Molecular basis of RNA guanine-7 methyltransferase (RNMT) activation by RAM

Maturation and translation of mRNA in eukaryotes requires the addition of the 7-methylguanosine cap. In vertebrates, the cap methyltransferase, RNA guanine-7 methyltransferase (RNMT), has an activating subunit, RNMT-Activating Miniprotein (RAM). Here we report the first crystal structure of the human RNMT in complex with the activation domain of RAM. A relatively unstructured and negatively charged RAM binds to a positively charged surface groove on RNMT, distal to the active site. This results in stabilisation of a RNMT lobe structure which co-evolved with RAM and is required for RAM binding. Structure-guided mutagenesis and molecular dynamics simulations reveal that RAM stabilises the structure and positioning of the RNMT lobe and the adjacent α-helix hinge, resulting in optimal positioning of helix A which contacts substrates in the active site. Using biophysical and biochemical approaches, we observe that RAM increases the recruitment of the methyl donor, AdoMet (S-adenosyl methionine), to RNMT. Thus we report the mechanism by which RAM allosterically activates RNMT, allowing it to function as a molecular rheostat for mRNA cap methylation.

: RAM binds surface grove on RNMT but does not alter the active site. a) RNMT electrostatic potential plotted on the surface as calculated by APBS from the crystal structure (-5/+5 KbT/e). Positive charge is depicted as blue and negative charge as red. b) Alignment of residues involved in AdoMet binding for RNMT 165-476 -RAM 2-45 in cyan and Ecm1 in pink (PDB: 1RI3). For simplicity, only residues involved in polar interactions are highlighted as sticks, residues in black correspond to RNMT-RAM and in pink correspond to Ecm1. c) Time evolution of the methyl-N7 distance and SD-CE-N7 angle during a 50ns MD simulation. d) Time evolution of the K180 interaction with AdoMet carboxylate groups during a 50ns MD simulation.

X-ray diffraction data collection
The RNMT-RAM X-ray diffraction data were collected at 100 K at beamline ID24 -(Diamond synchrotron). The data were processed using XDS (37) and scaled using SCALA (38) from the CCP4 suite (39). The crystal diffracts to 2.35 Å and belongs to the space group P1. The cut-off criteria used were I/σ(I)>2.0 and Rmerge<50% for the outer shell as the CC(1/2) was not available due to the low redundancy of the collected data. The initial phase information was obtained by molecular replacement using Phaser from the CCP4 suite and the human catalytic domain RNMT as model (PDB ID: 3BGV). The initial densities were further improved by solvent flattening and histogram matching using RESOLVE (40), as implemented in the Phenix suite (41). The RAM molecule was built manually by iterative cycles of model building and refinement using COOT (42) and Refine (Phenix suite).
Diffraction data from the RNMT lobe deletion mutant and the thermolysin-cleaved RNMT were collected at 100K at beamlines ID30 (ESRF synchrotron) and ID04 (Diamond synchrotron) respectively. The structure determination method previously described for RNMT-RAM was used to solve and refine the structures of the RNMT lobe deletion / thermolysin structures. The crystals diffract to respectively 3.47 Å and 2.28 Å and belong to the space group P2 1 2 1 2 1 with CC(1/2) for the outer shell of 55% and 95% respectively. comprising a 0.2 ng/µL solution of leucine encephalin infused at 50µL/min, generating a reference ion at 556.2771. 2µL was injected onto a Waters Acquity BEH C18 column (50x2.1mm, particle size 1.7mm) held at 40C. Eluents used were A -0.1% formic in H2O

Recombinant protein purification for biochemical assays
(acidic) or 0.1% NH3 in H2O (basic), B -0.1% formic in ACN (acidic) or 0.1% NH3 in ACN (basic). The gradient was held at 100% A for 0.5min, then a linear gradient to 65% A after 4 min, then to 5% A after 5 min, held for 1 min before returning to 100% A. The flow rate was 0.5mL/min.

Simulation system setup
Using In addition, for the RNMT-RAM complex, a simulation system with ligands -Gppp and AdoMet -was built. Gppp was introduced to RNMT by aligning the protein with a Gpppbound structure of the E. cuniculi mRNA cap methyltransferase, Ecm1 (15) (PDB: 1RI2) and superposing Gppp to the RNMT active site. The force field parameters for Gppp and AdoMet were generated using the Antechamber utility in combination with the general AMBER force field (GAFF) (43). Atomic charges were assigned using the AM1-BCC method (44,45).
Each simulation system was immersed in a cubic box of TIP3P water molecules that extended at least 15 Å away from protein, and then was neutralized by adding the appropriate number of Clor Na + ions (46).

MD details and simulation protocol
All MD simulations were performed using the pmemd module of AMBER14. Periodic boundary conditions were used. The cut-off distance for the non-bonded interactions was 12 Å. Electrostatic interactions were treated using the smooth particle mesh Ewald method. The SHAKE algorithm was applied to all bonds involving hydrogens, and an integration step of 2.0 fs was used throughout (47) . Each system was first minimized (3,000 steps of steepestdescent energy minimization) to remove steric clashes, and then heated to 300/400 K over a 50 ps interval using a weak coupling algorithm (48). A 200 ps equilibration was performed to allow the solvent to redistribute around the positionally restrained protein.
The system was subsequently allowed to evolve freely unrestrained at constant temperature (300 K or 400 K) and pressure (1 atm) using the weak-coupling algorithm (48). Atomic coordinates were collected every 20 ps for further analysis.
The activation mechanism of RNMT involves protein-protein binding and large-scale conformational changes. It is a challenging task for standard atomistic simulations to properly sample such processes because of their long timescales and/or high conformational energy barriers. Therefore, along with MD simulations at a 'standard' temperature (300 K), hightemperature calculations (at 400 K) were performed to induce larger conformational changes in RNMT within reasonable simulation time. A total of 13 MD simulations were carried out: six systems (RNMT, RNMT-RAM, RNMT Δ419-458, RNMT R450E P452E, RNMT W178C A417C, and RNMT K409E K413E) were simulated at both 'standard' (300 K) and 'high' (400 K) temperature for 400 ns and 60 ns, respectively (Table S2). In addition, a 50 ns MD simulation (at 300 K) was carried out for the RNMT-RAM system with the ligands (AdoMet and Gppp) bound in the active site.

Analysis of MD trajectories
Protein structures and MD trajectories were visually inspected using the molecular visualization programs PyMOL (49) and VMD (50). Interatomic distances and angles, as well as root-mean-square deviations (rmsd) with respect to a reference structure, were monitored using the cpptraj module in AmberTools 15 (27 (51). The heatmap plots were created using the gnuplot program.

MD simulation (First nucleotide binding).
m7Gppp was introduced to the RNMT structure as described above. The phosphate chain was methylated to neutralize the additional charge without significantly restricting the freedom of movement during the optimization of the binding mod e. Restraints were set between the N7-Me and AdoHcy sulphur (160˚-180 ˚, S -N7: 3.6Å-3.9Å, S -C: 1.8 Å -2.2 Å) with weight 30 in MOE (52). The structure was minimized (Amber12:EHT) in a water box with progressively weakening heavy atom restraints and relaxed further in a short 1.2 ns MD simulation. The first nucleotide guanine was added to m7Gppp using a restrained docking in Glide (53) and m7GpppG binding was optimized once more as described above.