Metabolic regulation of hepatic PNPLA3 expression and severity of liver fibrosis in patients with NASH

Abstract Background and Aims The genetic PNPLA3 polymorphism I148M has been extensively associated with higher risk for development and progression of NAFLD towards NASH. Methods PNPLA3 and α‐SMA expression were quantified in liver biopsies collected from NASH patients (n = 26) with different fibrosis stages and PNPLA3 genotypes. To study the potential mechanisms driving PNPLA3 expression during NASH progression towards fibrosis, hepatocytes and hepatic stellate cells (HSCs) were cultivated in low and high glucose medium. Moreover, hepatocytes were treated with increasing concentrations of palmitic acid alone or in combination with glucose. Conditioned media were collected from challenged hepatocytes to stimulate HSCs. Results Tissue expression of PNPLA3 was significantly enhanced in biopsies of patients carrying the I148M polymorphism compared to wild type (WT). In NASH biopsies, PNPLA3 significantly correlated with fibrosis stage and α‐SMA levels independently of PNPLA3 genotype. In line, PNPLA3 expression was higher in α‐SMA positive cells. Low glucose increased PNPLA3 in HSCs, whereas high glucose induced PNPLA3 and de‐novo lipogenesis‐related genes expression in hepatocytes. Palmitic acid induced fat accumulation and cell stress markers in hepatocytes, which could be counteracted by oleic acid. Conditioned media collected from lipotoxic challenged hepatocytes markedly induced PNPLA3 mRNA and protein levels, fibrogenic and autophagic markers and promoted migration in HSCs. Notably, conditioned media collected from hepatocytes cultivated with both glucose and palmitic acid exacerbated HSCs migration, PNPLA3 and fibrogenic gene expression, promoting release of cytokines from HSCs. Conclusions Collectively, our observations uncover the diverse metabolic regulation of PNPLA3 among different hepatic cell populations and support its relation to fibrosis progression.


| INTRODUC TI ON
Non-alcoholic fatty liver disease (NAFLD) is a growing and burgeoning health issue with high prevalence in western countries. 1,2 A portion of NAFLD patients (approximately 5%-10%) evolves into non-alcoholic steatohepatitis (NASH), which represents the advanced manifestation of NAFLD and predisposes to severe hepatic impairment, such as cirrhosis and ultimately cancer. 3 Current evidence suggests that the progression from steatosis towards NASH is a multi-hit process triggered by sustained hepatocytes injury resulting in inflammation and fibrosis. 4 A recognized hallmark of this process is the accumulation of excessive free fatty acids (FFAs) in hepatocytes stored as triglycerides (TGs) in lipid droplets (LDs). The source of this FFAs flux towards hepatocytes derives from adipose tissue lipolysis, increased hepatic de novo lipogenesis and external factors such as imbalanced diets rich in sugars and saturated FFAs (i.e. western diet). In particular, saturated fatty acids, like palmitic acid (PA; C16:0), are known to be lipotoxic and promote metabolic stress and apoptosis in hepatocytes. [5][6][7] The consequence of sustained hepatocellular injury is the activation of hepatic stellate cells (HSCs), the main contributors in driving liver fibrosis as wound healing response, characterized by the replacement of normal hepatic parenchyma with fibrotic scars. 8,9 Liver fibrosis is considered the major determinant of mortality in patients with NASH. 10 Emerging evidence indicates that also genetic factors play a key role in hepatic disease progression and fibrosis development. 11 In this regard a single base polymorphism of the human patatin-like phospholipase domain-containing 3 (PNPLA3) gene, known as I148M (rs738409 C>G) has been linked to increased risk of hepatic TG accumulation 12,13 and is considered as an independent risk factor predisposing to NASH. [14][15][16][17] In the liver, PNPLA3 expression is regulated by carbohydrate feeding via sterol regulatory element-binding protein-1c (SREBP-1c) in mouse and human hepatocytes. [18][19][20] Notably, PNPLA3 is highly expressed in HSCs 21 and recent findings uncovered a strong contribution of PNPLA3 to achieve a full myofibroblast-like phenotype in HSCs and recognized a significant impact of its genetic variant I148M to exacerbate pro-inflammatory and pro-fibrogenic features of human HSCs. 22,23 However, the regulation of PNPLA3 during human hepatic fibrosis development is largely unclear. Therefore, we aimed to investigate whether: (a) hepatic PNPLA3 expression correlates with fibrosis progression in human NASH and according to the I148M genotype; (b) PNPLA3 expression is regulated by glucose and FFAs in hepatocytes as well as in HSCs; (c) metabolic stress in hepatocytes affects PNPLA3 expression and fibrogenic features in HSCs. Cases with NAS score ≥3 and fibrosis stage from 1 to 4 were classified as NASH. According to the NASH Clinical Research network guidelines, fibrosis stages were classified as follows: F0 = none; F1 = perisinusoidal or periportal; F1a = mild, zone 3, perisinusoidal; F1b = moderate, zone 3, perisinusoidal; F1c = peri-portal sinusoidal fibrosis; F2 = Zone 3 sinusoidal fibrosis and peri-portal sinusoidal fibrosis; F3 = Bridging fibrosis; F4 = Cirrhosis. 24 For convenience, from 1 to 2 were considered as 'mild' fibrosis, whereas fibrosis stage from 3 to 4 as 'severe' fibrosis. PNPLA3 genotyping was performed by real-time PCR for the I148M single-nucleotide polymorphism (rs738409 C>G), as done routinely in our clinical research center. 22,23 Only homozygote genotypes were used in this study (C/C as WT and G/G as I148M).

| Human samples
Conclusions: Collectively, our observations uncover the diverse metabolic regulation of PNPLA3 among different hepatic cell populations and support its relation to fibrosis progression.

K E Y W O R D S
genetic polymorphism, lipase, lipid stress, non-alcoholic fatty liver disease

Lay summary
Hepatic PNPLA3 expression increases during progression of liver fibrosis in patients with NASH and mainly colocalizes with activated hepatic stellate cell marker α-SMA. High glucose and saturated free fatty acids increase PNPLA3 and induce lipid stress in hepatocytes, which in turn results in PNPLA3 induction in hepatic stellate cells along with exacerbated pro-fibrogenic features.

| Human tissue staining and image quantification
For immunofluorescence (IF) staining, slides were blocked for 1 hour, incubated overnight at 4°C with polyclonal rabbit antihuman PNPLA3 antibody (#ab81874, Abcam). The day after, slides were washed and incubated at room temperature for 1 hour with monoclonal mouse anti-human α-SMA antibody (#A5228, Sigma-Aldrich). Slides were then washed and incubated with the respective secondary antibody: Alexa Fluor 488 goat anti mouse IgG for α-SMA and Alexa Fluor 594 goat anti rabbit IgG for PNPLA3 (both Thermo Fisher Scientific). Nuclei were counterstained with DAPI (Sigma-Aldrich) and mounted (VECTASHIELD® Mounting Medium, Vector Laboratories) for microscope analysis (Olympus BX51).
Relative quantification of IF staining was performed using Image J software in an automated fashion as seen previously. 25 Channels were splitted for the respective fluorochrome (Red for PNPLA3 and Green for α-SMA), and total fluorescence was quantified using macros in parallel for all the pictures taken. A number of six random shots were taken for each slide and quantified accordingly.

| Luciferase assay
Luciferase assay was performed as previously reported. 22

| Western blot
Protein expression analysis was performed as described previously. 22,23

| Migration assay
Cell migration was measured in modified Boyden chambers, as described previously. 22,23 Cells were fixed in methanol and stained with hematoxylin/eosin solution before counting.

| Elisa cytokine assay
HSCs supernatant was collected after 24 hours of exposure to IHH c.m. pre-incubated with 25 mmol/L Glc alone or in combination with PA (300 μmol/L). Proteome profiler for human cytokine array (R&D) was performed according to manufacturer instructions, as previously described. 22

| Flow cytometry
LX-2 and IHH were previously treated with 5 or 25 mmol/L of Glc, trypsinized and resuspended in fixing and permeabilizing reagents (all Thermo Fisher Scientific). Thereafter, they were incubated with a 1:500 dilution of the polyclonal rabbit anti human PNPLA3 antibody (Abcam) for 1 hour, on a shaker platform. A fluorochrome-labelled Alexa Fluor 594 goat anti rabbit IgG was incubated for 1 hour at room temperature on a shaker, DAPI was added to each tube for 10 minutes before flow cytometric analysis. Flow cytometry was performed with BD FACSCanto™ II and BD FACSDiva™ software (Becton Dickinson).

| Statistics
Data are presented as mean values ± standard deviation (SD).
Correlations were analyzed non parametrically by Spearman's rank correlation. Statistical analysis was performed using GraphPad Prism. The unpaired Student t test was used when two groups were compared, and two-way analysis of variance was used for multiple groups. P < .05 was considered statistically significant.

| PNPLA3 correlated with fibrosis development in human NASH and is more expressed in α-SMA positive cells
Based on our previous observation that PNPLA3 expression was induced during primary human HSCs activation in vitro, 22 we evaluated PNPLA3 expression during liver fibrosis development in human NASH biopsies. Our study population involved 26 individuals (Table 1) and was classified according to PNPLA3 genotype (two groups, C/C -WT, n = 15 and G/G -I148M, n = 11). No significant differences were observed in clinical aspects between these two groups, apart from a trend in increased transaminases level of ALT and AST in carriers of the genetic variant I48M (G/G), compatible with previous reports involving larger sample size. 13,14 In addition to PNPLA3 genotype, the cohort of liver samples was divided and analyzed according to fibrosis stage, ranging from F = 1 and 2 for 'mild'/periportal to F = 3 and 4 for 'severe' hepatic fibrosis/ cirrhosis ( Table 2). Notwithstanding the low sample size, significantly higher steatosis percentage (%) was found in G/G genotypes independently of the fibrosis stage (Table 2).
PNPLA3 and α-SMA tissue expression were investigated in these samples by immunofluorescence (IF) staining (Figure 1 and Supp. Figure S1). Strikingly, we found that PNPLA3 expression increased during the progression of liver disease from mild to severe fibrosis These observations confirmed and supported our previous findings that PNPLA3 is proportionally induced during human NASH progression. 22 Moreover, PNPLA3 relative quantification correlated significantly with the fibrosis stage (R = 0.793, P < .001 for C/C and R = 0.579, P = .006 for PNPLA3 G/G) and with α-SMA tissue expression independently of the genotype (R = 0.843, P < .01 for C/C and R = 0.5423, P < .01 for G/G) ( Figure 2B).
As previously reported, PNPLA3 is known to be expressed in hepatocytes and HSCs. 21,22 Notably, our immunofluorescence staining analysis revealed that during fibrosis progression PNPLA3 liver expres-

| PNPLA3 is oppositely regulated by glucose in human hepatocytes and HSCs
To unravel the potential mechanisms determining the induction of PNPLA3 and to mimic the starving conditions of patients undergoing liver biopsy, we next measured PNPLA3 expression in LX-2 and IHH, human HSCs and hepatocyte cell lines respectively, using culture media with low (5 mmol/L) or high (25 mmol/L) glucose (Glc). Firstly, flow cytometry analysis showed that the amount of PNPLA3 was significantly lower in hepatocytes compared to HSCs (P < .001) under low Glc conditions (5 mmol/L); when cells were treated with high Glc (25 mmol/L), the response was reversed (P < .001, Figure 3A). These data were confirmed by western blot analysis ( Figure 3B). Secondly, sterol regulatory element-binding protein-1c (SREBP-1c) expression and its downstream targets coding for de-novo lipogenesis pathway (FASN, ACC-1, SCD-1 and PNPLA3) were significantly induced by 25 mmol/L Glc in IHH (P < .001, Figure 3C), but blunted in HSCs.
Downstream of SREBP-1c, 28 the expression of liver X receptor α (LXRα) was also upregulated upon 25 mmol/L Glc in IHH whereas there was no change in HSCs. Inversely, the β isotype (LXRβ) expression increased after high Glc treatment in LX-2, but not in hepatocytes (P < .001, Figure 3C), suggesting differences in pathways controlling lipid/cholesterol levels between hepatocytes and HSCs.
Interestingly, the expression of the CCAAT/Enhancer-Binding Protein alpha (CEBPα), a transcription factor known to control hepatic gluconeogenesis/de-novo lipogenesis, 29,30 was strongly induced by high Glc in hepatocytes. Furthermore, the expression of the glucose regulated protein 78KD (GRP78) showed opposite trends (P < .001, Figure 3C), confirming the dual role of this protein in inhibiting SREBP-1c-induced de-novo lipogenesis. 31,32 To support mRNA gene expression data, we performed luciferase co-transfection assays using LXR (LXRE) and SREBP-1c (SRE) response elements in IHH and LX-2 followed by Glc stimulation (5 or 25 mmol/L).
Collectively, our data suggest that PNPLA3 transcript and pro-

| FA loading induces metabolic stress in hepatocytes which results in increased PNPLA3, profibrogenic genes and autophagy in HSCs
Lipotoxic saturated FFAs impair lipid accumulation in hepatocytes and contribute actively to inflammation and activation of HSCs. 33 In order to create an in vitro model resembling lipid-stressed hepatocytes, 33 human immortalized hepatocytes and hepatoma cell lines (IHH and HuH7, respectively) were challenged with increasing concentrations of palmitic acid (PA) alone or in combination with oleic acid (OA). After 24 hours, fatty acid loading was confirmed by Oil Red O staining and quantified using BioPix iQ Imaging Software (P < .05, Figure 4A). To confirm palmitate-induced lipotoxicity in hepatocytes, 33 (autophagy marker) protein levels were also significantly increased, starting off at low concentrations (ie 50µM PA; Figure 4F). Inversely, c.m. collected from hepatocytes treated with the combination of PA and OA was protective 36 and prevented these effects in HSCs (Supporting Figure S2 and S3). In addition, HSCs fibrogenic markers Collagen1α1, α-SMA and TIMP-1 were induced, along with autophagy markers ATG5 and LC3a, thus contributing to sustain HSCs activation ( Figure 4F).
Altogether, these data evidenced that excessive exposure to lipotoxic PA in hepatocytes activates HSCs, thus resulting in upregulation of PNPLA3, pro-fibrogenic markers as well as autophagy and migration in HSCs.

| Hepatocytes conditioned medium containing glucose in combination with PA exacerbated profibrogenic features of HSCs
The above observations brought us to verify whether the combination of both Glc and PA might exacerbate these fibrogenic features in HSCs. Therefore, we challenged IHH cells with PA at different concentrations (0, 100, 300µM) in combination with low or high Glc (5 or 25 mmol/L) for 24 hours (Supporting Figure S4A).  Figure 5C).
Notably, activation of JNK pathway was observed when HSCs were exposed to c.m. containing 300µM of PA and 25 mmol/L of Glc (p-JNK vs total JNK; P < .01, Figure 5C), but the increase in autophagy markers LC3A/B and ATG5 was significant only with 5 mmol/L Glc (P < .05, Figure 5C).
Moreover, gene expression analysis revealed that along with PNPLA3, combination of PA (300 μmol/L) and Glc (25 mmol/L) in IHH c.m. strongly induced in HSCs the expression of two pro-fibrogenic markers, Collagen1α1 and TIMP-1 (P < .05, Figure 6A), whereas the adipogenic regulators of lipid and glucose metabolism GRP78 and CEBPα were strongly reduced (P < .001, Figure 6A), along with the expression of PPARγ and its downstream target CD36 ( Figure 6A, Figure S4B).
Collectively, our in vitro model shows that we observed distinct effects in HSCs when exposed to IHH c.m. containing sat-

| D ISCUSS I ON
Our first aim in this study was to evaluate and quantify PNPLA3 expression during different stages of liver fibrosis in NASH patients, according to the different PNPLA3 genotypes (WT = C/C or I148M = G/G). Hereby, we demonstrated that in NASH livers PNPLA3 increased during fibrosis development, from mild (F1) to severe fibrosis (F4), along with the induction of the HSCs activation marker α-SMA (Figures 1 and 2, Figure S1). Importantly, PNPLA3 also correlates with fibrosis stage and α-SMA hepatic tissue expression, independently from the genotype ( Figure 2B). In addition, PNPLA3 expression is higher in hepatic cells positive for α-SMA ( Figure 2C).
Although our patient cohort is relatively small (n = 26) compared to other previous studies, serum liver enzymes such as ALT and AST (Table 1)  ballooning and NAS score, compared to the C/C carriers (Table 2).
Over the last decade the genetic polymorphism of PNPLA3, known as I148M, was strongly linked to non-alcoholic fatty liver disease (NAFLD) [12][13][14][15][16]37,38 and to the progression towards severity of liver diseases, such as cirrhosis and liver cancer. 39,40 Notably, the present study strongly supports our previous findings in vitro with primary human HSCs, 22,23 and therefore contributes to highlight the importance to group patients according to the different PNPLA3 genotype.
As a second aim, we were interested to explore the potential mechanism underlying the modulation of PNPLA3 expression across different liver cell types, in order to better understand the regulation of this protein under different metabolic stresses (FFAs and Glc) and to better resemble the physiological conditions of patients undergoing liver biopsy. Since patients were starved prior biopsy, we wondered whether the regulation of PNPLA3 expression mediated by glucose already shown in hepatocytes 18,19 was valid for HSCs as well. Conversely to hepatocytes, here we reported that in HSCs PNPLA3 protein level is high in low glucose concentration (5 mmol/L Glc, Figure 3A increases and stabilizes PNPLA3 protein levels without altering its transcript. 20 Moreover, our findings are in line with previous studies, since hepatocytes exposed to 25 mmol/L glucose clearly showed increased amount of the transcription factor SREBP-1c ( Figure 3C) and its transcriptional activity ( Figure 3D), which is responsible for PNPLA3 upregulation, along with enhanced expression of SREBP-1c downstream targets FASN, ACC-1, SCD-1 and LXRα. Interestingly, GRP78 reduction observed only in hepatocytes treated with high glucose supports previous findings where reduced expression of GRP78 in the liver was related to hepatic fat accumulation due to decreased inhibition of insulin-dependent cleavage of SREBP-1c. 31,32 In line, hepatocytes CEBPα upregulation with high glucose corresponds to the metabolic induction of hepatic gluconeogenesis. 29 As third aim, PNPLA3 genetic variant I148M has been strongly associated to progression of liver diseases (NASH) and liver fibrosis development. 12,13,[15][16][17]41,42 In this regard, lipids mediate directly or indirectly activation of intracellular pathways responsible of hepatocellular stress and innate immune system-mediated inflammation, a condition called lipotoxicity. In particular, the latter term refers to a process by which accumulation of certain toxic lipids, such as saturated FFAs in hepatocytes, triggers various molecular pathways of cell stress (ie JNK pathway), eventually resulting in hepatocellular death and release of molecular mediators affecting neighbouring cells, such as HSCs. 33 JNK is a member of the mitogen-activated protein kinases (MAPK) family and it has been shown to play a critical role in dietary animal models of NASH and in human NASH. [43][44][45] Resembling in vivo data, JNK activation is a recognized hallmark for palmitate-induced lipotoxicity in hepatocytes in vitro. 36  ( Figure 4F). Of note, previous work has reported that TIMP-1 is highly secreted by LX-2 overexpressing the I148M variant compared to LX-2 overexpressing the WT PNPLA3. 46 This evidence strengthens our findings, since we report that lipid exposed hepatocytes induce PNPLA3 increase and TIMP-1 expression in HSCs and therefore we might speculate that TIMP-1 secretion would be more pronounced when HSCs express the I148M PNPLA3.
Conversely, HSCs exposed to c.m. collected from hepatocytes treated with combination of PA and OA (1:2) migrated less and showed reduced PNPLA3 and α-SMA protein content ( Figure S3).
The combination of saturated FA (PA) and different doses of glucose determined even more pronounced outcomes in HSCs, as reflected by increased migration, release of cytokines (GM-CSF, G-CSF and MIF), PNPLA3 and JNK pathway, along with pro-fibrogenic genes (Collagen1α1 and TIMP-1) and downregulation of GRP78, CEBPα and PPARγ (Figures 5 and 6). Interestingly, increased levels of GM-CSF and G-CSF were shown to stimulate endogenous mechanisms for hepatic regeneration and tissue repair, 47 while MIF has been shown to exert antifibrotic effects on HSCs. 48 In addition to activation of key intracellular pathways, overloaded and damaged hepatocytes release bioactive molecules that interact with target cells and are responsible of the cell-cell interactions and of the following biological response. Previous studies explored the molecular mechanisms through which lipid overloaded hepatocytes triggered activation of HSCs. Povero and co-authors demonstrated that lipotoxic-harmed hepatocytes produce and release extracellular vesicles (EVs) which could carry microRNA molecules able to target PPARγ and therefore promote HSCs activation. 34 Indeed, our results showed that in HSCs the expression of the PPARγ and its target gene CD36 were reduced under the high glucose and PA levels ( Figure 6A, Figure S4B), where pro-fibrogenic genes Collagen1α1, PNPLA3 and TIMP-1 were upregulated. Anyhow, further studies will be necessary to explore the mechanism of gene regulation in HSCs via EVs and how glucose and lipotoxic/saturated FFA modulate EVs production in our settings.
Interestingly, the metabolic roles of GRP78 and CEBPα as important regulators of lipid and glucose homeostasis in hepatocytes have been already explored. 29 In conclusion, these observations emphasize that PNPLA3 expression in human biopsies is closely linked with HSCs activation and liver fibrosis progression in NASH livers, being metabolically controlled by glucose and pro-steatotic conditions. Moreover, the presence of the genetic variant I148M contributes to characterize the stage and severity of hepatic fibrosis, thus supporting PNPLA3 genotype evaluation as a gold standard for better and specific prognosis. All together, we stress that our preliminary data on in vivo PNPLA3 expression during liver fibrosis progression require future validations also in non-metabolic hepatic disorders.

ACK N OWLED G EM ENTS
We thank Nicole Auer and Dr Emina Halilbasic for the excellent technical support and help with acquisition of clinical data, respectively.