A GAP‐GTPase‐GDP‐Pi Intermediate Crystal Structure Analyzed by DFT Shows GTP Hydrolysis Involves Serial Proton Transfers

Abstract Cell signaling by small G proteins uses an ON to OFF signal based on conformational changes following the hydrolysis of GTP to GDP and release of dihydrogen phosphate (Pi). The catalytic mechanism of GTP hydrolysis by RhoA is strongly accelerated by a GAP protein and is now well defined, but timing of inorganic phosphate release and signal change remains unresolved. We have generated a quaternary complex for RhoA‐GAP‐GDP‐Pi. Its 1.75 Å crystal structure shows geometry for ionic and hydrogen bond coordination of GDP and Pi in an intermediate state. It enables the selection of a QM core for DFT exploration of a 20 H‐bonded network. This identifies serial locations of the two mobile protons from the original nucleophilic water molecule, showing how they move in three rational steps to form a stable quaternary complex. It also suggests how two additional proton transfer steps can facilitate Pi release.

. Data  Crystallization and Structure refinement.
Crystallization procedures for RhoGAP-RhoA-GDP-H2PO4complex. The initial crystals of RhoGAP-RhoA-GDP-AlF4were obtained using the crystallization conditions that are similar to the ones used for pdb 1tx4: 20 % PEG2000 MME (Sigma), 110 mM (NH4)2SO4, 100 mM MES, pH 6.0, 10 mM MgCl2, 3 mM DTT, 3 mM NaN3, 20 mM Tris-HCl pH 7.4, with 1 mM AlCl3 and 10 mM NH4F. Crystals were then transferred to a drop of the washing buffer containing 20% PEG2000 MME, 110 mM (NH4)2SO4, 100 mM MES, pH 6.0, 3 mM DTT, 3 mM NaN3, 20 mM Tris-HCl, pH 7.4, and 3 mM deferoxamine (DFO) for 2 h. This step was repeated, as necessary for complete depletion of aluminum and minimization of magnesium fluoride binding in the crystal. At this stage, selected crystals were flash-cooled with liquid nitrogen for data collection after being dipped into cryoprotectant (20% PEG2000 MME, 110 mM (NH4)2SO4, 100 mM MES and 20% glycerol). Diffraction data latterly showed a RhoGAP-RhoA-GDP-MgF3complex formed in the crystal ( Figure S2a). The remaining crystals were then transferred into a drop of the washing buffer with supplement of 10 mM MgCl2, and 200 mM NH4H2PO4 at a final pH of 5.5 for further soak for 1 h. This step is required to form RhoGAP-RhoA-GDP-H2PO4complex. The same cryoprotectant was used (20% PEG2000 MME, 110 mM (NH4)2SO4, 100 mM MES, and 20% glycerol) before these crystals were flash-cooled with liquid nitrogen.
Diffraction data were collected at 100 K with wave length 0.91731 Å, integrated with XDSGUI [1] and scaled with Aimless in CCP4i. [2] The RhoGAP-RhoA-GDP-H2PO4structure (PDB 6r3v) was solved by molecular replacement with MOLREP [3] in Collaborative Computational Project Number 4 suite (CCP4i) [2] using the coordinates of RhoGAP-RhoA-GDP-AlF4 -(PDB 1tx4) as the search model and refined by REFMAC5 [4] using the maximum likelihood method. The model building was interactively done by using σA-weighted electron density maps with coefficients 2Fo-Fc and Fo-Fc in COOT. [5] To confirm the electron density is that for a tetrahedral phosphate, the electron density of the phosphate (PO4) was explored by alternative refinement using PDB ligands "MGF" and "HOH" (trigonal planar MgF3 with an in-line H2O) as the model with the maximum Mg-F bond length strain (1.95 Å). This showed a region of positive difference density in the Fo-Fc map over the Mg atom ( Figure S2b) with a much smaller B factor (32.6 for Mg versus 40.4, 36.3, 36.3 for the three fluorines). When the bond length strain on the MGF was set to minimum, after 10 further rounds of refinement, the MGF became distorted to tetrahedral geometry and all three 'Mg-F' bonds were refined as being shorter than 1.6 Å ( Figure S2c).
The RhoGAP-RhoA-GDP-MgF3structure from the DFO depletion procedure has a resolution of 1.87 Å and was refined in the same way, but only for the purpose of comparison without deposition to Protein Data Bank to avoid duplication of 1ow3 ( Figure S2a).

Crystallization procedures for Rho-GDP complex.
RhoA-GDP crystals were obtained at 293 K by the sitting-drop method from solutions consisting of 10 mg mL -1 RhoA, 50 mM MES, pH 5.5, 150 mM NaCl, 5 mM MgCl2, 1 mM DTT, 1 mM AlCl3, 10 mM NaF, 10 mM GDP equilibrated against 24% (w/v) PEG3350 and 0.3 M NaCl. Plate crystals appeared overnight and were mounted directly from the mother liquor using a mesh loop and cryo-cooling as described previously. [6] Diffraction data collection and refinement method for RhoA-GDP. Diffraction data were collected from cryocooled crystals at 1.3 Å on a PILATUS 6M detector on beamline ID29 at the European Synchrotron Radiation Facility, Grenoble, France. Data were processed with XDS [1] and programs from the CCP4. The previously published structure of RhoA-GDP (PDB 1ftn), [7] without removal of bound ligand and water, was used as a search model for MOLREP. [3] Refinement was carried out alternately with REFMAC5 [4] and by manual rebuilding with the program COOT. [5] Ligands were included in the final rounds of refinement.
Model validation was performed using MolProbity. [8] Data collection and refinement statistics are summarized in Table S1. Figures were produced with the PyMOL Molecular Graphics System, Version 1.5.0.4 Schrödinger, LLC. [9]. Hydrogen bond distances were also measured using PyMOL and all those reported below have X-H-Y angles >145° ( Figure S3 and S4). water ligands, methyl pyrophosphate, and H2PO4 -. We also included WAT698 as it serves to stabilize GDP. Non-active methyl groups were removed and carbons at which the model would extend beyond the QM zone were converted into methyl groups ( Figure 3). Given that the purpose is to study the "microsolvated" region around the phosphates, residues far away from the region of interest were truncated (the electronic environment will be stable). In the same fashion, the guanosine in the GTP is truncated to a methyl group. All locked branch points that represent methyl groups approximating the rest of the protein environment are depicted -CH3 in Figure 3. This procedure has been previously used to describe TSA crystal structures of RhoA-RhoGAP with KS-DFT to within the uncertainty of the experimental measurement. [10] Our models for the ground state (GS) of the GDP-Pi intermediate complexes were obtained using Kohn-Sham Density Functional Theory (KS-DFT). We used the M06-2X functional formulation of KS-DFT. [11] A cc-pVDZ basis set was used to represent single-particle wavefunctions [12] for carbon, hydrogen, oxygen, and nitrogen atoms, while the cc-pVTZ basis was used for magnesium and phosphorus atoms. [12] Initially, the computation utilized cc-pVDZ for all atoms and an integration grid consisting of 99 radial points and 590 solid-angle points in the Lebedev grid, applying standard optimization algorithms to find the ground state structure.
We added basis functions in the manner most conducive of better representing the active site and improved the phosphorus basis set to cc-pVTZ to represent its polarizable electron density with more accuracy. We also increased the magnesium basis set given its strong electron polarization. All oxygen atoms in the phosphate moieties used aug-cc-pVDZ, apart from inorganic phosphate, and O3B, which had aug-cc-pVTZ because of their importance. [13] The structure was considered optimized when the force on all nuclei fell below 1 μHartree/Bohr.
The SCF was considered converged when the density matrix residual was less than 10 -6 . We optimized the geometry of the resulting active site model (210 total atoms) to obtain the GS using standard algorithms, [14] as implemented in the Gaussian09 software package. [15] We considered 7 possible hydrogen atom placements. The simplest case considered is the "immediate proton migration" of one of the two protons to either O1G, O2G, or O3G. By immediate migration, we mean the rotamer of the proton following an immediate migration from the nucleophilic water to the relevant oxygen. [10] We refer to these rotamers as being the "thermodynamic protonation states," because these rotamers presumably are more thermodynamically favorable in that they are stabilizing the leaving group. In each of these 7 possible protonation states, we selected a viable initial hydrogen position, and proceeded with the optimization.
Having optimized these 7 structures, we then compared the position of the heavy atoms in the active site region to the crystal structure. In each case, we removed D59 from the comparison; as it is on the boundary of the simulation, it is more prone to vacuum boundary artefacts, and also for WAT698. We used this methodology to compute seven GS structures with the two 'mobile' hydrogens (from the initial water molecule) apportioned variously on oxygens of the PG phosphate. Six of these have data for the H-bonds involving the four PG oxygens and O3B listed in Table S2 Entries 2-7       d) ) Figure S3. New high resolution RhoA-GDP product complex. a) The electron density (green mesh) of the key residues in Switch-II in the high-resolution structure of RhoA-GDP (5c4m). Electron density maps in green mesh are σA-weighted 2Fo -Fc contoured at 1σ. b) Overlay of RhoA-GDP product structures for the high resolution RhoA-GDP (5c4m in gray) and RhoA-GDP (1ftn in cyan). switch II (residues 61 -78) is highlighted in lime in 5m4c, whereas the same region with the missing density in 1ftn is highlighted in blue. The difference between the Switch-II region position is evident.    1 Table S2). This is the computed transition state. [10] Both H 1 (gray) and H 2 (magenta) are located on O4G directed at Thr37(C=O) and Gln63(C=O) respectively. It shows good H-bonds between O3B and Arg85' and Gln63, and between O2B and Lys18 and Gly62. O3B accepts H-bonds from Ala15 and Arg85'. O3G and O3B are 2.6 Å apart. This DFT structure (red sticks) is introduced to initiate the mobile proton sequence that follows here. Rainbow coloring is used for the progression of migrations.  Table S2) has H 1 (gray) located on O4G directed at Thr37(C=O) and H 2 (magenta) located on O3G directed at Gln63(C=O) (orange sticks). It shows good H-Bonds between Pi and Arg85', Gln63, Gly62, Lys18, and Thr37. O3B accept H-bonds from Ala15 and Arg85'. This is designated the first DFT structure in the main text.  Table S2) has H 1 (gray) located on O4G directed at Thr37(C=O) and H 2 (magenta) located on O3G directed at O3B (yellow sticks). It shows good H-bonds between Pi and Arg85', Gln63, Gly62, Lys18, and Thr37. O3B accepts H-bonds from Ala15 and Arg85'. This is designated the second DFT structure in the main text.  4 Table S2) has H 1 (gray) located on O4G directed at Gln63(C=O) and H 2 (magenta) located on O3G directed at O3B (green sticks). It shows good H-bonds between Pi and Arg85', Gln63, Gly62, Lys18, and Thr37. O3B accept H-bonds from O3G, Ala15 and Arg85'. This is designated the third DFT structure in the main text. Figure S7e. O4-Q63 & O2B-O3B (Entry 5 Table S2) has H 1 (gray) located on O4G directed at Gln63(C=O) and H 2 (magenta) re-located on O3B directed at O3G (cyan sticks). It shows good H-bonds between Pi and Ala15, Arg85', Gln63, Gly62, and Lys18. But Arg85' has moved (N2 shifting 0.3 Å) and does not H-bond to O3B while Thr37 has moved 0.6 Å away from PG. O3B donates a H-bond to O3G and accepts one from Ala15. Thr37(C=O) is 4.50 Å from PG. This is designated the fourth DFT structure in the main text.  6 Table S2) has H 1 (gray) located on O4G directed at Gln63(C=O) while H 2 (magenta) located on O1G has no H-bond acceptor (indigo sticks). There are good H-bonds between Pi and Arg85', Gln63, Gly62, Lys18, and Thr37(N-H). O3B is 2.89 Å from O3G and accepts H-bonds from Ala15 and Arg85'. Thr37(C=O) is now over 6 Å from PG and has rotated about both phi-and psi-bonds. This is designated the final DFT structure in the main text. Figure S8. The H-bonding network highlighted through the whole catalytic cycle, completed by the intermediate proposed.