QM/MM Well-Tempered Metadynamics Study of the Mechanism of XBP1 mRNA Cleavage by Inositol Requiring Enzyme 1α RNase

Abstract

A range of in silico methodologies were herein employed to study the unconventional XBP1 mRNA cleavage mechanism performed by the unfolded protein response (UPR) mediator Inositol Requiring Enzyme 1α (IRE1). Using Protein-RNA molecular docking along with a series of extensive restrained/unrestrained atomistic molecular dynamics (MD) simulations, the dynamical behavior of the system was evaluated and a reliable model of the IRE1/XBP1 mRNA complex was constructed. From a series of well-converged quantum mechanics molecular mechanics well-tempered metadynamics (QM/MM WT-MetaD) simulations using the Grimme dispersion interaction corrected semiempirical parametrization method 6 level of theory (PM6-D3) and the AMBER14SB-OL3 force field, the free energy profile of the cleavage mechanism was determined, along with intermediates and transition state structures. The results show two distinct reaction paths based on general acid-general base type mechanisms, with different activation energies that perfectly match observations from experimental mutagenesis data. The study brings unique atomistic insights into the cleavage mechanism of XBP1 mRNA by IRE1 and clarifies the roles of the catalytic residues H910 and Y892. Increased understanding of the details in UPR signaling can assist in the development of new therapeutic agents for its modulation.

Figure 1

(A) hIRE1 back-to-back dimer in complex with mRNA XBP1 single stem-loop complex generated from molecular docking calculations. The kinase and RNase domains of IRE1 dimer are shown in blue and pink, respectively. The luminal and transmembrane domains of hIRE1 are not shown in the figure. XBP1 mRNA single stem-loop is represented in green. The zoomed-in picture illustrates the catalytic residues (H910 and Y892) within the active site of the hIRE1 RNase in which the XBP1 mRNA cleavage reaction takes place. (B) Schematic 2D illustration of a concerted general acid–general base (GA–GB) reaction mechanism. The catalytic Histidine (H910) acts as a GB and initiates the reaction by abstracting a proton (red) from the Guanosine O2′ atom. Tyrosine (Y892) acts as a GA by donating its phenolic proton (blue) to the Cytidine O5′ atom.

Figure 2

(A) 2D and (B) 3D structures of the modeled XBP1 mRNA single stem-loop and the template used (tRNA^Phe) for the modeling. IRE1 can target and unconventionally cleave both RNA molecules as they share the same consensus sequence (5′-CNGNNGN-3′) within the stem-loops. The mutated residues in the XBP1 mRNA are shown in purple. (C) Composition of the QM subsystem (161 atoms) specified for the QM/MM WT-MetaD simulation. XBP1 stem-loop and two IRE1 catalytic residues (Y892 and H910) are shown in blue and red, respectively. The interatomic distances used to formulate the collective variables are shown by green arrows. (D) Two active catalytic centers (shown by asterisks) within the cleft between the RNase loops of the hIRE1 dimer.

Figure 3

Best docking poses (Model_A1 and Model_B1) generated by molecular docking. The top panel shows the entire complexes where the Kinase and RNase domains of the IRE1 dimer are shown in blue and pink, respectively. The XBP1 mRNA single stem-loop is represented as ribbon (black) and tubes/slabs (dark blue). The cleavage site of the XBP1 mRNA is highlighted in red. The bottom panel is a zoomed-in illustration of the active sites in the RNase domain of each IRE1 monomer along with the residues important in the cleavage mechanism discussed in the text. The atoms in the active site are colored as carbon in gray, hydrogen in white, nitrogen in blue, oxygen in red, and phosphorus in orange. Hydrogen atoms bound to carbon are omitted for clarity. The luminal and transmembrane domains of hIRE1 are not shown in the figure.

Figure 4

Backbone (A) RMSD and (B) RMSF of the XBP1 mRNA bound to the hIRE1 back-to-back dimer during a consecutive series of 7 × 50 ns restrained MD simulations followed by a final 200 ns unrestrained run. (C) Heavy-atom RMSF of the XBP1 mRNA binding nucleotides (G34/C35) during the classical MD simulation. Backbone (D) RMSD and (E) RMSF of the hIRE1 dimer bound to XBP1 mRNA during the MD simulations. The interacting residues of the hIRE1 dimer with XBP1 mRNA are marked with orange vertical bars. (F) Free energy of binding (ΔG_bind) between the hIRE1 dimer and XBP1 mRNA calculated for 100 snapshots extracted from the 200 ns unrestrained MD trajectory.

Figure 5

(A) Last snapshot of the MD simulation (t = 550 ns) superposed on the initial structure. The XBP1 mRNA in the first and last snapshots are presented in dark blue and green colors, respectively. The zoomed-in picture illustrates the nucleotides G34 and C35 bound to the RNA binding site. The Kinase and RNase domains of the hIRE1 dimer are shown in light blue and pink, respectively. (B) Abundance of atomic interactions between the hIRE1 dimer and nucleotides G34/C35 of XBP1 mRNA evaluated during the 200 ns unrestrained MD simulation.

Figure 6

2D projected free energy surface (FES) contour map of (A) CV1/CV2, (B) CV1/CV3, and (C) CV2/CV3. In each panel, the CV coordinates of the reactant structure (after classical minimization and equilibration steps), products, local minima, and saddle points are highlighted by green, yellow, blue, and red crosses, respectively. The main and alternative reaction paths are shown by black and red dashed lines, respectively. The numbers within the parentheses indicate the relative free energy values at each point. 1D projected free energy profiles of (D) CV1, (E) CV2, and (F) CV3, calculated from the negative of the cumulative biasing potential (black line) compared with those obtained by the reweighting technique (red line).

Figure 7

(A) Unconventional cleavage mechanism of mRNA XBP1 by IRE1 elucidated from QM/MM WT-MetaD simulations. The IRE1 and XBP1 atoms involved in the catalytic reaction are shown in red and blue, respectively. (B) Relative free energy diagram of the reactant, products, intermediates, and transition states. (C) Error in free energy estimation for each CV calculated by the block-average technique. (D) Diffusive behavior of the CV sampling during the simulation trajectory.