Inhibiting Phosphatidylcholine Remodeling in Adipose Tissue Increases Insulin Sensitivity Free

Phospholipids are a major class of lipids on cell membranes (1) and in plasma lipoproteins (2). Polyunsaturated fatty acids are usually esterified at the sn-2 position, whereas saturated and monounsaturated fatty acids are usually esterified at the sn-1 position (3,4). Seventy percent of phospholipids are phosphatidylcholines (PCs). PCs are synthesized de novo by the Kennedy pathway (5). The asymmetrical distribution of fatty acids at the sn-1 and sn-2 positions is maintained in part by a deacylation–reacylation process first proposed by Lands 60 years ago (Lands cycle) (3,6). The deacylation step of this cycle is catalyzed by calcium-independent phospholipase A2, which removes saturated or monounsaturated fatty acids from the sn-2 position of PCs. The reacylation step of the cycle is catalyzed by lysophosphatidylcholine acyltransferase (LPCAT), which adds polyunsaturated free fatty acids at the sn-2 position of PCs.

PCs within the membranes of mammalian cells have considerable structural diversity (7). Polyunsaturated PCs have kinks that prevent the molecules from packing together, enhancing membrane fluidity. Alteration in cell membrane PC saturation has been implicated in a variety of metabolic disorders, including diabetes, obesity, and heart disease (8,9). Previous studies have demonstrated that membrane PC composition is regulated by LPCATs (10,11). Thus, LPCAT activities are important for maintaining cell membrane structure.

LPCAT has four isoforms (11), LPCAT1 to LPCAT4. LPCAT3 plays an important role in the function of the liver and small intestine (12,13), because it is the major LPCAT isoform expressed in those organs (14,15). We previously reported that systemic, intestine-specific, or liver-specific deletion of Lpcat3 in mice affects lipidemia, and the phenotypic variations are attributable to a reduced amount of polyunsaturated PCs on the plasma membranes of enterocytes and hepatocytes (14,15). LPCAT3 expression has been directly linked to the biosynthesis of inflammatory lipid mediators in humans (16). Notably, LPCAT3 expression was associated with a protective effect on endoplasmic reticulum stress in response to saturated fatty acids in vitro and in mouse models of metabolic diseases in which LPCAT3 expression was directly associated with a decrease in hepatic inflammation (17).

In this study, we identified LPCAT3 as the major isoform of LPCAT in adipose tissue. Adipose tissue is a metabolically active endocrine organ that plays a critical role in regulating systemic lipid and glucose homeostasis (18) and modulates insulin action by secreting adipokines (19). Although LPCAT3 function is well studied in various metabolic tissues, the effect of LPCAT3 in adipose tissue remains unknown.

In this study, we used a lipidomic analysis of white adipose tissue (WAT) from adipocyte-specific Lpcat3 (aLpcat3)–knockout (KO) mice and found that several species of polyunsaturated PCs were significantly reduced. With the guidance of RNA sequencing (RNA-seq) and adipokine profiling, we tested our hypothesis that LPCAT3-mediated PC remodeling affects insulin receptor (IR) phosphorylation by modulating adipocyte plasma membrane lipid organization and thus influencing insulin sensitivity and glucose tolerance. Here, we present evidence that genetic and diet-induced mouse models with high LPCAT3 activity can inhibit insulin signaling in adipose tissue. Specifically, supplementing mature adipocytes from humans or mice with polyunsaturated PCs led to insulin resistance in vitro. Our comprehensive mechanistic study supports our proposal that manipulating membrane phospholipid saturation by regulating LPCAT3 activity is a potential strategy to treat insulin resistance.

Lpcat3-Flox mice (14) were crossed with adiponectin-Cre transgenic mice (The Jackson Laboratory) to establish Lpcat3-Flox/adiponectin-Cre mice (Supplementary Fig. 1). All mice used had C57BL/6J genetic backgrounds, and both male and female mice were used in this study. Mice were fed a standard normal chow diet (cat. no. 5053; PicoLab Rodent Diet), a Western diet (cat. no. TD.88137; Envigo) for 8 weeks, or a high-fat/high-calorie (HFHC) diet (cat. no. D12331; Research Diets) for 16 weeks. All animal experiments were approved by the Institutional Animal Care and Use Committee of The State University of New York Downstate Health Sciences University.

Fresh human adipose tissue samples were obtained from a donor patient from the New Jersey Organ and Tissue Sharing Network who was screened to be free of diabetes and cancer, was not a cigarette or heavy alcohol user, and was not taking medications known to alter metabolism. Fresh WAT was used to isolate mature adipocytes.

For insulin signaling analysis by Western blot, mice were fasted overnight and then injected intraperitoneally with insulin (5 units/kg body weight; Sigma-Aldrich) or PBS as a positive control 30 min before sacrifice. Adipose tissue was immediately dissected and frozen separately using dry ice. Approximately 100 mg subcutaneous WAT (sWAT) was homogenized and lysed in radioimmunoprecipitation assay buffer with phosphatase inhibitors (Roche) and proteinase inhibitors (Roche). The following primary antibodies are summarized in Supplementary Table 1. The blotting membrane was stained with Ponceau S (Sigma-Aldrich) for total protein normalization.

Total RNA was extracted from cells or tissue using TRIzol (Thermo Fisher Scientific), and cDNA was synthesized with a reverse transcription kit (Applied Biosystems). Quantitative PCR was performed using the SYBR Select Master Mix Kit (Applied Biosystems). Relative gene expression was normalized to 36B4 or Gapdh and calculated using the ΔΔCt method (Supplementary Research Design and Methods). Primer sequences are summarized in Supplementary Table 2.

As previously described, PC subspecies were measured using 200 mg sWAT by liquid chromatography–coupled tandem mass spectrometry (20).

LPCAT3 activity was measured in accordance with a published protocol (20) (Supplementary Research Design and Methods).

Mice were fasted overnight before intraperitoneal injection of glucose (2 g/kg body weight) for the glucose tolerance test or insulin (0.75 units/kg body weight) for the insulin tolerance test. Blood glucose was measured by tail vein bleeding and analyzed using a blood glucose monitoring system at 0, 15, 30, 60, and 120 min after injection.

Plasma insulin was quantified by insulin mouse ELISA kit (cat. no. EMINS; Invitrogen) in accordance with the manufacturer’s instructions (Supplementary Research Design and Methods).

Mouse stromal vascular fractions (SVFs) were isolated from 5-week-old aLpcat3-KO mouse sWAT (21). Mouse SVFs or human white subcutaneous preadipocytes (cat. no. CSC-C4010×; Creative Bioarray) were differentiated into mature adipocytes as previously described (21) (Supplementary Research Design and Methods). Mature human adipocytes were isolated from fresh adipose tissue as previously described (22) (Supplementary Research Design and Methods). Mature adipocytes were cultured in maintenance medium containing DMEM, 0.5 μg/mL insulin, 10% FBS, and 1% penicillin-streptomycin.

For insulin sensitivity analysis, induced mature adipocytes or isolated mature human adipocytes were cultured in insulin-free serum-free DMEM and treated with 0, 10, 50, or 100 μmol/L polyunsaturated PCs including 18:0/20:4 PC (cat. no. 850469; Avanti) or 18:0/18:2 PC (cat. no. 850468; Avanti) for 24 h and then stimulated with 100 nmol/L insulin for 30 min. Cells were washed with cold PBS and prepared for immunoblotting.

PC-treated cells were stimulated with 100 nmol/L insulin for 30 min and then cultured with 1 mmol/L 2-deoxyglucose for 20 min. Next, we measured 2-deoxyglucose-6-phosphate using the bioluminescent Glucose Uptake-Glo Assay Kit (cat. no. J1341; Promega) in accordance with the manufacturer’s instructions. Luminescence was recorded with 0.3- to 1-s integration on a luminometer.

Adipokines were measured in 70 μL pooled plasma samples from HFHC diet–fed aLpcat3-KO and wild-type (WT) mice (n = 6; male) using the Mouse Adipokine Array Kit (cat. no. ARY013; R&D Systems) in accordance with the manufacturer’s instructions (Supplementary Research Design and Methods).

Mouse sWAT was fixed in 4% formalin overnight at 4°C, and 10 μm paraffin sections were prepared. Each slice was deparaffinized and stained for immunofluorescence staining and hematoxylin-eosin staining. Primary antibodies are summarized in Supplementary Table 1. The rabbit monoclonal antibody immunoglobulin G isotype was used as a negative control. To minimize observer error, images were chosen and analyzed in a blind manner (Supplementary Research Design and Methods).

Adipose tissue lipid rafts were isolated as previously described (23–25) (Supplementary Research Design and Methods).

sWAT isolated from 2-month-old mice (n = 4 per group; male) was sent to Azenta Life Sciences for RNA extraction, library preparation, and standard RNA-seq. The raw data of sequence reads were processed using Trimmomatic (version 0.36) and further mapped to the Mus musculus GRCm38 reference genome (Ensembl [version 89]) using the STAR aligner (version 2.5.2b) (26). Each sample had at least 28,300,000 unique mapped reads. Unique gene hit counts were calculated using featureCounts from the Subread package (version 1.5.2). Gene hit counts were extracted and compared between each group using DESeq2 (version 1.30.1) (27). The Wald test was used to generate P values and log twofold changes. P values were adjusted using the Benjamini-Hochberg method. Genes with an adjusted P value of <0.05 and absolute fold change of >1.5 were considered differentially expressed genes for each comparison. Gene ontology functional enrichment analysis was performed using the PANTHER GO-slim biological process database (https://geneontology.org/) with default parameters (28,29). The dot enrichment plot was plotted by SRplot (https://www.bioinformatics.com.cn/srplot), an online platform for data analysis and visualization (adjusted P < 0.05).

The single-cell suspension from the sWAT SVF was sent to the Columbia Genome Center at Columbia University for library construction and further sequencing. The raw single-cell RNA-seq (scRNA-seq) data were processed using Cell Ranger (version 6.1.2; 10× Genomics). Filtered feature-barcode matrix files were obtained from the Cell Ranger pipeline, which contained only barcodes associated with cells. Additional analyses were performed using the R package Seurat (version 4.10). Cells with <300 genes and a total of fewer than three molecules detected within the cell were removed from the data, and gene counts were normalized and scaled using sctransform. We integrated our data from two conditions into a single analysis using an algorithm implemented in Seurat (30). We then applied a graph-based clustering algorithm to group cells into different clusters using the FindClusters function. We used cluster tree analysis (clustree function) to assess relatedness between clusters and annotate the clusters using known markers and data from relevant single-cell studies.

To identify genes that were differentially expressed between clusters, we compared the expression of each gene in each cluster with the same gene in all other clusters using a Wilcoxon rank sum test implemented in the FindAllMarkers function of Seurat. We set the min.pct argument to 0.25 (to include only genes with at least a 25% difference in mean expression) and the logfc.threshold argument to 0.25 (log fold change of at least 0.25 between the clusters).

Each experiment was repeated at least three times. Statistical analyses were performed using GraphPad Prism software (version 8.0). Data between two groups were analyzed by the unpaired two-tailed Student t test. For multiple group comparisons, data were analyzed by one-way ANOVA followed by the Tukey post hoc multiple comparisons test. Data are presented as mean ± SEM. Statistical significance was set at P < 0.05. Standard symbols are presented as follows: P > 0.05, not significant; *P < 0.05; **P < 0.01; ***P < 0.001; and ****P < 0.0001.

RNA-seq data sets were submitted to the National Center for Biotechnology Information Gene Expression Omnibus database under the accession number GSE235304. All other data sets generated and/or analyzed in this study are available from the corresponding author upon request.

LPCAT3 is the major LPCAT isoform expressed in WAT and brown adipose tissue (BAT) (Fig.1A). We crossed Lpcat3-Flox mice with adiponectin-Cre transgenic mice to generate aLpcat3 KO mice (Supplementary Fig. 1). There were no statistically significant differences in mouse body mass, phospholipid, triglyceride, plasma cholesterol, or apolipoprotein levels between WT control mice and aLpcat3-KO mice when mice were fed a normal chow diet (Supplementary Fig. 2A–E). Expression levels of several genes involved in sWAT de novo lipogenesis, fatty acid oxidation, and lipolysis processes were unchanged between the two groups (Supplementary Fig. 2F). We measured total LPCAT activity in sWAT, epididymal WAT, and BAT from male aLpcat3-KO and WT mice. LPCAT activity was reduced by 65–70% in these adipose tissues compared with controls, whereas no changes were observed in the liver or skeletal muscle (Fig.1B and C). Similar results were obtained in female mice (data not shown). To evaluate the effect of LPCAT3 deficiency on PCs in adipose tissue, we measured PC subspecies using liquid chromatography–coupled tandem mass spectrometry. The concentrations of multiple sn-2 polyunsaturated PCs were significantly lower in aLpcat3-KO male mice compared with WT mice, including 16:0/20:4, 18:0/18:2, 18:0/20:3, 18:0/20:4, and 18:1/20:4 PC (Fig.1D). The concentrations of two major lysoPCs (16:0 and 18:0) were unchanged in aLpcat3-KO male mice compared with WT mice, whereas the concentrations of two minor lysoPCs (16:1 and 18:1) were greater in aLpcat3-KO male mice compared with WT mice (Fig.1E). The ratio of lysoPC to PC was elevated when LPCAT3 was absent (Fig.1F), suggesting that LPCAT3 can influence adipose tissue phospholipid saturation.

To investigate the role of LPCAT3 in adipose tissue, we performed RNA-seq of sWAT from mice fed a normal chow diet to assess the overall gene expression differences between aLpcat3-KO and WT mice. The absence of LPCAT3 in adipose tissue altered the expression of 594 genes (adjusted P < 0.05 and ∣FC∣ > 1.5); 414 were upregulated, and 180 were downregulated (Fig.2A). An overview of the top 30 up- and downregulated genes is given in Fig.2B. Notably, PANTHER GO-slim analysis revealed that lipid- and glucose metabolism–related biological processes were altered in aLpcat3-KO mice (Fig.2C), including response to insulin, the fatty acid catabolic process, and the glucose metabolic process. These results indicate that LPCAT3 may have a critical role in regulating insulin signaling.

To better understand how LPCAT3 affects cellular heterogeneity and function within adipose tissue, we profiled the nonadipocyte fraction, also known as the SVF, from sWAT using scRNA-seq. By clustering gene expression profiles, we identified seven cell types in the SVF: adipose-derived stem and progenitor cells (ASPCs), macrophages, dendritic cells, mesothelial cells, T cells, mast cells, and endothelial cells (31–33) (Fig.2D and Supplementary Fig. 3A). There was no major difference in abundance of each cell type when we compared the composition of the SVF from sWAT in aLpcat3-KO and WT mice (Supplementary Fig. 3B).

ASPCs play a critical role in adipose tissue homeostasis. Gene ontology enrichment analysis of genes in ASPC clusters that were differentially expressed between aLpcat3-KO and WT mice revealed over-representation of genes involved in pathways related to transmembrane receptor signaling and response to growth factor stimulus (Fig.2E). Single-cell transcriptome analysis showed that ASPCs are not homogeneous; they form three distinct clusters (ASPC groups 1–3) (Supplementary Fig. 3C). Group 1 cells expressed Dpp4 and Pi16, markers of adipose stem cells (Supplementary Fig. 3D) (34,35). Col4a1, the discriminative marker of matrix fibroblasts (that have properties of adipose stem cells), was also highly expressed in group 1 ASPCs (36). Group 2 cells expressed Fabp4, Col4a1, and Lpl, markers of preadipocytes (Supplementary Fig. 3D) (35,37,38). Preadipocytes are characterized by higher adipogenic capacity than adipose stem cells (34,39,40). Group 3 cells, comprising the smallest group, expressed the same canonical mesenchymal progenitor markers as cells in groups 1 and 2 and also highly expressed mt-Co1, mt-Co2, and Xist, genes characteristic of high mitochondrial activity that could be indicators of apoptotic processes (Supplementary Fig. 3A and D). However, group 3 did not show a strong association with previous classifications, and the function of these cells needs to be further studied. There were some differences in the compositions of these subpopulations between the aLpcat3-KO and WT groups (Supplementary Fig. 3E), indicating an increase in the fraction of preadipocytes in LPCAT3-deficient mice; gene ontology molecular function analysis of each group also indicated that the binding activity of insulin-like growth factor (IGF) is altered when LPCAT3 is deficient (Supplementary Fig. 3F). However, we need to confirm that these differences are caused by LPCAT3 deficiency and not by variation in sample preparation or other confounding factors.

We also assessed plasma adipokine expression using a mouse adipokine protein array in mice fed an HFHC diet for 8 weeks. Interestingly, the expression of fibroblast growth factor 21, IGFs, and IGF-binding proteins was higher in plasma samples from aLpcat3-KO mice compared with those from WT mice, but adiponectin and leptin levels were unaffected (Fig.2F). Fibroblast growth factor 21 protects against systemic insulin resistance (41). IGFs share significant structural homology with insulin (42), and both IGF-I and IGF-II promote glucose clearance (43). Taken together, the results of our plasma adipokine profiling imply that aLPCAT3 deficiency may influence insulin signaling and systemic glucose metabolism, which is consistent with our bulk RNA-seq analysis results.

We assessed fasting glucose levels in normal chow diet–fed aLpcat3-KO and WT mice, and no significant differences were found between the groups (Fig.3A). However, aLpcat3-KO mice showed improved glucose tolerance and clearance, as measured by glucose and insulin tolerance tests, respectively (Fig.3B and C), suggesting LPCAT3 deficiency in adipose tissue enhances systemic insulin sensitivity. To gain a better understanding of the role of LPCAT3 in insulin sensitivity, we next assessed insulin signaling by Western blot in various adipose tissues of aLpcat3-KO and WT mice after injection of insulin or saline as a basal control. Under the stimulation of insulin, LPCAT3 deficiency dramatically induced tyrosine-phosphorylated IR levels in sWAT without affecting total IR levels (Fig.3D). The downstream targets of insulin signaling pathways, phosphorylated AKT and caveolin-1, were also much higher in the sWAT of aLpcat3-KO mice compared with that of WT mice (Fig.3D). Similar to the results in sWAT, insulin-stimulated IR and AKT phosphorylation were also significantly higher in BAT, but expression of caveolin-1 was not affected (Fig.3E). Consistent with the results from LPCAT activity, insulin-stimulated AKT phosphorylation in the liver and skeletal muscle was not different between the groups (Supplementary Fig. 2G). Collectively, these results indicate that aLpcat3 deletion improves systemic insulin sensitivity caused by enhanced insulin signaling in adipose tissue alone.

To gain insight into the effect of LPCAT3 on insulin resistance, aLpcat3-KO and WT mice were fed an HFHC diet for 16 weeks, starting at age 8 weeks, to induce obesity and impaired glucose tolerance. Both groups of mice had a similar time-dependent increase in body mass (Supplementary Fig. 4A) and sWAT and epididymal WAT mass (Supplementary Fig. 4B). Moreover, the depletion of LPCAT3 in adipose tissue did not affect the size or number of adipocytes in either group of mice fed an HFHC diet (Supplementary Fig. 4C and D). No significant differences were observed in plasma levels of phospholipids, triglycerides, or cholesterol in aLpcat3-KO mice compared with WT mice fed an HFHC diet (Supplementary Fig. 4E–G), similar to the results for mice fed a normal chow diet. Although there were no significant changes in fasting plasma glucose (Fig.4A) or free fatty acid levels (Supplementary Fig. 4H) in aLpcat3-KO mice compared with WT mice, fasting plasma insulin levels were dramatically reduced in aLpcat3-KO mice compared with WT mice (Fig.4B). Furthermore, aLpcat3-KO mice displayed a marked improvement in glucose and insulin tolerance (Fig.4C and D), similar to the results for mice fed a normal chow diet.

We next examined the effect of LPCAT3 deficiency on adipose tissue insulin signaling by Western blot. As expected, we found a robust induction of IR and AKT phosphorylation in aLpcat3-KO mice compared with WT mice (Fig.4E). We also observed a significant induction of total IR but not AKT protein levels in aLpcat3-KO mice (Fig.4E). In line with this observation, the relative expression of caveolin-1 was significantly higher in aLpcat3-KO mice compared with WT mice (Fig.4E). Collectively, we suggest that LPCAT3 deficiency in adipose tissue attenuates HFHC diet–induced insulin resistance by increasing IR activation and expression.

To confirm our observations, we used immunofluorescence staining for IR and caveolin-1 on sWAT and found that the fluorescence intensities of IR and caveolin-1 were greatly enhanced on the adipocyte plasma membranes of aLpcat3-KO mice compared with on those of WT mice (Fig.4F). We also stained for glucose transporter 4 (GLUT4), a well-known downstream factor in insulin signaling (44,45). GLUT4 on adipocyte plasma membranes was highly induced in the absence of LPCAT3, suggesting the improvement of glucose uptake in adipose tissue in aLpcat3-KO mice (Fig.4F and Supplementary Fig. 4I).

To validate the role of LPCAT3 in adipocyte insulin action, we processed to knock down Lpcat3 expression using adenovirus-mediated shRNA against LPCAT3 in differentiated mature adipocytes from the WT mouse SVF (Fig.4G). LPCAT3 deficiency led to the significant induction of IR and AKT activation (Fig.4H). However, LPCAT3 knockdown in mature adipocytes did not affect lipogenesis or lipolytic pathways, which is consistent with observations in vivo (Supplementary Fig. 5A and B). These observations support our findings that inhibition of LPCAT3 activity improves adipose tissue insulin sensitivity.

Next, we investigated the mechanism of LPCAT3 deficiency–mediated induction of insulin sensitivity. We focused on the lipid rafts in which IR and caveolin-1 are located. We extracted lipid rafts from sWAT of aLpcat3-KO and WT mice fed a normal chow diet. As shown in Fig.4I, lipid rafts (fractions 3–5) and nonrafts (fractions 10–12) were distinguished based on the expression of caveolin-1 and Lyn kinase by Western blot. We next examined the phosphorylation of IR in lipid rafts (fractions 3–5) and found that higher amounts of activated IR and caveolin-1 were expressed in sWAT from aLpcat3-KO mice than in that from WT mice (Fig.4J). Moreover, LPCAT3 deficiency induced the enrichment of cholesterol and sphingomyelin in the lipid rafts on cell membranes (Supplementary Fig. 4J), suggesting an enlargement of the lipid rafts mediated with LPCAT3 deficiency.

To further evaluate the relationship between LPCAT3 activity and insulin resistance, we used genetic or diet-induced mouse models of insulin resistance: ob/ob (obese mice), db/db (diabetic mice), and HFHC diet–fed WT mice. First, we found that LPCAT3 mRNA levels and activity were upregulated in WAT and BAT of WT mice fed an HFHC diet compared with those of mice fed a normal chow diet (Fig.5A and B and Supplementary Fig. 6A), although the changes in LPCAT3 mRNA levels in BAT did not reach statistical significance (Fig.5A and B). Furthermore, we compared LPCAT3 mRNA levels and activity in sWAT in WT, ob/ob, and db/db mice and found that both ob/ob and db/db mice had significantly higher LPCAT3 mRNA levels and activity than WT mice (Fig.5C and D and Supplementary Fig. 6B). As expected, activation of IR and AKT in ob/ob and db/db adipose tissue was also dramatically inhibited (Fig.5E). Collectively, induction of LPCAT3 expression could be one of the reasons for the impaired insulin signaling observed in ob/ob and db/db mice and in mice with dietary-induced insulin resistance.

We cultured mature human or mouse adipocytes in vitro and treated them with polyunsaturated PCs to examine insulin signaling. First, we isolated the SVF from aLpcat3-KO mouse sWAT and induced preadipocyte differentiation. After the preadipocytes became mature adipocytes (Supplementary Fig. 7A), we treated the cells with different concentrations of 18:0/18:2 PC, a major product of LPCAT3 activity (46), and stimulated the cells with insulin or saline as a basal control. Under basal conditions, we did not detect phosphorylated IR or phosphorylated AKT. Under insulin-stimulated conditions, we clearly detected phosphorylated IR, phosphorylated AKT, and GLUT4, and all these factors were inhibited by 18:0/18:2 PC in a dose-dependent manner (Fig.6A). In line with this, 18:0/18:2 PC treatment also inhibited the ability of mature adipocytes to take up glucose in a dose-dependent manner (Fig.6B).

Next, we sought to determine the consequence of polyunsaturated PC treatment on mature human adipocytes, which were derived from human adipose tissue preadipocytes (Supplementary Fig. 7B) or directly isolated from human WAT. We found that 18:0/18:2 PC treatment significantly attenuated insulin signaling by inhibiting the phosphorylation of IR and AKT and by reducing total IR content (Fig.6C and D). We also treated two mature human adipocytes with different concentrations of 18:0/20:4 PC, another major product of LPCAT3 activity (46), and found that the PC also could suppress IR and AKT activation (Supplementary Fig. 8A and B).

Adipose tissue is a key regulator of energy balance, plays an important role in lipid storage, and secretes a wide range of adipokines into the circulation that influence systemic metabolism (47). In the current study, we demonstrated that blocking adipocyte PC remodeling in mice promotes systemic insulin sensitivity, primarily by reducing the amount of polyunsaturated PCs in adipocyte plasma membranes.

One of the key findings of this study is that LPCAT3 is the major isoform of the LPCAT enzyme in adipose tissue, contributing >90% of the PC-remodeling activity. Adiponectin-Cre–mediated Lpcat3 deletion significantly reduced adipose tissue LPCAT3 activity and reduced polyunsaturated PC levels on the plasma membranes of adipocytes. Many reports, including ours (48), have indicated that the adiponectin-Cre–mediated approach can reduce protein or enzyme activity by ∼90%; however, we achieved only a 65 to 70% reduction of LPCAT3 activity in adipose tissue. Nevertheless, this reduction built the foundation for this study.

RNA-seq analysis showed that LPCAT3 deficiency resulted in significant upregulation of 414 genes and downregulation of 180 genes in adipose tissue. Importantly, PANTHER GO-slim analysis revealed adipocyte enrichment of lipid- and glucose metabolism–related genes in the upregulated gene set, indicating that LPCAT3 activity plays a critical role in regulating insulin signaling. This was confirmed by the adipokine array. We also isolated the SVF from aLpcat3-KO and WT mice and performed unbiased scRNA-seq to explore cellular diversity in adipose tissue microenvironments. We did not observe dramatic changes in cell population structure, because LPCAT3 depletion may affect only mature adipocytes that are not part of the SVF. Although the influence of LPCAT3-deficient mature adipocytes on the SVF could be minor, there was a noticeable increase in the preadipocyte fraction in LPCAT3-deficient mice. We observed an increased tendency for transmembrane receptor kinase signaling by gene ontology analysis of ASPCs, which implies insulin signaling in ASPCs, especially in group 1 ASPCs. ASPCs play a critical role in adipose tissue homeostasis and influence adipose tissue expansion during the transition into mature adipocytes (49). Insulin and IR play important roles in adipose tissue development (50). Moreover, ASPCs can differentiate into insulin-producing cells (51). Some studies suggest that abnormal insulin activity negatively affects ASPCs, causing the loss of their proliferative potential and weakening their differentiation potential (52). ASPCs also have a close linkage with human obesity (53). It is possible that LPCAT3 deficiency–mediated polyunsaturated PC reduction on the membrane of ASPCs could influence insulin signaling. This aspect deserves further investigation.

The fatty acids of membrane PCs in mammalian cells exhibit considerable structural diversity (7,54). Saturated PCs promote membrane molecules to pack tightly, thereby decreasing membrane fluidity. The interaction between cholesterol and sphingolipids drives the formation of lipid rafts, which constitute a specific microdomain on the plasma membrane (55). Rafts containing caveolin proteins, called caveolae, are particularly abundant in adipocytes (56). We previously found that manipulation of lipid organization in lipid rafts by reducing sphingomyelin content improves tissue and whole-body insulin sensitivity (57). In addition, caveolin-1 can interact with the IR and enhance insulin-mediated phosphorylation of IR substrate 1 (58). Although sphingolipids and cholesterol are the components of lipid rafts, many PCs are still the major lipids found in raft regions. Saturated and monounsaturated PCs are dominant in the lipid raft regions, whereas polyunsaturated PCs are dominant in nonraft regions (59,60). We hypothesize that increased concentrations of saturated PCs in adipose tissue result in enlargement of lipid rafts where IR is located, thereby enhancing insulin signaling. However, our results were separated by diet conditions. In mice fed a normal chow diet, we could clearly see activation of IR and AKT by their phosphorylation, but their total protein content was unchanged. This could be caused by a reduction in polyunsaturated PC amounts on lipid rafts, which could provide a microenvironment conducive to more IR activation. We indirectly confirmed this hypothesis by treatment of cultured mature human adipocytes with 18:0/18:2 PC. In contrast, the increased lipid content of the HFHC diet leads HFHC diet–fed mice to store more fat in adipose tissue. LPCAT3 deficiency can result in lipid rafts that are enlarged in size and that have greatly reduced amounts of polyunsaturated PCs and more tightly packed IRs. In agreement with these observations, we saw greater protein levels of activated IRs, IRs, and caveolin-1 in LPCAT3-deficient mice.

GLUT4 sits downstream of the insulin signaling pathway and promotes the uptake of glucose when it moves from intracellular storage vesicles to the plasma membrane after insulin signaling is activated (44,45). We confirmed the LPCAT3 deficiency–mediated induction of insulin signaling by GLUT4 immunostaining and polysaturated PC–associated glucose uptake. It is also possible that LPCAT3 deficiency–mediated enlargement of lipid rafts could provide more space for GLUT4 (61). Our glucose and insulin tolerance studies also confirmed that LPCAT3 depletion can promote insulin signaling.

Another key finding of this study is that LPCAT3 is upregulated under three metabolic disorders. Mice fed an HFHC diet had significantly higher LPCAT3 mRNA expression and adipose tissue LPCAT activity (LPCAT3 being the major isoform) than mice fed a normal chow diet. Liver X receptor is one of the important transcription factors that control LPCAT3 expression in liver tissue via a liver X receptor response element in the proximal promoter region of the Lpcat3 gene (17). Furthermore, Lpcat3 is a direct target of peroxisome proliferator–activated receptor δ, which suggests peroxisome proliferator–activated receptor δ functions as a regulator of phospholipid metabolism through LPCAT3 (62). Our results provide further evidence for both regulations. Leptin-deficient ob/ob mice and leptin receptor-deficient db/db mice are widely used as animal models to study human-like obesity and related metabolic disorders (63). We found adipose tissue LPCAT activity was significantly greater in ob/ob and db/db mice than in WT mice, and blockage of insulin signaling observed in these animals could be partially attributed to induction of LPCAT3 activity. The mechanisms of these phenomena also deserve further investigation.

We used two sources of human mature adipocytes to evaluate the human relevance of this study. Results from a previously published in vitro assay determined that 18:0/18:2 PC and 18:0/20:4 PC are major products of human LPCAT3 (46). We found that treatment of both types of mature human adipocytes with 18:0/18:2 PC and 18:0/20:4 PC significantly attenuated insulin signaling by reducing IR activation, IR content, and AKT activation (Fig. 6 and Supplementary Fig. 8). However, 18:0/20:4 PC was not as effective as 18:0/18:2 PC. This could be because 1) 18:0/18:2 PC is the most abundant PC in adipose tissue and could be the most abundant PC in lipid rafts, so its changes could have larger effects than 18:0/20:4 PC, and 2) exogenous PC supplementation may affect insulin signaling in a different manner than endogenous PC supplementation.

Although we focused our study on WAT, we also found that LPCAT3 is the major isoform in BAT and that LPCAT3 deficiency enhanced insulin signaling. Although both BAT and WAT secrete factors that act as hormones to affect liver and skeletal muscle metabolism, BAT differs fundamentally from WAT, because BAT can directly take up large quantities of glucose (64). Thus, BAT and WAT differ mechanistically in their glucose uptake and metabolism pathways. The insulin-sensitive GLUT4 translocation pathway operates in BAT, but in the context of a more active GLUT1 translocation pathway (GLUT1 being regulated by norepinephrine) (65). Nonetheless, insulin substantially stimulates glucose uptake in BAT (66). Silencing of GLUT4 and GLUT1 in brown adipocytes shows that both contribute extensively to overall glucose uptake (67). Thus, LPCAT3 deficiency in BAT can also promote systemic insulin sensitivity, as we observed in this study.

There are some limitations to this study. Although a 65 to 70% reduction of LPCAT3 activity in adipose tissue built the foundation for the current study, we still do not know why the approach could not achieve >90% reduction, as in other gene KO studies. Also, it should be important to study why 18:0/18:2 PC was more effective than 18:0/20:4 PC at blocking insulin signaling in mature adipocytes, although both PCs are the products of LPCAT3 activity.

Can inhibition of LPCAT3 be used as a treatment for metabolic diseases, such as insulin resistance and dyslipidemia? The current aLPCAT3 deficiency study, together with a previous skeletal muscle–specific LPCAT3 deficiency study (68), clearly indicates depletion of LPCAT3 in both tissues can significantly increase systemic insulin sensitivity. In our previous inducible small intestine– and liver-specific studies (14,15), we found that 1) small intestine LPCAT3 deficiency has a bigger effect on plasma lipid levels than that of liver deficiency, 2) the deficiency has an effect on lipid retention in both tissue, and 3) the deficiency has a marginal effect on endoplasmic reticulum stress. Taken together, we believe that inhibition of LPCAT3 activity could be a novel approach for treating insulin resistance and hyperlipidemia.

Funding. This work was supported by Veterans Affairs Merit Award 000900-01, National Institutes of Health (NIH) grants RO1 HL139582 and RO1 HL149730 grants to X.-C.J., and NIH grants P30 AG013319 and P30 AG044271 to the Lipidomics Core (X.H.).

Duality of Interest. No potential conflicts of interest relevant to this article were reported.

Author Contributions. M.H. and Z.L. performed 85% of the experiments, analyzed data, and modified the manuscript. V.S.K.T. and O.E. performed single-cell sequence analyses and modified the manuscript. M.P. and X.H. measured lipids using liquid chromatography–coupled tandem mass spectrometry and modified the manuscript. X.-C.J. conceived the ideas, designed and discussed the experiments, supervised progress, and wrote the manuscript. X.-C.J. 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.