Sphingosine 1-phosphate mediates adiponectin receptor signaling essential for lipid homeostasis and embryogenesis

Abstract

Cells and organisms require proper membrane composition to function and develop. Phospholipids are the major component of membranes and are primarily acquired through the diet. Given great variability in diet composition, cells must be able to deploy mechanisms that correct deviations from optimal membrane composition and properties. Here, using lipidomics and unbiased proteomics, we found that the embryonic lethality in mice lacking the fluidity regulators Adiponectin Receptors 1 and 2 (AdipoR1/2) is associated with aberrant high saturation of the membrane phospholipids. Using mouse embryonic fibroblasts (MEFs) derived from AdipoR1/2-KO embryos, human cell lines and the model organism C. elegans we found that, mechanistically, AdipoR1/2-derived sphingosine 1-phosphate (S1P) signals in parallel through S1PR3-SREBP1 and PPARγ to sustain the expression of the fatty acid desaturase SCD and maintain membrane properties. Thus, our work identifies an evolutionary conserved pathway by which cells and organisms achieve membrane homeostasis and adapt to a variable environment.

Fig. 1. Membrane-lipid composition defects precede embryonic lethality in DKO mice.

A Mouse crosses and genotype frequencies at E12.5 (n = 218) and E15.5 (n = 46). See also Supplementary Data 1. B E12.5 embryo showing parts used. C Sections of WT and DKO embryos at E12.5 stained with H&E. (1) surface ectoderm; (2) fourth ventricle; (3) basal plate of metencephalon; (4) trigeminal (V) ganglion; (5) lumen of primitive nasal cavity; (6) olfactory epithelium; (7) third ventricle; (8) optic recess; (9) region of optic chiasma. More sections in Supplementary Fig. 1A–H. D Lipidomics-based PCA of E12.5 heads. Quantities analyzed: mol% of fatty acids species and type (SFA, MUFA, PUFA), and total amount of PE, Cer, DiCer, GlcCer, LacCer, TAG and FC normalized to total PC. Variables used are shown in Supplementary Fig. 1I and Supplementary Data 2. E Volcano plot of lipid species differing between WT and DKO at E12.5, with a p < 0.01 based on t-tests (two sided and assuming normality). Species differing by ≥1% mol are labeled; species measured are in Supplementary Data 2. F Number of proteins identified in WT and DKO embryos at E12.5 and with altered levels (p < 0.01). See Data S3. G GSEA Enrichment plot of the KEGG Fatty Acid Metabolism Pathway. H Gene Set Enrichment Analysis (GSEA) of proteomics data: 16 KEGG pathways are over-represented in DKO embryos at E12.5. See Supplementary Fig. 1K and Supplementary Data 3. I, J Abundance of apolipoproteins in WT and DKO embryos at E12.5. n = 3 biologically independent replicates per condition. K, L PC 32:0 and S1P abundance in brains at E15.5. n = 4, 3, 3 and 1 biologically independent replicate for WT, R1KO, R2KO and DKO, respectively. Lipidomics in Supplementary Fig. 1L and Supplementary Data 2. (n.d. non-detected) I–L shows mean ± SEM and t-tests (two sided and assuming normality) were used to identify significant differences between treatments. **p < 0.01 and *p < 0.05. See Supplementary Fig. 1 and Supplementary Data 1–3. Source data provided as a Source Data file.

Fig. 2. S1P rescues membranes in AdipoR2 KO MEFs.

A, B PA (16:0) and OA (18:1) abundance in PC of MEFs ± PA 200 µM ± OA 200 µM for 18 h. n = 4 biologically independent replicates per condition. Additional lipidomics in Supplementary Fig. 2A–C and Supplementary Data 4. C Laurdan dye method. D GP index of Laurdan-stained MEFs. n = 15 images, except for R2KO on basal media where n = 14. E–G Pseudocolor images from D. Packed structures (white arrows). Enlarged examples in Supplementary Fig. 2F. H Confocal image of R2KO MEFs stained with C1-Bodipy-C12. I R2KO MEFs stained with Laurdan (green) and ER-tracker (purple); merged signals in white. Originals in Supplementary Fig. 2G. J Color-coded image of Z-stack of R2KO MEFs stained with C1-Bodipy-C12. K, L Electron microscopy of spiral membranes (red arrows) and nuclear envelope blebbing (yellow arrows). n = 27 and 22 sections for WT and DKO, respectively. “N” indicates nuclei. Larger images and quantifications in Supplementary Fig. 2L, M, Q, R. M, N S1P and Sph abundance in MEFs ± PA 200 µM for 6 h. n = 4 biologically independent replicates except for DKO + PA where n = 3. From Supplementary Data 4. O PCA based on varying fatty acid species in PC and PE of MEFs. P– R PA, OA and PUFA levels in PE of MEFs + PA 50 µM ± S1P 1 µM. n = 3 biologically independent replicates per condition. From Supplementary Data 4. S GP index of MEFs + PA 50 µM ± S1P 1 µM and Laurdan-stained. n = 14, 16 and 20 separate images for WT, DKO and DKO + S1P. T, U qPCR in MEFs + PA 50 μM ± S1P 10 μM. n = 3 biologically independent replicates per condition. Data shows mean ± SEM, except ±SD in T, U. t-tests (two sided and assuming normality) identified significant differences between treatments. *p < 0.05, **p < 0.01, ***p < 0.001. See Supplementary Figs. 2, 3 and Supplementary Data 4, 5. Source data in Source Data file.

Fig. 3. S1P promotes resistance against saturated fatty acids by improving membrane homeostasis in mammalian cells.

A S1P abundance in WT MEFs ± PA 400 µM for 12 h. n = 4 biologically independent replicates for each condition. Related lipidomics in Supplementary Data 4. B, C Pseudocolor images and GP index of WT MEFs ± PA 400 µM ± S1P 10 or 1 µM ± OA 400 µM for 18 h and stained with Laurdan. Note the large number of “small” and highly packed lipid droplets induced by PA + OA (white arrows). In C, n = 15 images analyzed per condition, except for PA + OA where n = 14. D, E PA (16:0) and PalOA (16:1) abundance in the PC and PE of HEK293 cells + PA 400 µM ±S1P 10 µM for 24 h. n = 5 biologically independent replicates per condition. Related lipidomics are in Data S5. F Average GP index of the rat insulinoma INS-1E ± PA 400 µM ± S1P 10 or 1 µM ± OA 400 µM and stained with Laurdan. For each condition, from left to right, n = 19, 17, 16, 15, and 6 images analyzed. G Glucose-stimulated insulin secretion (GSIS) assay: Amount of insulin secreted (in 2 h) by INS-1E cells pre-exposed ± PA 400 µM ± S1P for 18 h. n = 4 biologically independent replicates for each condition. H Schematics of the Zn²⁺ atom inside human AdipoR2. PDB 5LX9 entry was used as model. The Zn²⁺ is coordinated by three histidines. I Western-Blot of AdipoR2 and Tubulin in HEK293 cells transfected with human AdipoR2: WT sequence, H348A mutant and H202,348,350A mutations. Uncropped blots in Source Data. J Average GP index from several images of HEK293 cells transfected with human AdipoR2: WT sequence, H348A mutant and H202,348,350A mutations and challenged with PA 400 µM for 24 h. For each condition, from left to right, n = 10, 15, 10, 15 and 15 images analyzed. Data are represented as mean ± SEM. t-tests (two sided and assuming normality) were used to identify significant differences between treatments. *p < 0.05, **p < 0.01, ***p < 0.001. See also Supplementary Fig. 4 and Supplementary Data 4–6. Source data are provided as a Source Data file.

Fig. 4. Sphingosine kinases and S1PR3 are required to maintain membrane homeostasis in human/mouse cells.

A Average GP index from several images of NT, Sphk1 and Sphk2 siRNA-treated HEK293 cells challenged with PA 200 µM and stained with the Laurdan dye. n = 20 separate images analyzed for each condition. B, C Pseudocolor images and average GP index from several images of NT and Sphk1 + 2 siRNA HEK293 cells treated with PA 200 µM ± S1P 1 µM or ± OA 200 µM. In C, n = 20 separate images analyzed for each condition. D, E Schematic representation of S1P signaling and S1PR pharmacological modulators used in this work. F Relative expression of S1PRs in WT MEFs measured by qPCR. n = 3 technical replicates per condition. G, H Pseudocolor images and average GP index from several images of WT MEFs treated with vehicle, PA 400 µM, PA 200 µM ± JTE-013 5 µM (S1PR2 antagonist), ± TY52156 5 µM (S1PR3 antagonist). In H, for each condition from left to right, n = 15, 5, 15, 15 and 15 separate images analyzed. I Average GP index from several images of NT, AdipoR2, S1PR2, S1PR3 siRNA HEK293 cells treated with PA 200 µM ± S1P 1 µM. For each condition from left to right, n = 10, 10, 15, 10, 14 and 15 separate images analyzed. Data are represented as mean ± SEM, except F: ± SD. t-tests (two sided and assuming normality) were used to identify significant differences between treatments. *p < 0.05, **p < 0.01, ***p < 0.001. See also Supplementary Fig. 5 and Supplementary Data 5. Source data are provided as a Source Data file.

Fig. 5. S1P signaling via SREBP1 and PPARγ maintains membrane homeostasis in human/mouse cells.

A Average GP index from several images of WT MEFs treated with vehicle, betulin 1 μM (inhibitor of SREBP maturation) and PA 200 μM ± betulin 1 μM. n = 15 separate images analyzed for each condition. B Average GP index from several images of NT, AdipoR2 and SREBF1 siRNA HEK293 cells treated with PA 200 µM ± S1P 1 µM and ± OA 200 µM. For each condition from left to right, n = 10, 10, 10, 15, 15 and 10 separate images analyzed. C Western-Blot and quantification (n = 3 experiments) of NT and SREBF1 siRNA HEK293 cells treated with vehicle and PA 400 µM ± S1P 1 µM. The red arrow points to the precursor SREBP1 band. Note the presence of an unspecific band just below SREBP1 band. n = 3 biologically independent replicates per condition. Uncropped blots in Source Data. D, E Average GP index from several images of WT and R2KO MEFs treated with vehicle, PA (200 or 50 µM) ± T0070907 1 µM (PPARγ antagonist). In D, n = 10 separate images analyzed. In E, n = 15 for vehicle and 18 for T0070907. Representative images of D are shown in Supplementary Fig. 7A. F Average GP index from several images of NT and PPARG siRNA HEK293 cells treated with vehicle and PA 200 µM ± S1P 10 or 1 µM. n = 20 separate images analyzed for each condition. G Average GP index from several images of NT and RXRA siRNA HEK293 cells treated with vehicle and PA 200 μM ± S1P 1 μM. Data are represented as mean ± SEM. t-tests (two sided and assuming normality) were used to identify significant differences between treatments. *p < 0.05, **p < 0.01, ***p < 0.001. See also Supplementary Figs. 6, 7 and Supplementary Data 5. Source data are provided as a Source Data file.

Fig. 6. SCD is the final protein effector of the AdipoR2-S1P pathway.

A, B Western-Blot and quantification of WT and DKO MEFs in A and NT and AdipoR2/PPARγ/SCD siRNA in HEK293 cells in full media or basal media ± PA 200 µM. Uncropped blots in Source Data. C–F PA (16:0) and OA (18:1) abundance (mol%) and in the PC of HEK293 cells treated with different siRNA and PA 200 μM ± S1P 1 μM. For C and D, n = 4 independent biological replicates per condition. For E and F, n = 3 independent biological replicates per condition. Related lipidomics are shown in Supplementary Fig. 8M, N and in Data 6. G Average GP index from several images of NT and SCD siRNA HEK293 cells treated with vehicle and PA 200 µM ± S1P 10 µM and ± OA 200 µM. n = 15 separate images analyzed for each condition. Data are represented as mean ± SEM. t-tests (two sided and assuming normality) were used to identify significant differences between treatments. *p < 0.05, **p < 0.01, ***p < 0.001. See also Supplementary Fig. 8 and Supplementary Data 5–7. Source data are provided as a Source Data file.

Fig. 7. Model.

Membrane rigidification activates a ceramidase activity intrinsic to AdipoR2, resulting in the generation of sphingosine 1-phosphate (S1P) that separately activates the SREBP1 and PPARγ transcription factors. SREBP1 and PPARγ together promote increased transcription of SCD, resulting in increased levels of unsaturated fatty acids and restored membrane fluidity.