was eluted with buffer B (50 mM Tris-HCl, pH 7.5, 300 mM KCl, 250 mM imidazole, 5 mM β-mercaptoethanol). Eluted samples were buffer-exchanged into buffer A prior to the next purification step. Samples (in buffer A) were further purified by an additional IMAC purification step using a fast protein liquid chromatography system (GE Healthcare) and a 5 ml HisTrap column (GE Healthcare) with linear gradient elution up to 250 mM imidazole. Protein was further purified using size exclusion chromatography on a Superdex75 column (GE Healthcare). Gel filtration buffer contained 50 mM sodium diphosphate, pH 7.5, 300 mM KCl, 10% glycerol, 5 mM β-mercaptoethanol. Purified recombinant protein was concentrated using a 10-kDa cutoff centrifugal filter concentrator (Millipore). Prior to in vitro experiments, protein was dialyzed overnight in buffer C (50 mM sodium phosphate, pH 6.8, 200 mM KCl, 5% glycerol, 1 mM β-mercaptoethanol) to remove endogenously bound SAM or other ligands. Protein was stored at 4 °C.
Site-directed mutagenesis was performed using QuikChange II site-directed mutagenesis protocol (Stratagene, La Jolla, CA, USA). Seven mutants of RlmH were generated (H129A, E138Q, E138A, R142A, Y152F, H153F and R154A) by PCR using plasmid pET24b-RlmH as a template in combination with mutagenic primers (Table S2). Mutations were confirmed by DNA sequencing. Mutant proteins were expressed and purified as the wild-type protein.
Size exclusion chromatography coupled with static laser light scattering. Dialyzed recombinant proteins were buffer-exchanged into the size-exclusion chromatography/multi-angle light scattering (SEC/ MALS) running buffer (50 mM potassium phosphate, pH 7.0, 300 mM KCl, 2 mM DTT and 5% glycerol). Two amounts (0.35 mg/ml in 50 µl and 3.85 mg/ml in 550 µl) of each protein sample were subjected independently to SEC/MALS analysis. Each filtered protein sample was subjected to a Superdex S-200 HR10/30 column (GE Healthcare) connected to an Agilent 1200 High Performance Liquid Chromatography System equipped with an auto sampler (Agilent Technologies, Wilmington, DE, USA). The eluate from SEC was monitored by a photodiode array (PDA) UV/VIS detector, UV, (Agilent Technologies), Optilab rEX differential refractometer, (Wyatt Technology, Santa Barbara, CA, USA), static and dynamic, HELEOS II multi-angle laser light scattering detector with QELS capability (Wyatt Technology). The system was equilibrated in SEC/MALS running buffer (50 mM potassium phosphate, pH 7.0, 300 mM KCl, 2 mM DTT and 5% glycerol) at a flow rate of 0.5 ml/min. Weighted average molecular weights were determined across the entire elution profile using 1 sec intervals as previously described 66  4,000 × g for 20 min. The resulting cell pellet was resuspended in 30 ml of buffer A (20 mM Tris-HCl, pH 7.0, 10.5 mM MgCl 2 , 100 mM NH 4 Cl, 0.5 mM EDTA, 6 mM β-mercaptoethanol). Cells were lysed in a microfluidizer (Microfluidics, Westwood, MA, USA) at 18,000 psi and clarified by centrifugation at 38,397 × g for 20 min at 4 °C. The resulting supernatant was centrifuged again to yield the clarified lysate. The lysate was layered onto an equal volume of 1.1 M (37.7%) sucrose cushion in buffer B (20 mM Tris-HCl, pH 7.0, 15 mM MgCl 2 , 500 mM NH 4 Cl, 0.5 mM EDTA, 6 mM β-mercaptoethanol), and centrifuged at 214,573 × g for 16 h in a Ti45 rotor (Beckman Coulter, Danvers, MA, USA). Ribosome pellets were drained and quickly rinsed with 5 ml buffer C (20 mM HEPES · KOH pH 7.0, 6 mM magnesium acetate, 30 mM NH 4 Cl, and 6 mM β-mercaptoethanol). Each pellet was then resuspended in 5 ml of buffer C, transferred to 1.5 ml tubes, and clarified by centrifuging at 17,000 × g for 5 min. The supernatant was diluted with 40 ml buffer B, and the final concentration of NH 4 Cl was adjusted to 500 mM. This solution was loaded into Ti70 tubes and topped up with buffer B, centrifuged at 351,046 × g for 2 h in a Beckman Ti70 rotor. The pellet was resuspended in 0.5 ml of buffer C and layered onto a 10-35% (w/w) sucrose gradient in buffer C, then centrifuged at 71,999 × g for 13 h in a SW-28 rotor (Beckman Coulter). Gradients were analyzed by continuous monitoring of the absorbance at 254 nm. The 70S ribosome peak (~5 ml) was collected and buffer exchanged to buffer C using a 100-kDa cutoff centrifugal filter concentrator (Millipore, Billerica, MA, USA). The resulting partially purified ribosomes were subjected to a second round of 10-35% (w/w) sucrose gradient purification to remove contaminating 50S ribosomes. The purity of the ΔrlmH 70S ribosomes was verified by analytical sucrose gradient (centrifugation at 273,865 × g for 2 h in a Beckman SW-41 rotor). Purified ribosomes in buffer C were flash frozen in liquid nitrogen and stored as small aliquots at −80 °C. Substrate binding competition assay. Substrate binding competition assays were performed as described for the methylation assay except that a range of concentrations of mutant protein was incubated with activated ribosomes for 20 min at 25 °C before wild-type RlmH · [ 3 H]-SAM was added.

In vitro methylation assays. ΔrlmH
Rate analyses. The rate of methyl group incorporation into ΔrlmH E. coli 70S ribosomes was calculated by fitting to y = y 0 + A(1 − e −kt ), where dy/dt = Ake −kt . The initial velocity dy/dt 0 = Ak 67 . Michaelis-Menten kinetics were determined using GraphPad Prism 6.0 g (GraphPad Software, San Diego CA USA) and values are reported in Table 1. Since the fitted K M for RlmH is lower than the concentration of the substrate (≤0.3 uM), the <or> symbols are used to reflect the uncertainties in k cat /K M or Fold Changes. Burst kinetics data from the substrate binding competition assays were fit using Igor Pro 6.11 (WaveMetrics, Inc., Portland, OR, USA).
Protein crystallization and data collection. Purified RlmH was concentrated to 10 mg/ml in the buffer containing 50 mM sodium diphosphate, pH 7.5, 300 mM KCl, 10% glycerol, 5 mM β-mercaptoethanol. Initial crystallization hits were obtained using Crystal Screens 1-2 (Hampton Research, Aliso Viejo, CA, USA) and Morpheus Screen (Molecular Dimensions Ltd., Newmarket, Suffolk, United Kingdom) using sitting-drop vapor diffusion in 96-wells plates. SAM-and sinefungin-bound RlmH crystals were obtained with Morpheus screen solution 21 containing 0.1 M Tris-bicine buffer (pH 8.5), 0.09 M NaF, 0.09 M NaBr, 0.09 M NaI, 20% v/v PEG 550 MME, 10% w/v PEG 20,000 in the presence of 2 mM SAM or sinefungin. With this initial hit condition, different protein to crystallization solution ratios were screened in 24-well plates. Larger crystals suitable for data collection were obtained in 24-well plates, via hanging-drop vapor diffusion, after 1 week at 16 °C. Drops were formed by mixing 2 µl protein and 2 µl crystallization solutions. Crystals were cryo-protected with 20% glycerol in reservoir solution and flash-cooled in liquid nitrogen.
Crystals were screened at beam line 24-ID-B and diffraction data were collected at beam line 24-ID-C at the Advanced Photon Source (APS) of Argonne National Laboratory (Argonne, IL, USA). Data were processed and integrated using XDS 68 , scaled with SCALA 69 or AIMLESS 70 in CCP4 71 . Crystals belonged to space group P2 1 and the asymmetric unit contained 4 molecules (2 dimers).
Structure determination and refinement. -Structures were solved by molecular replacement using MOLREP 72 in CCP4 71 and PHASER 73,74 in PHENIX 75 using the apo E. coli RlmH crystal structure (PDB ID 1NS5) as the starting search model. Structural models were visualized and remodeled with COOT 76 . Refinement was performed using REFMAC5 77 and PHENIX 1.9_1692 75 . Individual B-factors were refined together with each chain as TLS group. Final models were evaluated using the comprehensive validation option in PHENIX 75 and yield good crystallographic and stereochemical statistics 78 (Table 3) Table 3. Data collection, processing and refinement statistics. Statistics for the highest-resolution shell are shown in parentheses. a R merge and R meas from PHENIX are reported 75 . b CC 1/2 is the correlation coefficient of the mean intensities between two random half-datasets 83 . CC* is an estimate of the "true" CC of the data under examination to the (unknown) true intensities 83 . c 5% of the reflections were selected for free R calculation. d Root-mean-square deviation from ideal values 84 . e MolProbity score was calculated as defined in ref. 78.