Harmonizing Experimental Data with Modeling to Predict Membrane Protein Insertion in Yeast

Abstract

Membrane proteins must adopt their proper topologies within biological membranes, but achieving the correct topology is compromised by the presence of marginally hydrophobic transmembrane helices (TMHs). In this study, we report on a new model membrane protein in yeast that harbors two TMHs fused to an unstable nucleotide-binding domain. Because the second helix (TMH2) in this reporter has an unfavorable predicted free energy of insertion, we employed established methods to generate variants that alter TMH2 insertion free energy. We first found that altering TMH2 did not significantly affect the extent of protein degradation by the cellular quality control machinery. Next, we correlated predicted insertion free energies from a knowledge-based energy scale with the measured apparent free energies of TMH2 insertion. Although the predicted and apparent insertion energies showed a similar trend, the predicted free-energy changes spanned an unanticipated narrow range. By instead using a physics-based model, we obtained a broader range of free energies that agreed considerably better with the magnitude of the experimentally derived values. Nevertheless, some variants still inserted better in yeast than predicted from energy-based scales. Therefore, molecular dynamics simulations were performed and indicated that the corresponding mutations induced conformational changes within TMH2, which altered the number of stabilizing hydrogen bonds. Together, our results offer insight into the ability of the cellular quality control machinery to recognize conformationally distinct misfolded topomers, provide a model to assess TMH insertion in vivo, and indicate that TMH insertion energy scales may be limited depending on the specific protein and the mutation present.

Figure 1

The design of a model substrate to measure transmembrane domain insertion in yeast. (A) A cartoon depicting the topology of Ste6p and Ste6p^∗ and the two topologies adopted by Chimera N^∗ is given. Each construct has a 3× HA-tag in the lumenal loop between TMH1 and TMH2. Chimera N^∗ was made by internal deletion of amino acids 141–1042 in the coding sequence for 3× HA Ste6p^∗. Chimera N^∗ displays dual topologies, with TMH2 residing inside the ER lumen (left, red), which consequently deposits NBD2^∗ into the ER lumen and allows for N-glycan addition. Alternatively, TMH2 can insert properly into the membrane (right, green), which prevents acquisition of N-linked glycans. (B) S. cerevisiae strain BY4742 expressing Chimera N^∗ under the control of a low-expression ADH promoter (pCG28) was grown to log phase, and cellular protein was extracted and incubated in the absence or presence of Endo H. Chimera N^∗ was detected by SDS-PAGE and subsequent immunoblotting with anti-HA antibody. Red and green arrows to the right of the immunoblot mark the positions of the glycosylated and nonglycosylated forms, respectively.

Figure 2

Chimera N^∗ resides in the ER and is degraded by ERAD. (A) The residence of Chimera N^∗ in the cell was investigated by indirect immunofluorescence using mouse anti-HA antibody (Chimera N^∗, green), rabbit anti-Kar2p (ER lumen, red), and 4′,6-diamidino-2-phenylindole (nuclei, blue). Primary antibodies were decorated with Alexa goat anti-mouse 488 and goat anti-rabbit 568. Images were captured using a confocal microscope, and a slice through the plane of the ER is shown. Scale bars, ∼5μm. (B) Chimera N^∗ was expressed at high copy in pdr5Δ (circles) or pdr5Δpep4Δ (triangles) yeast. Before CHX chase analysis, cells were preincubated at 26°C with either dimethylsulfoxide (filled symbols) or 100 μM MG132 (open symbols) for 20 min and chased for the indicated times. Graphed data represent the means ± SD from a representative experiment of n = 4 independent experiments for pdr5Δ and n = 1 performed with four technical replicates for pdr5Δpep4Δ.

Figure 3

The predicted and measured insertion of Chimera N^∗ in yeast. (A) S. cerevisiae expressing Chimera N^∗ variants was grown to log phase, cellular proteins were extracted, and the Chimera N^∗ species were resolved by SDS-PAGE and immunoblotted to detect the HA tag. Variants were loaded from left (negative ΔG, green triangle) to right (positive ΔG, red triangle) in the order listed in Table 2. The red and green arrows to the right mark the positions of the noninserted and inserted forms, respectively. ^∗ denotes a background band that migrates beneath the inserted form. A representative HA-HRP blot is shown, and the corresponding glucose-6-phosphate dehydrogenase blot serves as a loading control from n = 2–4 independent experiments. The percentage insertion was determined by the following equation: % inserted = (f_inserted/(f_inserted + f_uninserted)) × 100. (B) The data from part (A), displayed in Table 2, were plotted against ΔG_pred and ΔG_app and then fitted to a five parameter sigmoidal equation using SigmaPlot.

Figure 4

Calculation of optimal insertion orientation. (A) Definition of angles describing helix orientation in the membrane is given. (B) Insertion energy of the physics-based continuum model based on Eq. 1 of the supplement for the wild-type helix is shown as a function of helix rotation (ψ) and tilt angle (ϕ). The minimal energy value (circle) used to compute energies in Fig. 5A is highlighted. The energy barrier corresponding to embedding the end of the helix in the membrane can be seen from the high energy values from ϕ = 45–70° (yellow).

Figure 5

Physical models of insertion stability and molecular simulations. (A) Energy values were derived from our physics-based energy model computed for each TMH2 variant and plotted against the measured percent insertion (green dots). ΔG_pred data are from Fig. 3B (black circles). Percent insertion predicted for a two-state Boltzmann distribution with energy difference ΔG bounded between 15 and 92% is shown (black curve). Insets show representative MD snapshots of the wild-type (WT) and P15L and G10L variants with mutated residues shown as licorice, protein backbone in a new cartoon, and headgroup phosphate atoms as red spheres. (B–E) Simulation details for the wild-type (B), P15L (C), G10L (D), and S13L (E) simulations are shown. Top: the average kink angle θ (defined in last panel) along the helix was computed, revealing large kinks at position 14 for wild-type and G10L and a straight helix for P15L. Middle: the angle ϕ that the helix makes with the membrane normal (defined in last panel) is shown. The average over the last half of each simulation is shown in the upper left corner. Bottom: a histogram of total backbone hydrogen bonds through the membrane span over the trajectory is shown. The average value computed over the last half of the simulation is shown in the upper left corner.