Differential Lipid Binding Specificities of RAP1A and RAP1B are Encoded by the Amino Acid Sequence of the Membrane Anchors

Abstract

RAP1 proteins belong to the RAS family of small GTPases that operate as molecular switches by cycling between GDP-bound inactive and GTP-bound active states. The C-terminal anchors of RAP1 proteins are known to direct membrane localization, but how these anchors organize RAP1 on the plasma membrane (PM) has not been investigated. Using high-resolution imaging, we show that RAP1A and RAP1B form spatially segregated nanoclusters on the inner leaflet of the PM, with further lateral segregation between GDP-bound and GTP-bound proteins. The C-terminal polybasic anchors of RAP1A and RAP1B differ in their amino acid sequences and exhibit different lipid binding specificities, which can be modified by single-point mutations in the respective polybasic domains (PBD). Molecular dynamics simulations reveal that single PBD mutations substantially reduce the interactions of the membrane anchors with the PM lipid phosphatidylserine. In summary, we show that aggregate lipid binding specificity encoded within the C-terminal anchor determines PM association and nanoclustering of RAP1A and RAP1B. Taken together with previous observations on RAC1 and KRAS, the study reveals that the PBD sequences of small GTPase membrane anchors can encode distinct lipid binding specificities that govern PM interactions.

Figure 1

RAP1A and RAP1B proteins form PM nanoclusters. (A) Representative EM images (1 μm²) of 4.5 nm gold particles coupled to anti-GFP antibodies on PM sheets. Intact PM sheets were prepared from BHK cells expressing GFP-RAP1A-G12V or GFP-RAP1B-G12V and labeled with 4.5 nm gold-anti-GFP. Coordinates of each gold particle were used to generate digitalized images of the gold point patterns that are color-coded in red. (B–D) PM sheets of BHK cells expressing GFP-RAP1A-G12V, -wild-type (WT) or S17N were immunogold labeled and imaged by EM. Univariate K-functions were used to analyze the gold patterns. (B) Plots of weighted mean standardized univariate K-function are shown. Values of L(r) – r above the 99% confidence interval (C.I) indicate significant clustering. (C) PM binding of GFP-RAP1A-G12V, -WT or S17N was quantified as mean gold labeling intensity (±SEM; (n = 12–17) for each condition). Student t tests were used to evaluate statistical differences between mean gold labeling density of GFP-RAP1A-G12V and GFP-RAP1A-WT or S17N (***p < 0.001). (D) Peak value of L(r) – r (=L_max) was used to summarize the extent of nanoclustering. L_max values are means ± SEM (n = 10–15). The significance of differences from control L_max values were analyzed using bootstrap tests. (E–G) The EM experiments in (B–C) were repeated in BHK cells expressing GFP-RAP1B-G12V, -WT or S17N (n = 16–27). Weighted mean univariate K-functions (E), mean gold labeling densities (F) and L_max values (G) for GFP-RAP1B-G12V, -WT or S17N are shown. (H) PM sheets from BHK cells coexpressing GFP- and RFP-RAP1A/B constructs were labeled with 6 nm gold-anti-GFP and 2 nm gold-anti-RFP and visualized by EM. Colocalization of GFP- and RFP-RAP1 proteins were analyzed by integrated bivariate K-functions (=LBI). The green dotted lines indicate 95% C.I. LBI values above the C.I (>100) indicate significant coclustering (±SEM; (n ≥ 12) for each condition). (I) The bivariate EM experiment in (H) was repeated in BHK cells coexpressing GFP-RAP1A and RFP-RAP1B, or GFP-RAP1B and RFP-RAP1B. LBI values are means +SEM [(n = 10–14) for each condition]. Bootstrap tests were used to evaluate the significance of statistical differences (*p < 0.05).

Figure 2

Differential lipid sorting specificities are encoded in the membrane anchors of RAP1 proteins. (A,B) PM sheets of BHK cells coexpressing RFP-tagged wild-type (RFP-RAP1-WT) or GTP-bound mutant RAP1 (RFP-RAP1-G12V) with a GFP-tagged lipid probe for PS (GFP-LactC2), PIP2 (GFP-PLCδ), PIP3 (GFP-AKT), PA (GFP-PASS), or cholesterol (GFP-D4H) were labeled with 6 nm gold-anti-GFP and 2 nm gold-anti-RFP and imaged by EM. Bivariate coclustering analysis of the two gold populations yields LBI values [±SEM, (n ≥ 12) for each condition] which reflect the lipid binding preferences of RAP1A (A) or RAP1B (B). (C,D) Details of mutations made in the RAP1A and RAP1B membrane anchor domains. (E,F) Heatmaps were generated using mean LBI values to quantify coclustering between each RFP-PBD mutant and each GFP-lipid probe. For each lipid probe, the LBI value for wild-type anchor construct (RAP1-G12V) was assigned as midpoint (marked in white) with lower or higher LBI values marked in blue or red, respectively. Bootstrap tests were used to evaluate the significance of statistical differences (*p < 0.05, and **p < 0.01).

Figure 3

RAP1 PBD mutants exhibit different PM binding affinities. (A–D) PM sheets of BHK cells expressing GFP-RAP1A or RAP1B anchor mutants were labeled with 4.5 nm gold-anti-GFP and analyzed by EM. PM localization was quantified as mean gold labeling intensity per 1 μm² and is shown as mean ± SEM (A,C), and the extent of nanoclustering is summarized as L_max (mean ± SEM, n ≥ 12) (B,D). (A–D) Significant differences between L_max values for RAP1 PBD wild-type (RAP1-G12V) and each anchor mutants were evaluated in bootstrap tests (*p < 0.05, and **p < 0.01), and differences in gold labeling density were evaluated in two-tailed t tests (*p < 0.05, and **p < 0.01).

Figure 4

Simulation setup and bilayer adsorption of RAP1A and RAP1B membrane anchor mutants. (A) Sequence of the RAP1A and RAP1B membrane anchors with basic residues in blue, acidic in red, polar in green, and hydrophobic residues in black. The underlined Lys or Arg residues were mutated to Gln, and the C-terminal Cys residue is carboxymethylated following geranylgeranylation. (B) Example of MD simulation setup, with three peptides embedded in the mixed-lipid leaflet of an asymmetric model membrane composed of POPC (gray) and POPS lipids (red). Lipid phosphorus atoms are shown in vdW spheres and the peptides (in this case RAP1A WT) in licorice colored as shown in panel A except for the prenylated Cys residue, which is in yellow. Water and ions are shown as a blue surface. (C) Snapshots at the start (t = 0 μs) and end (t = 5 μs) of the MD simulations of membrane-embedded wild-type and indicated mutant RAP1A and RAP1B membrane anchors, showing complete adsorption of the peptides on the bilayer surface at the end of the simulation. Only lipid phosphorus atoms, peptide backbone, and prenyl chains are shown colored as in panel B while the rest of the atoms are omitted for clarity.

Figure 5

Interactions of RAP1A and RAP1B with lipids. (A) Distribution of peptide-lipid HB (N_HB) and van der Waals (vdW) contacts (N_C) per peptide separately for POPC (gray) and POPS (black). HB was defined with a donor–acceptor distance cutoff of 3.1 Å and a donor-hydrogen-acceptor angle cutoff of 30° and included all polar and charged side chains and lipid headgroup oxygen atoms. N_C was computed using a carbon–carbon distance cutoff of 4 Å and included all nonpolar side-chain carbons (including the prenylated cysteines) and lipid acyl chain carbon atoms. (B) Heatmap of normalized HB frequency (HB) between the PBD Lys or Arg side chains with POPS headgroup oxygen atoms.