AGPAT3 deficiency impairs adipocyte differentiation and leads to a lean phenotype in mice

Abstract

Acylglycerophosphate acyltransferases (AGPATs) catalyze the de novo formation of phosphatidic acid to synthesize glycerophospholipids and triglycerides. AGPATs demonstrate unique physiological roles despite a similar biochemical function. AGPAT3 is highly expressed in the testis, kidney, and liver, with intermediate expression in adipose tissue. Loss of AGPAT3 is associated with reproductive abnormalities and visual dysfunction. However, the role of AGPAT3 in adipose tissue and whole body metabolism has not been investigated. We found that male Agpat3 knockout (KO) mice exhibited reduced body weights with decreased white and brown adipose tissue mass. Such changes were less pronounced in the female Agpat3-KO mice. Agpat3-KO mice have reduced plasma insulin growth factor 1 (IGF1) and insulin levels and diminished circulating lipid metabolites. They manifested intact glucose homeostasis and insulin sensitivity despite a lean phenotype. Agpat3-KO mice maintained an energy balance with normal food intake, energy expenditure, and physical activity, except for increased water intake. Their adaptive thermogenesis was also normal despite reduced brown adipose mass and triglyceride content. Mechanistically, Agpat3 was elevated during mouse and human adipogenesis and enriched in adipocytes. Agpat3-knockdown 3T3-L1 cells and Agpat3-deficient mouse embryonic fibroblasts (MEFs) have impaired adipogenesis in vitro. Interestingly, pioglitazone treatment rescued the adipogenic deficiency in Agpat3-deficient cells. We conclude that AGPAT3 regulates adipogenesis and adipose development. It is possible that adipogenic impairment in Agpat3-deficient cells potentially leads to reduced adipose mass. Findings from this work support the unique role of AGPAT3 in adipose tissue.

NEW & NOTEWORTHY AGPAT3 deficiency results in male-specific growth retardation. It reduces adipose tissue mass but does not significantly impact glucose homeostasis or energy balance, except for influencing water intake in mice. Like AGPAT2, AGPAT3 is upregulated during adipogenesis, potentially by peroxisome proliferator-activated receptor gamma (PPARγ). Loss of AGPAT3 impairs adipocyte differentiation, which could be rescued by pioglitazone. Overall, AGPAT3 plays a significant role in regulating adipose tissue mass, partially involving its influence on adipocyte differentiation.

INTRODUCTION

Acylglycerophosphate acyltransferases (AGPATs), also known as lysophosphatidic acid acyltransferases (LPAATs), are a group of homologous enzymes that catalyze the de novo formation of phosphatidic acid (PA) from lysophosphatidic acid (LPA) to produce phospholipids (PLs) and triglycerides (TGs). Eleven AGPAT enzymes (AGPATs 1–11), each encoded by a different gene, have been identified in mice and humans based on their sequence homologies (1). Canonical AGPATs, including AGPATs1, 2, 3, 4, and 5, specifically prefer LPA to form PA, whereas AGPATs 6–11 are classified as lysophospholipid acyltransferases (LPLATs) or glycerophosphate acyltransferases (GPATs) based on their substrate specificities (2). Although the enzymatic activities of canonical AGPATs are known, their exact biological functions are still unclear.

Several knockout (KO) studies have demonstrated the unique physiological roles for different canonical AGPATs. AGPAT1 is widely distributed; homozygous loss of Agpat1 in mice results in high embryonic lethality and widespread disturbances of metabolism, sperm development, and neurological function (3). AGPAT2 is highly abundant in adipose tissue, whose mutations cause human type 1 congenital generalized lipodystrophy (4). Targeted gene deletion of Agpat2 in mice recapitulates human Type 1 Congenital Generalized Lipodystrophy (CGL1), characterized by the near total absence of adipose tissue from birth (5). AGPAT4 is most highly expressed in the brain (6, 7). Agpat4-deficient mice show severe dysfunctions, including cognitive impairment, impaired force contractility, and altered white adipose tissue (WAT) (6–8). The biological functions of other canonical AGPATs in lipid metabolism remain to be characterized.

Human AGPAT3 is highly conserved and ubiquitously expressed in multiple tissues, including the liver, kidney, heart, brain, and testis, with moderate expression in the adipose tissue [(9) and www.gtexportal.org]. Like AGPAT1 and AGPAT2, AGPAT3 is localized to the endoplasmic reticulum (2, 10). It was recently shown to also reside in the Golgi complex to control the structure and trafficking functions of the Golgi (11). Recent topological analyses of human AGPAT3 reveal two transmembrane domains, one separating the conserved acyltransferase motifs I and II, which are thought to form a functional unit critical for enzymatic activity (12). AGPAT3 prefers arachidonic acid (AA, C20:4) and docosahexaenoic acid (DHA, C22:6) fatty acyl-CoA donors (9, 12, 13). Indeed, AGPAT3 is required to produce DHA-containing phospholipids (PL-DHA) in various tissues, such as the testes and retina (14, 15). Male Agpat3 knockout (Agpat3-KO) mice are infertile due to spermatogenesis failure, likely attributed to the drastic reduction of PL-DHA in spermatids (14). However, DHA-rich diets did not restore sufficient PL-DHA to improve male infertility in Agpat3-KO mice (16). Meanwhile, AGPAT3 was shown to be upregulated during in vitro myoblast differentiation and may contribute to the enhanced PL-DHA content of endurance-trained muscles (17). Recently, a genome-wide association study identified the rs4819351-A variant in AGPAT3 associated with Alzheimer’s disease-related metabolic decline in the brain (18). AGPAT3 gene polymorphisms have also been associated with milk production traits in Chinese Holstein cows (19, 20). However, whether AGPAT3 plays a role in adipose tissue and whole body energy balance remains unknown.

In this study, we examined the metabolic phenotype of global Agpat3-KO mice and explored the potential role of AGPAT3 in adipocyte differentiation. We identified that Agpat3-KO mice are leaner with reduced adipose tissue concomitant with slight growth retardation. They generally maintained insulin and glucose homeostasis and whole body energy balance besides increased water intake. Further in vitro studies identified upregulation of AGPAT3 during mouse and human adipocyte differentiation and revealed an intrinsic role of AGPAT3 in adipocyte differentiation. These studies emphasize the importance of AGPAT3 in adipose tissue development.

MATERIALS AND METHODS

Animals

Agpat3/Lpaat3-KO mice harboring neomycin and LacZ cassette (Agpat3^{tm1(EUCOMM)Wtsi}, also called Agpat3-KO) were purchased from the European Conditional Mouse Mutagenesis Project. Agpat3-KO mice were generated from C57BL/6N mouse embryonic stem cells but were backcrossed to C57BL/6J (PRID:IMSR_JAX:000664) for at least four generations. Detailed targeting strategies and genotyping protocols have been previously reported (14). Mice were housed at a 12-h light/12-h dark cycle with ad libitum access to food and water. Both male and female age-matched littermates were used. Four-hour fasted mice were euthanized by cervical dislocation. All animal experiments followed guidelines issued by Augusta University’s institutional animal care and use committee (Protocol No. 2012-0462).

Body Composition, Metabolic Cage Studies, and Thermogenesis

Body composition was assessed by nuclear magnetic resonance (NMR) spectroscopy (Minispec LF90II, Bruker) as per the manufacturer’s instructions. Whole body energy balance was evaluated using the Comprehensive Lab Animal Monitoring System with Oxymax (Columbus Labs). Mice were housed singly in metabolic cages. After 24-h adaptation, food intake, water drinking, locomotor activity, and indirect calorimetry were monitored for 3 days as per the manufacturer’s instructions. For cold exposure experiments, mice were housed at 4 ± 1°C (individually housed without food). Rectal body temperature of the mice was monitored at 0, 1, 2, 3, 4, 5, and 6 h using a BAT-12 thermometer (Physitemp).

Plasma Biochemistry and Glucose and Insulin Tolerance Tests

Blood was collected in Microvette CB 300 EDTA K2E (Sarstedt 16.444.100). Plasma levels of insulin growth factor 1 (IGF1) (R&D system, MG100) and insulin (Millipore Sigma, EZRMI-13k) were measured by ELISA according to the manufacturer’s instructions. Plasma nonesterified fatty acid (NEFA) (WAKO), glycerol (Sigma), triglyceride, and cholesterol (ThermoFisher) were measured using colorimetric kits. Glucose tolerance tests were performed in mice after a 6-h fast followed by an injection of a glucose bolus at 1.5 g/kg body wt. Insulin tolerance tests were performed in mice after a 4-h fast. Mice were injected with humulin at 0.75 units/kg body wt. Blood glucose levels were measured by tail vein bleeds before and at 15, 30, 60, and 120 min after administration using a OneTouch Ultra blood glucose meter.

Histology and Determination of Adipocyte Cell Size

Tissues were fixed with neutral-buffered formalin and embedded in paraffin. Histology slides were prepared by the Histology Core at Augusta University (AU). Images were analyzed using a Zeiss Axioplan-2 imaging system. To measure the adipocyte cross-sectional area, eight random microscope fields (×20 magnification) in hematoxylin-eosin (H&E)-stained adipose tissues from three representative mice within each group were digitally recorded. Adipocyte cell sizes were analyzed using QuPath V. 0.5.1 as previously described (21).

Adipose DNA Measurements

Adipose DNA contents were analyzed as previously reported (22) with slight modification. In brief, ∼100 mg of adipose tissue was homogenized in 10 volumes of high-salt phosphate-buffered saline pH 7.4 (HS-PBS; 15 mM NaH2PO4, 81 mM Na2HPO4, and 2 M NaCl). After 3-h incubation on ice, the homogenate was centrifuged (13,000 rpm, 1 min at 4°C), and the supernatant was recovered. About 5 µL of serial diluted DNA standard (Sigma, D-7656) or sample (in duplicate) was added to 94.5 μL of HS-PBS and 0.5 μL of 200 μg/mL Hoechst 33342(Tocris Bioscience, 51-17) and read on a BioTek Synergy H1 Hybrid Multi-Mode Microplate Reader (BioTek) at 461 nm (emission). The number of cells in the adipose tissue was expressed as total DNA (µg) per fat depot.

Cell Culture, Generation of Retroviral Constructs and Retroviral Infections, and Adipocyte Differentiation

NIH/3T3 cells (ATCC CRL-1658) and 3T3-L1 cells (ATCC CL-173) were cultured in high-glucose Dulbecco’s modified Eagle’s medium (DMEM) supplemented with 10% bovine calf serum (BCS, ATCC 30–2030) and penicillin-streptomycin (Pen/Strep). 3T3-L1 differentiation was induced by standard hormone protocol [dexamethasone, IBMX (3-isobutyl-1-methylxanthine), and insulin (DMI)], as we previously described (23). pBABE puro peroxisome proliferator-activated receptor gamma 2 (PPARγ2) was a gift from Bruce Spiegelman (Addgene Plasmid # 8859; http://n2t.net/addgene:8859; RRID:Addgene_8859). pBABE-puro SV40 LT was a gift from Thomas Roberts (Addgene Plasmid # 13970; http://n2t.net/addgene:13970; RRID:Addgene_13970). Two small interfering RNA sequences targeting mouse Agpat3 were obtained from Dharmacon, Inc. (Lafayette, CO), synthesized as short hairpin RNA (shRNA) using TTCAAGAGA as a linker and subcloned into the RNAi-Ready pSiren-RetroQ vector (CLONTECH, Cat. No. 631526). shRNA against luciferase was used as a control. Retroviral packaging Bosc-23 (ATCC, CRL-11270) cells were cotransfected with the targeting plasmids and packaging vector pCL-eco (Imgenex, Cat. No. 10045 P). Forty-eight hours after transfection, the culture media containing the virus particles were collected. 3T3-L1 cells were infected with virus media mixed with DMEM and 10% BCS media at 1:2 in the presence of 8 μg/mL polybrene. Cells were then selected with 2 µg/mL puromycin for at least 48 h and cultured for DMI-induced adipocyte differentiation.

Human Adipose Tissue, Adipose Tissue Fraction, Isolation, and Differentiation of Stromal Vascular Cells and Mouse Embryonic Fibroblasts

Discarded human mediastinal adipose tissue was collected from patients undergoing cardiothoracic surgery in the operating rooms of Augusta University Health. All procedures were approved by the Augusta University Biosafety Office (Protocol No. 1210). Human or mouse adipose tissues were thoroughly minced, digested with collagenase type IV (Worthington Biochemical Corporation), filtered, and centrifuged to separate floating mature adipocyte fraction from the pelleted stromal vascular cells (SVCs). SVC pellets were resuspended, plated, and grown in a preadipocyte growth medium (Cell Applications, Inc.). Human SVCs were differentiated as previously described (24). MEFs were isolated from 12.5- to 14.5-day-old embryos (25) and cultured in high-glucose DMEM supplemented with 10% fetal bovine serum (FBS, Thermo Scientific, Cat. No. SH3091003HI) and Pen/Strep for standard DMI-induced adipocyte differentiation. In MEF experiments, 1 µM pioglitazone (Sigma-Aldrich) was added on days 0–4 (D0–D4) after differentiation with the vehicle in the control cells.

Tissue and Intracellular TG Analyses and Oil Red O Staining

Tissues were homogenized in standard PBS buffer followed by lipid extraction and quantification, as previously detailed (26). Cultured cells were directly lysed in 1% Triton X-100-PBS. TG concentrations were measured using a triglyceride assay kit (Thermo Scientific, TR-22421) and normalized to tissue weights or protein concentrations (cell). Oil red O staining was performed, and the results were photographed with a camera as described previously (26).

Western Blots

As previously described, tissue and cells were homogenized in lysis buffer with freshly added protease and phosphatase inhibitors (23). Protein concentrations were analyzed by Bradford assay kit (Bio-Rad). Standard immunoblot analyses were performed using the following antibodies: adipose TG lipase (ATGL) (CST#2138, RRID:AB_2167955), GAPDH (Proteintech #60004-1-Ig, RRID:AB_2107436), hormone-sensitive lipase (HSL) (CST#4107, RRID:AB_2296900), phosphorylated hormone-sensitive lipase (p-HSL) (CST#4139, RRID:AB_2135495), PPARγ (CST#2435, RRID:AB_2166051), Perillipin 1 (PLIN1) (CST#9349, RRID:AB_10829911), α tubulin (Proteintech #66031-1-Ig, RRID:AB_11042766), and Uncoupling protein 1 (UCP1) (Abcam #10983, RRID:AB_2241462). The blots were visualized using the ECL chemiluminescence system by Amersham Imager 600 (Cytiva).

RNA Isolation and qRT-PCR

RNA was extracted from tissue and cells using Trizol (ThermoFisher Scientific, Cat. No. 15-596-018), followed by reverse transcription using MLV-V reverse transcriptase with random primers (ThermoFisher). qRT-PCR was performed using SYBR Green dye in AriaMx (Agilent). Data were normalized to housekeeping genes (Rplp0 and/or Ppia) based on the geNorm algorithm (RRID:SCR_006763) and expressed as fold changes relative to control tissue or cells. The primers used were listed in Table 1.

Table 1.

RT-PCR primer sequences for murine (m) and human (h) genes

Gene Name	5′ Forward Primer Sequence	3′ Reverse Primer Sequence
mAgpat1	GCTGGCTGGCAGGAATCAT	GTCTGAGCCACCTCGGACAT
mAgpat2	TTTGAGGTCAGCGGACAGAA	AGGATGCTCTGGTGATTAGAGATGA
mAgpat3	CCAGTGGCTTCACAAGCTGTAC	CCCTGGGAATACACCCTTCTG
mAgpat4	GCTGCTGAAGTCCCAGTTTC	CACTCCAGAAGCATCACCAA
mAgpat5	GCTATGTGAACGCAGGAACA	AATGCCTGACTGGCTGAAA
mPnpla2	GATGTGCAAACAGGGCTACA	CTTCCTCTGCATCCTCTTCC
mC/ebpa	GAACAGCAACGAGTACCGGGTA	GCCATGGCCTTGACCAAGGAG
mDgat1	TGGCCAGGACAGGAGTATTT	CACAGCTGCATTGCCATAGT
mDgat2	CTTCCTGGTGCTAGGAGTGG	CCAGCTGGATGGGAAAGTAG
mElovl3	GATGGTTCTGGGCACCATCTT	CGTTGTTGTGTGGCATCCTT
mFabp4	AGCCCAACATGATCATCAGC	TCGACTTTCCATCCCACTTC
mFasn	GCTGCTGTTGGAAGTCAGC	AGTGTTCGTTCCTCGGAGTG
mGhr	AGCCTCGATTCACCAAGTGT	AGGGCATTCTTTCCATTCCT
mHnf4a	GCATGGATATGGCCGACTACA	ACCTTCAGATGGGGACGTGT
mIgf1	GGCATTGTGGATGAGTGTTG	GGCTCACCTTTCCTTCTCCT
mIgf1r	ATGAACCCGAAACCAGAGTG	TCACCGCGTGTCATTAGTTC
mIgfbp3	CAAAGCACAGACACCCAGAA	GGACTCAGCACATTGAGGAA
mLipin1	CAGCCTGGTAGATTGCCAGA	GCAGCCTGTGGCAATTCA
mPlin1	CACCTGCGGCTGTGCTGG	CGATGTCTCGGAATTCGCT
mPpara	CCACGAAGCCTACCTGAAGA	ACTGGCAGCAGTGGAAGAAT
mPparγ2	TCTCCTGTTGACCCAGAGCA	GTGGAGCAGAAATGCTGGAG
mPpia	CTGTTTGCAGACAAAGTTCCA	AGGATGAAGTTCTCATCCTCA
mRplp0	CGCTTTCTGGAGGGTGTCCGC	TGCCAGGACGCGCTTGTACC
mScd1	TTCCCTCCTGCAAGCTCTAC	CAGAGCGCTGGTCATGTAGT
mSrebp1c	GGAGCCATGGATTGCACATT	AGGCCCGGGAAGTCACTGT
mUcp1	ACTGCCACACCTCCAGTCATT	CTTTGCCTCACTCAGGATTGG
hAGPAT1	CTGGCAGGAGTCATCTTCATC	AGCCATTGTGGTTTCTCGTT
hAGPAT2	GGACGGTGGAGAACATGAG	GGTTGGAGACGATGACACAG
hAGPAT3	GCAGTCATCATCCTCAACCA	CGCTTGCAGAACACAATCTC
hAGPAT4	ATCGGCTGGATGTGGTACTT	ATGCTTCTTCTCCGTGAACC
hAGPAT5	TGGATTGTTGCTGACATCTTG	TGCAACTTGTTTCGCATCTC
hRPLP0	GCAATGTTGCCAGTGTCTGT	AGATGGATCAGCCAAGAAGG
hPPARγ	TCCATGCTGTTATGGGTGAA	ACGGAGCTGATCCCAAAGT

hPPARγ, human peroxisome proliferator-activated receptor gamma; mAgpat1, murine acylglycerophosphate acyltransferase 1; mFabp4, murine fatty acid-binding protein 4; mGhr, murine GH receptor; mIgf1, murine insulin growth factor 1; mIgf1r, murine IGF-1 receptor; mIgfbp3, murine IGF-binding protein 3.

Statistics

All in vitro experiments were performed in triplicates at least three times. Detailed sample sizes for animal studies were provided in each figure legend. Data were presented as means ± SE. Statistical significance was tested using unpaired t tests, multiple t tests, and one-way or two-way ANOVA as dictated by the experiments using GraphPad Prism Software (RRID:SCR_002798). SPSS 29 (RRID:SCR_002865) was used to perform analysis of covariance (ANCOVA). A P value less than 0.05 was considered significant in all statistical analyses.

RESULTS

Agpat3-Knockout Mice Are Short and Lean

We first investigated the role of Agpat3 in vivo by obtaining global Agpat3-KO mice. As previously reported (14), Agpat3-KO male mice were infertile since we could not get any pups from male Agpat3-KO mice by natural mating. We focused on characterizing the metabolic phenotype of Agpat3-KO mice. We followed the growth curves and found that male Agpat3^Het mice displayed lower body weights starting from 6 wk old, which were more evident in Agpat3-KO mice when compared with the age-matched littermate wild-type mice (Fig. 1, A and B). Agpat3-KO mice exhibited a 4% decrease in tibia length compared with wild-type mice, suggesting growth retardation (Fig. 1C). They tended to have lower fat contents with no differences in lean mass percentage (Fig. 1, D and E). In line with the lower body weights, the masses of epididymal white adipose tissue (eWAT), subcutaneous white adipose tissue (sWAT), and liver were significantly lower in male Agpat3-KO mice (Fig. 1F), albeit with comparable heart weights (Fig. 1G). Female Agpat3-KO mice exhibited no differences in growth retardation by 16 wk old (Fig. 1H). They demonstrated a lower body weight trend when they reached 9 mo old (Fig. 1I). However, we did not detect changes in their tibia lengths compared with female wild-type mice (Fig. 1J). Interestingly, 9-mo-old female Agpat3-KO mice also exhibited ∼50% reduction of gonadal white adipose tissue (gWAT) and sWAT with no changes in liver masses (Fig. 1K).

Figure 1.

AGPAT3 deficiency reduces body weight and adipose mass. A: body weights from birth until postnatal 18 wk for male wild-type (WT) (n = 11–14), Agpat3^Het (Het) (n = 8–14), and Agpat3-KO (KO) (n = 11–15) mice. Two-way ANOVA with Tukey’s multiple comparisons test. B: representative pictures. C: tibia length (n = 6–8 for each group). D: % of fat mass. E: % of lean mass (n = 10 or 11 for each group). Unpaired t test for C–E. F: tissue mass of eWAT, sWAT, and liver. n = 7–10/group. Multiple t tests correct with the Holm–Sidak method. G: ventricle weights. n = 7–10/group. Four-hour fasted male 18-wk-old WT and Agpat3-KO mice were used for B–G. H: body weights from birth until postnatal 16 wk for female WT and KO (n = 11) mice. Body weights (I) and tibia length (J) in 9-mo-old female mice. Unpaired t test for I and J. K: tissue mass of gWAT, sWAT, and liver in 9-mo-old female mice. Multiple t tests correct with the Holm–Sidak method. I–K: WT, n = 6; KO, n = 7. Plasma IGF-1 levels (WT: n = 13; KO: n = 10) (L), unpaired t test, and mRNA analyses (M) of the genes in the liver (n = 7/group), multiple t tests correct with the Holm–Sidak method, in 4-h fasted 18-wk-old male mice. AGPAT3, acylglycerophosphate acyltransferase 3; eWAT, epididymal white adipose tissue; gWAT, gonadal white adipose tissue; KO, knockout; sWAT, subcutaneous white adipose tissue; VW, ventricle weight. *P < 0.05; **P < 0.005 vs. WT mice.

Growth hormone (GH) mediates the postnatal growth and metabolism primarily through IGF-1 (27, 28). Consistent with the decreased tibia length, male Agpat3-KO mice displayed lower levels of plasma IGF-1 (Fig. 1L). Hepatocyte is recognized as the primary source of circulating IGF1 (29). However, we did not observe a significant reduction in hepatic mRNA expression of Igf-1 in Agpat3-KO mice. There were also no significant changes in the expression of the GH receptor (Ghr), IGF-1 receptor (Igf1r), and IGF-binding protein 3 (Igfbp3) between the two genotypes (Fig. 1M). These data suggest that deletion of Agpat3 results in more male-specific growth retardation and reduced adipose tissues in both male and female mice.

AGPAT3 Deletion Reduces Adipose Tissue Mass in Mice

We next examined whether AGPAT3 plays an intrinsic role in regulating adipose tissue mass. After normalizing body weights, male Agpat3-KO mice still exhibited ∼50.1% and 34.7% reductions in eWAT and sWAT masses, respectively (Fig. 2A). Brown adipose tissue (BAT) masses were also reduced by 26% in male Agpat3-KO mice (Fig. 2A). When normalized to body weights, the liver weights were increased by 10.9% in Agpat3-KO mice (Fig. 2A). Histological analyses identified reduced adipocyte sizes in male Agpat3-KO eWAT, which were further confirmed after assessment of their adipocyte diameters (Fig. 2, B and C). sWAT of male Agpat3-KO mice demonstrated similar reductions in adipocyte sizes (Fig. 2, B and D). The cell number in the eWAT trended lower and was significantly reduced in the sWAT of Agpat3-KO mice (Fig. 2, E and F). Agpat3-KO BAT exhibited reduced lipid droplet sizes compared with wild-type BAT (Fig. 2B). In accordance, there was about a 20% reduction of TG contents in Agpat3-KO BAT (Fig. 2G). Despite an enlarged liver in Agpat3-KO mice, there was no sign of changes in cellular structures and lipid deposition between the two genotypes (Fig. 2B). Assessment of the hepatic triglyceride contents in these animals also revealed no differences between the two genotypes (Fig. 2H). This was further supported by the comparable expressions of genes involved in lipid synthesis downstream of PA production (such as Lipin1, Dgat1, and Dgat2) and de novo lipogenesis (Srebp1c, Scd1, and Fasn) (Fig. 2I). Together, these data suggest that AGPAT3 contributes to the regulation of adipose mass.

Figure 2.

Agpat3-KO mice are lean with reduced white and brown adipose tissue. Male 18-wk-old wild-type (WT) and Agpat3-KO (KO) mice were used after a 4-h fast. A: tissue masses of eWAT, sWAT, liver, and BAT presented as % of body weight (BW). WT: n = 11; KO: n = 13. Multiple t tests correct with the Holm–Sidak method. B: representative H&E staining. Scale bar = 100 µm. Morphometric analysis of adipocyte cell size (based on a total of about 1,200 cells/genotype) of eWAT (C) and sWAT (D) (n = 3/genotype). Adipose tissue cell number expressed as total DNA content (µg) per fat depot in eWAT (E) (n = 12–14) and sWAT (F) (n = 8). TG contents in BAT (G) and liver (H) after normalized to tissue wet weights (n = 6–9/group). Unpaired t tests for C–H. I: mRNA analyses of genes in the liver of male WT (n = 7) and KO (n = 6) mice. Multiple t tests correct with the Holm–Sidak method. AGPAT3, acylglycerophosphate acyltransferase 3; BAT, brown adipose tissue; eWAT, epididymal white adipose tissue; H&E, hematoxylin-eosin; KO, knockout; TG, triglyceride. *P < 0.05; **P < 0.005 vs. WT mice.

Agpat3 Deficiency in Adipose Tissues Did Not Cause Changes in Adipocyte Marker and Lipid Metabolic Genes

Next, we examined gene expressions in eWAT and BAT of wild-type and Agpat3-KO mice. As expected, the Agpat3 gene was almost nondetectable in Agpat3-KO eWAT and BAT (Fig. 3, A and B, respectively). However, we found no changes in Agpat1, 2, 4, and 5 expression, suggesting a lack of compensatory regulation of other canonical AGPATs in the Agpat3-KO eWAT and BAT (Fig. 3, A and B, respectively). Meanwhile, the expressions of genes involved in lipid synthesis downstream of PA production (Lipin1, Dgat1, and Dgat2), de novo lipogenesis (Srebp1c, Scd1, and Fasn), and lipid β-oxidation and hydrolysis (Pparα and Pnpla2) were not altered in both eWAT and BAT between two genotypes (Fig. 3, A and B, respectively). In eWAT, the expression of white adipocyte marker genes (Pparγ and C/ebpa) was maintained between wild-type and Agpat3-KO mice. There were also no changes in the expression of Ucp1 and Elovl3, suggesting that Agpat3 deficiency did not induce browning in eWAT (Fig. 3A). Similarly, we did not detect overt differences in the mRNA expression of brown adipocyte marker genes (Prdm16, Ucp1, and Elovl3) in BAT of wild-type and Agpat3-KO mice (Fig. 3B). The lack of changes in the expression of adipocyte marker proteins between the two genotypes was further confirmed by Western blot in both eWAT (Fig. 3C) and BAT (Fig. 3D). We also found no differences in the expressions of adipose TG lipase (ATGL, encoded by Pnpla2) and total and phosphorylated hormone-sensitive lipase (HSL) in both eWAT and BAT, suggesting that the reduced adipocyte sizes in eWAT and BAT were not due to changes in TG catabolic enzymes (Fig. 3, C and D, respectively). These data indicate that AGPAT3 deficiency reduced white and brown fat masses without altering the expression of adipocyte markers in Agpat3-KO mice.

Figure 3.

Agpat3-KO mice display minimal gene changes in white and brown adipose tissue. mRNA analyses of genes in eWAT (A) and BAT (B) of wild-type (WT) (n = 11) and Agpat3-KO (KO) (n = 13) mice. Multiple t tests correct with the Holm–Sidak method. Western blot in WAT (C) and BAT (D) in WT and KO mice (n = 6/group). Male 18-wk-old mice were used. AGPAT3, acylglycerophosphate acyltransferase 3; eWAT, epididymal white adipose tissue; KO, knockout; WAT, white adipose tissue. **P < 0.005 vs. WT mice.

Agpat3-KO Mice Display Minimal Changes in Metabolic Homeostasis and Energy Balance

We next examined whether altered adipose masses in Agpat3-KO mice perturb metabolic homeostasis. Fasting plasma levels of TG, NEFA, and glycerol, but not cholesterol, were lower in Agpat3-KO mice (Fig. 4, A–D, respectively). Agpat3-KO mice also exhibited reduced insulin levels (Fig. 4E). However, glucose tolerance and insulin sensitivity were maintained when comparing both genotypes (Fig. 4, F and G). Metabolic cage studies further revealed that Agpat3-KO mice exhibited similar food intake during light and dark phases, suggesting their lower body weights were not due to the decreased food intake (Fig. 4H). Interestingly, Agpat3-KO mice did exhibit an increased water intake compared with wild-type mice (Fig. 4I). However, their oxygen consumption and heat production were comparable between the two genotypes, demonstrated by analysis of covariance (ANCOVA) between oxygen consumption/heat production and body weight (Fig. 4, J and K, respectively). The respiratory exchange ratios (RER) [carbon dioxide production (V̇co₂)/oxygen consumption (V̇o₂)) were also similar between the two genotypes (Fig. 4L), and so were their physical activities (Fig. 4M). Since Agpat3-KO mice displayed reduced BAT mass, we further evaluated the adaptive thermogenesis by exposing animals to a cold environment (4°C) without access to food. Wild-type and Agpat3-KO mice maintained similar body temperatures before and after cold exposures for up to 6 h (Fig. 4N). These data suggest that Agpat3 deficiency reduces circulating lipid metabolites and increases water intake without altering glucose homeostasis, energy balance, and adaptive thermogenesis.

Figure 4.

Agpat3-KO mice maintained insulin and glucose homeostasis, energy balance, and adaptive thermogenesis, except for increased water intake. Plasma levels of TG (A), cholesterol (B), NEFA (C), glycerol (D), and insulin (E) in 18-wk-old male wild-type (WT) and Agpat3-KO (KO) mice after a 4-h fast. n = 11–13/group. Unpaired t tests. Glucose tolerance test (n = 6/group) (F) and insulin tolerance test (n = 5 or 6/group) (G). Food intake (H), water intake (I), oxygen consumption (V̇o₂) (J) as presented as mL/h/mouse and regression plot of dark period oxygen consumption vs. body weight (BW) analyzed by one-way ANCOVA. K: heat production and regression plot of dark period heat production vs. body weights analyzed by one-way ANCOVA. L: RER is assessed by the ratios of carbon dioxide production vs. oxygen consumption. M: physical activities measured by infrared beam breaks. N: rectal body temperature after exposure to 4°C without food. WT: n = 6; KO: n = 8. All experiments were performed in 18-wk-old male mice. H–M: two-way ANOVA with Tukey’s multiple comparisons test. Agpat3, acylglycerophosphate acyltransferase 3; ANCOVA, analysis of covariance; NEFA, nonesterified fatty acid; RER, respiratory exchange ratio; TG, triglyceride. *P < 0.05; **P < 0.005.

AGPAT3 Is Upregulated during Adipocyte Differentiation

To identify what underlies the reduced body fat masses in Agpat3-KO mice, we next determined whether AGPAT3 expression is regulated during adipogenesis compared with other canonical AGPATs. Not surprisingly, Agpat2 was the dominant Agpat that was upregulated by >20 folds in adipocyte fractions. Agpat1 and Agpat3 were also elevated in adipocyte fractions, albeit to a lesser extent than Agpat2 (Fig. 5A). In day 10 (D10) differentiated mature human adipocytes, AGPAT2 was dominantly upregulated compared with D0 SVCs. There was an approximately twofold upregulation of AGPAT3 in D10 differentiated human adipocytes, and so was AGPAT1 (Fig. 5B). Regulation of Agpat4/AGPAT4 was more complex as it was reduced in mouse adipocytes but increased in differentiated human adipocytes (Fig. 5, A and B). No changes were observed in the expression of Agpat5/AGPAT5 in mouse primary adipocytes and differentiated human adipocytes (Fig. 5, A and B). Upregulation of PPARγ2 was used as a mature adipocyte marker (Fig. 5, A and B). Furthermore, we found that Agpat2 and Agpat3 expression became elevated until 24 h after induction, whereas Pparγ expression showed significant upregulation as early as 4 h after DMI treatment in 3T3-L1 cells (Fig. 5C). By day 8, when the adipocytes reached full maturity, there were 22.2-, 3.5-, and 209-fold increases in the expression of Agpat2, Agpat3, and Pparγ, respectively (Fig. 5C). Interestingly, PPARγ2 overexpression predominantly upregulated the expression of Agpat2 with minimal effect on Agpat3. However, both Agpat2 and Agpat3 genes exhibited a further increase in the expression following treatment with the PPARγ agonist pioglitazone, similar to the regulation observed in the well-known PPARγ target fatty acid-binding protein 4 (Fabp4) (Fig. 5D). These data suggest that Agpat3, like Agpat2, could be regulated by PPARγ and contribute to adipocyte differentiation.

Figure 5.

AGPAT3 is upregulated during human and mouse adipocyte differentiation, and PPARγ2 overexpression upregulates AGPAT3. A: qRT-PCR in SVCs and adipocytes fractionated from eWAT of C57BL/6J mice (n = 5/group). B: qRT-PCR in human preadipocytes at day 0 (D0, 2 days of confluence) and 10 days (D10) after adipocyte differentiation. Multiple t tests correct with the Holm–Sidak method. *P < 0.05; **P < 0.005. C: qRT-PCR in 3T3-L1 cells at 0, 2, 4, 9, and 24 h or 8 days (D8) after DMI-induced adipocyte differentiation. Unpaired t tests vs. 0 h. **P < 0.005 vs. 0 h. D: mRNA expression of Agpat2, Agpat3, and Fabp4 in NIH/3T3 cells stably transduced with retroviruses overexpressing SV40-LT or PPARγ2. Cells overexpressing PPARγ2 were treated with vehicle (Veh) or 1 µM pioglitazone (Pio) for an additional 24 h. One-way ANOVA with Tukey’s multiple comparisons test. *P < 0.05; **P < 0.005 vs. SV40-LT cells; ##P < 0.005 vs. Veh-treated PPARγ2-overexpressing cells. AGPAT3, acylglycerophosphate acyltransferase 3; DMI, dexamethasone, 3-isobutyl-1-methylxanthine, and insulin; eWAT, epididymal white adipose tissue; PPARγ2, peroxisome proliferator-activated receptor gamma 2; SVC, stromal vascular cell.

Agpat3 Knockdown Impairs Adipocyte Differentiation in Vitro

To determine whether AGPAT3 plays a direct role in adipocyte differentiation, we transduced 3T3-L1 preadipocytes with retroviruses expressing shRNA-targeting murine Agpat3 (shAgpat3#1 and shAgpat3#2), using shLuc as a control. Stably transduced cells were selected, expanded, and induced to differentiation by DMI. Basal Agpat3 mRNA was downregulated by 74% (shAgpat3#1) and 60% (shAgpat3#2), respectively, on D0. On D8 after DMI treatment, Agpat3 mRNA was upregulated by 3.3-folds in shLuc cells, and so were Agpat2 and Pparγ, which were elevated by 18.2- and 212-folds, respectively (Fig. 6A). The expression of Agpat3 stayed knocked down in both shAgpat3 cells at D8 (Fig. 6A). shLuc cells underwent DMI-induced adipocyte differentiation and triglyceride accumulation normally. In contrast, the knockdown of Agpat3 expression led to significant inhibition of TG accumulation, as revealed by oil red O staining and direct measurement of TG content on D8 (Fig. 6, B and C). Impaired adipogenesis in two independent Agpat3-knockdown cell lines was further confirmed by their impaired expressions of Agpat2 and Pparγ (Fig. 6A) as well as reduced levels of adipocyte marker proteins (PPARγ and PLIN1) compared with shLuc cells at D8 (Fig. 6D). These data suggest that AGPAT3 also plays a role in adipocyte differentiation.

Figure 6.

Agpat3 knockdown in 3T3-L1 cells impairs adipocyte differentiation. 3T3-L1 cells were stably infected with retroviruses expressing control shLuc or shAgpat3#1 or shAgpat3#2 and induced to differentiate using a standard hormone cocktail. A: expression of mRNA encoding Agpat3, Agpat2, and Pparγ2 in D0 and D8 differentiated cells. Data were expressed as fold of expression by normalizing to control cells at D0 relative to Ppia expression. Two-way ANOVA with Tukey’s multiple comparisons test. *P < 0.05; **P < 0.005 vs. shLuc cells at the same day; ##P < 0.005 vs. shLuc at D0. Lipid accumulation was assessed by oil red O staining (B) and quantified by a triglyceride analysis kit (C) after lipid extraction on D8. One-way ANOVA with Tukey’s multiple comparisons test. **P < 0.005 vs. shLuc cells. D: Western blot in D8 differentiated adipocytes. Three independent experiments. Agpat3, acylglycerophosphate acyltransferase 3.

AGPAT3 Deletion Impairs Adipocyte Differentiation, Which Is Rescued by PPARγ Agonist Pioglitazone

To further analyze the importance of AGPAT3 for adipocyte differentiation, we evaluated the capacity of the postconfluent Agpat3^−/− MEFs to differentiate into adipocytes in vitro. Ten days after adipocyte differentiation, Agpat3 mRNA was reduced by 90% in Agpat3^−/− MEFs (Fig. 7A). Although Agpat3^+/+ MEFs primarily accumulated neutral lipids as determined by oil red O staining (Fig. 7B), only 50% of Agpat3^−/− MEFs differentiated into adipocytes, which were further confirmed by oil red O staining and significantly lower accumulation of TG in these cells (Fig. 7, B and C, respectively). Accordingly, the mRNA levels of Agpat2, master transcription regulators (Pparγ and C/eboα), and lipid droplet-associated proteins (Plin1) either trended lower or were significantly inhibited in adipogenically differentiated Agpat3^−/− MEFs (Fig. 7A). The inhibited expression of PPARγ and PLIN1 was further confirmed by Western blot (Fig. 7C). These data further emphasize that AGPAT3 deficiency attenuates full adipogenesis.

Figure 7.

Pioglitazone rescues white adipocyte differentiation blockade of Agpat3 deficiency. MEF cells were isolated from WT (^+/+) and Agpat3-KO (^−/−) E14 embryos and subjected to adipocyte differentiation in the presence of insulin, dexamethasone, and IBMX (DMI). mRNA expression (A) (n = 3/group), oil red O staining (B), Western blot analyses (C) at D10 of adipocyte differentiation. Multiple t tests correct with the Holm–Sidak method. D: 1 µM exogenous PPARγ ligand pioglitazone was added at D0 and kept at that level until D4. D10 cells were stained with oil red O, and lipids were extracted and quantitated by triglyceride analysis (E) (data were normalized to cellular protein levels). n = 4 or 5/group. Multiple t tests correct with the Holm–Sidak method. F: Western blots in WT and Agpat3-KO D10 MEFs. Agpat3, acylglycerophosphate acyltransferase 3; D0, day 0; KO, knockout; MEF, mouse embryonic fibroblast; PPARγ, peroxisome proliferator-activated receptor gamma; WT, wild type. *P < 0.05; **P < 0.005. Three independent experiments.

We next tested whether PPARγ activation could rescue the adipogenic capacity of Agpat3^−/− MEFs. We treated differentiating Agpat3^+/+ and Agpat3^−/− MEFs with pioglitazone and DMI adipogenic cocktail. As expected, pioglitazone enhanced the adipogenic potential of Agpat3^+/+ MEFs compared with DMI treatment alone. Pioglitazone treatment significantly increased the adipogenesis of Agpat3^−/− MEFs to the levels of DMI-treated Agpat3^+/+ MEFs as revealed by similar oil red O staining, cellular TG concentration, and protein levels of adipocyte markers (PPARγ and PLIN1) (Fig. 7, D, E, and F, respectively). These data suggest that PPARγ defects are the main driving force for abnormal adipogenesis in Agpat3 deficiency.

DISCUSSION

We show here that AGPAT3 modulates adipocyte differentiation and adipose tissue mass in vitro and in vivo. Gene ablation of Agpat3 in mice reveals a male-specific growth retardation. Both male and female Agpat3-KO mice demonstrate reduced white and brown adipose tissue, which did not result in altered metabolic homeostasis and adaptive thermogenesis. AGPAT3 was upregulated during adipocyte differentiation, potentially through PPARγ. Loss of AGPAT3 impairs adipocyte differentiation, and such deficiency could be rescued by activation of PPARγ, the master transcription regulator of adipocyte differentiation. Our study emphasizes that, among canonical AGPATs, AGPAT3 also plays a physiological role in regulating adipogenesis and maintaining adipose tissue mass.

Thus far, several abnormalities have been associated with global Agpat3-KO mice, including infertility and impaired visual function. These abnormalities have been primarily attributed to the deficiency in PL-DHA syntheses in Agpat3-KO mice, leading to spermatogenic failure, abnormal retinal layers, and disordered disk morphology in photoreceptor cells (14, 15). Here, we examined the metabolic phenotype of Agpat3-KO mice. We identified a male-specific growth retardation potentially associated with the reduced serum IGF-1 levels. Decreased serum IGF-1 concentration was a common finding in several mice models with growth retardation (30, 31). However, reduced IGF-1 levels are not caused by an altered GH-IGF1 axis in the liver, as the expression of these signaling proteins was not altered in Agpat3-KO mice. Although the mechanism underlying the reduced IGF-1 remains unknown, we postulate that the sex-specific growth retardation observed in male Agpat3-KO mice may be associated with reduced testosterone levels due to impaired spermatogenesis. Nevertheless, such a possibility requires further investigation.

Interestingly, despite the reduced IGF-1 and insulin levels, the leaner Agpat3-KO mice largely maintained normal glucose homeostasis. It remains unknown whether this is because the reduction in insulin levels was not severe enough to significantly disrupt glucose homeostasis or simply other compensatory mechanisms or factors may be at play to ensure glucose regulation in Agpat3-KO mice. Agpat3-KO mice also did not demonstrate changes in food intake, energy expenditure, and physical activities. Therefore, the lower circulating lipid metabolites in Agpat3-KO mice could not be due to the altered food intake; instead, it reflects reduced adipose mass. Similarly, since Agpat3-KO mice eat a similar amount of food as wild-type mice, we also cannot attribute the reduced insulin levels to their food intake. At this time, we do not know whether β-cells of Agpat3-KO mice have altered glucose-stimulated insulin secretion. Notably, Agpat3-KO mice display increased water intake. Although the normal glucose homeostasis in Agpat3-KO mice argues against a diabetic mellitus phenotype, it is plausible to postulate that Agpat3-KO mice may have kidney dysfunction, considering its high expression in the kidney (9). In fact, a recent study demonstrates an association of AGPAT3 single-nucleotide polymorphisms (SNPs) with kidney function in African ancestry (32). More studies are needed to characterize the phenotype associated with AGPAT3 deficiency in the pancreas and kidney. This will yield exciting insights into the diverse effects of AGPAT3 on downstream lipid metabolism, which can vary even from one tissue to another tissue.

Although we confirmed a dominant upregulation of Agpat2 during mouse and human adipocyte differentiation, our data also identified elevated expression of other Agpats, including Agpat1 and Agpat3, during adipogenesis, supporting their potential contribution to adipose development. Indeed, different AGPATs seem to exert differential effects on WAT mass in vivo. AGPAT1 and AGPAT2 are highly homologous isoenzymes expressed in adipocytes; deficiency of each causes lipodystrophy, suggesting AGPAT1 and AGPAT2 are crucial for adipogenesis, and they cannot compensate for each other in adipocytes (3, 5). Agpat4^−/− mice exhibit a surprisingly higher epididymal WAT mass than wild-type littermates, albeit with no changes in other WAT depots and BAT (8, 33). Homologs of Agpat4, such as Agpat1, 3, 5, are significantly upregulated in Agpat4^−/− mouse brains, indicating potential functional compensation in the absence of Agpat4 (7). In our study, reduced adipose mass in Agpat3-KO mice highlights the intrinsic role of Agpat3 in modulating adipose tissue development. Notably, we also did not observe any compensatory upregulation of other Agpats in the adipose tissue of Agpat3-KO mice, including the dominant adipose-expressed Agpat2. These data emphasize the distinct role of each AGPAT in the adipose tissue despite a similar biochemical function.

The exact mechanisms underlying the reduced adipose mass in Agpat3-KO mice remain elucidated. Our in vitro study found that AGPAT3 deficiency attenuated adipocyte marker gene expression and adipogenesis. The observation of markedly reduced adipocyte numbers, particularly in sWAT of Agpat3-KO mice, strongly suggests that decreased adipogenesis plays a significant role in the reduced sWAT mass observed in vivo. Currently, only the role of AGPAT2 in adipogenesis has been intensively studied. AGPAT2 deficiency has been reported to impair adipogenesis through the modulation of the lipidome, thus altering normal activation of phosphatidylinositol 3-kinase (PI3K)/Akt and PPARγ to suppress adipogenic program (34–36). However, the PPARγ overexpression or its agonist pioglitazone can only partially rescue the adipogenic defect in Agpat2^−/− cells (34, 36). We identified no differences in the hormone cocktail-induced PI3K/AKT signaling activation during the early stage of adipogenesis in Agpat3^−/− MEFs (data not shown), and PPARγ activation can fully restore adipogenesis in Agpat3^−/− cells. It is conceivable that AGPAT3 deficiency intrinsically impairs adipogenic capacity by regulating PPARγ activity. Alternatively, PPARγ activation may increase the expression of genes involved in lipid metabolism, such as AGPAT2, which could compensate for the loss of AGPAT3 during adipocyte differentiation. A limitation of our study is that we did not analyze the phospholipid and TG species, including PA, LPA, phosphatidylinositol species, and PPARγ inhibitor cyclic PA in differentiating or differentiated Agpat3^−/− MEFs, as we reasoned that it would be difficult to attribute the observed differences to the consequences or the causes of impaired adipocyte differentiation. Although AGPAT2 is necessary for normal brown adipose differentiation by regulating lipid deposition, interferon-stimulated genes, and mitochondrial morphology (37), we do not know whether AGPAT3 regulates brown adipogenesis, which warrants further investigation.

Notably, the residual Agpat3-KO WAT depots demonstrated reduced adipocyte size and the absence of obvious changes in the expression of adipocyte marker genes. The lack of AGPAT3 also reduced the lipid deposition in brown fat. These findings suggest that beyond adipogenesis, AGPAT3 may be involved in TG synthesis to impact lipid droplet size within adipocytes and thus adipose mass. Future endeavors could be invested in analyzing the in vivo lipid compositions in adipose tissues of Agpat3-KO mice. Meanwhile, the generation of adipose-specific Agpat3 knockout mice could further explain whether reduced adipose mass in global Agpat3-KO mice is due to impaired AGPAT3-mediated TG synthesis and whether adipose-specific deletion of Agpat3 is resistant to high-fat diet-induced adipose expansion and obesity.

In summary, the Agpat3-KO mice revealed reduced adipose mass with generally maintained metabolic homeostasis. Agpat3 deficiency impairs adipogenesis in vitro, partially explaining the lean phenotype in vivo. These results illustrate that AGPAT3 also serves a specific biological function(s) with no redundancy from AGPAT1 and AGPAT2 in regulating adipose mass.

DATA AVAILABILITY

The data that supports the findings of this study are available upon request from the corresponding author.

GRANTS

This work was supported by the National Institute of Diabetes and Digestive and Kidney Diseases (1R56 DK135657) and the National Heart, Lung and Blood Institute (2R01HL132182) (to W.C.).

DISCLOSURES

No conflicts of interest, financial or otherwise, are declared by the authors.

AUTHOR CONTRIBUTIONS

W.C. conceived and designed research; H.Z., K.F., and W.C. performed experiments; H.Z., K.F., and W.C. analyzed data; H.Z., V.P., L.R.H., H.W.K., and W.C. interpreted results of experiments; H.Z. and W.C. prepared figures; W.C. drafted manuscript; H.Z., K.F., V.P., L.R.H., H.W.K., Z.B., N.L.W., and W.C. edited and revised manuscript; H.Z., K.F., V.P., L.R.H., H.W.K., Z.B., N.L.W., and W.C. approved final version of manuscript.