Elevated Adipocyte Membrane Phospholipid Saturation Does Not Compromise Insulin Signaling

That saturated fatty acid (SFA) intake impacts insulin sensitivity was proposed nearly 100 years ago (1). When humans consume saturated fat, insulin function is impaired within hours (2), and higher levels of circulating SFAs correlate with an increased risk of developing insulin resistance and type 2 diabetes (3). Mechanistically, increased phospholipid acyl chain saturation is associated with decreased insulin sensitivity (4–6), and modulating the fatty acid composition of the plasma membrane bilayer alters insulin receptor accessibility, insulin binding, and insulin action in several cell types (7–10).

Similarly, it has been proposed that in obesity, white adipose tissue (WAT) insulin resistance is caused by an increased SFA incorporation in adipocyte membranes, leading to the activation of intracellular stress pathways that in turn inhibit insulin receptor signaling (11). Indeed, the levels of SFA-containing phosphatidylcholines (PCs), the most abundant mammalian phospholipid (12), decrease in healthy differentiating adipocytes, while the opposite is seen with unsaturated fatty acid (UFA)-containing PCs (13,14). Inhibiting the fatty acid stearyl coenzyme A desaturase (SCD) during differentiation leads to an increase in SFA-containing phospholipids and a decrease in insulin-mediated protein kinase B (AKT) phosphorylation (13). Conversely, genetic induction of ELOVL6 decreases the SFA-to-UFA ratio in adipocytes, leading to increased membrane fluidity and enhanced insulin signaling (15), while modulating phospholipid composition in rat adipocytes by increasing polyunsaturated fatty acid (PUFA) content in the diet improves all insulin-related functions (16).

Interestingly, diets varying in fat content and composition rarely impact SFA levels in adipocyte membranes, while PUFA-containing phospholipid levels can vary extensively (17). In fact, adipocyte phospholipid saturation remains tightly regulated, even during insulin resistance (18). During obesity, expansion of WAT leads to several changes in adipocyte membrane composition, namely, increases in PUFA-containing phospholipids (19).

We hypothesized that increased adipocyte phospholipid saturation could decrease adipocyte insulin sensitivity. To investigate this, we leveraged genetic models lacking adiponectin receptor 2 (ADIPOR2). ADIPOR2 has been described as a receptor for the adipokine adiponectin (20,21), acting as a ceramidase that regulates cellular ceramide levels (22). Importantly for the current study, loss of ADIPOR2 markedly increases fatty acyl saturation in membrane phospholipids in several cultured cell models (23–26). Here, we found that human and mouse ADIPOR2-knockout (KO) preadipocyte, adipocyte, and mouse WATs are excessively rich in SFA-containing phospholipids but show no defect in insulin signaling.

Human preadipocytes were isolated from subcutaneous WAT (scWAT) of healthy women undergoing elective fat removal at Sahlgrenska University Hospital in Gothenburg, Sweden, using an established protocol (27). Primary adipose-derived stromal vascular fraction cells were then immortalized with human telomerase reverse transcriptase, as previously described (28).

Human immortalized preadipocyte ADIPOR2-KO line was generated by electroporating ribonucleoproteins containing recombinant Streptococcus pyogenes Cas9 (TrueCut Cas9 Protein v2, A36498, Thermo Fisher) and single-guide (sg)RNA pair (CGAGCCAACAGAAAACCGAT and CAACTGGATGGTACACGAAG, TrueGuide Synthetic guide RNA, Thermo Fisher) targeting exon 1 of the human ADIPOR2 gene, following a recently described protocol (29).

Human adipose-derived stem cells (ASCs) were seeded at 56,000 cells/cm² in culture plates and kept in maintenance medium consisting of Endothelial Cell Growth Medium MV2 (PromoCell) supplemented with 0.05 mL/mL tetracycline (TET)-free serum, 5 ng/mL human recombinant epidermal growth factor, 10 ng/mL recombinant human basic fibroblast growth factor, 20 ng/mL insulin-like growth factor (long R3 IGF), 0.5 ng/mL human recombinant vascular endothelial growth factor 165, 1 µg/mL ascorbic acid, and 1% penicillin/streptomycin. At ∼90% confluency, cells were washed once in PBS before being switched to an induction medium: BM-1 (ZenBio) supplemented with 3% TET-free FBS (Thermo Fisher), 1% penicillin/streptomycin (Thermo Fisher), 1 μmol/L dexamethasone (Sigma-Aldrich), 0.5 mmol/L IBMX (Sigma-Aldrich), 100 nmol/L insulin (Actrapid, Novo Nordisk, Bagsvaerd, Denmark), and 1 μmol/L pioglitazone (Sigma-Aldrich) for 7 days, with medium change every second day. Cells were then switched to maintenance medium: BM-1 supplemented with 3% TET-free FBS, 1% penicillin/streptomycin, 1 μmol/L dexamethasone, and 100 nmol/L insulin (Actrapid) for 23 days, with medium change every second day. Cells were differentiated for 30 days, followed by experiments.

Termination of all mice (9:00–12:00 a.m.) was done under isoflurane anesthesia, with organs first being weighed, followed by immediate snap-freezing in liquid nitrogen and storage at −80°C until further analysis. Approximately 90 mg of adipose tissues were used for lipidomic analyses.

For insulin sensitivity studies, 12-week-old chow-fed male (four wild-type [WT], eight ADIPOR2-KO) and female mice (eight WT, eight ADIPOR2-KO) were fasted overnight for 12 h. The next morning, mice were acclimatized in the study room for 2 h and then injected intraperitoneally with saline (PBS, 10010023, Thermo Fisher) or 2.5 units/kg of insulin in saline (Actrapid), followed by termination after 10 min. Livers and adipose tissues were isolated, snap frozen in liquid nitrogen, and stored at −80°C until analysis. Group randomization was done based on body weight to obtain similar body weight averages between groups. End point signal was used to check for potential confounders, of which there were none.

The Supplementary Material includes additional methods, Appendix A includes supplemental material, and Appendix B includes all lipidomics data used in this study.

Metabolomics data are deposited at European Molecular Biology Laboratory-European Bioinformatics Institute (EMBL-EBI) MetaboLights (30) with identifier MTBLS6080 and can be accessed here: https://www.ebi.ac.uk/metabolights/MTBLS6080. Raw data are available from the corresponding authors upon reasonable request. Requests for human cell lines and genetic reagents should be addressed to K.P.

Primary human ASCs were isolated from the stromal vascular fraction of subcutaneous adipose tissue obtained during surgery. Human ASCs were immortalized with human telomerase reverse transcriptase, followed by clonal isolation, to generate a cell line for further studies (Fig. 1A). Our immortalized clone, termed “immortalized human preadipocytes,” retained the ability to differentiate (Fig. 1B). To generate a cell model with membrane phospholipids containing elevated levels of SFAs, we disrupted the ADIPOR2 gene using clustered regularly interspaced short palindromic repeats (CRISPR)/Cas9 (Supplementary Fig. 1A and B), resulting in a heterogenous pool of edited cells lacking ADIPOR2 (Fig. 1C and Supplementary Fig. 1C). This pool also showed vulnerability to the saturated palmitic acid (PA) (Fig. 1D), a hallmark phenotype of cells lacking ADIPOR2 (25), and did not upregulate ADIPOR1, even when challenged with PA (Fig. 1E). To avoid clonal isolation-derived artifacts (31), we opted to perform further experiments on this pool of edited cells.

As with other cell types (25,26), ADIPOR2-deficient (ADIPOR2-KO) human preadipocytes showed elevated levels of SFA-containing phospholipids compared with control cells and showed a much stronger increase in SFA-containing phospholipids than control cells in response to PA treatment (Fig. 2A). Total phospholipid amounts were also strongly elevated in ADIPOR2-KO cells treated with PA (Supplementary Fig. 2A); this was not due to any change in cell number (Supplementary Fig, 2B). This increase in phospholipids was mostly driven by an increase in SFA-containing PCs (Fig. 2B), and specifically PCs containing two PA moieties (16:0/16:0) (Fig. 2C). A similar increase was also observed in phosphatidylethanolamines (PEs), the second most abundant phospholipid class (Fig. 2B and C). Elevation of PA-containing PCs in the ADIPOR2-KO pool was already evident in normal culture conditions (Supplementary Fig. 2C) but was exacerbated with ≥100 μmol/L PA treatments (Fig. 2C). There was also upregulation of SFA-containing diacylglycerols (DAGs) in the ADIPOR2-KO pool, which was not aggravated by PA treatment (Supplementary Fig. 2D).

We then used the Laurdan dye method to evaluate membrane packing (32). As expected, we observed increased membrane rigidity in the ADIPOR2-KO cells challenged with ≥100 μmol/L PA compared with control cells (Fig. 2D and E). Furthermore, PA-treated ADIPOR2-KO cells exhibited decreased basal respiration, which is consistent with previously observed SFA toxicity (Fig. 2F and G) (25). This was not seen in normal culturing conditions (Supplementary Fig. 2E), indicating a specific mitochondrial respiratory defect in PA-treated ADIPOR2-KO cells. An established response to disrupted membrane homeostasis is activation of the endoplasmic reticulum (ER) unfolded protein response (ER-UPR) (33). Several of the ER-UPR response genes were strongly upregulated in the ADIPOR2-KO preadipocytes treated with ≥50 μmol/L PA (Fig. 2H), with WT cells showing similar increases only at higher PA doses (Supplementary Fig. 2F). To summarize, ADIPOR2-KO human preadipocytes contain an excess of SFA-containing phospholipids and are more susceptible to palmitate-induced lipotoxicity.

Next, we investigated whether increased SFAs in phospholipids affects insulin signaling in differentiated adipocytes. To do so, we differentiated ADIPOR2-KO and control preadipocytes for 1 month using a standard adipogenic cocktail (Fig. 3A). We found no differences in the differentiation capacity between genotypes, as assessed by lipid accumulation (Supplementary Fig. 3A) and the expression of adipocyte marker genes (Supplementary Fig. 3B). Several adipogenic markers showed differential expression between genotypes in PA-treated conditions, such as an upregulation of PPARG and downregulation of SCD and ELOVL6 (Supplementary Fig. 3C), as well as an increased level of ER-UPR markers in the ADIPOR2-KO cells (Supplementary Fig. 3D). Importantly, we verified that the ADIPOR2 locus remains ablated in the differentiated cells (Supplementary Fig. 3E).

Having established that the ADIPOR2-KO adipocytes differentiated as well as WT cells, we next tested whether their lipidome displayed increased SFA-containing phospholipids. In line with the preadipocyte findings, the ADIPOR2-KO adipocytes showed strong increases in SFA-containing phospholipids both in normal culture conditions and when 500 μmol/L PA was added compared with control cells (Fig. 3B and C). As with preadipocytes, the increase in phospholipid SFAs was mostly attributed to PC and PE that contained two PA chains being increased two- and threefold, respectively (Fig. 3D). Increasing levels of PA treatment also increased total phospholipids compared with controls (Supplementary Fig. 3F). In normal culture conditions, PEs, but not PCs, showed increased saturation compared with the control (Supplementary Fig. 3G). SFA-containing DAGs were upregulated in control and in PA-supplemented media, again as in preadipocytes, compared with control cells (Supplementary Fig. 3H).

We then used the nanoscale secondary ion mass spectrometry (NanoSIMS) method to spatially resolve exogenous [¹³C]palmitate in control and ADIPOR2-KO adipocytes. By combining electron micrographs and NanoSIMS images, we found that exogenously supplied palmitate accumulates at significantly higher levels within the cytoplasm of ADIPOR2-KO cells, suggesting that cytoplasmic organelle membranes (e.g., ER) are enriched in SFA-containing phospholipids. No differences were found between nuclei or lipid droplets (Fig. 3E and Supplementary Fig. 4A and B), and this method could not resolve the plasma membrane.

Considering these findings, we wanted to see whether the increased SFAs in phospholipids had an impact on insulin signaling in the ADIPOR2-KO adipocytes. Surprisingly, the ADIPOR2-KO adipocytes remained insulin sensitive, as determined by monitoring insulin-induced phosphorylation of AKT (Fig. 3F), and total insulin-mediated glucose uptake was also unchanged in ADIPOR2-KO adipocytes compared with controls (Fig. 3G). Basal (noninsulin-mediated) glucose uptake was, however, increased in ADIPOR2-KO adipocytes pretreated with 600 μmol/L PA compared with control adipocytes (Fig. 3H). Furthermore, other components of the insulin-signaling pathway, namely, extracellular signal–regulated kinase, insulin receptor, insulin receptor substrate 1, and insulin-like growth factor 1-receptor (34), all displayed the expected increase in phosphorylation in response to insulin in both control and ADIPOR2-KO cells, both in normal media and when pretreated with 600 μmol/L PA (Supplementary Fig. 3K). Importantly, PA uptake in ADIPOR2-KO adipocytes remained similar to that of control cells (Supplementary Fig. 3I), suggesting that observed differences in SFA-containing phospholipids in the ADIPOR2-KO cells were not due to changes in SFA uptake. Also, no differences could be measured with respect to cellular respiration between genotypes (Supplementary Fig. 3J), indicating that increased membrane saturation in adipocytes did not affect mitochondrial respiration, in contrast to the preadipocytes, where basal respiration was impaired in the ADIPOR2-KO cells challenged with PA. Adiponectin, a proposed ligand for ADIPOR2 (35), had no effect on lipid composition, insulin signaling, or expression of adipogenic markers in control or ADIPOR2-KO adipocytes (Supplementary Fig. 5A–C).

Next, we attempted to increase adipocyte phospholipid saturation by acutely inhibiting SCD. Treating adipocytes with an SCD inhibitor increased SFA in PC and PE (Supplementary Fig. 5D), resulting in the induction of ER-UPR response genes (HSPA5 and DDIT3) (Supplementary Fig. 5F). Importantly, SCD inhibitor treatment did not affect insulin-mediated AKT phosphorylation in adipocytes (Supplementary Fig. 5E). Altogether, these results show that insulin signaling is not impaired in ADIPOR2-KO human adipocytes despite elevated levels of SFA-containing phospholipids and ER-UPR induction.

To extend our findings to WAT, we leveraged a previously described ADIPOR2-KO mouse model (36). As in previous studies, ADIPOR2-KO mice had enlarged brains (Supplementary Fig. 6A), and males were sterile. We isolated, cultured, and differentiated stromal vascular fraction cells from scWAT of ADIPOR2-KO mice (Fig. 4A). To determine whether the effects of ADIPOR2 KO were conserved between human and mouse preadipocytes, we evaluated whether the mouse preadipocytes were also vulnerable to PA. Indeed, we observed increased membrane rigidification in ADIPOR2-KO mouse preadipocytes treated with PA compared with control preadipocytes (Supplementary Fig. 6B). Adipor2 gene expression was absent in these cells (Supplementary Fig. 6C). As in the human cells, ER-UPR markers were strongly elevated in the ADIPOR2-KO preadipocytes after PA pretreatment compared with controls (Supplementary Fig. 6D).

Like the human cells, the ADIPOR2-KO mouse preadipocytes differentiated normally (Fig. 4A) and contained comparable levels of lipid droplets (Fig. 4B) and similar expression of the adipogenic markers to WT controls under normal culturing conditions (Fig. 4C). Transcriptional changes in response to PA treatment were similar between genotypes, except for an upregulation of Adipoq and a downregulation of Lep in the ADIPOR2-KO model (Supplementary Fig. 6E). Having established that the ADIPOR2-KO mouse preadipocytes differentiated similar to WT, we evaluated the lipidome of these cells. The ADIPOR2-KO mouse adipocytes displayed strikingly elevated levels of SFA-containing phospholipids (Fig. 4D). The phospholipid species being elevated the most were SFA-containing PCs and PEs, with PC 16:0/16:0 being most increased (Fig. 4E). Total phospholipid levels were unaffected by genotypes and treatments (Supplementary Fig. 6F), and the total levels of DAGs were decreased in the ADIPOR2-KO mouse adipocytes irrespective of treatment (Supplementary Fig. 6G and H).

Next, we measured the response of differentiated mouse adipocytes to insulin. We found that insulin-induced AKT phosphorylation was similar between ADIPOR2-KO and control mouse adipocytes, with and without PA treatment (Fig. 4F). Noninsulin-mediated glucose uptake was also similar between genotypes irrespective of treatment (Fig. 4G), as was total insulin-mediated glucose uptake (Fig. 4H). Treating with PA did not increase ER-UPR markers in the ADIPOR2-KO mouse adipocytes compared with controls (Supplementary Fig. 6I). To summarize, mouse ADIPOR2-KO adipocytes were excessively rich in SFA-containing phospholipids, showed normal adipogenic gene expression, normal glucose uptake, and retained fully functional insulin signaling even in the presence of PA.

We next investigated whether our observed effects of Adipor2 deletion were present in WAT. We isolated WAT from ADIPOR2-KO mice, which had similar bodyweight (Fig. 5A) but lower WAT weight compared with WT controls (Fig. 5B), as previously reported (36). Consistent with the in vitro findings, the ADIPOR2-KO WAT had higher levels of SFA-containing phospholipids in both scWAT and epididymal (eWAT) fat compared with WT controls (Fig. 5C). The strongest effect could be seen in scWAT PC saturation (Fig. 5D), in which SFA levels were strikingly increased twofold compared with the WT group. No changes in total PE composition were observed between genotypes, but PE and PC 16:0/16:0 were increased in both scWAT and eWAT of ADIPOR2-KO mice compared with WT (Fig. 5D and E). Similar to the human adipocytes, total phospholipid levels were elevated in scWAT and eWAT of ADIPOR2-KO mice compared with the WT cohort (Supplementary Fig. 7A), and the excess SFA remained even after normalizing for the total phospholipid levels (Supplementary Fig. 7B). An increase in DAGs in scWAT was also observed in the ADIPOR2-KO mice compared with controls (Supplementary Fig. 7C), with changes in DAG composition in both adipose tissues (Supplementary Fig. 7D and E). Importantly, the composition of triacylglycerols (TAGs), which accumulate in lipid droplets and are not membrane components, did not differ between the ADIPOR2-KO and WT mice (Fig. 5F), indicating a specific role of ADIPOR2 in membrane homeostasis. Taken together, the lack of ADIPOR2 leads to markedly elevated levels of SFA-containing phospholipids in vivo, mirroring our in vitro findings.

Lipogenic, desaturation, and inflammatory markers were expressed at similar levels in ADIPOR2-KO and WT scWAT and eWAT. There were no changes in the expression of ER-UPR markers between genotypes (Fig. 6A), in contrast to what we observed in human adipocytes. Separately, we found that 12-week-old chow-fed male ADIPOR2-KO mice had a greater number of smaller eWAT adipocytes compared with controls (Fig. 6B–D), similar to what has previously been observed (36).

Next, we found that both fasted male and female ADIPOR2-KO mice displayed similar levels of AKT phosphorylation in response to exogenous insulin in scWAT, eWAT, and liver as the control group (Fig. 6E), despite being slightly leaner (Supplementary Fig. 8A). The ADIPOR2-KO mice also showed similar expression of lipogenic, desaturase, and inflammatory genes in liver and muscle compared with controls, with the only observed difference being the upregulation of Acsl4 in muscle (Supplementary Fig. 6B). ER-UPR markers showed no changes between genotypes (Supplementary Fig. 8B). Brown adipose tissue (BAT) and liver were similar in weight between genotypes (Supplementary Fig. 8C). In summary, the impact of ADIPOR2 KO on membrane saturation was remarkably consistent between human adipocytes, mouse adipocytes, and mouse WAT, and did not impair WAT insulin signaling.

Next, we attempted to further elevate SFA levels in WAT phospholipids in the ADIPOR2-KO mice by housing them at thermoneutrality to minimize SFA oxidation by BAT and feeding them a high-fat diet (HFD) rich in SFA for 3 months. ADIPOR2-KO mice weighed less than WT controls at an early age, but the weight of both genotypes became similar as the mice matured (Supplementary Fig. 9A). Body weight was comparable between the genotypes at the end of the study (Fig. 7A), with a decrease in scWAT weight in ADIPOR2-KO mice compared with controls (Fig. 7B). BAT weight was similar, while liver weight was decreased and brain weight was increased in the ADIPOR2-KO mice (Supplementary Fig. 9B). Importantly, we again observed the upregulation of SFA-containing phospholipids in ADIPOR2-KO adipose tissues compared with WT (Fig. 7C). Increased saturation was strongest in scWAT PCs (Fig. 7D), with PC 16:0/16:0 predominantly mediating this increase (Fig. 7E). Total phospholipids were similar between genotypes in both adipose tissues (Supplementary Fig. 9C). DAGs were slightly elevated in ADIPOR2-KO scWAT (Supplementary Fig. 9D), with increased saturation compared with controls (Supplementary Fig. 9E), while eWAT was largely unaffected (Supplementary Fig. 9F). SFA-containing TAG species showed no difference between genotypes (Supplementary Fig. 9G). The differences in lipid composition between control and ADIPOR2-KO mice were diminished after the HFD compared with the chow diet (Figs. 5A–D and 7A–D). Control and ADIPOR2-KO mice showed no differences in gene expression in WAT depots (Fig. 7F) or liver (Supplementary Fig. 9H). Similar to what was observed on chow, the ER-UPR was not elevated in adipose tissue (Fig. 7F), liver, or muscle (Supplementary Fig. 9H) of ADIPOR2-KO mice after the HFD feeding and thermoneutral housing (Fig. 7H and Supplementary Fig. 9H).

Lastly, no differences in oral glucose tolerance tests were observed between ADIPOR2-KO and WT mice in glucose clearance or insulin levels (Fig. 7G). Overall, and in agreement with the in vitro studies presented above, the marked excess in membrane phospholipid saturation in the ADIPOR2-KO mice did not impair WAT insulin sensitivity.

In this work, we used the ADIPOR2 gene KO model to increase SFA-containing phospholipids in cellular membranes of adipocytes in vitro and in vivo, to assess whether this increase would affect insulin signaling. ADIPOR2 KO resulted in an approximately twofold increase in SFA-containing PCs in human and mouse adipocytes as well as in mouse WAT, exhibiting a remarkably consistent effect across species in vitro and in vivo. Further, the membrane SFA content can be increased three- to fourfold by pretreating ADIPOR2-KO cells with PA. We found that human and mouse adipocytes, as well as mouse WAT, respond normally to insulin irrespective of SFA content in phospholipids, suggesting that the increase in adipocyte membrane phospholipid saturation per se is not causing insulin resistance in adipocytes.

In mouse and human preadipocytes, increased SFA in phospholipids resulted in ER-UPR activation, which is a consequence of increased membrane lipid saturation observed in many different cell types (37–44). Furthermore, activated ER-UPR is a feature of obesity and insulin resistance in human tissues (45–47). In our study, human ADIPOR2-KO adipocytes showed ER-UPR activation accompanied by an excess of SFA-containing phospholipids. In contrast, the ER-UPR and lipid-remodeling genes showed no differences between genotypes in mouse adipocytes or whole WAT, suggesting that mouse adipocytes respond more mildly to the membrane saturation. The ER-UPR results are consistent with the findings regarding cellular respiration: preadipocytes (Fig. 2F and G) but not differentiated adipocytes (Supplementary Fig. 3J) showed impaired respiration when treated with PA. Altogether, these results suggest that adipocytes are especially resilient to PA challenges in spite of the accumulation of SFA-rich phospholipids.

ADIPOR2-KO adipocytes showed no differences in SFA levels in TAGs compared with controls (Fig. 5F and Supplementary Fig. 9G). This contrasts with previous studies using SCD and ELOVL6 genetic models to modulate adipocyte membrane lipid composition (13,48,49). SCD and ELOVL6 are both key enzymes of the de novo lipogenesis pathway, and their reaction products are major constituents of adipocyte lipid droplets. As such, the ADIPOR2-KO model is unique in its way to specifically modulate the fatty acid composition of cell membranes but not lipid droplets.

An important limitation in the current study is that the lipidomics analyses were performed on whole cells and tissues, making it impossible to determine the subcellular location of the SFA-rich phospholipids. It is possible that the SFA-rich PC increases seen in the ADIPOR2-KO mice are located primarily in the ER, consistent with the NanoSIMS data (Fig. 3E and Supplementary Fig. 4B), while the plasma membrane, in which the insulin receptor is located, may successfully maintain composition homeostasis. Another limitation is that only a fraction of the cellular lipidome was measured. Therefore, there could potentially be other lipid species balancing out the increased SFAs in phospholipids.

Feeding the ADIPOR2-KO mice with an HFD in thermoneutrality did not exacerbate the excess of SFA-containing phospholipids. In fact, the largest differences in membrane composition in vivo were observed when the ADIPOR2-KO mice were kept on a chow diet. While the reasons for dietary differences in observed phenotypes are unclear, it is noteworthy that diet-induced changes in phospholipid SFA content is difficult to achieve in WT animals, even with diets ranging from 8 to 88% SFA in energy content (50). Therefore, the ADIPOR2-KO mouse represents a good model to study the consequences of excess membrane SFAs in vivo.

Previous studies have examined the metabolic consequences of Adipor2 ablation or overexpression in mice, although never with a focus on adipocytes. A first ADIPOR2-KO model was described in 2007 (51). These ADIPOR2-KO mice were generally healthy but had higher plasma insulin levels than control mice, suggesting some degree of insulin resistance. In contrast, and in agreement with our findings, three studies of the same ADIPOR2-KO mouse model that we used here showed that loss of Adipor2 protects against diet-induced obesity and insulin resistance (36,52,53). More recently, mice carrying floxed versions of the AdipoR2 locus have been generated to study hematopoiesis, with the key finding that ADIPOR2 promotes hematopoietic stem cell quiescence by suppressing inflammatory cytokine expression (54). In the future, it would be interesting to use this floxed Adipor2 model to study the consequence of Adipor2 loss in adipose tissue. Mouse models overexpressing Adipor2 in the adipose tissue have been generated and show increased ceramidase activity associated with reduced hepatic lipid accumulation (55). This study did not examine phospholipid composition but found that overexpression of Adipor2 results in a marked decreased in SFA-containing ceramides, which is consistent with the primary function of ADIPOR2 being to regulate SFA levels in membrane lipids.

In conclusion, we have shown that ADIPOR2-KO in cells and mouse WAT leads to an increased SFA content and rigidity in cell membranes and that insulin response and glucose handling are not impaired in adipocytes and mouse WAT in which there is a substantial increase in SFA-containing phospholipids.

Acknowledgments. The authors thank the senior staff at AstraZeneca Early Cardiovascular, Renal and Metabolism Department, Jacqueline Naylor, Bader Zarrouki, and David Baker, for their advice, comments, and resource management. The authors thank AstraZeneca Animal Sciences & Technologies staff for their in vivo support. The authors thank Therese Admyre, Johan Forsström, Sara Torstensson, and Renata Leke (AstraZeneca, Gothenburg, Sweden) for their support in performing the in vivo studies. Histology slides were prepared by Histocenter (Sweden), which the authors thank for their excellent services. The authors acknowledge the Center for Cellular Imaging at the University of Gothenburg (Sweden) and Stephanie Ling (AstraZeneca, Cambridge, U.K.) for their assistance in NanoSIMS sample preparation.

Funding. NanoSIMS data were obtained at the Chemical Imaging Infrastructure, joint infrastructure of Chalmers University of Technology and University of Gothenburg, supported by the Knut and Alice Wallenberg Foundation and hosted by the AstraZeneca BioVentureHub. This project has received funding from Stiftelsen för Strategisk Forskning (ID16-0049). C.B. is supported by Stiftelsen för Strategisk Forskning (ID18-0082).

Duality of Interest. H.P., K.P., S.B., A.A., R.N., L.L., M.G., A.-C.A., L.A., C.B., M.K., B.K., S.W., D.K., S.H., M.M., M.B.-Y., and X.-R.P. are presently employed by AstraZeneca and may be AstraZeneca shareholders. No other potential conflicts of interest relevant to this article were reported.

Author Contributions. H.P. prepared the manuscript. H.P., K.P., S.B., A.A., M.R., R.N., L.A., C.B., M.K., S.W., and D.K. performed experiments and analyzed the data. H.P., K.P., R.N., L.L., and M.S.G. performed and analyzed lipidomics. H.P., K.P., M.M., M.B.-Y., X.-R.P., and M.P. conceptualized the study. K.P., M.M., M.B.-Y., X.-R.P., and M.P. supervised the study. A.-C.A. performed and analyzed histopathology. M.K., B.K., and S.H. contributed technology. S.H. supervised lipidomics experiments. Other authors contributed to the review and editing of the manuscript. M.P. is the guarantor of this work and, as such, had full access to all the data in the study and takes responsibility for the integrity of the data and the accuracy of the data analysis.