Dynamin-2 R465W mutation induces long range perturbation in highly ordered oligomeric structures

High order oligomers are crucial for normal cell physiology, and protein function perturbed by missense mutations underlies several autosomal dominant diseases. Dynamin-2 is one of such protein forming helical oligomers that catalyze membrane fission. Mutations in this protein, where R465W is the most frequent, cause dominant centronuclear myopathy, but the molecular mechanisms underpinning the functional modifications remain to be investigated. To unveil the structural impact of this mutation in dynamin-2, we used full-atom molecular dynamics simulations and coarse-grained models and built dimers and helices of wild-type (WT) monomers, mutant monomers, or both WT and mutant monomers combined. Our results show that the mutation R465W causes changes in the interactions with neighbor amino acids that propagate through the oligomer. These new interactions perturb the contact between monomers and favor an extended conformation of the bundle signaling element (BSE), a dynamin region that transmits the conformational changes from the GTPase domain to the rest of the protein. This extended configuration of the BSE that is only relevant in the helices illustrates how a small change in the microenvironment surrounding a single residue can propagate through the oligomer structures of dynamin explaining how dominance emerges in large protein complexes.


Supplemental Figure
: Dynamin-2 dimer models. Three replicates were run for each dimer system. G domains are in blue, bundle signaling elements (BSEs) in yellow, stalks in red, and PH domains in green. Final dynamin-2 dimer conformations of each replicate after 200 ns of simulation of the WT (A), HT (B), and FM (C) systems. The residue 465 is displayed as cyan dot. Measurement of the G-BSE (D) and BSE-stalk (E) angles showed that there are no significant structural differences between the three conditions (One-way ANOVA). The vector used to calculate these angles were the same that those we used in the Coarse-grained analysis to measure the G-BSE and BSE-stalk angles (see Results and Discussion section).

Full-atom molecular dynamics simulations
For full-atom MD, we mutated the WT protein by using the Mutator plugging of Visual Molecular Dynamics 6 . The arginine 465 was exchanged for a tryptophan. Next, we built the WT HT and R465W dimer systems. Each system was added to a water box. The hydrated systems were neutralized with NaCl at a concentration of 150 mM. The full systems were relaxed through molecular dynamics (MD) simulations using the AMBER suit 7 . MD was performed with the ff14SB force field 8 . The temperature was set at 310 K using the Langevin thermostat with the isobaric-isothermal (NPT) ensemble. We fixed the pressure at 1 atm. The integration of the equation of motion of Newton was performed with the verlet algorithm using a 2 ps timestep. For non-bonded interactions, a cutoff of 10 nm was used. Each system was subjected to energy minimization using periodic boundary conditions. Three replicas of each system were run for 200 ns.

Free binding energy calculation
Free Binding Energy calculations were obtained with the MM/GBSA by using AmberTools18 7 .

Coarse-grained Molecular Dynamics simulations
We used the human dyn-1 coordinates (PDB ID: 6DLU) 9 along with a dyn-1 helix kindly provided by Dr. Jenny Hinshaw as a template to build the dyn-2 helices. We built: a dyn-2 helix with 56 WT proteins (WT helix); a helix composed of 28 WT dynamins and 28 mutated intercalated dynamins (HT helix), and a mutated helix consisting of 56 mutant dynamins (R465W helix). To reduce the number of particles, after vacuum minimization, we converted full-atom helices into coarse-grained (CG) models 10 using Martini force field v2.2. These CG models were reduced to ~500.000 particles. We applied elastic networks to ensure the CG dyn-2 proteins behave properly. We added a GTP and a magnesium ion on each dynamin-2 G-domain. Each GTP and Mg+2 were under distance restraints to avoid the separation from the G-domain. GTP molecules were composed of SC1 (TN0), SC2 (TG2), SC3 (TG3), SC4 (TNa), BB2 (SN0), BB3 (SC2), and three BB1 (Q0) coarse-grained beads (Uusitalo et al., 2015), and a MG+ (Qd) beads for Mg +2 . We also added a nanotube with an outer diameter of 13.7 nm composed of lipid heads to reduce computational cost and simulate the neck of the vesicle during endocytosis. The glycerol molecules from the lipid nanotube were under position restraints of 1000 kJ/mol/nm 2 during the simulations. We minimized each system in vacuum for 50.000 steps. Next, each system was solvated with non-polarizable CG water molecules BP4 11 and ionized them with 150 mM of NaCl. To prevent freezing of water molecules, we added anti-freeze BP4 beads 11 . Then, each system was submitted to 5 μs of MD simulation in an NPT ensemble. We fixed the temperature at 310 K using the v-rescale thermostat and the pressure was fixed at 1 atm using the Parrinello-Rahman barostat 12 . Electrostatics interactions were calculated using the Reaction-field method. Van der Waals interactions were calculated using the Verlet cutoff-scheme. The Newton's equation of motion was integrated with the leapfrog algorithm with a timestep of 10 fs using Gromacs software v.5.0 13 for all the CG simulations carried out with periodic boundary conditions.

Solvent Accesible Surface Area Calculation
To calculate the SASA, the radius of the amino acid, that is Arg465 or Trp465, were used to find the points of the residue exposed to the solvent. Only the points near the residue Arg465 or Trp465 were considered. To prevent that internal pockets or protein voids affects SASA calculation, we used the restrict option. The SASA calculations use as a reference the SASA of a specific residue between 2 alanine by each side, which means to use the sequence AAWAA, which represents the 100% of the exposition of a residue in water.

Statistics
For comparing the three different conditions we used ANOVA analysis and Turkey posttest for data with a normal distribution, and Kruskal-Wallis and Dunn's post-test for non-normal distribution datasets. We plotted the data with R Studio software.