The PNPLA3‐I148M variant increases polyunsaturated triglycerides in human adipose tissue

The I148M variant in PNPLA3 is the major genetic risk factor for non‐alcoholic fatty liver disease (NAFLD). The liver is enriched with polyunsaturated triglycerides (PUFA‐TGs) in PNPLA3‐I148M carriers. Gene expression data indicate that PNPLA3 is liver‐specific in humans, but whether it functions in adipose tissue (AT) is unknown. We investigated whether PNPLA3‐I148M modifies AT metabolism in human NAFLD.


| INTRODUC TI ON
A common non-synonymous single nucleotide polymorphism (rs738409; c.444C>G, p.I148M) in the patatin like phospholipase domain containing 3 (PNPLA3, adiponutrin) gene was found in the Dallas Heart Study to significantly increase liver fat content in three different ethnic groups. 1 This finding has since been extensively replicated. 2 The I148M allele is found in 30%-50% of all subjects 3,4 and increases the risk of both alcoholic and non-alcoholic fatty liver disease (NAFLD), including cirrhosis and hepatocellular carcinoma. 5 In contrast to NAFLD associated with insulin resistance and metabolic syndrome, in which the steatotic liver mainly consists of saturated fat, the human liver lipidome is characterized by absolute and relative increases in polyunsaturated triglycerides (TGs) in PNPLA3-I148M variant carriers compared with non-carriers. 6 The I148M variant increases polyunsaturated fatty acid (PUFA) retention in liver TGs and decreases incorporation of PUFAs into phospholipids. 7 These data closely resemble those of knock-in mice expressing a catalytically inactive form of PNPLA3 in the liver (PNPLA3-S47A). 8 Non-esterified fatty acids (NEFAs) resulting from adipose tissue (AT) lipolysis are the main source of intrahepatic triglycerides (IHTGs) in NAFLD. 9 There are no data on whether PNPLA3-I148M exerts changes in the lipid composition of AT, as it does in the liver. 6,7 Moreover, the potential impact of the I148M variant on AT lipolysis or the composition of NEFAs released from AT has not been studied.
Of interest, PNPLA3 (previously known as adiponutrin) was initially discovered in mice as a nutritionally regulated transmembrane protein thought to be specific to the adipocyte lineage. 10,11 In humans, the PNPLA3 transcript is, in contrast to findings in mice and rats, [12][13][14] much more abundant in the liver than in AT. 15,16 Concentrations of the PNPLA3 protein in the human liver or AT have not, however, been previously studied. This would be important as efforts are currently ongoing to find therapeutic targets for the treatment of advanced NAFLD in genetically predisposed patients. [17][18][19] In the present study, we investigated whether the human AT lipidome is modified in a polyunsaturated direction in carriers of PNPLA3-I148M compared with non-carriers, as it is in the liver.
Since this was found to be the case, we next examined whether the variant affects AT lipolysis or the composition of circulating NEFAs. In addition, we compared PNPLA3 mRNA and protein levels between human liver and subcutaneous AT in a subset of the volunteers.

| Effects of PNPLA3-I148M on AT TG and serum NEFA composition
We profiled the AT lipidome and fasting serum NEFA composition in 125 consecutively recruited patients undergoing laparoscopic and PNPLA3 mRNA and protein levels were measured in subcutaneous AT and liver biopsies in a subset of the volunteers.
Results: PUFA-TGs were significantly increased in AT in carriers versus non-carriers of PNPLA3-I148M. The variant did not alter the rate of lipolysis or the composition of fasting serum NEFAs. PNPLA3 mRNA was 33-fold higher in the liver than in AT (P < .0001). In contrast, PNPLA3 protein levels per tissue protein were three-fold higher in AT than the liver (P < .0001) and nine-fold higher when related to wholebody AT and liver tissue masses (P < .0001).
Conclusions: Contrary to previous assumptions, PNPLA3 is highly abundant in AT.
PNPLA3-I148M locally remodels AT TGs to become polyunsaturated as it does in the liver, without affecting lipolysis or composition of serum NEFAs. Changes in AT metabolism do not contribute to NAFLD in PNPLA3-I148M carriers.

K E Y W O R D S
adipose tissue, fatty acids, lipidomics, lipolysis, non-alcoholic fatty liver disease, triglycerides

Key points
• The common I148M variant in the gene PNPLA3 is the main genetic risk factor for fatty liver disease, but whether the variant protein exists or alters lipid metabolism in human adipose tissue is unknown.
• We found that the PNPLA3 protein is found at high concentrations in human adipose tissue and that carriers of the PNPLA3-I148M variant have changes in their adipose tissue lipid composition that mirror those seen in the liver. bariatric surgery who fulfilled the following inclusion criteria: (a) age 18-75 years; (b) no known acute or chronic disease except for obesity, type 2 diabetes, NAFLD or hypertension on the basis of history, physical examination, electrocardiogram and standard laboratory tests (complete blood count, serum creatinine and electrolyte concentrations); (c) alcohol consumption <20 g/d for women and <30 g/d for men; (d) no clinical or biochemical evidence of liver disease other than NAFLD (such as hepatitis B or C), or clinical signs or symptoms of inborn errors of metabolism; (e) no history of use of drugs or toxins influencing liver steatosis and (f) not pregnant or lactating. We have previously reported data on the liver lipidome in a cohort that mostly consisted of the same volunteers. 6,7 The present cohort differs slightly from that pub-

| Effects of PNPLA3-I148M on AT fatty acid composition and inflammation
In addition to the AT lipidome profiling described above, we ex- The inclusion criteria were as listed above. The volunteers participated in a clinical research visit during which needle biopsies of abdominal AT were also obtained (vide infra). In addition, on a separate visit, liver IHTG content was measured by proton magnetic resonance spectroscopy ( 1 H-MRS).

| Effects of PNPLA3-I148M on in vivo AT lipolysis
We recruited 28 non-diabetic volunteers by contacting participants of prior metabolic studies who were known to be homozygous (PNPLA3 148II , n = 19; PNPLA3 148MM , n = 9) based on previous genotyping results. The inclusion criteria were as listed above. These volunteers participated in a clinical research visit as well as in a metabolic study during which whole-body lipolysis was measured using

| Adipose tissue and liver biopsies
Immediately at the beginning of the laparoscopic bariatric surgery procedure, a wedge biopsy of the liver was taken in addition to a subcutaneous abdominal AT biopsy. The AT sample and approximately one-half of the liver sample were immediately snap frozen in liquid nitrogen and stored at −80°C until subsequent analysis of molecular lipids. The time from obtaining the biopsies until freezing of the samples in liquid nitrogen was approximately 1 minute. The remainder of the liver biopsy was sent to the pathologist for routine histopathological assessment using the criteria proposed by Brunt et al. 22 For the non-surgical volunteers, needle aspiration biopsiesof subcutaneous abdominal AT were taken under local anaesthesia with 1% lidocaine at the clinical research visit as previously described. 23

| Lipidomic analysis
The AT lipidome was analyzed using an ultra-high performance liquid chromatography-quadrupole time-of-flight mass spectrometry system (UHPLC-QTOF-MS; Agilent Technologies). In addition to TGs, the analysis covered most of the major molecular lipids including ceramides, sphingomyelins, phosphatidylcholines, phosphatidylethanolamines and lysophosphatidylcholines. For detailed methodology, see Supporting Information.

| Composition of AT FAs
The analysis of AT FA composition was performed using gas chromatography (GC). AT lipids were extracted according to the method of Folch et al. 24 The TG fraction was separated by solid-phase extraction 25 and fatty acid methyl esters (FAMEs) prepared and analyzed by GC. 26 FAs were identified using a standard containing FAMEs ranging from chain length 6 to 24 (Sigma-Aldrich). A FAME standard of known composition (AOCS std#6, Thames Restek) and a quality control sample (mixture of fatty acids [Sigma-Aldrich] and TGTG [MaxEPA fish oil, Seven Seas]) were run alongside each batch of samples to check correct peak identification and instrument performance. GC results were converted into mol%.

| Composition of fasting serum NEFAs
The analysis of NEFAs was done using comprehensive two-dimensional gas chromatography time-of-flight mass spectrometry (GC×GC/TOFMS; Pegasus 4D, LECO Corporation), as described previously in detail 27 and outlined in Supporting Information.

| Insulin sensitivity of whole-body AT lipolysis
The rate of whole-body lipolysis was measured basally after an overnight fast and during intravenously maintained euglycaemic hyperinsulinaemia by infusing [ 2 H 5 ]glycerol as previously described. 28 The basal and insulin infusion periods both lasted 120 minutes, and the rate of the continuous insulin infusion was 0.4 mU•kg −1 •min −1 . The low insulin infusion rate was chosen to maximize the likelihood of detecting changes in lipolysis. 29

| Measurement of IHTG content
In the 28 volunteers in whom in vivo lipolysis was measured, and in the 50 volunteers from whom a needle biopsy of AT was obtained, IHTG content was measured by 1 H-MRS, as described. 30 To facilitate comparison between spectroscopic and histological IHTG measurements, spectroscopic fat percentages were converted to correspond to those obtained by liver biopsy using an equation we have previously published. 31

| Messenger RNA expression
Real-time quantitative polymerase chain reaction (RT-qPCR) was performed on reverse-transcribed mRNA isolated from liver and AT samples, as described in Supporting Information.

| Statistics
Analyses were performed with Statistical Package for the Social Sciences (SPSS) version 25 (IBM Corporation) and GraphPad Prism version 7.04 (GraphPad Software). The Shapiro-Wilk test was used to assess continuous variables for normality. We compared two independent groups using the unpaired Student's t test or the Mann-Whitney U test for normally and non-normally distributed variables respectively. We used the Pearson's χ 2 test or the Fisher's exact test as appropriate to evaluate if distribution of categorical variables differed between two groups. To compare gene and protein expression in AT and liver biopsies from the same volunteers, we used the paired t test. ΔC t values were used in statistical analyses of the RT-qPCR data. For statistical analysis of AT lipidomic and serum NEFA composition data, missing values were imputed using half mean plus a very small amount of random noise. Lipid species with missing values in more than 50% of samples were excluded from analyses. Lipidomic data were log 2 -transformed before statistical hypothesis testing, and the Benjamini-Hochberg procedure 33 was applied to control false discovery rate (FDR) at a preselected level of Q = 20%. We report unadjusted P values for findings that are determined as discoveries.
Otherwise, a P < .05 was considered statistically significant.
We have previously shown highly significant differences in lipidomic profiles of the liver between PNPLA3-I148M carriers (PNPLA3 148MM/MI ) and non-carriers (PNPLA3 148II ) in a sample of 125 volunteers. 6 This justifies the similar sample size used for the AT and serum analyses, and the comparison of AT FA composition between 25 homozygous carriers and 25 non-carriers. Regarding the lipolysis study, interindividual variability in insulin suppression of glycerol rate of appearance (R a ) was determined based on data we have previously acquired in obese volunteers. 34 Based on these data, we calculated that 9 homozygous carriers and 19 non-carriers are needed to detect a 14% between-group difference in insulin suppression of glycerolR a using a 2-sided t test with a β value of 0.80 and an α value of 0.05. Power calculations were performed using G*Power 3.1.9.6. 35

| The AT lipidome is enriched with polyunsaturated TGs in PNPLA3-I148M variant carriers
Clinical characteristics of the 125 volunteers in whom lipidomic analyses of AT were conducted are shown in Table 1  when excluding volunteers with type 2 diabetes from analyses ( Figure S1). Previous lipidomic analysis of the liver in mostly the same volunteers showed similar PUFA enrichment in liver TGs of PNPLA3-I148M carriers. 6 We did not observe changes in concentrations of ceramides, sphingomyelins, lysophosphatidylcholines, phosphatidylcholines or phosphatidylethanolamines between the groups (Table S3).
We conducted a further analysis of the composition of medium-to very long-chain FAs in AT samples of homozygous volunteers (PNPLA3 148II , n = 25; PNPLA3 148MM , n = 25). The groups were similar with respect to age, sex, BMI and metabolic parameters (Table S4). As a whole, there were no significant changes in saturated or monounsaturated FAs between the groups. We and ADIPOQ (1.00 ± 0.10 vs 1.57 ± 0.38 AU, P = .04) were significantly increased in AT of PNPLA3 148MM as compared to PNPLA3 148II volunteers.

| In vivo AT lipolysis or fasting serum NEFA composition is not affected in PNPLA3-I148M variant carriers
Clinical characteristics of the 28 volunteers in whom whole-body lipolysis was measured are shown in Table 1. The 9 homozygous carriers  3.02 ± 0.14 μmol•kg −1 •min −1 in the PNPLA3 148II group, with no significant difference between the groups (P > .05, Figure 2A).
We profiled the composition of fasting serum NEFAs in the same 125 volunteers in whom lipidomic studies of AT were conducted, as described above. After correcting for multiple testing, we found  (Table S5).

| Expression of PNPLA3 mRNA is markedly higher in the liver compared to AT, but the PNPLA3 protein is more abundant in AT
The change we observed in the AT lipidome in carriers of PNPLA3-I148M was unexpected as mRNA expression has been shown to be very low in human AT as compared to the liver. 15,16 There are, however, no protein data available. Therefore, we investigated PNPLA3 protein levels in tissue samples of AT and the liver in a subset of 20 volunteers (mean age 46.0 ± 1.9 years, mean BMI 45.6 ± 1.4 kg/m 2 ).
Quantitative PCR analysis showed that PNPLA3 mRNA expression was markedly higher in the liver compared to AT. Normalized to the mRNA levels of the reference genes 36B4 and ACTB, expression of PNPLA3 mRNA was on average 33-fold higher in the liver than in AT (P < .0001; Figure 3A).
PNPLA3 antibody specificity was confirmed by immunoblotting mock-transfected HuH7 cell lysates and cells transfected with hu-manPNPLA3 ( Figure S4). Immunoblotting ( Figure 3B) revealed that the level of PNPLA3 protein was three-fold higher in AT than the liver (P < .0001; Figure 3C), and two-fold higher when normalized to β-actin levels (P < .0001). Total protein concentration was eight-fold higher in the liver samples than the AT samples (P < .0001). Thus, per milligram of tissue, the concentration of PNPLA3 was three-fold higher in the liver than in AT (P < .0001) ( Figure 3D). We estimated whole-body levels of PNPLA3 in AT and the liver by multiplying the concentration of PNPLA3 per milligram of tissue by the estimated organ weight. Average liver mass was 2.3 ± 0.2 kg, and average AT mass was 54.4 ± 2.9 kg. Assuming homogenous levels of PNPLA3 in the liver and in AT depots, whole-body levels of PNPLA3 were ninefold higher in AT than the liver (P < .0001) ( Figure 3E).

| D ISCUSS I ON
The present series of studies were undertaken to investigate whether the PNPLA3-I148M variant changes AT TG composition, as it does in the liver. Since this was found to be the case, we next de- Thus, the excess of polyunsaturated TGs in AT cannot be secondary to their transfer from the liver to AT in VLDL.
Adipose tissue is chronically inflamed in obese subjects, which may contribute to insulin resistance and the development of NAFLD. [36][37][38] We have previously shown that AT inflammation is absent in PNPLA3-I148M carriers with NAFLD compared with non-carriers and suggested that this may contribute to the lack of insulin resistance in carriers of PNPLA3-I148M. 39 In the present study, the omega-6-to omega-3-PUFA ratio was lower in AT of carriers versus non-carriers of PNPLA3-I148M, reflecting lower concentrations of omega-6 AA and higher concentrations of omega-3 DPA ( Figure S2). Expression of anti-inflammatory genes was increased and pro-inflammatory genes unchanged in variant carriers compared with non-carriers. These changes are anti-rather than pro-inflammatory. Arachidonic acid is a precursor of eicosanoids that mediate the production of pro-inflammatory cytokines, 40 while DPA is synthesized from a precursor of anti-inflammatory eicosanoids. 40 An increased omega-6 to omega-3 ratio is associated with pro-inflammatory states and impaired function of metabolically active tissues such as the liver and AT. 40 The present data thus suggest that carriers of the PNPLA3-I148M variant possess metabolically healthy, PUFA-enriched AT that does not harbour pro-inflammatory properties. Interestingly, AT enriched with PUFA was shown in the Scottish Heart Health Extended Cohort study to decrease the risk of future cardiovascular events, independent of other known risk factors. 41 The PNPLA3-I148M variant has also been associated with protection against cardiovascular sequelae in NAFLD. [42][43][44][45] The composition of subcutaneous AT is affected by long-term changes in dietary FA intake. 46  and PNPLA3 148MM/MI groups (Table S5). Overall, these data imply that the increase in polyunsaturated IHTGs in I148M variant carriers is not as a result of increased NEFA delivery from AT to the liver.
Because we unexpectedly found the AT lipid composition to differ between carriers and non-carriers of the PNPLA3-I148M variant, we compared gene and protein expression of PNPLA3 between liver and AT samples in a small subset of the volunteers. PNPLA3 mRNA was markedly higher in the human liver than in subcutaneous AT (Figure 3), consistent with previous studies. 15,16 This is in stark contrast to mice in which PNPLA3 mRNA expression is unequivocally highest in AT depots and only small amounts of mRNA can be detected in other tissues, including the liver. [12][13][14] These contradictory results between human and mouse studies are yet to be explained but may reflect physiological differences between species. Despite higher gene expression in the liver, the PNPLA3 protein was much more abundant in AT than the liver (Figure 3). Importantly, we extrapolated AT protein expression from one subcutaneous AT biopsy to the whole body, which assumes homogenous expression in all compartments of these tissues. This finding challenges the previous liver-centric view of the protein and its function in humans and raises the question as to whether polymorphisms in PNPLA3 could introduce significant alterations in human lipid metabolism via extrahepatic pathways.
We conclude that the PNPLA3 protein is found not only in the human liver but also highly abundantly in AT. This is contrary to previous assumptions, according to which PNPLA3 is a liver-specific protein in humans. The PNPLA3-I148M variant alters AT lipid composition in a similar fashion as in the liver, 6 that is, the lipidome is significantly enriched with polyunsaturated TGs. This change in AT lipid composition cannot explain the higher polyunsaturated IHTG content in PNPLA3-I148M carriers, since the variant does not affect the rate of AT lipolysis or the composition of NEFAs released from AT. We propose that the PNPLA3-I148M variant remodels TG composition in both the liver and AT independently, with the enrichment of PUFAs. This human knowledge is relevant as efforts are currently ongoing to develop novel pharmaceuticals to treat NAFLD caused by PNPLA3-I148M. [17][18][19] On the basis of our findings, we suggest that therapies aimed at ameliorating NAFLD due to PNPLA3-I148M should be liver-specific.

ACK N OWLED G EM ENTS
We gratefully acknowledge volunteers for their help. We thank Aila Karioja-Kallio, Päivi Ihamuotila and Mia Urjansson for their excellent technical assistance.

LH is a British Heart Foundation Senior Research Fellow in Basic
Science.