Human GDPD3 overexpression promotes liver steatosis by increasing lysophosphatidic acid production and fatty acid uptake

The glycerol phosphate pathway produces more than 90% of the liver triacylglycerol (TAG). Lysophosphatidic acid (LysoPA), an intermediate in this pathway, is produced by glycerol-3-phosphate acyltransferase. Glycerophosphodiester phosphodiesterase domain containing 3 (GDPD3), whose gene was recently cloned, contains lysophospholipase D activity, which produces LysoPA from lysophospholipids. Whether human GDPD3 plays a role in hepatic TAG homeostasis is unknown. We hypothesized that human GDPD3 increases LysoPA production and availability in the glycerol phosphate pathway, promoting TAG biosynthesis. To test our hypothesis, we infected C57BL/6J mice with adeno-associated adenovirus encoding a hepatocyte-specific albumin promoter that drives GFP (control) or FLAG-tagged human GDPD3 overexpression and fed the mice chow or a Western diet to induce hepatosteatosis. Hepatic human GDPD3 overexpression induced LysoPA production and increased fatty acid uptake and incorporation into TAG in mouse hepatocytes and livers, ultimately exacerbating Western diet-induced liver steatosis. Our results also showed that individuals with hepatic steatosis have increased GDPD3 mRNA levels compared with individuals without steatosis. Collectively, these findings indicate that up-regulation of GDPD3 expression may play a key role in hepatic TAG accumulation and may represent a molecular target for managing hepatic steatosis. increases LysoPA availability and hepatic FA flux, promoting TAG biosynthesis and accumulation through the ER glycerol in of GDPD3 with steatosis GDPD3 Western diet-induced


Introduction
Nonalcoholic fatty liver disease (NAFLD) was first described in 1980 (1). About 75% of individuals with NAFLD have nonalcoholic fatty liver (NAFL), characterized by isolated hepatic steatosis or combined steatosis and mild nonspecific inflammation. The remaining 25% have nonalcoholic steatohepatitis (NASH), with steatosis, mild nonspecific inflammation, and hepatocellular injury with or without hepatic fibrosis (2). Hepatic steatosis is the hallmark of NAFL/NAFLD resulting from increased triacylglycerol (TAG) acquisition and disposition in hepatocytes. Increased hepatic TAG can result from: 1) increased uptake of plasma non-esterified fatty acid (FA) derived from adipose tissue or diet, 2) elevated de novo FA synthesis, 3) increased FA esterification to form TAG, 4) reduced FA oxidation, and/or 5) impaired very low-density lipoprotein (VLDL) assembly and secretion (3). There are two pathways involved in TAG synthesis, the glycerol phosphate pathway and the monoacylglycerol pathway; the former is used by most cells for TAG synthesis, whereas the latter predominates in the small intestine (4). The first step of TAG synthesis in the glycerol phosphate pathway (Fig. 1) is FA esterification of glycerol 3-phosphate by glycerol-3-phosphate acyltransferase (GPAT) to form lysophosphatidic acid (LysoPA) (5). An additional FA is subsequently incorporated into LysoPA, forming phosphatidic acid (PA) by a family of 1acylglycerol-3-phosphate acyltransferases (AGPAT) (6). The lipin phosphatidic acid phosphatase enzymes dephosphorylate PA to synthesize diacylglycerol (DAG) (7). Finally, DAG is converted to TAG through the action of diacylglycerol acyltransferase (DGAT) (8). The glycerol phosphate pathway primarily occurs in the endoplasmic reticulum (ER) and to a lesser extent on lipid droplets, at the mitochondrial membrane, and at the nuclear envelope (9).
Bacterial and fungal glycerophosphodiester phosphodiesterases (GDPD also known as GDE) 4 synthesis in liver ( Fig. 1) (14). In addition, our previous study identified mouse GDPD3 gene as one of the top 5 genes involving in liver lipid metabolism and hepatic steatosis (15). However, mammalian GDPD3 has not previously been implicated in the regulation of hepatic TAG metabolism. Therefore, this study was designed to examine whether human GDPD3 overexpression in mouse liver increases LysoPA availability for TAG synthesis through the ER glycerol phosphate pathway, leading to hepatic steatosis and determine what the underlying mechanisms are. We reported that hepatic human GDPD3 overexpression induced LysoPA production and increased FA uptake and incorporation into TAG in mouse hepatocytes and livers, ultimately exacerbating Western diet-induced liver steatosis.

Mice
C57BL/6J mice (The Jackson Laboratory, Stock No. 000664) were housed in standard caging at 22°C in a 12-hour light and 12-hour dark cycle (dark from 6 p.m. to 6 a.m.) at standard temperature and humidity conditions with ad libitum access to water and a standard chow diet (Purina-LabDiet, Prolab RMH 3000).
All experiments were performed using a protocol approved by the Institutional Animal Care and Use Committee at Wake Forest School of Medicine in facilities approved by the American Association for Accreditation of Laboratory Animal Care.

Gdpd3 mRNA quantification in mouse tissues
Twelve-week old mice were fasted for 24 hours (3 male and 3 female mice), anesthetized with ketamine/xylazine, and then transcardially perfused with ice-cold 1x Dulbecco's Phosphate Buffered Saline (DPBS) (15). Liver, visceral white adipose tissue (Epididymis for male mice and periovarian for female mice), brown adipose tissue, heart, brain (Cerebrum), kidney, duodenum, jejunum, ileum, and colon were snap-frozen in liquid nitrogen and stored in a -80˚C freezer. mRNA abundance of Gdpd3 was measured by real-time PCR. For a fasting/refeeding study (15), 12-week old male mice were fasted for 24 hours (n=5), and another group was fasted for 24 hours followed by refeeding a standard chow diet for 12 by guest, on July 31, 2020 www.jlr.org Downloaded from hours (n=5); fasting start times were staggered between groups so that both groups of mice were sacrificed together. Livers were collected to measure TAG levels (Wako Diagnostics L-Type TG M) and Gdpd3 mRNA abundance by real-time PCR.

Overexpression of hepatic human GDPD3 in chow-fed mice
Six-week old male mice were injected (retro-orbital; 3x10 11 genome copies per mouse) with adenoassociated adenovirus (AAV) serotype 8 (Vector Biolabs) encoding a hepatocyte-specific promoter (i.e., albumin) driving green fluorescent protein (AAV-GFP) expression (controls; n=6) or FLAG-tagged human GDPD3 (AAV-hGDPD3) to overexpress human GDPD3 in mouse liver (n=6). Mice were terminated 18 weeks after AAV injection, after an overnight fast, and plasma and livers were harvested for lipid measurement (15). Livers were lipid extracted using a modified Folch method for measurement of LysoPA content (lipid class analysis) as reported previously (16). Liver (50 mg) was homogenized in 20 volumes of extraction solvent (chloroform:methanol, 2:1, vol:vol) using a Precellys homogenizer (Montigny-le-Bretonneux, France) at 6,700 RPM for 15 seconds (repeated 3 times at 30-second intervals for a total of four times). The liver homogenate was incubated at 4°C for 2 hours and then the supernatant was harvested and stored at -20°C prior to analysis. Mixed mode HPLC was performed on an Agilent (Palo Alto, CA) 1200 HPLC system consisted of a quaternary pump, an autosampler, a column oven and an evaporative light scattering detector. A Chromolith Performance Si column (100mm × 4.6mm, macropores size 2.1µm and mesopores size 13nm, Merck, Darmstadt, Germany) was used for class-based lipid separation following previously described gradient and separation conditions (16). Peaks for LysoPA were manually integrated. The samples were analyzed randomly with blank runs in between.
LysoPA (18:1) was purchased from Avanti Polar Lipids (Alabaster, AL) and used as a standard to establish the retention time profiles in mixed mode HPLC analysis. Six calibrator concentrations ranging from 0.003125 to 0.1 mg/ml for LysoPA were employed to construct standard curves. LysoPA concentration in each sample was determined using the standard curve. by guest, on July 31, 2020 www.jlr.org

Downloaded from
Chow-fed mice (16-18 weeks of age) were used for comprehensive systemic metabolic phenotyping including measurement of blood glucose, plasma insulin, and plasma β-hydroxybutyrate levels after a 24-hour fast followed by a 3-hour chow refeeding (15). We also measured oxygen consumption (VO2, ml/hr/kg), carbon dioxide production (VCO2, ml/hr/kg), heat production (kcal/hr/kg), food intake (g/12hr), and respiratory exchange ratio (VCO2/VO2) of mice during a 12-hour light and 12hour dark cycle for 3 consecutive days using a 6-chamber indirect calorimetry system (PhenoMaster: TSE Systems, Bad Homburg, Germany). Data collected during the second day are presented (17).
Plasma and extracted liver lipids were used to measure TAG (Wako Diagnostics L-Type TG M), free cholesterol (Wako Free Cholesterol E), total cholesterol (Wako Cholesterol E), and total phospholipids (Wako Phospholipids C) using enzymatic assays (15). Cholesteryl ester was calculated using the equation of (TAG -free cholesterol) x 1.67 (18). The non-extractable liver residue was air-dried and solubilized with 0.1 N of NaOH to quantify liver protein content using the Lowry method (19).

Hepatic LysoPA production
Twelve-week old male mice were administered AAV-GFP or AAV-hGDPD3.

Primary hepatocyte isolation
The procedure of primary hepatocyte isolation was described previously (15). Mouse primary hepatocytes were seeded at 0.5-1 x 10 6 per well in a 6-well plate containing William's E medium (Life technologies #12551-032) supplemented with antibiotics, including 100 IU/ml of Penicillin and 100 μg/ml of Streptomycin (Life technologies #15140-122). After 2-3 hours of cell attachment at 37°C under 5% CO2, hepatocytes were cultured in William's E medium supplemented with antibiotics, 10% fetal bovine serum, and indicated tracers labeled with radioactive isotopes.

Overexpression of hepatic human GDPD3 in mice on a Western diet
Six-week old male mice were injected with AAV-GFP (n=15) or AAV-hGDPD3 (n=15). Two weeks later, the mice were switched from a standard rodent chow to a Western diet (ENVIGO #TD88137; 42% fat calories, 0.2% cholesterol) for 12 weeks to induce obesity, insulin resistance, and hepatic steatosis (15). Mice were used for a comprehensive systemic metabolic phenotyping between 8 to 10 weeks of a Western diet feeding as described above. Mice were also fasted for 16 hours for an intraperitoneal glucose (Sigma-Aldrich #G7528) tolerance test (1 g/kg body weight) or fasted for 4 h for an intraperitoneal insulin (Eli Lilly and Company Humulin RU-100) tolerance test (0.75 U/kg body weight).
AAV-injected mice were terminated after 12 weeks of Western diet feeding and an overnight fast and plasma and liver were harvested for lipid measurements as described above. Untargeted lipidomic analysis of liver LysoPA species was also performed. Liver (100 mg) was homogenized in 500 μl of Vanquish UHPLC system (Thermo Scientific) (20). Lipids were separated on an Accucore C30 column (2.6µm, 3mm x 150mm) using a linear gradient with acetonitrile:water (60:40, vol:vol) (mobile phase A) and isopropyl alcohol:acetonitrile (90:10, vol:vol) (mobile phase B), both of which contain 0.1% formic acid and 10mM ammonium formate. LysoPA (16:0), LysoPA (18:0), and LysoPA (18:1) standards were used to verify the LysoPA peaks based on the peak fragmentation and retention times. LipidSearch 4.2 software (Thermo Scientific) was used for lipid identification and quantitation. The LC-MS data containing high resolution MS and data dependent MS2 were searched using the parameters as follows: precursor mass tolerance, 5 ppm; product mass tolerance, 5 ppm; phospholipid species selection, LysoPA.
Primary hepatocytes were isolated from Western-diet-fed mice (8 to 10 weeks of feeding) and used to study de novo TAG biosynthesis using [1-14 C]-acetic acid (PerkinElmer #NEC084A001MC), FA esterification into TAG using [9,10-3 H(N)]-oleic acid (PerkinElmer #NET289005MC), FA uptake using [1-14 C]-palmitic acid as described previously (15). In each study, 3 to 5 biological experiments were conducted on separate days. In each experiment, primary hepatocytes were isolated from 1 mouse per genotype and 3 incubation time points were investigated (n=3 per time point).

Patient liver samples
This study was approved by the Institutional Review Board (IRB00010545) at Wake Forest School of Medicine. Patients undergoing Roux-en-Y gastric bypass, lap band, cholecystectomy, hernia repair, or other abdominal surgery were approached for informed consent to collect fasting blood samples (just prior to surgery) and one intraoperative liver biopsy during their procedure. The main inclusion criteria for participation were age 21-80 years and intent to undergo an elective abdominal surgery with the visualization of the liver at Wake Forest University Baptist Medical Center (Winston-Salem, NC).
Exclusion criteria were positive hepatitis C antibody or hepatitis B surface antigen, cirrhosis or chronic liver disease (other than NAFLD), past or present malignancy, other than non-melanoma skin cancer, or need for chronic anticoagulation. Over the course of the study, 54 subjects were recruited; no vulnerable populations were included. Fasting blood samples were used for glucose and insulin levels as well as by guest, on July 31, 2020 www.jlr.org Downloaded from plasma lipid profiles as described above. The liver biopsy was a resection of hepatic tissue taken with a scalpel as well as a core needle biopsy. The liver tissue was divided in the operating room into 2 samples.
One portion (~0.5 -1 g) was immediately snap frozen in liquid nitrogen for liver lipid profiles and RNA/cDNA extraction for GDPD3 gene expression by real-time PCR. The second portion of the liver sample (needle core biopsy) was fixed 10% buffered formalin and submitted to surgical pathology for histopathologic characterization by a certified pathologist at Wake Forest University Baptist Medical Center. Briefly, hematoxylin and eosin and trichrome slide preparations were performed per center standards (21). A semiquantitative measure of macro-steatosis was made by point counting hepatocytes at 4009 optical magnification. An overlying grid of 35 µm 2 boxes was used to assist counting. The hepatocytes of five, non-overlapping fields that did not include a portal tract were hand counted, yielding percentages of hepatocytes as macro-steatotic (lipid droplets large enough to displace the nucleus and/or ≥15 µm) and normal (no evidence of lipid droplets) (22). Degrees of macro-steatosis were graded according to convention established by Kleiner et al. (23) including <5% of macro-steatosis, grade 1; ≥5% to <30% of mild macro-steatosis, grade 2; ≥30% to ≤60% of moderate macro-steatosis, grade 3; >60% of severe macro-steatosis, grade 4.

Real-time PCR Analysis
Total RNA was isolated from hepatocytes and tissues using QIAzol lysis buffer (Qiagen #79306) and RNeasy Plus Universal Kit (Qiagen #73404). cDNA synthesis was performed with an Omniscript RT kit (Qiagen #205113) using 1µg of RNA. Real-time PCR was performed using TaqMan® Gene Expression (Mm00487205_m1). Gene expression was normalized to the housekeeping gene Gapdh (Mm99999915_g1) or 18S ribosomal RNA (Thermo Scientific #4352930E) and analyzed using the ΔΔ-CT method with 95% confidence (24).

Statistics
Results are presented as mean ± SEM. Data were analyzed using a two-tailed Student's unpaired t test for cell and mouse studies or unpaired t test with Welch's correction for the human study, or two-way ANOVA with Sidak multiple comparisons. All analyses were performed using GraphPad Prism 7 software. A p-value of 0.05 is used as the cutoff for statistical significance.

Endogenous GDPD3 gene expression
GDPD belongs to a highly conserved family of enzymes (12). The GDPD3 gene is conserved in human, chimpanzee, rhesus monkey, dog, cow, rat, mouse, zebrafish, and frog. Human GDPD3 is a 318 amino acid protein that shares 78.9 % sequence identity with mouse (330 amino acids) and 76.9% with rat (331 amino acids) orthologues. We measured Gdpd3 expression across multiple tissues from male and female C57BL/6J mice after a 24-hour fasting. Gdpd3 mRNA abundance was relatively low (Ct 29-33) in all tissues examined and was not affected by sex of the mice (Fig. 2A). We also found that livers displayed relatively low endogenous Gdpd3 mRNA expression ( Fig. 2A). Previous studies reported that livers of C57BL/6J mice accumulated large amounts of TAG (a 4-14 fold increase) after fasting for 6-24 hours (25,26). Indeed, we observed that fasting significantly increased liver TAG content (Fig. 2B), but did not modify Gdpd3 mRNA levels compared to refeeding (Fig. 2C). These data suggest that GDPD3 does not play a role in liver TAG homeostasis in response to fasting and fasting/refeeding status.

Overexpression of human GDPD3 in mouse liver increases hepatic LysoPA content but does not affect hepatic TAG content in chow-fed mice.
Previous studies reported that recombinant mouse and human GDPD3 is an ER membrane-associated enzyme that contains lysophospholipase D activity, hydrolyzing lysophospholipids to produce LysoPA in non-hepatic cell lines (13,27). Because LysoPA is an intermediate in the ER glycerol phosphate pathway for TAG synthesis (14), we hypothesized that human GDPD3 increases LysoPA substrate, promoting hepatic TAG biosynthesis and accumulation. To test our hypothesis, we performed a lipid class analysis using livers isolated from chow-fed mice injected with adeno-associated adenovirus encoding a hepatocyte-specific promoter (i.e., albumin) driving FLAG-tagged human GDPD3 (AAV-hGDPD3) or green fluorescent protein (AAV-GFP) as control. There was no difference in mouse endogenous liver Gdpd3 mRNA expression between AAV-GFP and AAV-hGDPD3 injected mice (Fig. 3A). Human GDPD3 overexpression in mouse liver was verified at protein levels (Fig. 3B). No expression of human by guest, on July 31, 2020 www.jlr.org Downloaded from GDPD3 was observed in adipose tissue and muscle, indicating the specificity of human GDPD3 expression in mouse liver (Fig. 3C). Hepatic LysoPA content was increased ~2-fold with human GDPD3 overexpression compared to GFP controls (Fig. 3D).
To confirm the data of liver lipid class analysis (Fig. 3D), we isolated primary hepatocytes from chow-fed mice and treated the cells with [ 14 C]-palmitic acid. The production of [ 14 C]-LysoPA from [ 14 C]palmitic acid increased in primary hepatocytes overexpressing human GDPD3 compared to GFP controls in a time-dependent manner (Fig. 3E). Overexpression of human GDPD3 in mouse primary hepatocytes did not affect mouse endogenous Gdpd3 mRNA levels (Fig. 3F) and was verified by increased human GDPD3 protein levels (Fig. 3G).
We then examined whether increased hepatic LysoPA content affected hepatic lipid content and systemic metabolism because our previous study suggested an association between dysregulated hepatic lipid metabolism and impaired systemic metabolic phenotype (15). We observed that mice overexpressing human GDPD3 in liver had increased plasma LysoPA levels ( Fig. 4A) but had normal circulating (Fig.   4B) and hepatic (Fig. 4C) lipid concentrations and content, respectively, compared to GFP control mice.

Overexpression of human GDPD3 increases LysoPA species and promotes TAG accumulation in liver of Western diet-fed mice.
LysoPA produced by human GDPD3 can serve as substrate for TAG synthesis in liver, which is stimulated by high-fat Western diets (28). We investigated whether hepatic overexpression of human GDPD3 in Western diet-fed mice would stimulate hepatic TAG synthesis and accumulation and worsen systemic metabolism. Although the body weight (Fig. 5A) was not statistically different from Western diet-fed GFP control mice, liver weight (Fig. 5B) and TAG levels (Fig. 5C) increased in Western diet-fed mice with hepatic human GDPD3 overexpression. Hepatic cholesteryl ester (CE) and free cholesterol (FC), but not phospholipid (PL), contents were also increased in Western diet-fed mice with hepatic human GDPD3 overexpression (Fig. 5D-F). We then conducted an untargeted lipidomic analysis and found one of LysoPA fatty acyl species (16:0) was significantly elevated in the livers of Western diet-fed mice consuming the Western diet (Fig. 5G). However, similar to results with chow-fed mice (Fig. 4), systemic phenotype (blood glucose, plasma insulin, plasma β-hydroxybutyrate, oxygen consumption, carbon dioxide production, respiratory exchange ratio, heat production, and food intake; Data not shown), including blood glucose (Fig. 5H) and insulin (Fig. 5I) homeostasis, was not significantly affected by hepatic GDPD3 overexpression, despite elevated hepatic neutral lipid content (Fig. 5C-E).

Hepatic GDPD3 overexpression increases lipogenesis in Western diet-fed mice.
To determine whether increased hepatic TAG in Western-diet fed mice overexpressing human GDPD3 was due to increased FA uptake or de novo hepatic lipogenesis, we incubated primary hepatocytes isolated from Western diet-fed mice with: 1) [ 14 C]-palmitic acid to measure FA uptake, 2) [ 3 H]-oleic acid to measure FA incorporation into hepatic neutral lipids, and 3) [ 14 C]-acetic acid to examine de novo FA synthesis and incorporation into neutral lipids. Hepatocyte [ 14 C]-palmitic acid uptake (Fig. 6A) was significantly higher in AAV-hGDPD3 mice versus AAV-GFP control mice consuming a Western diet.
[ 3 H]-oleic acid incorporation into hepatocyte TAG (Fig. 6B), CE (Fig. 6C) and, to a lesser extent PL (Fig.   6D) was increased in Western diet-fed AAV-hGDPD3 mice versus AAV-GFP control mice. However, hepatocyte TAG synthesis from [ 14 C]-acetic acid was not affected by GDPD3 overexpression (Fig. 6E), suggesting the primary effect of GDPD3 overexpression was on FA uptake and esterification into TAG ( Fig. 6A-B). Newly synthesized CE (Fig. 6F) from [ 14 C]-acetic acid was significantly increased in hepatocytes from Western diet-fed AAV-hGDPD3 mice versus AAV-GFP control mice, whereas newly synthesized PL (Fig. 6G) from [ 14 C]-acetic acid was reduced in Western diet-fed AAV-hGDPD3 hepatocytes versus control AAV-GFP hepatocytes, likely due to feedback inhibition of PL synthesis (29,30). Indeed, we did not observe any change of liver PL levels in the steady state between genotypes by guest, on July 31, 2020 www.jlr.org Downloaded from (Fig.5F). Human GDPD3 expression in mouse primary hepatocytes used for these experiments was determined by immunoblotting (Fig. 6H).
Studies have shown that hepatic steatosis may be present in up to 70% of patients with type 2 diabetes (32,33). In our study, 15 out of 33 patients with macro-steatosis were diabetics and only 1 out of 21 patients without steatosis was diabetic (Fig. 7E). Indeed, patients with macro-steatosis (n=33, 15 were diabetics) had increased fasting blood glucose (Fig. 7E) compared to patients without hepatic steatosis (n=21, 1 was diabetic). However, there was no difference in fasting plasma insulin (Fig. 7F)

Discussion
NAFLD encompasses a wide spectrum of liver abnormalities, ranging from benign hepatocellular accumulation of lipids (NAFL or hepatic steatosis), through NASH, to fibrosis and cirrhosis in the absence of excessive consumption of alcohol and hepatitis viral infection (2). The etiology and pathogenesis of NAFLD are poorly understood, reflecting the genetic and environmental heterogeneity of the disease (34). The molecular basis of the metabolic response to dietary fat and its role in NAFLD development remains to be elucidated. In this study, we reported that human GDPD3 overexpression leads to enhanced hepatic steatosis in diet-induced obese mice, and GDPD3 mRNA expression is associated with NAFL in human livers. GDPD3 catalyzed production of LysoPA, a precursor of TAG, and increased hepatic FA uptake and incorporation into esterified neutral lipids, ultimately promoting TAG formation and accumulation, resulting in hepatic steatosis.
Mouse and human GDPD3 were recently cloned and identified as an ER membrane-associated enzyme containing lysophospholipase D activity, converting lysophospholipid to LysoPA in non-hepatic cells (13,27). Since LysoPA is an intermediate in the glycerol phosphate pathway for TAG biosynthesis (14) and 90% of hepatic TAG is produced through this pathway, we explored whether GDPD3 overexpression increased TAG synthesis. We found that human GDPD3 overexpression stimulated LysoPA production in livers (Fig. 3D) and hepatocytes (Fig. 3E) of chow and Western diet-fed mice (Fig. 5G). Although increased hepatic LysoPA mediated by human GDPD3 did not affect TAG hemostasis in chow-fed mice (Fig. 4C), it resulted in hepatic lipid accumulation in Western diet-fed mice ( Fig. 5C-E). GDE1 (35), another GDPD (also known as GDE) family member, hydrolyzes glycerophosphoinositol (36) and glycerophosphocholine (37) to produce glycerol 3-phosphate which is also an intermediate in the glycerol phosphate pathway for TAG biosynthesis (14). Hepatic Gde1 overexpression in mice fed a high-fat, high-sucrose diet for 7 days induced hepatic TAG accumulation possibly by increasing glycerol 3-phosphate availability for de novo TAG synthesis (35), supporting the concept that the GDPD/GDE enzymes may contribute to lipogenesis by providing additional substrate for TAG synthesis from lysophospholipid precursors. The production of LysoPA from glycerol 3-phosphate is controlled by GPAT in the glycerol phosphate pathway for TAG biosynthesis (14). GPAT1 isoform is an outer mitochondrial membrane protein and contributes 30-50% of total GPAT activity in the liver (14).
The ER isoform GPAT4 also contributes about 50% of total GPAT activity in the liver (38). Because GPAT4 resides in the ER, where the glycerol phosphate pathway occurs for TAG synthesis, we examined endogenous liver GPAT4 gene expression in chow-and Western diet-fed mice. There was no difference in hepatic Gpat4 mRNA expression between AAV-GFP and AAV-hGDPD3 injected mice consuming chow or Western diet (Data not shown). Thus, we believe that the increased production of hepatic LysoPA in mice injected with AAV-hGDPD3 compared to AAV-GFP controls consuming chow (Fig.   3D) or Western diet (Fig. 5G) was mainly caused by human GDPD3 activity, not by increased GPAT expression.
To determine the relative contribution of exogenous versus de novo synthesized FA used for hepatic TAG biosynthesis in hepatocytes from Western diet-fed mice overexpressing human GDPD3, we used [ 14 C]-acetic acid to trace the de novo FA and TAG synthesis, [ 14 C]-palmitic acid to monitor cellular FA uptake, and [ 3 H]-oleic acid to follow FA esterification into TAG. Newly synthesized TAG from [ 14 C]acetic acid was not affected by human GDPD3 overexpression (Fig. 6E), indicating that de novo lipogenesis is unlikely causing hepatic TAG accumulation in AAV-hGDPD3 mice on Western diet. In addition, mRNA expression of acetyl-CoA carboxylase 1 and FA synthase, two enzymes that regulate de novo FA synthesis (39,40), was unaffected in Western diet-fed mouse liver expressing human GDPD3 (Data not shown). However, we observed an increase in hepatocyte [ 14 C]-palmitic acid uptake (Fig. 6A) and [ 3 H]-oleic acid incorporation into TAG (Fig. 6B) and CE (Fig. 6C), in human GDPD3overexpressing hepatocytes isolated from Western diet-fed mice. Donnelly et al. (41) used a multiplestable-isotope approach demonstrating that liver fatty acyl groups consisted of 60% of free FA influx from adipose tissue (lipolysis), 15% from the diet, and 25% of de novo lipogenesis (FA synthesis from acetate). In addition, Duarte et al. (42) used a deuterated water NMR approach and reported that a high-fat diet induced hepatic TAG accumulation was mediated through re-esterification of existing or ingested lipids, not de novo lipogenesis. Therefore, two components are required for our model: 1) GDPD3 increases LysoPA flux, and 2) a high-fat diet provides FA substrate. Through acylation and dephosphorylation reactions in the glycerol phosphate pathway, LysoPA is converted to TAG for hepatic neutral lipid storage and accumulation (Fig. 1).
How does overexpression of GDPD3 increase hepatic FA uptake (Fig. 6A)? Peroxisome proliferator-activated receptors (PPARs), such as PPARα and PPARγ, are ligand-activated nuclear receptors that transcriptionally regulate glucose and lipid metabolism in multiple metabolically active organs (43). PPARγ expression is low in liver, but elevated in NAFLD (44)(45)(46). Deletion of PPARγ in mouse hepatocytes protects against hepatosteatosis (45,47). Prior studies demonstrated that PPARγ activation induces the expression of genes involved in FA uptake, such as CD36/FA translocase (48) and FA transporter proteins (FATPs) (49), leading to increased hepatic TAG synthesis and storage. Because LysoPA is a well-known PPARγ ligand (50)(51)(52), we hypothesize that PPARγ is the functional receptor for LysoPA produced by human GDPD3, enhancing hepatic steatosis via increased FA uptake through CD36 and/or FATPs (Fig. 1). We found that total and membrane associated hepatic CD36 protein expression was not affected by overexpression of human GDPD3 (Data not shown). We will focus future studies on investigating the effect of human GDPD3 overexpression on liver PPARγ activation and FA uptake via FATPs, such as FATP2 (53) and FATP5 (54). Lastly, PPARα is the predominant PPAR isoform in the liver and regulates hepatic FA metabolism by inducing expression of genes involved in FA oxidation including carnitine palmitoyltransfersase 1 (Cpt1) and Cpt2 (55). CPT1α is the primary Cpt1 isoform expressed in liver (55). LysoPA has not previously been identified as a PPARα ligand and there was no difference in the gene expression of hepatic PPARα, Cpt1α, and Cpt2 between Western diet-fed AAV-GFP and AAV-hGDPD3 injected mice (Data not shown), suggesting that human GDPD3 overexpression did not affect hepatic FA oxidation.
Our study has potential limitations. We did not perform loss-of-function studies because minimal endogenous hepatic Gdpd3 gene expression in chow-fed C57BL/6J mice ( Fig. 2A) reduces the dynamic range to observe functional differences between the wild-type and Gdpd3 knockdown livers. In addition, global Gdpd3 knockout mice have no apparent visible abnormalities as assessed by hematoxylin-and by guest, on July 31, 2020 www.jlr.org Downloaded from eosin-stained liver sections compared to age-and gender-matched controls (56). Feeding global Gdpd3 knockout mice a high-fat diet is required to determine whether inhibition of GDPD3 contributes significantly to prevention of hepatic steatosis. Another limitation is that hepatic steatosis in Western dietfed mice was achieved with supraphysiological hepatic overexpression of human GDPD3. It remains unknown if the relatively small increase in hepatic GDPD3 mRNA in humans with liver steatosis (Fig.   7C) is sufficient to contribute to enhanced liver TAG accumulation.
In summary, our findings indicate that GDPD3 overexpression increases LysoPA availability and hepatic FA flux, promoting TAG biosynthesis and accumulation through the ER glycerol phosphate pathway in mouse hepatocytes/livers. Although our results rely on GDPD3 overexpression, our human data suggest that increased dietary FA availability may be responsible for increased transcription of hepatic GDPD3 in humans with hepatic steatosis and identify GDPD3 as a potential contributor to Western diet-induced hepatic steatosis.

Data availability
The data supporting the findings of this study are available within the article. Derived data supporting the findings of this study are available from the corresponding author (C.C.K.) on request.   In a second study, 12-week old male mice were fasted for 24 hours (n=5), and another group was fasted for 24 hours followed by refeeding a standard chow diet for 12 hours (n=5); fasting times were staggered so that both groups of mice were sacrificed together. Livers were collected to measure (B) triacylglycerol (TAG) levels and (C) Gdpd3 mRNA levels by real-time PCR. All results are mean ± SEM and analyzed using a two-tailed Student's unpaired t test. Ct, cycle threshold; WAT, white adipose tissue; BAT, brown adipose tissue.  Six-week old male mice were treated with the indicated adeno-associated adenovirus (AAV; n=6 per genotype). Eighteen weeks after AAV injection, plasma and livers were collected from chow-fed mice for the measurement of (A) plasma lysophosphatidic acid (LysoPA), (B) plasma triacylglycerol (TAG), total cholesterol (TC), and free cholesterol (FC), and (C) liver TAG, TC, cholesteryl ester (CE), FC, and phospholipid (PL). (D) Chow-fed mouse body weight was measured at 8, 16 and 24 weeks of age. Chow-fed mice (16-18 weeks of age) were used for the measurement of (E) blood glucose, (F) plasma insulin, (G) β-hydroxybutyrate, (H) oxygen consumption (VO2, ml/hr/kg), (I) carbon dioxide production (VCO2, ml/hr/kg), (J) respiratory exchange ratio (VCO2/VO2), (K) heat production (kcal/hr/kg), and (L) food intake (g/12hr). All results are mean ± SEM and analyzed using a two-tailed Student's unpaired t test.

Fig. 5. Overexpression of human GDPD3 in Western diet-fed mice increases hepatic lipid content.
Six-week old male mice were treated with the indicated adeno-associated adenovirus (AAV; n=10 per genotype) for 2 weeks and then fed a Western diet for 12 weeks. Western diet-fed mice were used for the measurement of (A) body weight at 8, 12, 16, and 20 weeks of age. Livers of Western diet-fed mice were collected for the measurement of (B) wet weight, (C) triacylglycerol (TAG), (D) cholesteryl ester (CE), (E) free cholesterol (FC), (F) phospholipid (PL), and (G) LysoPA (16:0). (H) Glucose tolerance test and (I) insulin tolerance test were performed between 8 to 10 weeks of a Western diet feeding. All results are mean ± SEM and analyzed using a two-tailed Student's unpaired t test. ]-acetic acid, as well as expression of (H) human GDPD3 protein, were measured. All results are mean ± SEM and analyzed using two-way ANOVA with Sidak multiple comparisons (A-G). DPM, disintegrations per minute.

Fig. 7. Liver and plasma lipid profiles in individuals without or with macro-steatosis.
Liver biopsy and fasting blood samples were collected from 54 individuals for the measurement of (A) liver triacylglycerol (TAG), (B) liver cholesteryl ester (CE), (C) liver GDPD3 mRNA, (D) fasting plasma TAG, (E) fasting blood glucose, and (F) fasting plasma insulin levels. All results are mean ± SEM and analyzed using unpaired t test with Welch's correction.