Structurally detailed coarse-grained model for Sec-facilitated co-translational protein translocation and membrane integration

Abstract

We present a coarse-grained simulation model that is capable of simulating the minute-timescale dynamics of protein translocation and membrane integration via the Sec translocon, while retaining sufficient chemical and structural detail to capture many of the sequence-specific interactions that drive these processes. The model includes accurate geometric representations of the ribosome and Sec translocon, obtained directly from experimental structures, and interactions parameterized from nearly 200 μs of residue-based coarse-grained molecular dynamics simulations. A protocol for mapping amino-acid sequences to coarse-grained beads enables the direct simulation of trajectories for the co-translational insertion of arbitrary polypeptide sequences into the Sec translocon. The model reproduces experimentally observed features of membrane protein integration, including the efficiency with which polypeptide domains integrate into the membrane, the variation in integration efficiency upon single amino-acid mutations, and the orientation of transmembrane domains. The central advantage of the model is that it connects sequence-level protein features to biological observables and timescales, enabling direct simulation for the mechanistic analysis of co-translational integration and for the engineering of membrane proteins with enhanced membrane integration efficiency.

Fig 1. 3D-CG model geometry.

(A) Components of the 3D-CG model overlaid on a high-resolution cryo-EM structure of the ribosome-translocon complex [25]. 3D-CG model beads are represented by opaque spheres and are labeled according to their color. The region representing the implicit membrane is drawn as a grey background. (B) 3D-CG model snapshots of the two possible translocon conformations, with a closed lateral gate (top) and with an open lateral gate (bottom). In each case, a NC is shown emerging from the ribosome exit channel and interacting with the translocon. (C) Coordinate system for the 3D-CG model. Coordinates for the translation insertion point at the ribosome exit channel, the origin, and four points illustrating the bounds of the implicit membrane are indicated. (D) Simulation snapshots showing representative states during a simulation trajectory, including: (i) the start of translation, (ii) topological inversion of a TMD during integration, (iii) release of the C-terminus at the end of translation, and (iv) the end of a simulation in which the TMD has integrated into the membrane, the lateral gate is closed, and all polypeptide segments have exited the channel.

Fig 2. Bottom-up parameterization of NC bead-translocon interactions.

(A) Simulation snapshot of the residue-based coarse-grained simulation system using the MARTINI force field. The translocon is in its closed conformation, a tripeptide substrate is shown in red, lipids are shown with head groups in white and tail groups in grey, water is represented as a transparent surface, and ions are shown as yellow spheres. (B) PMFs for translocating homogeneous tripeptides across the closed (left) and open (right) channel conformations. PMFs calculated using MARTINI for all four tripeptides are plotted as transparent lines, with shaded regions indicating the estimated error. The MARTINI PMFs are scaled by a factor of 0.25 and are vertically shifted such that the average value for 4.0σ ≤ z ≤ 4.5σ is 0. Best-fit PMFs calculated using the 3D-CG model are plotted as opaque dashed lines, and are fit in the range z ≥ −2σ (dashed vertical line). All PMFs are presented as a function of z, rather than d_z, since these values differ only by an offset of 0.1σ. (C) Piecewise linear interpolation relating values of λ^c and λ^o to the substrate hydrophobicity g. The endpoints of the piecewise linear interpolation correspond to the four substrates in B. (D) PMFs calculated using the 3D-CG model and the best-fit parameters, for the same four peptides as in B, but with the ribosome and translocon plug domain included.

Fig 3. Example sequence mapped to 3D-CG model representation.

(A) An input amino-acid sequence (AA) and secondary structure assignments (SS; H for helix and C for coil) are mapped to 3D-CG beads and assigned values of q_i, g_i, λ^c(g_i), and λ^o(g_i) based on the properties of sequential amino-acid triplets. (B) Visualization of heterogeneous NC properties and correspondence with structural elements. Left, a snapshot of a NC with each CG bead colored by g_i; red beads are hydrophobic, while cyan beads are hydrophilic. Right, the same snapshot colored by assigning each NC bead to a domain.

Fig 4. 3D-CG model predictions of membrane integration versus secretion.

(A) Snapshots of the initial system configuration, an intermediate state in which the H-segment (yellow) enters the channel, and two possible simulation products. Simulations are initialized with the TMD upstream of the H-segment (red) integrated into the membrane. (B) Probability of membrane integration (p(integration)) as a function of the number of leucine residues in the H-segment. Experimental results from Hessa et al. [51] are reproduced in black, while results from the 3D-CG model are shown in red. Each point for the 3D-CG model is the average of all three frameshifts. The solid lines are sigmoidal fits to each data set. (C) Schematic representation of three possible 3D-CG representations of the same sequence (i.e., frameshifts). The example sequence is the Lep construct with a 7 leucine H-segment (identified in yellow region). Each triplet is colored according to its value of g. (D) Probability of membrane integration as a function of the number of leucine residues in the H-segment for each individual frameshift.

Fig 5. Experimental versus simulated predictions of the single-residue apparent free energy of integration.

Each point corresponds to a different amino acid, with the character of the amino acid indicated by its plotted color. Each 3D-CG calculated $\Delta G_{\text{app}}^{\text{aa}}$ value the average of three frameshifts, the error bars indicate the standard error of the mean.

Fig 6. 3D-CG model predictions for TM topology.

(A) Snapshots of the initial system configuration and the two possible TM topologies. (B) 3D-CG model simulation results showing the fraction of trajectories that reach the Type 2 topology as a function of the number of C-terminal loop residues, plotted for a normal translational rate (solid black) and a slowed translation rate (dashed red). (C) Experimental results from Göder et al [14], with a normal translation rate (solid black) and with the addition of cyclohexamide, a translation rate inhibitor (dashed red).