Structural determinants and functional consequences of protein affinity for membrane rafts

Abstract

Eukaryotic plasma membranes are compartmentalized into functional lateral domains, including lipid-driven membrane rafts. Rafts are involved in most plasma membrane functions by selective recruitment and retention of specific proteins. However, the structural determinants of transmembrane protein partitioning to raft domains are not fully understood. Hypothesizing that protein transmembrane domains (TMDs) determine raft association, here we directly quantify raft affinity for dozens of TMDs. We identify three physical features that independently affect raft partitioning, namely TMD surface area, length, and palmitoylation. We rationalize these findings into a mechanistic, physical model that predicts raft affinity from the protein sequence. Application of these concepts to the human proteome reveals that plasma membrane proteins have higher raft affinity than those of intracellular membranes, consistent with raft-mediated plasma membrane sorting. Overall, our experimental observations and physical model establish general rules for raft partitioning of TMDs and support the central role of rafts in membrane traffic.

Fig. 1

Raft-targeting features distributed over the TMD of LAT. a Sequences of TMDs used to determine the localization of the raft-targeting feature. Mutations were made to either the entire TMD sequence (trLAT-allL) or to individual amino acid quartets (e.g. 1pL). b  Example of quantification of raft partitioning in GPMVs. A protein of interest (trLAT; red) is expressed in cells, which are then stained with a lipid dye (F-DiO; green) with known phase preference (non-raft phase for F-DiO). The dye is used to identify the non-raft phase in GPMVs, and the relative fluorescence intensity of the protein in the raft versus non-raft phase gives the raft partition coefficient, K_p,raft. Mutating all TMD residues to Leu (trLAT-allL) decreases the raft affinity relative to the native LAT TMD (trLAT-wt). Vesicles in images are 5–10 μm in diameter. c The TMD was divided into 6 parts, that were mutated individually to Leu to identify the location of the raft-targeting features. None of these partial mutations reproduced the lack of raft affinity of the all-Leu construct, suggesting a distributed feature responsible for the raft affinity of the LAT TMD. d The 16 core residues of the LAT TMD were randomized to create trLAT-scr. This construct partitioned at parity with the wild-type TMD, suggesting amino acid properties, rather than sequence, as the key determinant of raft affinity. Average±SD for 3–5 independent trials, each with > 10 vesicles/condition; **p<0.01; ***p<0.001, one-way ANOVA. Sequences and partitioning values for all variants are given in Supplementary Table 1

Fig. 2

TMD surface area determines raft partitioning. a For 56 TMD constructs, computed side chain surface area correlates strongly with raft affinity (r = 0.74, p < 0.001), with TMDs with smaller areas preferentially partitioning to the raft phase. b Schematic model for accessible surface area (ASA)-dependent raft affinity based on differential interfacial tension (Δγ_TMD,Lo-Ld) between a TMD and the membrane matrix in raft versus non-raft phases. c Apparent raft affinity ΔG_raft,app for 12 constructs whose TMDs are comprised solely of Ala and Leu residues shows a clear dependence on TMD side chain surface area. The dotted line shows the model fit, which yields a prediction for Δγ_TMD,Lo-Ld of 1.1 pN/nm (r = 0.93, p<0.001). Average±SD for > 10 vesicles/condition representative of 3–5 independent trials. Sequences and partitioning values for all variants are given in Supplementary Table 2. d Endpoint snapshots of CG MD simulations of model TMDs (yellow) with small (all-Ala) and large (all-Phe) surface areas in a phase separated membrane composed of 50% DPPC (blue), 30% DLiPC (green), and 20% cholesterol (white). e Quantification of contacts between TMDs and lipids reveals that all-Ala interacts more avidly with L_o lipids (DPPC+Chol) than L_d lipids (DLiPC). f Umbrella sampling calculation of relative free energy (PMF) difference resulting from translating a TMD from the center of the L_o domain (x = 0 nm) to the L_d (x = 6 nm). all-Ala has much higher relative L_o affinity than all-Phe

Fig. 3

Construction of a comprehensive, predictive raft partitioning model. a TMD length and palmitoylation have been previously identified as determinants of raft partitioning. These have been formalized together with the surface area model from Fig. 2 into a comprehensive description of raft partitioning for TMDs. b Using parameters derived from published data and Fig. 2, raft affinity predicted from the tripartite model (ΔG_raft,pred) shows excellent agreement with measured values (ΔG_raft,app) for all tested TMDs (i.e., the constructs from Fig. 2a plus 4 palmitoylation mutants and 4 TMD truncations) without any floating fit parameters. c The model was used to predict raft affinity for the 729 single-pass PM proteins in the human proteome, from which three were chosen as representative raft-preferring proteins and two as non-raft preferring. d The partitioning of constructs containing TMDs from these five proteins followed predicted trends, with all three raft-predicted TMDs (CD4, LIME, PAG) partitioning efficiently to the raft domains, while both predicted non-raft proteins (LAX and LDLR) showed minimal raft partitioning. Vesicles in images are 5–10 μm in diameter. e-g The characteristics determining the raft partitioning of these constructs were dissected by mutating the TMDs to modify palmitoylation, TMD length, and ASA. e Mutating either of two palmitoylation sites reduced raft partitioning for LAT and PAG; f truncating the TMD length by 6 amino acids reduced raft partitioning for LAT, PAG, and LIME (but not LIME-Δ6exo); and g increasing ASA reduced raft partitioning for PAG and LIME, while decreasing ASA significantly increased raft partitioning for LDLR. Average±SD for 3–5 independent trials, each with > 10 vesicles/condition; *p<0.05; **p<0.01; ***p<0.001, one-way ANOVA. Sequences and partitioning values for all variants are given in Supplementary Tables 3, 4

Fig. 4

PM proteins have higher raft partitioning, and mutations of raft affinity reduced PM localization. Proteome-wide bioinformatic predictions of a palmitoylation, b TMD length, and c TMD side chain surface area reveal that all features associated with raft partitioning are predicted to be significantly enhanced in single-pass PM proteins versus a combined set of single-pass ER and Golgi proteins. d Consistently, predicted raft affinity is significantly higher for PM proteins than non-raft (3 outliers omitted for clarity). Significances are one-way ANOVA relative to PM; ***p<0.001; **p<0.01. The distribution of ΔG_raft is clearly not unimodal; histogram with Gaussian fits in Supplementary Fig. 6. e–g Experimental evaluation of bioinformatics predictions. PM localization of two predicted raft proteins (LAT and PAG) was significantly reduced by mutating any of the structural features determining raft partitioning, including e palmitoylation, f TMD length, and g TMD surface area. Scale bars are 5 μm. Avg ± SD for 20–40 cells/condition. Significances are one-way ANOVA relative to the wild-type TMD; ***p<0.001; **p<0.01