LPIAT1/MBOAT7 depletion increases triglyceride synthesis fueled by high phosphatidylinositol turnover

Objective Non-alcoholic fatty liver disease (NAFLD) is a common prelude to cirrhosis and hepatocellular carcinoma. The genetic rs641738 C>T variant in the lysophosphatidylinositol acyltransferase 1 (LPIAT1)/membrane bound O-acyltransferase domain-containing 7, which incorporates arachidonic acid into phosphatidylinositol (PI), is associated with the entire spectrum of NAFLD. In this study, we investigated the mechanism underlying this association in mice and cultured human hepatocytes. Design We generated the hepatocyte-specific Lpiat1 knockout mice to investigate the function of Lpiat1 in vivo. We also depleted LPIAT1 in cultured human hepatic cells using CRISPR-Cas9 systems or siRNA. The effect of LPIAT1-depletion on liver fibrosis was examined in mice fed high fat diet and in liver spheroids. Lipid species were measured using liquid chromatography-electrospray ionisation mass spectrometry. Lipid metabolism was analysed using radiolabeled glycerol or fatty acids. Results The hepatocyte-specific Lpiat1 knockout mice developed hepatic steatosis spontaneously, and hepatic fibrosis on high fat diet feeding. Depletion of LPIAT1 in cultured hepatic cells and in spheroids caused triglyceride accumulation and collagen deposition. The increase in hepatocyte fat content was due to a higher triglyceride synthesis fueled by a non-canonical pathway. Indeed, reduction in the PI acyl chain remodelling caused a high PI turnover, by stimulating at the same time PI synthesis and breakdown. The degradation of PI was mediated by a phospholipase C, which produces diacylglycerol, a precursor of triglyceride. Conclusion We found a novel pathway fueling triglyceride synthesis in hepatocytes, by a direct metabolic flow of PI into triglycerides. Our findings provide an insight into the pathogenesis and therapeutics of NAFLD.


AbsTrACT
Objective Non-alcoholic fatty liver disease (NAFLD) is a common prelude to cirrhosis and hepatocellular carcinoma. The genetic rs641738 C>T variant in the lysophosphatidylinositol acyltransferase 1 (LPIAT1)/ membrane bound O-acyltransferase domaincontaining 7, which incorporates arachidonic acid into phosphatidylinositol (PI), is associated with the entire spectrum of NAFLD. In this study, we investigated the mechanism underlying this association in mice and cultured human hepatocytes. Design We generated the hepatocyte-specific Lpiat1 knockout mice to investigate the function of Lpiat1 in vivo. We also depleted LPIAT1 in cultured human hepatic cells using CRISPR-Cas9 systems or siRNA. The effect of LPIAT1-depletion on liver fibrosis was examined in mice fed high fat diet and in liver spheroids. Lipid species were measured using liquid chromatography-electrospray ionisation mass spectrometry. Lipid metabolism was analysed using radiolabeled glycerol or fatty acids. results The hepatocyte-specific Lpiat1 knockout mice developed hepatic steatosis spontaneously, and hepatic fibrosis on high fat diet feeding. Depletion of LPIAT1 in cultured hepatic cells and in spheroids caused triglyceride accumulation and collagen deposition. The increase in hepatocyte fat content was due to a higher triglyceride synthesis fueled by a non-canonical pathway. Indeed, reduction in the PI acyl chain remodelling caused a high PI turnover, by stimulating at the same time PI synthesis and breakdown. The degradation of PI was mediated by a phospholipase C, which produces diacylglycerol, a precursor of triglyceride. Conclusion We found a novel pathway fueling triglyceride synthesis in hepatocytes, by a direct metabolic flow of PI into triglycerides. Our findings provide an insight into the pathogenesis and therapeutics of NAFLD.

InTrODuCTIOn
In parallel with the obesity epidemics, non-alcoholic fatty liver disease (NAFLD) is an emerging threat for human health. NAFLD comprises a spectrum of conditions, going from excess in hepatic triglyceride (TG) content in the absence of alcohol consumption, leading to cirrhosis and hepatocellular carcinoma. Increase in TG synthesis is one of the mechanisms leading to NAFLD. 1 2 Lipogenesis may be mediated by at least two mechanism: (1) a high flux of carbon down glycolysis (as during excess in fructose significance of this study What is already known on this subject? ► Non-alcoholic fatty liver disease (NAFLD), the most common chronic liver disease worldwide, is characterised by an increase in hepatic synthesis of triglycerides classically driven by carbohydrate. ► The genetic rs641738 C>T variant in the membrane bound O-acyltransferase domain-containing 7 (MBOAT7)/ lysophosphatidylinositol acyltransferase 1 (LPIAT1) associated with the entire spectrum of NAFLD. However, the mechanism underlying this association remains unclear.
What are the new findings? ► The hepatocyte-specific Lpiat1 knockout mice developed steatosis and LPIAT1 depletion in cultured human hepatocytes upregulates triglyceride synthesis. ► LPIAT1 depletion in human and murine hepatocytes caused a high phosphatidylinositol (PI) turnover, which continuously produced diacylglycerol, a substrate for triglyceride synthesis. ► LPIAT1 depletion exacerbated the fibrotic phenotype in liver spheroids and in mice fed high fat diet, recapitulating the human phenotype.
How might it impact on clinical practice in the foreseeable future? ► Classically, increased hepatic fat content may be due to higher de novo lipogenesis due to insulin action or high fructose intake. In this work, we show a novel pathway where lipogenesis is due to a direct metabolic flow of PI into triglycerides caused by downregulation of MBOAT7/LPIAT1. Ultimately, this novel de novo lipogenesis pathway may be used as therapeutic target to treat NAFLD. intake 3 4 ), supplying excess triose phosphates, which serve as the glycerol backbone for TG synthesis; (2) a sterol regulatory element-binding protein 1c (SREBP1c)-mediated production of fatty acids due to fatty acid synthesis. 2 5 NAFLD has also a strong genetic susceptibility, and among the genes of susceptibility 6 the single nucleotide polymorphism (rs641738 C>T) in the membrane bound O-acyltransferase domain-containing 7 (MBOAT7) gene increases the susceptibility to the entire spectrum of NAFLD, 7 8 by inducing a reduction in the liver MBOAT7 mRNA expression. [8][9][10][11] MBOAT7 encodes lysophosphatidylinositol acyltransferase 1 (LPIAT1), which preferentially incorporates arachidonic acid (AA; 20:4 n-6) into phosphatidylinositol (PI). [12][13][14] PI is a constitutive of membrane phospholipids and the precursor of signalling lipid phosphoinositides, 15 but so far no direct role of PI in TG synthesis has been shown.
Unlike other major membrane phospholipids containing various unsaturated fatty acids, PI contains most exclusively AA as acyl chain in the sn-2 position. 16 17 Individuals homozygous for rare disruptive mutations in the LPIAT1/MBOAT7 gene (hereafter called LPIAT1) have severe intellectual disability, epilepsy and autistic features. 18 Consistently, Lpiat1-deficient mice exhibits aberrant brain development accompanied by disordered lamination and neuronal processes in the cortex, suggesting that AA-containing PI produced by LPIAT1 is essential for neonatal neural development. 13 However, as Lpiat1deficient mice die within a month after birth, the mechanism by which Lpiat1 deficiency causes liver steatosis has not been investigated.

Depletion of Lpiat1 in adult mice causes steatosis
To bypass the neonatal lethality of Lpiat1-deficient mice, we first generated the tamoxifen-inducible Lpiat1 knockout mice (Ubc-CreER T2 ; Lpiat1 f/f ), by mating Lpiat1-floxed mice (Lpiat1 f/f ) with the transgenic mice carrying tamoxifeninducible Cre transgene (Ubc-CreER T2 ) 19 (online supplementary figure 1A). In Ubc-CreER T2 ; Lpiat1 f/f mice, Lpiat1 f alleles were almost completely converted to deleted alleles (Lpiat1 Δ ) within 1 week after tamoxifen treatment (online supplementary figure 1B). LPIAT1 protein measured by western blotting was absent in all of the tissues examined, including brain, kidney and liver, after 2 weeks after tamoxifen treatment (figure 1A, online supplementary figure 1C). Tamoxifentreated Ubc-CreER T2 ;Lpiat1 f/f (Lpiat1 Δ/Δ ) mice fed on chow diet grew normally for at least 2 months. As expected, liquid chromatography-electrospray ionisation mass spectrometry (LC/ESI-MS) analysis showed that the amount of 38:4 PI (most probably 18:0/20:4 PI) was dramatically reduced in Lpiat1 Δ/Δ liver, whereas other PI species, such as 34:1 and 36:2, were slightly increased (online supplementary figure  1D). In contrast, molecular species of phosphatidylcholine (PC) in Lpiat1 Δ/Δ liver were similar to those in Lpiat1 f/f liver (online supplementary figure 1E). Contrary to conventional Lpiat1 knockout mice, 13 Lpiat1 Δ/Δ mice showed a histologically normal brain morphology (figure 1B). Consistently with the carriers of the MBOAT7 rs641738 C>T variant, Lpiat1 Δ/Δ livers showed accumulation of neutral lipids (figure 1C,D, online supplementary figure 1F). These results suggest that the depletion of Lpiat1 in adult mice results in an increased hepatic TG content.
Hepatocyte-specific Lpiat1-deficient mice develop steatosis Next, to examine whether Lpiat1 specific depletion in hepatocytes causes steatosis, we generated hepatocyte-specific Lpiat1 knockout mice (LKO), by crossing the Lpiat1 f/f mice with the albumin-Cre (Alb-Cre) mice. 20 LPIAT1 was depleted in the liver in LKO mice as shown by western blotting ( figure 1E). Similarly to Lpiat1 Δ/Δ liver, the amount of AA-containing PI (38:4 PI) was dramatically reduced in LKO liver, whereas other PI species, such as 34:2, 34:1, 36:3, 36:2 and 36:1 PI, were significantly increased ( figure 1F). There was no accumulation of lysoPI, the major lipid substrate of LPIAT1, in LKO liver (online supplementary figure 2A). PI content was slightly decreased in LKO liver ( figure 1G). There was not difference in the amount of other phospholipids (PC, phosphatidylethanolamine (PE), phosphatidylserine (PS), sphingomyelin (SM), cardiolipin (CL) and phosphatidylglycerol (PG)) (figure 1G, online supplementary figure 2B-E), except the 34:2 PG and 36:2 PG that were increased in LKO liver (online supplementary figure 2E). LKO mice displayed higher liver weights compared with control mice (Lpiat1 f/f ) ( figure 1H). Hepatocyte-specific depletion of Lpiat1 also caused significant increase in neutral liver fat content in the form of TGs, as revealed by Oil red O staining and biochemical measurement (figure 1I,J). There was no difference in hepatic cholesterol content between LKO and Lpiat1 f/f liver (figure 1K). mRNA levels of several genes involved in lipogenesis, including Srebf1, were unchanged (figure 1L). mRNA levels of genes involved in fatty acid β-oxidation and gluconeogenesis were not significantly changed, except for Pklr that was increased in LKO mice ( figure 1L). Taken all together, these results show that Lpiat1 depletion in hepatocytes causes (1) remodelling of PI with an overall depletion of AA, and (2) increased liver TG content.
Next, to understand the mechanism underlying the increase in liver fat content, we then examined the rate of hepatic TG secretion in the form of very low-density lipoprotein (VLDL) in LKO mice. Fasted mice were injected intraperitoneally poloxamer 407, a lipoprotein lipase inhibitor 21 and TG concentration in plasma were assessed. There was no decrease in the rate of hepatic TG secretion between LKO mice and Lpiat1 f/f mice (figure 1M), indicating an absence of VLDL retention in the liver. However, total plasma TG was decreased to approximately half in LKO mice (online supplementary figure 2F), and VLDL-TG was reduced among the plasma lipoprotein fractions in LKO plasma (online supplementary figure 2G). This might be explained by an increase in the catabolism of VLDL-TG mediated by LPL and other lipases. The total plasma cholesterol levels and plasma cholesterol distribution among the plasma lipoproteins were comparable between LKO and Lpiat1 f/f mice (online supplementary figure 2H,I).

LPIAT1 depletion causes cellular TG accumulation in cultured human hepatocytes
To elucidate the molecular mechanism causing steatosis in LKO mice, we generated hepatoma cell line (Huh-7) deficient in LPIAT1 using CRISPR-Cas9 system (figure 2A). In Huh-7 cells, a substantial amount of 20:3-containing PI (38:3 PI) was detected. This is probably because Mead acid (20:3) is synthesised from oleic acid as a substitute of 20:4 in the cells cultured with foetal bovine serum (FBS)-containing medium, which contains much less essential fatty acids than murine plasma and tissues. 22   Consistently with our data in Huh-7 cells, LPIAT1 downregulation results in a twofold increase in intracellular lipid content (figure 2E,F). Taken all together, these data demonstrate that LPIAT1 deficiency causes intracellular TG accumulation in a hepatocyte-autonomous manner.

Depletion of LPIAT1 in cultured human hepatocytes results in higher TG synthesis
To determine how LPIAT1 deficiency increased cellular TG levels, cells were incubated with [ 14 C]oleic acid for 24 hours, and the radioactivity in TG fractions from cells and culture medium were measured. The radioactivity in both cellular TGs, and the sum of cellular and medium TGs, corresponding to the total synthesised TGs, were increased by 30% after LPIAT1 depletion (figure 2G). To confirm our data, we incubated HepG2 cells with [ 3 H]glycerol for 5 hour, lipids were extracted and separated by one-dimensional thin layer chromatography (TLC), and radioactivity corresponding to TGs was measured (figure 2H). Consistently with our previous experiment, siRNA-mediated LPIAT1 knockdown resulted in a 30% increase in TG synthesis (figure 2H). These results indicate that either TG synthesis or TG degradation are affected by the depletion of LPIAT1, without affecting TG secretion. Our results are consistent with a recent study focusing on the role of LPIAT1 on hepatic insulin resistance showing depletion of LPIAT1 in Huh-7 results in higher TG synthesis. 23 To assess the TG degradation, cells were incubated with [ 14 C] glycerol for 12 hours, and subsequently chased with label-free medium and triacsin C to inhibit incorporation of [ 14 C]glycerol into TG. TG degradation rate was determined by monitoring [ 14 C]-labelled TGs at each time point. TG degradation rate was not different between wild-type (WT) and LPIAT1 KO cells (figure 2I). Consistent with this observation, autophagymediated TG degradation was comparable between WT and LPIAT1 KO Huh-7 cells (figure 2J). β-oxidation was also not different in LPIAT1 KO Huh-7 cells (figure 2K) and in LPIAT1depleted HepG2 cells ( figure 2L). Taken all together, data suggest that LPIAT1 downregulation/depletion results in higher intracellular TG content, specifically by increasing TG synthesis without affecting TG degradation or secretion.
Expression of most of the genes involved in TG metabolism (figure 2M) including SREBF1 was not changed in Huh-7. Moreover, phosphorylation of Akt, which accelerates lipid synthesis ( figure 2N,O), was not affected by the depletion of LPIAT1 in this model. Finally, genes involved in TGs synthesis were not changed after siRNA-mediated LPIAT1 knockdown in HepG2 cells (data not shown).

Depletion of LPIAT1 in cultured human hepatocytes results in higher PI turnover
To elucidate the mechanism underlying upregulated TG synthesis in LPIAT1 KO Huh-7 cells, we analysed phospholipid metabolism. Phospholipid synthesis rate was measured by the incorporation of [ 14 C]glycerol into each phospholipid fraction. The rate of [ 14 C]glycerol incorporation into PI was increased by LPIAT1 depletion, while incorporation into PC and PE did not change ( figure 3A,B). PI synthesis was also upregulated in LKO mice liver (figure 3C,D). PI is produced from phosphatidic acid by sequential action of CDP-diacylglycerol (CDP-DAG) synthase (CDS) and PI synthase. mRNA levels of CDS1, CDS2 and CDIPT (encoding PI synthase) were not significantly changed (online supplementary figure 3G). CDS2 is inhibited by AA-containing PI, 24 and in our model the prediction would be an activation of CDS2 due to the lack of AA-containing PI. To test this hypothesis, cells were incubated with [ 3 H]CTP for 2 hour, and the radioactivity in CDP-DAG fractions from cell lysates were measured. As expected, the rate of CDP-DAG synthesis was highly increased by LPIAT1 depletion ( figure 3E). Similarly, levels of CDP-DAG were strikingly increased both in LPIAT1 KO Huh-7 cells and the liver of LKO mice, reflecting the upregulation of CDS activity (figure 3F,G, online supplementary figure 3H and I). These data suggest that depletion of AA-containing PI causes increased activity of CDS2, leading to enhanced PI synthesis in LPIAT1 KO Huh-7 cells.
In spite of accelerated PI synthesis, cellular PI levels were comparable between WT and LPIAT1 KO Huh-7 cells (online supplementary figure 3J), suggesting that LPIAT1 depletion boosted at the same time PI synthesis and degradation. To measure PI degradation rate, we performed pulse-chase experiments with [  PI are degraded by phospholipase C (PLC) into diacylglycerol, which is a substrate for TG synthesis. Thus, we hypothesised that PLC activity is upregulated in the LPIAT1 KO cells and LKO liver. Inositol monophosphate (IP 1 ) was accumulated in LPIAT1 KO cells and accumulation of IP 1 was enhanced by inhibiting IP 1 phosphatase with Li + (figure 3L). The accumulation of IP 1 was also observed in the liver of LKO mice (figure 3M). These data suggest that PLC activity is upregulated by the absence of LPIAT1. Thus, we concluded that depletion of AA-containing PI causes activation of both PI synthesis by upregulating CDS2 activity, and PI degradation into diacylglycerol by a protein with PLC activity to generate TG ( figure 3N).

lPIAT1 depletion results in liver inflammation and fibrosis
MBOAT7 rs641738 C>T sequence variant is associated with liver fibrosis. To investigate the effect of LPIAT1 downregulation on fibrogenesis, we downregulated LPIAT1 by siRNA in spheroids composed by hepatocytes (HepG2 cells) and hepatic stellate

Hepatology
(LX-2) cells. 25 26 LPIAT1 knockdown did not affect cell viability (figure 4A) and resulted in higher collagen synthesis by hepatic stellate cells in spheroids (figure 4B,C). We next examined mRNA expression of markers of fibrogenesis, fibrosis remodelling and inflammation. LPIAT1 knockdown resulted in higher collagen expression, activation of stellate cells and upregulation of inflammatory and profibrotic cytokines ( figure 4D). To confirm our data in vivo, Lpiat1 f/f and LKO mice were fed high fat diet (HFD). LKO mice showed no differences in the relative oxygen consumption (VO 2 ) (online supplementary figure 4A,B). LKO mice gained body weight similar to Lpiat1 f/f mice fed chow diet during 18 weeks from 6 weeks old ( figure 4E). However, when fed HFD, LKO mice gained less body weight than their Lpiat1 f/f littermates. Despite the lower weight, liver weight and TG content were higher in LKO mice than in Lpiat1 f/f mice ( figure 4F,G). Furthermore, LKO mice showed elevated circulating levels of alanine aminotransferase (ALT) and aspartate aminotransferase (AST) when fed HFD (figure 4H,I), suggesting hepatic damage in these mice. On the other hand, there was no difference in Akt phosphorylation in the liver between Lpiat1 f/f and LKO mice on insulin stimulation (online supplementary figure 4C-E).
To understand the changes associated with hepatic injury in LKO mice, hepatic gene expression from Lpiat1 f/f and LKO mice fed either chow or HFD for 18 weeks were measured by microarray analysis. Consistently with our results in spheroids, categorisation of the differentially expressed genes between HFD-fed Lpiat1 f/f and LKO mice, in relation to their functions in biological pathways, revealed an enrichment of genes involved in inflammation and fibrosis ( figure 4J,K). Real-Time qPCR confirmed that the expression of inflammatory and fibrogenic genes were significantly higher in LKO liver than in the Lpiat1 f/f liver on HFD feeding (figure 4L). Consistently, marked accumulation of extracellular collagen and inflammatory cells were observed in the section of HFDfed LKO mice liver ( figure 4M).
Taken all together, these results suggest that LPIAT1 depletion results in NAFLD with fibrosis, recapitulating the human phenotype. Since it was proposed that phospholipid acyltransferases promote the reacylation of free AA and reduce inflammatory eicosanoid production, 6 we investigated eicosanoid levels. Consistently with the recent work by Helsley et al, 23 the levels of inflammatory bioactive lipid mediators were not affected, indicating that enhanced inflammation in LKO mice liver fed HFD were independent of inflammatory bioactive lipid mediators (online supplementary figure 4F-I).

LPIAT1 rs641738 variant decreases levels of LPIAT1 mrnA and AA-containing PI in the human liver, without affecting expression of genes for de novo lipogenesis
Finally, we investigated effect of LPIAT1 rs641738 variant on phospholipid composition and mRNA expression of genes for de novo lipogenesis and fibrosis in human liver samples. Similarly to in vitro and in vivo data, lipidomic analysis of the liver from 115 morbidly obese individuals from Finland 27 showed that rs641738 T allele was associated with a reduction in 36:4 PI and 38:3 PI (figure 5A), probably composed by AA and mead acid in its acyl chains, respectively. On the other hand, there were no differences in other phospholipids, namely PC and PE (data not shown).
Next, we examined differences in hepatic mRNA levels of genes involved in de novo lipogenesis and fibrosis among the rs641738 genotypes in 123 morbidly obese individuals from Italy. 28 As expected, rs641738 T allele was associated with the reduction of LPIAT1 mRNA (figure 5B). Consistently with the results in mice, there was no difference in the liver mRNA levels of genes for de novo lipogenesis and the other pathways involved in hepatic TG handling tested among the LPIAT1 rs641738 genotypes (figure 5C). However, there was no difference in the expression of genes related to fibrosis among the genotypes ( figure 5D). This may probably due to the presence of a low degree of fibrosis in this cohort.

DIsCussIOn
In the present study, we demonstrated that depletion of LPIAT1 in hepatocytes causes TG accumulation fueled by an elevated PI turnover, and it promotes liver fibrosis in vivo and in in vitro liver spheroids. Although phospholipids and TGs are synthesised from a common precursor, namely phosphatidic acid, metabolic interference between phospholipids and TGs has never been shown. In this work, we show for the first time a direct metabolic flow from PI to TGs mediated by LPIAT1 depletion in hepatocytes, causing an increase in the total hepatic TG content.
Given that PI is a minor component of membrane phospholipids (~5% of total phospholipids in the mouse liver) and that the PI amount (~63 nmol/mg protein) was far less than TG amount (~428 nmol/mg protein) in LKO liver, how does the PI breakdown match the TG accumulation in LPIAT1-depleted liver in terms of quantity? The metabolic labelling study in hepatocytes suggests that it is the increased PI turnover fueling quantitatively the accumulation of TGs, most probably by supplying diacylglycerol in LPIAT1-depleted hepatocytes.
LPIAT1 depletion caused liver inflammation and fibrosis in mice and in human organoids without changes in inflammatory bioactive lipid mediators related to AA. A Mendelian randomisation study using common genetic variants increasing liver fat content, 29 including the MBOAT7 rs641738 C>T variant, showed that increase in liver fat content causes hepatic inflammation and fibrosis. The results of the present study are consistent with this model, where excess in liver fat is per se deleterious to the liver.
NAFLD is characterised by increase in hepatic TG synthesis, mediated by SREBP1c dependent or fructose driven lipogenesis, 2 and therapeutic targets are in development to treat NAFLD acting on this pathway. 30 31 With our findings, we have identified a novel pathway for TG synthesis mediated by a faster PI metabolism. Future studies are warranted to understand whether the downregulation of this pathway is a viable strategy to treat NAFLD.

MeTHODs Mice
To generate Lpiat1 floxed mice on a C57BL6/J genetic background, a~2 kb fragment of genomic DNA containing exon 2-4 of the mouse  Ubc-CreER T2 and Alb-Cre mice to generate Ubc-Cre; Lpiat1 f/f and Alb-Cre; Lpiat1 f/f (LKO), respectively. Tamoxifen was solubilised at a concentration of 20 mg/mL in corn oil (Wako) and 200 µL per 20 g body weight was injected intraperitoneally into 8-week-old male mice for 5 days to delete Lpiat1 gene in Ubc-Cre Lpiat1 f/f . All mice were housed in climate-controlled (23℃) pathogen-free facilities with a 12-hour light-dark cycle, with free access to standard chow (CE2; CLEA Japan) and water. A 6-week-old LKO and littermate control mice were placed on a HFD (HFD-60; Oriental Yeast) for 18 weeks. All animal experiments were performed in accordance with protocols approved by Animal Committees of the University of Tokyo in accordance with the Standards Relating to the Care and Management of Experimental Animals in Japan.

Histological analysis
Tissues were fixed in Bouin's fluid (picric acid/37% formaldehyde/acetic acid=75/25/5) or 10% neutral-buffered formalin and embedded in paraffin. The sections (5 µm thickness) were cut using a microtome and stained with H&E or stained using Picro-Sirius Red Stain kit (ScyTek Laboratories). For immunohistochemistry, the deparaffinised sections were immunostained using the antibody against F4/80 (1/50, Bio-Rad) and VECTASTAIN Elite ABC kit (Vector Laboratories) with biotinylated Rabbit Anti-Rat IgG antibody (Vector Laboratories). The sections were also counterstained with hematoxylin. For Oil Red O staining, tissues were frozen in O.C.T. Compound (Sakura Finetek) and cut using a cryostat. The sections (10 µm thickness) were allowed to dry overnight at room temperature, washed with water, rinsed with 60% isopropanol, and stained with Oil Red O solution for 15 min.

lipid extraction
Lipids of each tissue and cells were extracted by the method of Bligh and Dyer. 32 The extracted solutions were dried up with centrifugal evaporator, dissolved in methanol : isopropanol=1:1, and stored at −20°C. Fatty acid metabolites were further purified from tissues by solid-phase extraction using InertSep NH2 columns (GL Science) with deuterium-labelled internal standard (11(12)-EET-d11). Briefly, InertSep NH2 columns were preconditioned with 6 mL of hexane and lipids extracted from tissues by the method of Bligh and Dyer were applied with 500 µL of chloroform. Columns were then washed with 6 mL of chloroform/isopropanol (2/1, v/v), followed by the elution with diethyl ether/acetic acid (98/2, v/v).The extracted solutions were dried up with centrifugal evaporator, dissolved in methanol:isopro-panol=1:1, and stored at −20°C.

electrospray ionisation mass spectrometry
For the detection of phospholipids, LC/ESI-MS-based lipidomics analyses were performed on a Shimadzu Nexera UPLC system (Shimadzu) coupled with a QTRAP 4500 hybrid triple quadrupole linear ion trap mass spectrometer (AB SCIEX). Lipids extracted from tissues or cells were injected by an autosampler; typically, 10 µL (3 nmol phosphorous equivalent) of the sample was applied. Chromatographic separation was performed on a SeQuant ZIC-HILIC PEEK coated column (250 mm × 2.1 mm, 1.8 µm; Millipore) maintained at 50°C using mobile phase A (water/acetonitrile (95/5, v/v) containing 10 mM ammonium acetate) and mobile phase B (water/acetonitrile (50/50, v/v) containing 20 mM ammonium acetate) in a gradient programme (0-22 min: 0% B→40% B; 22-25 min: 40% B→40% B; 25-30 min: 0% B) with a flow rate of 0.5 mL/min. The instrument parameters for negative ion mode were as follows: curtain gas, 10 psi; collision gas, 7 arb. unit; ionspray voltage, −4500 V; temperature, 700°C; ion source gas 1, 30 psi; ion source gas 2, 70 psi. The instrument parameters for positive ion mode were as follows: curtain gas, 10 psi; collision gas, 7 arb. unit; ionspray voltage, 4500 V; temperature, 700°C; ion source gas 1, 30 psi; ion source gas 2, 50 psi. Precursor ion scan of 184 Da with a scan range of m/z 510-960 for Q1 in the positive ion mode were used to detect PC and SM. Neutral loss scans of 141 Da with a scan range of m/z 510-960 for Q1 in the positive ion mode were used to detect PE. Precursor ion scan of m/z 241 and 153, and neutral loss scans of 87 Da with a scan range of m/z 550-1000 for Q1 in the negative ion mode was used to detect PI (LPI), PG and PS, respectively. Detection of CDP-DAG species was performed by multiple reaction monitoring (MRM) as described in online supplementary table 1.

Determination of phospholipid composition
Phospholipids were separated by TLC on silica gel 60F254 plates (Merck) in chloroform:methanol:acetic acid=65:25:13. The area of silica gel corresponding to each phospholipid (SM, PC, PS+PI, PE, PG and CL) was scraped off the plates. Isolated phospholipids were then extracted by the method of Bligh and Dyer described above. The fraction containing PS and PI was further separated by TLC in chloroform:methanol:formic acid:water=60:30:7:3, then PS and PI were extracted separately. To determine the amount of each phospholipid, lipid phosphorus was measured by Bartlett's method. 34

Measurement of cellular or hepatic TG and cholesterol level
TG E Test Wako and Cholesterol E Test Wako (Wako) were used respectively according to manufacturer's instruction. In brief, lipid extracts from tissues or cells were dried in a glass and added 15% TritonX-100 in acetone followed by drying up and incubated with reaction solution for 5 min. Absorption at 600 nm was measured and TG and cholesterol levels were calculated.

real-time PCr
Total RNA was extracted from tissues and cells using Isogen II (Nippon-gene) and reverse-transcribed using the high-capacity complementary DNA (cDNA) reverse-transcriptase Kit (Applied Biosystems). Quantitative real-time PCR was carried out using SYBR Green PCR Master Mix or KAPA CYBR FAST qPCR Master mix (TaKaRa) and LightCycler 480 or LightCycler 96 (Roche Diagnostics). All measurements were normalised to GAPDH, β-actin or 18S ribosomal RNA.
For gene expression analysis of spheroids, cells were subjected to RNA extraction by RNeasy Plus kit (Qiagen, Hilden, Germany). RNA was retrotranscribed by high-capacity cDNA reverse transcription kit (ThermoFisher Scientific). Gene Hepatology expression was detected by real-time qPCR (Bio-Rad Laboratories), using TaqMan probes (Life Technologies).
The probe-primer sets and the sequences of the oligonucleotides are listed in online supplementary tables 2 and 3.

Measurement of hepatic TG secretion in vivo
Measurement of hepatic TG secretion rate was performed as previously reported 21 with slight modification. Mice were fasted from AM 10:00 for 6 hours and intraperitoneally injected with 200 µL of 1 mg/mL Poloxamer 407 (P-407) (Sigma) solution in sterile PBS. Blood was collected prior to injection (0 hour) and at 1.5 and 3 hours after injection. Serum TG was measured using TG E Test Wako.

PCr cloning
Mouse Lpiat1 were amplified by PCR using cDNA derived from brain of C57BL/6J mice as templates.

Measurement of plasma biomedical markers and lipoprotein profiles
Blood samples were collected with EDTA from the mice fasted for 6 hour. Plasma lipoprotein profiles were analysed by an online dual enzymatic method using high-performance liquid chromatography (Skylight Biotech). Serum ALT and AST levels were quantified using the transaminase CII-test Wako kit (Wako). Serum TG or cholesterol levels were quantified using the TG E Test Wako and Cholesterol E Test Wako (Wako), respectively. Serum glucose level was quantified using Glucose CII-test Wako kit (Wako). Serum insulin level was quantified using mouse insulin ELISA kit (Mercodia).

Microarray
Total RNA was extracted from LKO or F/F mice liver and purified using the RNeasy Mini Kit (Qiagen). Microarray was performed by SurePrint G3 Mouse gene expression 8×60K Ver. 2.0 (Agilent Technologies).

Oxygen consumption
VO 2 was measured with an O 2 /CO 2 metabolism measurement system every 3 min for 24 hours in the animals under the fasting condition (Model MK-5000; Muromachikikai). Each mouse was placed in a sealed chamber (560 mL volume) with an air flow of 0.5 L/min for 24 hours at room temperature. The oxygen consumed was converted to millilitres per minute by multiplying it by the flow rate.

Insulin sensitivity
HFD-fed mice were fasted for 6 hour and injected with or without insulin (Eli Lilly) (5 mU/g body weight) for 10 min. Liver were homogenised and Akt phosphorylations were analysed by western blot.

Cell culture
Huh-7 cells and HepG2 were obtained from JCRB cell bank and ATCC, respectively. LX-2 cells were purchased form Millipore. Huh-7 cells and LX-2 cells were maintained in Dulbecco's modified Eagle's medium (DMEM) supplemented with 10% foetal calf serum and 100 units/mL penicillin 100 µg/mL streptomycin, and 2 mM L-glutamine (PSG). HepG2 cells were maintained in minimum essential medium (MEM) supplemented with 10% FBS, 2 mM L-glutamine, 1 mM sodium pyruvate, 1x non-essential amino acids, 100 units/mL penicillin and 100 µg/ mL streptomycin. These cell lines have been tested for mycoplasma contamination.

Generating lPIAT1 KO Huh-7 cells
Single guide RNA (sgRNA) sequences targeting the region adjacent to the exon 3 of LPIAT1 were designed using the online CRISPR design tool from the Zhang lab at the Broad Institute. The sgRNA sequence, listed in online supplementary table 4, was then cloned into the pSpCas9 (BB)−2A-Puro (PX459) vector (a gift from Feng Zhang, Addgene plasmid # 48139). 35 These vectors were transfected to Huh-7 cells using Lipofectamine 2000 (Invitrogen). Forty-eight hours after transfection, the cells were selected with fresh medium containing 1 µg/mL puromycin.

retrovirus-mediated gene expression and selection
For retroviral gene expression, GFP-tagged LPIAT1 was introduced into a pMXs-Puro Retroviral Vector (Cell Biolabs). These retroviral vectors and VSVG encoding vector were transfected to a transient retrovirus packaging cell line PlatE (provided by T. Kitamura, Institute of Medical Science, University of Tokyo) using Lipofectamine 2000 (Invitrogen). Forty-eight hours after transfection, retrovirus was collected by centrifugation. For infection into Huh-7 cells, the cells were incubated with the retrovirus for 24 hours in the presence of 10 µg/mL polybrene (Sigma-Aldrich). Infected Huh-7 cells were selected with fresh medium containing 1.5 µg/mL puromycin.

bODIPY staining
Cells grown on coverslips were fixed with 4% paraformaldehyde in PBS at room temperature for 15 min, permealised with 0.1% Triton X-100 in PBS at room temperature for 5 min. Lipid droplets were stained with BODYPY 493/503 (1/100 in 3% BSA in PBS) (Invitrogen) for 1 hour. Confocal microscopy was performed using a TCS SP8 (Leica) with a 63 × 1.2 Plan-Apochromat oil immersion lens.

Oil red O staining
HepG2 cells were transiently transfected with Scramble siRNA 5 nM or LPIAT1 siRNA 5 nM. Cells were grown in MEM +10% FBS at 37°C in a humidified atmosphere of 5% CO 2 . After 48 hours, HepG2 cells were harvested and subjected to Oil Red O fluorescent staining (Sigma-Aldrich). Nuclei were staining using DAPI (Sigma-Aldrich). Neutral lipid content was normalised to the number of nuclei, and quantified by ImageJ.

Collagen content determination by immunofluorescence
Spheroids were cultured as described above, and collected after 96 hours. Spheroids were subjected to immunofluorescence for the determination of the collagen content, as previously described. 26 Briefly, Sections were blocked with 4% BSA-PBS for 1 hour and then stained with anti-COL1A1 (Sigma-Aldrich, HPA011795) (1/100) in 4% BSA-PBS for 1 hour, followed by incubation with fluorescent secondary antibody (anti-rabbit IgG Alexa Fluor 594, invitrogen) (1/1000) for 1 hour at room temperature. Nuclei were stained by DAPI (Sigma-Aldrich) (1/8000 in PBS) for 5 min. Pictures were obtained using Axioplan 2 (Zeiss) with AxioVision 4.8 Software (Zeiss). Collagen content was normalised to the number of nuclei, and quantified by ImageJ.

Analysis of TG synthesis
Huh-7 cells were incubated in DMEM supplemented with 10% foetal calf serum and PSG containing 0.4 µCi per well of [ 14 C] oleic acid for 24 hours, followed by lipid extraction and separation by TLC in hexane/diethyl ether/acetic acid (70/30/1, v/v). Radioactivity of TG fraction was measured.
HepG2 cells were incubated with MEM w/o FBS +6 µCi/ mL [ 3 H]glycerol (PerkinElmer) +1.5 mM glycerol (Sigma-Aldrich)+1% BSA (Sigma-Aldrich) for 5 hour. Cells were collected and washed with PBS to remove the excess of medium. Lipids were extracted adding 3 mL of chloroform:methanol (2:1 v/v) (Sigma-Aldrich) and 1 mL of acidified solution (17 mM NaCl, 1 mM H 2 SO 4 ) (Sigma-Aldrich). Samples were centrifuged at 3000 rpm for 10 min. The lower organic phase was saved and dried under a gentle nitrogen stream. Lipids were reconstituted in 50 µL of chloroform (Sigma-Aldrich), and separated by onedimensional TLC using TLC silica gel plates (Merck-Millipore). Triolein (Sigma-Aldrich) was used as a marker. Petroleum ether:diethyl ether:acetic acid (40:60:1, vol/vol) (Sigma-Aldrich) was used as the mobile phase. The area of TLC silica gel corresponding to the newly synthesised radio-labelled TG were cut out. The rate of released radio-labelled TG was measured, using liquid scintillation counting (PerkinElmer). Data were normalised for number of cells.

Analysis of phospholipid synthesis
Huh-7 cells were incubated in DMEM supplemented with 10% foetal calf serum and PSG containing 0.4 µCi per well of [ 14 C] glycerol for 2 hours. For analysis of CDP-DAG synthesis, cells were incubated in inositol depleted DMEM with 2% dialyzed FBS and PSG containing 10 µCi [ 3 H]CTP for 2 hours. Cell lipids were extracted and separated by TLC in chloroform/1-propanol/ ethyl acetate/methanol/0.25% KCl (25/25/25/10/9, v/v) and radioactivity of each phospholipid fraction was measured.

Analysis of phospholipid synthesis in vivo
Eighteen-to-twenty-week-old mice were fasted for 6 hours before intraperitoneal injection of [ 14 C]glycerol for 30 min (0.133 mCi/ kg body weight). Mice were killed by cervical dislocation and liver was immediately frozen. Radioactivity of each phospholipid fraction was measured as mentioned above. β-oxidation assay β-oxidation ability was determined as described previously. 36 After incubation of Huh-7 cells with 1 µCi [ 14 C]palmitic acid for 4 hour, cells were harvested and the cell pellets were suspended with 1 M perchoric acid, vortex-mixed for 5 min and centrifuged at 14 000 g for 10 min at 4°C. Radioactivity of the water soluble fraction was quantified by liquid scintillation counting.

Analysis of lipid degradation speed
HepG2 cells were incubated with MEM w/o FBS +8.3 µCi/ mL [ 3 H]palmitate (PerkinElmer) +2.2 mM PA (Avanti Polar Lipids)+1% BSA for 2 hours. Cell media were collected and palmitate β-oxidation rate was assessed, as previously described by Hansson et al. 37 The rate of oxidised [ 3 H]palmitate was measured using liquid scintillation counting (PerkinElmer). Data were normalised for number of cells.

Fatty acid composition of TG
TG were separated from lipid extracts of Huh-7 cells by TLC in hexane/diethyl ether/acetic acid (70/30/1, v/v). The area corresponding to TG was scraped off the plates and TG was extracted from silica gel. Isolated TG were methylated with 2.5% H 2 SO 4 in methanol. The resulting fatty acid methyl esters were then extracted with hexane and subjected to GC-MS analysis as described previously. 38 GC-MS analysis was performed by using an Agilent 7890A-5975C GC-MS network system (Agilent Technologies, Wilmington, Delaware, USA) equipped with a DB-23 capillary column (60 m × 250 μm × 0.15 μm; Agilent Technologies). The oven temperature programme was as follows; the initial temperature was 50°C for 1 min, then raised to 175°C at 25°C/min, to 235°C at 5 °C/min, and held for 5 min. The injector and detector temperatures were both set at 250°C. extraction of inositol phosphate IP 1 extraction was performed as previously reported 39 40 with brief modification. Cell pellets were suspended with 3.2 M acetic acid in methanol and vortex-mixed for 30 min at room temperature and centrifuged at 5000 g for 20 min at 4°C. Supernatant was dried up with centrifugal evaporator and resuspended in 100 µL 5% aqueous methanol. 10 µL were used to determine the content of inositol phosphate. Mice liver were homogenised in extraction buffer (1 M HClO 4 , 3 mM EDTA) using handy micro homogeniser (Physcotron NS-310E). After centrifugation at 15 000 rpm for 5 min at 4°C, the supernatants neutralised with neutralisation buffer (1 M K 2 CO 3 , 3 mM EDTA) and incubated on ice for 2 hours. After centrifugation at 15 000 rpm for 10 min at 4°C, supernatants were collected and 5 µL were used to determine the content of inositol phosphate.

Autophagy flux assay
For induction of autophagy, the culture medium was changed to glucose-free and amino acid-free EBSS for 4 hours with or without 150 nM Bafilomycin A1 (Sigma). LC3 protein was detected by western blotting.

lipidomic and gene expression analysis in the human liver
The cohort from Finland where the genetic association between the MBOAT7 rs641738 and lipidomic data were examined has been previously described. 27 Lipidomic data were obtained from 125 human liver biopsy using a quadrupole time-of-flight (Q-TOF) Premier mass spectrometer combined with an Acquity Ultra Performance LC (Waters, Milford, Massachusetts, USA) as previously described. 27 Briefly, immediately at the beginning of Hepatology the laparoscopic bariatric surgery, wedge biopsies of the liver were obtained from obese adults. Part of the biopsy was snapfrozen in liquid nitrogen for subsequent analysis of molecular lipids. MBOAT7 rs641738 genotype was performed by TaqMan PCR method (Applied Biosystems, Foster City, California, USA) according to the manufacturer's instructions followed by post-PCR allele-specific fluorescence measurements by an ABI Prism Sequence Detection System (ABI 7900HT, Applied Biosystems). In the present study, only results on liver phospholipids composition from a subset of 115/125 individuals with complete required data were included. Each participant provided written informed consent after being explained the nature and potential risks of the study.
The cohort from Italy where the genetic association between the MBOAT7 rs641738 and liver mRNA expression was examined has been previously described. 28 Gene expression data were obtained from human liver biopsy of a subset of adult obese individuals as previously described. 25 41 Briefly, RNA sequencing was performed using the Illumina HiSeq 4000 platform (Novogene, Hong Kong, China). RNA reads were mapped against the human genome, and the gene read count (Ensembl human transcript reference assembly, V.75) was determined using RSEM software. To quantify gene expression, the RSEM per gene count data were normalised using the DESeq2 package. MBOAT7 rs641738 genotype was performed by TaqMan PCR method (Applied Biosystems, Foster City, California, USA) as previously described. 8

statistical analysis
For in vivo and in vitro studies, data were converted to means±SEM., and the unpaired Student's t-test or nonparametric Mann-Whitney test were applied to determine significant differences between two samples. Statistical differences between multiple treatment groups and a control group were determined using analysis of variance and the Tukey-Kramer post hoc test. Sample sizes were chosen on the basis of previous experience in our laboratory. The experiments were performed and analysed in a non-randomised and non-blinded fashion. No data were excluded from the analysis. Variance was similar between the groups that were statistically compared.
Data from lipidomic composition and gene expression in human liver tissues stratified by MBOAT7 rs641738 genotype were shown as mean and SD. P values were calculated by linear regression analysis under an additive genetic model adjusted for age, gender and body mass index. Values were log transformed before entering the model. Statistical analyses were performed using IBM Statistical Package for Social Sciences (IBM SPSS, V.19.0).

Data availability
The data supporting the findings of this study are available from the corresponding author on reasonable request. The microarray data can be accessed at the Gene Expression Omnibus repository (GSE141920). Primer sets for real-time PCR, probes for qPCR by the TaqMan system and sgRNA target sequences are available in online supplementary tables 2-4, respectively.