Phosphatidylethanolamines Are Associated with Nonalcoholic Fatty Liver Disease (NAFLD) in Obese Adults and Induce Liver Cell Metabolic Perturbations and Hepatic Stellate Cell Activation

Pathogenesis roles of phospholipids (PLs) in nonalcoholic fatty liver disease (NAFLD) remain incompletely understood. This study investigated the role of PLs in the progression of NAFLD among obese individuals via studying the alterations in serum PL composition throughout the spectrum of disease progression and evaluating the effects of specific phosphatidylethanolamines (PEs) on FLD development in vitro. A total of 203 obese subjects, who were undergoing bariatric surgery, were included in this study. They were histologically classified into 80 controls (C) with normal liver histology, 93 patients with simple hepatic steatosis (SS), 16 with borderline nonalcoholic steatohepatitis (B-NASH) and 14 with progressive NASH (NASH). Serum PLs were profiled by automated electrospray ionization tandem mass spectrometry (ESI-MS/MS). HepG2 (hepatoma cells) and LX2 (immortalized hepatic stellate cells or HSCs) were used to explore the roles of PL in NAFLD/NASH development. Several PLs and their relative ratios were significantly associated with NAFLD progression, especially those involving PE. Incubation of HepG2 cells with two phosphatidylethanolamines (PEs), PE (34:1) and PE (36:2), resulted in significant inhibition of cell proliferation, reduction of mitochondrial mass and membrane potential, induction of lipid accumulation and mitochondrial ROS production. Meanwhile, treatment of LX2 cells with both PEs markedly increased cell activation and migration. These effects were associated with a significant change in the expression levels of genes involved in lipogenesis, lipid oxidation, autophagy, apoptosis, inflammation, and fibrosis. Thus, our study demonstrated that elevated level of PEs increases susceptibility to the disease progression of obesity associated NAFLD, likely through a causal cascade of impacts on the function of different liver cells.


Introduction
Non-alcoholic fatty liver disease (NAFLD) is the most common chronic liver disease with a global prevalence of about 25% [1]. NAFLD comprises a spectrum of diseases

Elevation of Serum PLs in SS, B-NASH, and NASH
A total of 203 species of PLs and sphingomyelin (SM) were identified and quantified in serum by using ESI-MS/MS. In our study, we investigated a relation between the levels of several PL species and NAFLD. A significant increase was observed in PE and cholesterol ester (CE) species including PE (34:1), (36:2), (34:2) and C16:1 CE in NASH, B-NASH and SS compared to control subjects ( Figure 1A-D). The ratio of lysophosphatidylcholine (LPC)/PEs and LPC/C16:1 CE showed a dramatic reduction between NAFLD and control groups ( Figure 1E-I).

Elevation of Serum PLs in SS, B-NASH, and NASH
A total of 203 species of PLs and sphingomyelin (SM) were identified and quantified in serum by using ESI-MS/MS. In our study, we investigated a relation between the levels of several PL species and NAFLD. A significant increase was observed in PE and cholesterol ester (CE) species including PE (34:1), (36:2), (34:2) and C16:1 CE in NASH, B-NASH and SS compared to control subjects ( Figure 1A-D). The ratio of lysophosphatidylcholine (LPC)/PEs and LPC/C16:1 CE showed a dramatic reduction between NAFLD and control groups ( Figure 1E-I).

PEs Increased Lipid Droplet Accumulation in HepG2 cells
The intracellular content of neutral lipids such as cholesterol esters and trigl that are packaged in lipid droplets (LDs) was estimated by fluorescence intensity orescence microscopy after BODIPY staining of the cytoplasmic LDs.

PEs Increased Lipid Droplet Accumulation in HepG2 Cells
The intracellular content of neutral lipids such as cholesterol esters and triglycerides that are packaged in lipid droplets (LDs) was estimated by fluorescence intensity and fluorescence microscopy after BODIPY staining of the cytoplasmic LDs. Treatment of HepG2 cells with 1 mM PE for 48 h significantly increased the staining of LDs compared to the control as observed by fluorescence microscopy. A highly significant difference in fluorescence intensity (p < 0.001) was detected in both PE (34:1) and PE (36:2) treated cell cultures as compared to the control group ( Figure 3).  represent the means ± SDs of three independent experiments *** p < 0.001 vs. control.

PEs Increased Mitochondrial Dysfunction in HepG2 cells
To ascertain whether there are alterations in mitochondrial functions in respon PE treatment in HepG2 cells, we evaluated mitochondrial mass, membrane potenti well as mitochondrial reactive oxygen species (ROS) production. Mitochondrial mass membrane potential were measured by using MitoTracker green and MitoTracker respectively. A significant reduction in both mitochondrial mass and membrane pote was observed under the fluorescence microscope after treatment with 1 mM PE for ( Figure 4A). A reduction in fluorescence intensity of mitochondrial mass was detecte PE (34:1) and PE (36:2) treated cell cultures (p ˂ 0.001) ( Figure 4B). Additionally, mitoc drial membrane potential was also significantly reduced in HepG2 cells treated wit (34:1) and PE (36:2) as measured by fluorescence intensity (p ˂ 0.001) ( Figure 4C). M

PEs Increased Mitochondrial Dysfunction in HepG2 Cells
To ascertain whether there are alterations in mitochondrial functions in response to PE treatment in HepG2 cells, we evaluated mitochondrial mass, membrane potential as well as mitochondrial reactive oxygen species (ROS) production. Mitochondrial mass and membrane potential were measured by using MitoTracker green and MitoTracker red, respectively. A significant reduction in both mitochondrial mass and membrane potential was observed under the fluorescence microscope after treatment with 1 mM PE for 48 h ( Figure 4A). A reduction in fluorescence intensity of mitochondrial mass was detected in PE (34:1) and PE (36:2) treated cell cultures (p < 0.001) ( Figure 4B). Additionally, mitochondrial membrane potential was also significantly reduced in HepG2 cells treated with PE (34:1) and PE (36:2) as measured by fluorescence intensity (p < 0.001) ( Figure 4C). Mitochondrial reactive oxygen was measured by MitoSox red as the mitochondrial superoxide indicator. Mitochondria in both PEs-treated cells produced more superoxide than control cells as evidenced by more than two times higher fluorescence intensity compared to the control [

PEs Altered the Expression of Mitophagy and Apoptosis/Cell Proliferation-Related Genes in HepG2 Cells
To determine whether these PEs have an impact on mitophagy and apoptosis pathways, we examined the mRNA levels of genes involved in both processes after treatment of  2). However, the expression of carbohydrate response element binding protein (ChREBP) was not significantly altered in response to PE treatment compared to the control group ( Figure 6).

PEs Induced LX2 Cell Migration
The hepatic stellate cell (HSC) is the key cell type responsible for fibrosis in acute and chronic liver injuries. To examine the activation of LX2 by PE, we performed the cell migration assay using transwell by incubating cells with 1 mM PE for 24 h. The assay showed dramatic increments in the number of stained migrated cells after treatment with PE (34:1) and PE (36:2) ( Figure 7A). The data showed a significant~2-fold elevation in absorbance after incubation with PE (34:1) and PE (36:2) (p < 0.001 and p < 0.01, respectively) ( Figure 7B).  2). However, the expression of calcium homeostasis endoplasmic reticulum protein (ChREBP) was not significantly altered in response to PE treatment compared to the control group ( Figure  6).

PEs Induced LX2 Cell Migration
The hepatic stellate cell (HSC) is the key cell type responsible for fibrosis in acute and chronic liver injuries. To examine the activation of LX2 by PE, we performed the cell migration assay using transwell by incubating cells with 1 mM PE for 24 h. The assay showed dramatic increments in the number of stained migrated cells after treatment with PE (34:1) and PE (36:2) ( Figure 7A). The data showed a significant ~2-fold elevation in absorbance after incubation with PE (34:1) and PE (36:2) (p < 0.001 and p ˂ 0.01, respectively) ( Figure  7B).  Results are depicted as relative change of control at OD595. Bars represent the means ± SDs of three independent experiments. ** p < 0.01, and *** p < 0.001 vs. control.

PE Treatments Increased mRNA Expression of Inflammation and Fibrosis-Related Genes in LX2 Cells
The activation of LX2 detected in the present study by the cell migration assay after incubation with PE indicated the induction of inflammation and fibrosis. To validate these actions, we examined the effect of PE on the expression of genes involved in both processes. The results showed overexpression of fibrotic marker genes including α-smooth muscle actin (α-SMA) [~2.9-fold, p < 0.001 and ~2.4-fold, p < 0.001 for PE (34:1) and PE   (TNFα and IL6). Results are depicted as relative change of control. Data represent the mean ± SD of three independent experiments. * p < 0.05, ** p < 0.01, and *** p < 0.001 vs. control. α-SMA, alpha-smooth muscle actin; COL1A1, collagen type I alpha 1, COL3A1, collagen type III alpha 1; TIMP1, TIMP metallopeptidase inhibitor 1; TIMP3, TIMP metallopeptidase inhibitor 3; TGFꞵ, transforming growth factor beta-1; TNFα, tumor necrosis factor-alpha; and IL6, interleukin-6.

Discussion
Phospholipids (PLs) play fundamental roles in numerous biochemical reactions and different metabolic pathways. PL composition alterations in both blood and liver tissue of NAFLD patients as compared to healthy individuals have been observed previously [15,29]. However, it remains less investigated how the blood PL composition might be associated with NAFLD progression among obese individuals. More importantly, the actual role of PLs, especially PEs, in NAFLD as to whether they are considered as a causal Relative mRNA expression Results are depicted as relative change of control. Data represent the mean ± SD of three independent experiments. * p < 0.05, ** p < 0.01, and *** p < 0.001 vs. control. α-SMA, alpha-smooth muscle actin; COL1A1, collagen type I alpha 1, COL3A1, collagen type III alpha 1; TIMP1, TIMP metallopeptidase inhibitor 1; TIMP3, TIMP metallopeptidase inhibitor 3; TGFβ, transforming growth factor beta-1; TNFα, tumor necrosis factor-alpha; and IL6, interleukin-6.

Discussion
Phospholipids (PLs) play fundamental roles in numerous biochemical reactions and different metabolic pathways. PL composition alterations in both blood and liver tissue of NAFLD patients as compared to healthy individuals have been observed previously [15,29]. However, it remains less investigated how the blood PL composition might be associated with NAFLD progression among obese individuals. More importantly, the actual role of PLs, especially PEs, in NAFLD as to whether they are considered as a causal factor, or just a consequence of the disease remains incompletely studied. In the current study, we investigated the role of PLs in obesity-associated NAFLD progression in vivo and explored the mechanism underlying this role in vitro.
Based on our results, a remarkable change was observed in serum PLs that varied significantly between SS, B-NASH and NASH subjects in obese individuals. The most dramatically changed PL species and their ratios observed in our study were CE, PE, LPC/PE, and LPC/CE. As demonstrated in Figure 1, increased levels of three specific PEs and CE while decreased ratios of LPC/PE were associated with the disease progression. Our data showed a distinguishable increase in PE species containing saturated fatty acid (SFA) and monounsaturated fatty acid (MUFA) in NAFLD subjects compared to control. The elevation of MUFA/SFA containing PLs may be due to increased insulin secretion as well as high dietary intake in obese subjects [30] which consequently activates de novo lipogenesis and accumulation of MUFA/SFA PLs fraction in blood [25,31].
Lysophosphatidylcholines (LPCs) are lipids derived from PCs mediated by phospholipase A2 (PLA2) which cleaves PC to produce LPC and free fatty acid, and/or by lecithin-cholesterol acyltransferase (LCAT) which transfers fatty acids from PC to free cholesterol to produce LPC and CE. LPC can also be converted back to PC mediated by lysophosphatidylcholine acyltransferase (LPCAT) in the presence of acyl-CoA [32]. Therefore, our findings may suggest an imbalanced activity of these enzymes in obesityassociated NAFLD development. Indeed, altered function of both LCAT and LPCAT3 have been found to be associated with NAFLD [33,34]. Unlike previous studies, we did not observe a significant association between PC/PE ratio and NAFLD disease severity in our cohort, which may be related to the obese background of our cohort, since PC/PE ratio seemed to be more likely associated with NAFLD in lean individuals [20,21], though animal studies might not support this notion [35]. Further studies are needed to validate and clarify the role of PC, PE, and LPC in NAFLD in both lean and obese individuals.
Our data also suggest that the LPC/CE and LPC/PE ratios that are significantly associated with NAFLD disease progression in our samples may be primarily driven by the increased CE and PE levels, given that CE or PE alone was associated with NAFLD, while LPC was not. This suggests that accumulated CE and PE are likely more causal to the disease progression. Compared to what have been broadly studied on the role of cholesterol homeostasis in NAFLD and NASH, the causal role of PE in the development and progression of NASH remains largely elusive. We thus set out to investigate the impact of PE on underlying features of NASH histology, i.e., steatosis, cell proliferation, and hepatic stellate cell (HSC) activation in vitro.
Since HepG2 and LX2 have been widely used to model NAFLD in vitro [36], we utilized these two cell models in our study to elucidate the effect of PE on the development of the disease. In the current study, HepG2 cells were used to study lipid accumulation, apoptosis, and mitochondrial dysfunction, while LX2 cells were used to assess inflammation and fibrosis [37]. PEs containing SFA/MUFA side chains with 16 and 18 carbons are the most abundant lipid molecular species in the human liver [27]. Moreover, previous investigators showed that PE species play a basic role in the pathology of several diseases [38]. Oleic acid (OA) and palmitic acid (PA) are the most abundant MUFA and SFA in human tissues and are commonly used in vitro and in vivo to stimulate steatosis [28,39,40]. Based on our lipidomic data, the PEs that mostly showed significant changes in NASH were PE species containing SFA and MUFA. Accordingly, we performed the in vitro study with PE (34:1) and (36:2) that contain OA (C18 MUFA) and PA (C16 SFA) moieties as examples. Treating HepG2 cells with these two representative PEs significantly decreased cell viability but increased intracellular neutrolipid accumulation coupled with increased impairments in mitochondrial function. The mitochondrial dysfunction was characterized by reduced mi-tochondrial mass and their membrane potential. This dysfunction was also associated with increased ROS production. Previous studies using HepG2 and other cells also consistently demonstrated that PE induces cell apoptosis and mitochondrial dysfunction. However, these studies did not specify the acyl composition of PEs used to treat the cells [41,42]. Our study confirmed these previous findings with two specific PEs, and further demonstrated that HepG2 cells treated with these two PEs also developed increased neutrolipid accumulation and ROS production. In correspondence with these results, our data showed significantly altered expression of genes involved in mitophagy (LC3, and mTOR), apoptosis/cell proliferation (BAX, BCL2, PAK2, CYCS, and ChREBP), beta-oxidation (CPT and PPARα) and lipogenesis (DGAT, FASN, SREBP and FXR). Again, these gene expression patterns also confirmed the previously observed impairment in apoptosis/cell proliferation associated with PE treatment [41,42], while the increased expression of lipogenesis genes following PE treatments may combine with the impaired mitochondrial function and together lead to neutrolipid accumulation and ROS production. These cellular impacts of the two PEs well match the pathogenic features, i.e., cell death, oxidative stress, lipid accumulation, and mitochondrial dysfunction that are underlying NAFLD [43].
Hepatic stellate cells (HSC) activation plays a central role in hepatic fibrosis during the development of NASH. Upon activation, they differentiate into proliferative and migratory myofibroblasts that accumulate in areas of liver injury producing extracellular matrix (ECM) components and cytokines [44]. The data presented herein for the cell migration assay revealed that incubation of LX2 cells with both PEs resulted in significantly increased cell stimulation and subsequently migration. In parallel, gene expression analysis indicated a relative upregulation of genes involved in fibrosis (α-SMA, Col3A, Col1A, TIMP1, TIMP3 and TGF-beta) and inflammation (TNFα and IL6).
These impacts of PE again match the pathogenic changes underlying NASH. These observations, coupled with the positive correlation between these PEs and the progression of NAFLD, highlighted that elevated PEs are causal factors involved in the development of NASH. The consistent results between these two PEs suggested that SFA and MUFA in PEs may exert a similar impact on liver injury and disease progression of NAFLD.
Our study has a few limitations. First, our study does not have an independent obese NAFLD sample set to further confirm our findings. Therefore, our observations may only be limited to our sample set, and further replications would help to validate our results. Second, we only explored the potential causal role of two common PEs containing SFA and MUFA in vitro. HepG2 and LX2 may not be ideal in vitro models for NAFLD and NASH as well. It is unclear whether other PEs, e.g., those with very long chain polyunsaturated fatty acid (PUFA) moieties play different roles in the development of NASH, especially in vivo. Third, further characterizing the protein level of marker genes in the model may provide additional information for the impact of PEs. Fourth, our observations at the causal consequences of PEs on liver cell injuries and HSC activation should be further validated by inhibiting the action of PEs. Unfortunately, key mediators underlying the biological impact of PE signaling are currently unknown. Future studies should further examine the impact of different PE species in animal models and critical signaling mediators for the function of PEs. Nevertheless, our study provides novel insights into the mechanism of action of PE in NAFLD development and generated new hypotheses warranting further investigations.
In conclusion, the current study revealed that elevated levels of PEs containing SFA and MUFA are associated with the development of NAFLD and NASH, while these PEs are highly likely involved in the development of liver damage and disease progression processes by inducing lipid accumulation, mitochondrial dysfunction, cell growth inhibition, oxidative stress as well as activation of HSCs.

Subjects
In the present investigation, a retrospective study was conducted on samples obtained from 203 obese subjects who were undergoing bariatric surgery. The detailed description of these subjects and the process of sample collection process were reported previously [45]. Collection of these samples was originally approved by the Institutional Review Board (IRB) of the Medical College of Wisconsin Froedtert Hospital (PRO ID: PRO00005335).
Based on their liver histology, subjects were classified into 80 apparently healthy controls (C), 93 patients with simple steatosis (SS), 16 patients with borderline NASH (B-NASH) and 14 patients with NASH [45]. NAFLD was diagnosed based on histological assessment after exclusion of other causes of liver disease. Either SS, B-NASH or NASH was diagnosed according to steatosis grade, steatosis distribution, microvesicular steatosis, ballooning, lobular inflammation, portal inflammation, fibrosis, NAFLD activity score (NAS) and elevated liver enzymes. The NAS score is the sum of scores for steatosis, lobular inflammation, and ballooning, and it ranges from 0 to 8 [26].

Lipidomic Analysis
Targeted lipidomic analysis of lipids was conducted using an automated electrospray ionization (ESI)-tandem mass spectrometry approach according to a protocol described previously [46]. Data acquisition was performed based on a modified procedure by Devaiah et al. [47]. Briefly, serum lipids were extracted according to Folch method [48], then lipid extracts were dissolved in chloroform/methanol (9:1) prior to analysis. Precise amounts of internal standards for various phosphatidylcholines (PCs), lysophosphatidylcholines (LPCs), phosphatidylethanolamines (PEs), lysophosphatidylethanolamines (LPEs), phosphatidic acids (PAs), phytanoyl PAs, phosphatidylserines (PSs), phytanoyl PSs and phosphatidylinositols (PIs) were added. After mass spectrometry analysis, the data were normalized, and the background of each spectrum was deducted. Peaks for the target lipids in these spectra were then identified, followed by data correction for isotopic overlap, and molar amount calculation by comparing to the internal standards in the same lipid class. Data were expressed as the percentage of total lipids signal.

Cell Culture
Hepatoma cells HepG2 and human hepatic stellate cells LX2 were maintained in high glucose Dulbecco's modified eagle medium (DMEM) supplemented with 10% fetal bovine serum (FBS) and low glucose DMEM supplemented with 2% FBS, respectively, 100 U/mL penicillin G and 100 mg/mL streptomycin sulfate in a 5% CO 2 -humidified incubator at 37 • C. Both cells were periodically authenticated using microsatellite markers.

Phospholipid Treatment
The cell study comprised three groups: a control group, a PE (34:1) group and a PE (36:2) group. Concentrated PE stocks were prepared in 2:1 (v/v) chloroform/methanol mixtures and stored in glass tubes under nitrogen at −20 • C. The desired amount was dried in glass tubes under a stream of nitrogen. Cell lines were treated with different concentrations of PEs [0.25 mM, 0.5 mM, and 1 mM of each of PE (34:1) and PE (36:2)]. The control group received the same amount of the vehicle. PEs were conjugated with 2% fatty acid-free bovine serum albumin.

Cell Viability Assay
HepG2 cells were plated at an initial density of 5 × 10 3 cells/well in a 96-well plate and treated with various concentrations (0.25, 0.5 and 1 mM) of PE (34:1), PE (36:2), or vehicle, for 24, 48 and 72 h. After incubation, 10 µL of Cell Counting Kit-8 (CCK8) (ApexBio, Houston, TX, USA) was added to each well of the 96-well microplate. The plate was placed in a CO 2 incubator for 1-4 h to react. The absorbance was measured at 450 nm by a microplate reader.

Lipid Droplet/Nucleus Staining with BODIPY/Hoechst
Lipid accumulation was assessed using fluorescent detection with BODIPY 493/503 (Invitrogen/Molecular Probes, Eugene, OR, USA) according to the method of Baumann et al. [49]. HepG2 cells were seeded in a 96-well plate at a density of 5 × 10 3 cells/well, then the cells were incubated at 37 • C overnight. Thereafter, the cells were treated with vehicle or 1 mM PE (34:1 and 36:2) and incubated for 48 h. Fixation of cells was performed by adding 100 µL of 5% paraformaldehyde (PFA) solution in phosphate buffer saline (PBS) into the culture media to achieve a final PFA concentration of 2.5%, then incubation was allowed for 15 min at room temperature (RT). PFA was carefully removed, and the cells were washed twice with 100 µL PBS. BODIPY/Hoechst stock solutions were prepared to a working concentration of 5 µg/mL BODIPY and 1 µg/mL Hoechst, then 100 µL of both dyes were added in parallel and incubated for 15 min at RT. After incubation, the cells were washed twice with PBS, then 130 µL of PBS were added, and the resulting fluorescence intensity was measured at 493/503 for BODIPY and 352/454 for Hoechst by a microplate reader. To visualize the lipid droplets (LD) under the fluorescence microscope, the cells were cultured in a 35 mm culture dish overnight, then treated as above and thereafter stained with 5 µg/mL BODIPY and 1 µg/mL Hoechst. The LD were observed under the trinocular phase contrast fluorescence microscope (ZEISS Axiovert 200M, Carl Zeiss, Göttingen, Germany).

Mitochondrial ROS Production/Nucleus Staining by MitoSox™/Hoechst
Mitochondrial reactive oxygen species (ROS) production was assessed by a red mitochondrial superoxide indicator MitoSox (Invitrogen/Molecular Probes, Eugene, OR, USA) according to the manufacturer's instructions. HepG2 cells were cultured at a density of 5 × 10 3 cells/well in a 96-well plate, then incubated at 37 • C overnight. The cells were incubated with vehicle or 1 mM PE (34:1 and 36:2) for 48 h. The culture medium was carefully removed and PBS containing 1 µg/mL Hoechst and 5 µM MitoSox was added to the cells, followed by incubation at 37 • C for 30 min. The staining solution was aspirated, the cells were washed twice with PBS, then the resulting fluorescence was measured using a plate reader at excitation/emission of 510/580 nm for MitoSox and 352/454 nm for Hoechst. For visualization of ROS production underneath the fluorescence microscope, the cells were cultured in a 35 mm culture dish overnight, treated as above, stained with 5 µM MitoSox and 1 µg/mL Hoechst, then observed under the trinocular phase contrast fluorescence microscope (ZEISS Axiovert 200M, Göttingen, Germany).

Mitochondrial Membrane Potential (∆Ψm) and Mitochondrial Mass/Nucleus Staining
Mitochondrial membrane potential and mitochondrial mass were determined by staining with Mitotracker green and Mitotracker red (Invitrogen/Molecular Probes, Eugene, OR, USA), respectively, according to the manufacturer's instructions. HepG2 cells were cultured at a density of 5 × 10 3 cells/well in a 96-well plate overnight, treated with vehicle or 1 mM PE (PE 34:1 and PE 36:2) for 48 h. Mitochondrial mass and mitochondrial membrane potential were determined by staining cells with 100 nM, 200 nM, and 1 µg/mL Mitotracker red, green and Hoechst, respectively. Fluorescence intensity of Mitotracker red, green and Hoechst was assessed by a plate reader at excitation/emission of 579/599 nm, 490/516 nm and 352/454, respectively. To examine alteration in mitochondrial mass and membrane potential under the trinocular phase contrast fluorescence microscope (ZEISS Axiovert 200M, Göttingen, Germany), the cells were plated in a 35 mm culture dish overnight, treated as above, then stained.

Cell Migration Assay
Cell migration was assayed with a Transwell plate according to the method described by Justus et al. [50]. In a 24-well Transwell with 8.0 µm pore polycarbonate membrane inserts (CORNING, Corning, NY, USA), 2 × 10 4 LX2 cells were plated in serum-free low glucose DMEM with vehicle, 1 mM PE (34:1 and 36:2), or 0.04 µg/mL lipopolysaccharide (LPS) as a positive control. An aliquot of 500 µL of low glucose DMEM containing 10% FBS as chemoattractant was added to the lower chamber. The cells were incubated at 37 • C and 5% CO 2 for 24 h. Migrated cells were fixed with 70% ethanol, stained with 0.2% crystal violet and pictures were captured underneath an upright digital microscope (SeBa™-Laxco, Mill Creek, WA, USA). Crystal violet was extracted by 33% acetic acid, transferred to a 96-well clear microplate, and the absorbance at 595 nm was measured using a plate reader.

RNA Extraction and Reverse Transcription Polymerase Chain Reaction (RT-PCR)
RNA was extracted using TRIzol ® Reagent (Life Technologies Corporation, Carlsbad, CA, USA) from HepG2 and LX2 cell suspensions pretreated with vehicle or 1 mM PE (34:1 and 36:2) for 48 h prior to extraction according to the manufacturer's instructions. Concentration and purity of RNA were measured spectrophotometrically at A280 and A260. For quantification of mRNA expression levels, RNA was reversely transcribed using reverse transcription cDNA Synthesis kit (ThermoFisher Scientific, Waltham, MA, USA). The cDNA was used as template in qPCR Applied Biosystems QuantStudio™ 7 Flex Real-Time PCR System using Universal SYBR Green (BioRad, Irvine, CA, USA) and gene-specific primer pairs listed in Supplement Table S1.

Statistics
Statistical analysis was performed by the statistical software R and GraphPad Prism 7.4 software. Continuous data were checked for normality by Shapiro-Wilk's test. For patient and clinical characteristics, count and percentage were used to summarize categorical variables and, for continuous variables, median and range were used. Other continuous data (such as % of total lipids signal, cell viability, staining, fluorescence intensity, mRNA expression, etc.) were expressed as mean ± standard deviation (SD). Parametric data were analyzed by unpaired t-test, one-way ANOVA, or two-way ANOVA. Unpaired t-test was performed to compare between two groups. One-way ANOVA was carried out to compare among three or more groups, followed by Tukey's or Holm's post-hoc pairwise comparisons. Correspondingly, two-way ANOVA was used to analyze differences between three or more independent groups that had been split on two variables (time-dependent data), followed by Tukey's or Holm's post-hoc analysis. Non-parametric data were analyzed using Kruskal-Wallis test, followed by Dunn's or Holm's post-hoc pairwise comparisons.