Protein–lipid map in yeast. The table displays all interactions color coded according to the number of experimental supports: reproducibility with proteins expressed in homologous and heterologous systems, binding to artificial (liposome) and biological (live-cell imaging) membranes and genetic interactions. The data set is provided as Supplementary information (Supplementary Tables S2C and S4; see also Supplementary Data 1). The bar charts show the fraction of novel interactions per lipid or protein classes for which supporting evidence is available. The total numbers of novel interactions are displayed next to each bar. The pie charts represent the fraction of conserved proteins. PtdInsP, phosphatidylinostitol phosphate; GPL, glycerophospholipid; SL, sphingolipid; FA, fatty acid; GL, glycerolipid; S&P, sterol and prenol.

Identification of a new cryptic lipid-binding domain in Ecm25. (A) Structure-based alignment of Ecm25 and five known CRAL/TRIO domain. Conserved hydrophobic sites are colored blue, polar sites are red, invariant residues are boxed, sites occupied by small residues are magenta and the ligand-binding sites of the structures are colored gray. Secondary structure elements are numbered according to the canonical C-terminal CRAL/TRIO domain structure of Sec14 (PDB:1aua), and grayed out when disrupted or absent in some of the structures. Sequences prefixed with ‘h' are from human; all others are from S. cerevisiae. (B) The truncated Ecm25-CRAL/TRIO binds to artificial membranes in flotation assays. Liposomes contain 1 mM 1-palmitoyl-2-oleoyl-snglycero-3-phosphocholine (POPC) and phosphatidylserine (PS), PtdIns or PtdIns(3,4,5)P₃, as indicated (3%, molar). Ecm25-CRAL/TRIO was analyzed by western blot.

The PH domain of Slm1 cooperatively binds PtdInsPs and phosphorylated sphingolipids. (A) A significant fraction of proteins-binding phosphorylated LCBs also contained a PH domain. For each sphingolipid-binding group, the pie chart shows the fraction of proteins with a PH domain (^*P<0.00001). (B) Lipid-binding specificities of PH-domain containing proteins; more than half of those binding PtdInsPs also interacted with LCB-P. Proteins that did not show any binding were omitted. (C, D) Recruitment of Slm1-PH and PLCδ-PH to liposome membranes of different compositions: 3% (molar) PtdIns(4,5)P₂ and/or 3% (molar) DHS-1P. (C) Quantitative measurement of liposome binding by isothermal titration calorimetry. Points represent heat evolved per mol of injected protein. ND: not determined. (D) Monitoring of Slm1-PH and PLCδ-PH co-eluting with liposomes by size exclusion chromatography and western blot. (E) Role of PtdInsPs and sphingolipids in the recruitment of Slm1 and PLCδ-PH to biological membranes in vivo. On the left, Slm1 fused to GFP (Slm1-GFP) co-localizes with eisosomes (Pil1-RFP). Interference with PtdIns(4,5)P₂ metabolism using a temperature-sensitive mutant of the phosphatidylinositol 4-phosphate 5-kinase (mss4^ts). Intensity plot profiles were extracted from the area represented by the respective white lines. On the right, localization of PLCδ-PH fused to GFP. Cell treatments are as in Figure 5A.

Lipid-binding fingerprints as predictors of protein localization, domains and functions. (A) Proteins clustered according to their lipid-binding profiles. The wedge shapes highlight some groups that are enriched (P<0.05) in a particular annotation. The black oblong represents the group of proteins that did not bind any lipid in this study. This group is deficient in proteins annotated by SGD (Nash et al, 2007) as lipid binding. The violet cluster of 12 proteins from Caf120 to Kes1 is also enriched for myriocin-insensitive proteins (P=0.01). (B) Live-cell imaging of Mss4 fused to GFP reveals localization in punctate membrane structures.

Identification of protein–lipid interactions in S. cerevisiae using lipid–arrays and live-cell imaging. (A) Charting protein–lipid interactions with lipid–arrays. (B) Summary of the different lipids analyzed. Lipids are grouped according to their metabolic pathways. Numbers of different lipids analyzed for each pathway are indicated. (C) Summary of the different proteins analyzed. Proteins are grouped according to their LBDs. For each category, the numbers of proteins analyzed are indicated. (D) Summary of the validation procedure. The number of proteins analyzed in each assay is represented. Asterisk points to the series of 29 proteins that bound PtdInsPs (and not sphingolipids) and that we used as control.

Assessment of the lipid–array data quality. (A, B) Comparison with sets of literature curated and genetic interactions. (A) Summary of the two reference data sets and strategies used to assess quality. Genetic coverage is defined as the percentage of physical protein–lipid interactions covered by the data set of genetic interactions. (B) Estimation of accuracy based on PtdInsPs metabolic pathway. The left column shows the number of proteins in the literature-derived reference set that are covered (dark blue) or not (magenta) by the data set of genetic interactions. The right column shows the fraction of proteins devoid of LBD and that did not bind PtdInsPs in the lipid–array that interact genetically with enzymes involved in the synthesis of these lipids (light blue, background coverage by genetics). The central column shows the estimation of accuracy (fraction of true positives) in the lipid–array data set assuming that genetic coverage results from the combination of true positives (that will have the same genetic coverage as the literature-derived reference data set, in dark blue) and false positive (with a background coverage, in light blue). The estimated fraction of true positive that is not covered by genetics is shown in magenta. Numbers on top of the columns are overall genetic coverage for each set of proteins (^*P=0.015 and ^**P=0.035).

Lipid–arrays and live-cell imaging identify new series of sphingolipid effectors. (A) The effect of the perturbation of sphingolipid metabolism on GFP-fusion localization was monitored in yeast cells treated with myriocin, myriocin and dihydrosphingosine (DHS) or untreated (Blank). (B) The proteins-binding sphingolipids in vitro are grouped according to their sphingolipid-binding specificities, and ordered by function. The symbol # points to proteins with a PH domain. Interactions are color coded as in Figure 2, but excluding live-cell imaging. The green boxes label the proteins selected for further analysis by live-cell imaging. Those sensitive to 2 h myriocin treatments are represented with filled green boxes. For each sphingolipid-binding group, the histogram next to the matrix summarizes the overall sensitivities to myriocin. For comparison, myriocin sensitivity of 32 controls is also shown in magenta. Non-physiological lipids in S. cerevisiae are in gray and italic. DHS-1P, dihydrosphingosine-1-phosphate; PHS-1P, phytosphingosine-1-phosphate; Sph-1P, sphingosine-1-phosphate; 3-ketoDHS, 3-ketodihydrosphingosine; DHS, dihydrosphingosine; PHS, phytosphingosine; Sph, sphingosine; DHCer, dihydroceramide; PHCer C8, phytoceramide C8; PHCer C18, phytoceramide C18.

The 2.0 Å X-ray structure of Slm1-PH reveals two putative binding sites required for Slm1 function in vivo. (A) Comparison of the structures of Slm1-PH with that of PLCδ-PH (Protein Data Bank (PDB) entry 1mai). The two structures are represented in the same orientations. PLCδ-PH binds lipid within a positively charged canonical site conserved in Slm1 (orange circle). Slm1 presents an additional positively charged cavity (gray circle). Two key residues (R477 and R478) were identified that contribute to both positively charged binding sites in Slm1. Side chain of analogous residues in PLCδ (K30 and K32) point exclusively to the canonical binding site. Electronegative regions are colored red and electropositive regions blue. (B) Sequence alignment of Slm1-PH and PLCδ-PH. Mutated amino acids in Slm1 (R477 and R478) and their analogous in PLCδ are boxed in black. Conserved residues are blue if positively charged and green if hydrophobic. (C) The two putative Slm1-PH-binding sites are required for Slm1 membrane association with the eisosome. Kymograph representations of Slm1-GFP and Pil1-RFP (1 frame per 1.5 s). Green lines above the kymograph represent the association dynamics of Slm1-GFP with the Pil1-RFP (red lines). Mutations affecting clusters of positive charges in one (R477A or R478A) or both (R478A/R477A) resulted in specific destabilization of Slm1 association with Pil1-containing microdomains. (D) The two putative Slm1-PH-binding sites are required for Slm1 function in vivo. Strains carrying Slm1 mutations in one (R477A or R478A) or both (R477A/R478A) putative lipid-binding sites show cumulative phenotypic defects. Upper panels show growth defect in the presence of myriocin; lower panels during heat stress. The bar chart shows actin polarization defects following heat stress. * decrease for the double mutants is significant compared with wild-type (P<0.01), R477A (P<0.01) and R478A (P<0.01) single mutants. The mutations have no effect on Slm1 expression levels (western blot not shown).