AQP3 is regulated by PPARγ and JNK in hepatic stellate cells carrying PNPLA3 I148M

Aquaglyceroporins (AQPs) allow the movement of glycerol that is required for triglyceride formation in hepatic stellate cells (HSC), as key cellular source of fibrogenesis in the liver. The genetic polymorphism I148M of the patatin-like phospholipase domain-containing 3 (PNPLA3) is associated with hepatic steatosis and its progression to steatohepatitis (NASH), fibrosis and cancer. We aimed to explore the role of AQP3 for HSC activation and unveil its potential interactions with PNPLA3. HSC were isolated from human liver, experiments were performed in primary HSC and human HSC line LX2. AQP3 was the only aquaglyceroporin present in HSC and its expression decreased during activation. The PPARγ agonist, rosiglitazone, recovered AQP3 expression also in PNPLA3 I148M carrying HSC. When PNPLA3 was silenced, AQP3 expression increased. In liver sections from patients with NASH, the decreased amount of AQP3 was proportional to the severity of fibrosis and presence of the PNPLA3 I148M variant. In PNPLA3 I148M cells, the blockade of JNK pathway upregulated AQP3 in synergism with PPARγ. In conclusion, we demonstrated profound reduction of AQP3 in HSC carrying the PNPLA3 I148M variant in parallel to decreased PPARγ activation, which could be rescued by rosiglitazone and blockade of JNK.

Hepatic stellate cells (HSC) represent central players in the pathogenesis of liver fibrosis 1,2 . Quiescent HSCs store vitamin A in the liver. However, in response to hepatic injuries, HSCs may undergo activation and transdifferentiate into a highly proliferative and myofibroblast-like phenotype responsible for hepatic fibrogenesis 3,4 .
Lipid content is a crucial factor in HSC pathophysiology 5,6 and lipogenic activity decreases in parallel with vitamin A content during HSC activation 7 . In a recent study we showed that primary HSC carrying the human genetic variant of PNPLA3 (adiponutrin), known as PNPLA3 I148M, lack peroxisome proliferator-activated receptor gamma (PPARγ, NR1C3) expression and activity 8 , which is closely linked to HSC activation 9 . This PNPLA3 I148M variant has been associated to higher accumulation of fat in liver, steatohepatitis and inflammation, progression to fibrosis/cirrhosis and liver cancer 10,11 . PPARγ, the master regulator of adipogenesis, is a ligand activated nuclear receptor, which is mainly expressed in adipose tissue and plays a crucial role in adipogenesis and energy metabolism. In liver, PPARγ is involved in HSC lipid storage, conveys their quiescence and, is reduced in activated HSCs as a direct effect of JNK activation 9,12 .
Aquaglyceroporins (AQPs) are channel proteins, facilitating glycerol diffusion in cells 13,14 . Glycerol represents the backbone structure for triglyceride synthesis which is a fundamental factor in cell metabolism 15 . In HSC, AQPs expression and function are poorly understood. To our knowledge, only one study showed that changes in AQPs expression and water permeability may increase resistance to apoptosis in activated HSCs, although the metabolic impact of glycerol diffusion in activation and quiescence was not analyzed 16 . Importantly, previous studies revealed repression of AQP3 by LPS exposure in human colon epithelial HT-29 cells 17 but induction in murine 3T3-L1 adipocytes 18 , in line with AQP3 upregulation by arachidonic acid or prostaglandin in human retinal pigment epithelial cells, already suggesting a role of JNK 19 . Since in PNPLA3 I148M expressing HSCs the JNK/AP-1 pathway is strongly induced 20 , the subsequent PPARγ downregulation, led us to hypothesize that PPARγ could regulate AQPs expression in HSC in connection to the I148M variant 20 .
Therefore, we aimed to uncover (i) which AQPs were expressed in primary hepatic stellate cells, (ii) how AQP expression is regulated at the molecular level and (iii) whether PNPLA3 mutations modulated AQP levels in HSC.

Results
AQP3 is the only adequately expressed aquaglyceroporin in human primary HSC and down-regulates during their activation. We explore the expression of all known aquaglyceroporin AQP3, -7, -9 and -10 present in primary HSC expressing the WT PNPLA3 variant by quantitative RT-PCR. Interestingly, only AQP3 was detectable in human primary HSC, but not in hepatocytes (HepG2 cell line) (Fig. 1A). Therefore, we measured its expression in stellate cells during their activation. Over time, there was a profound downregulation of AQP3 paralleled by an increased expression of the senescence marker, p21 ( Fig. 1B and D) and the profibrogenic marker α-SMA (not shown). Moreover, protein levels of AQP3 (Fig. 1C) and AQP3 surface staining quantified by a flow cytometric assay ( Fig. 1E and Suppl. Figure 1) were reduced. Collectively, these data demonstrated the presence of AQP3 in primary human HSC and its down regulation during HSC activation.  PPARγ agonist increases AQP3 expression in LX2 cells in association with enhanced lipogenesis and reduced PNPLA3 expression. Since AQP3 is a PPARγ target gene 21 , and PPARγ is strongly down-regulated in fully activated HSCs 20 , we further explored whether a PPARγ agonist, rosiglitazone (RSG), may reverse the repression of AQP3 during HSC activation. AQP3 mRNA expression increased after RSG stimulation in LX2 cells ( Fig. 2A), as well as lipogenesis (demonstrated by increased gene expression of FASN, SREBP1c, SCD1 and PPARγ) while expression of PNPLA3 was significantly reduced (Fig. 2B). These results demonstrate that the regulation of AQP3 is strongly PPARγ dependent in HSC; to such an extent that administration of a PPARγ agonist up-regulates AQP3 expression.
AQP3 expression is reduced in PNPLA3 I148M cells and patients but it is restored in vitro by the PPARγ agonist rosiglitazone. In order to further explore the dependence of AQP3 on PPARγ in HSC, we used LX2 overexpressing PNPLA3 wild type (WT) and I148M, since the latter have previously shown 8 to lack PPARγ. As seen in Fig. 3A, AQP3 was present only in LX2 cells overexpressing PNPLA3 WT, with a dramatic down regulation in LX2 cells overexpressing the PNPLA3 I148M variant. Notably, RSG treatment of I148M PNPLA3 strongly restored AQP3 expression while WT cells displayed a relatively milder up-regulation of AQP3 (Fig. 3B). Conversely, when we stably silenced PNPLA3 in LX2 cells (Fig. 4A) AQP3 expression was strongly up-regulated (Fig. 4B). In order to explore the relevance for AQP3 in humans, we performed an immunofluorescence double staining of AQP3 and α-SMA of either healthy liver patient tissues (all fibrosis stage 0 and WT for PNPLA3, n = 4) or NASH patients with fibrosis stage 1c and 4 (n = 5 for PNPLA3 I148M, n = 5 for PNPLA3 WT). Our data showed that AQP3 is relatively more abundant in normal WT PNPLA3 liver (C/C) than in PNPLA3 I148M livers (G/G) (Fig. 5A,B) and inversely correlates with the stage of fibrosis. In conclusion, we demonstrated that PPARγ robustly up-regulates AQP3 expression in HSCs in a clear PNPLA3 isoform-dependent manner and that in line PNPLA3 I148M patients show lower AQP3 expression.

AQP3 is regulated by PPARγ through JNK pathway and shows a tight dependence on PNPLA3.
Since we previously demonstrated that PNPLA3 I148M activates JNK pathway via blocking PPARγ action, which could be partially reversed by a PPARγ agonist 8 , we next explored whether blockade of JNK pathway with a specific antagonist (SP600125) affected AQP3 expression. LX2 cell line overexpressing PNPLA3 I148M treated with JNK antagonist showed increased AQP3 expression with an additive effect in combination with RSG (Fig. 6A). Moreover, cells with inhibited JNK pathway had higher vitamin A content (Fig. 6B, Suppl. Figure 2), cell size and subsequently became quiescent as reflected by loss of p21 (not shown). Collectively as summarized schematically in Fig. 6C, the impact of PNPLA3 I148M on AQP3 expression could be blocked by JNK inhibition and PPARγ stimulation.

Discussion
Aquaporins are interesting but extremely elusive targets in human physiology 22 . They are expressed in many organs but their metabolic function, especially in adipose tissue and liver, is still unclear 14,[22][23][24] . AQP9 in hepatocytes has been by far the most studied AQP in liver [25][26][27] , being a key player in the development of liver steatosis 28 . AQP3 expression was originally found in colon, kidney and liver in humans and had in vitro the capacity to transport water, urea and glycerol 29,30 . While AQP3 deletion in kidney resulted in polyuria 31 , deletion in skin reduced glycerol and water content 32 showing that in vivo AQP3 transports water as well as glycerol. Despite its  hepatic expression, AQP3 cell type localization in the liver was not investigated so far. In HSC, a few studies from our group 33,34 explored the metabolic relevance of AQP3 in liver fibrosis and its tight hormonal regulation by adiponectin. However, our experiments took place only in LX2 cell line, therefore some questions remained open on the pathway involved and its meaning in vivo. We tested AQP3 expression in human primary HSC, discovering the molecular mechanism involved also in connection to a disease model.
The PNPLA3 I148M variant is key risk factor for development of hepatic steatosis and its progression to more severe liver disease, with development of more advanced fibrosis, cirrhosis and cancer 11,35,36 . A recent study from our group 8 , uncovered how PNPLA3 I148M in HSC results in an intrinsic reduction of PPARγ, the main gene involved in lipid metabolism and AQPs regulation 21,25,27 . Importantly, several AQPs have been shown to be regulated by PPARγ agonists. More specifically, RSG induced AQP7 expression in adipose tissue from OLETF rats and AQP3 in a hepatoma cell line 37 . In line, in leptin deficient mice treated with RSG, the levels of AQP3 in subcutaneous adipocytes and AQP9 in hepatocytes increased 38 . Therefore, we aimed to explore a putative connection between AQP3 expression and PNPLA3 I148M within HSC activation.
We here demonstrate that AQP3 is the only aquaporin expressed in human primary HSC consistent with previous findings in LX2 cells 33 , and repressed during their activation. Treatment with rosiglitazone, a PPARγ agonist and upstream transcription factor known to regulate AQPs expression in other organs 20 , induced AQP3 expression and therefore may modulate glycerol influx. In line with reduced PPARγ activity in PNPLA3 I148M 8 , AQP3 expression decreased dramatically. This highlights the interplay between AQP3 and PNPLA3, consistent with their roles in glycerol uptake and triglyceride synthesis or degradation, respectively, further emphasized by the increased expression of genes involved in lipogenesis. Importantly, downregulation of AQP3 in PNPLA3 I148M could be strongly counteracted with the PPARγ agonist rosiglitazone in comparison to WT cells, underlying a strong PPARγ dependence. Moreover when PNPLA3 was silenced in LX2 cells, AQP3 strikingly upregulated. These data indicate that PPARγ is not only required for stimulation of AQP3, but may also drive its baseline expression. Interestingly, PPARγ agonists are known to promote white adipose tissue differentiation and expansion in pre-adipocytes and stimulate lipogenesis 39 , similar to what we now observed in HSC. In human  Figure 2. C. Schematic summary of the proposed molecular mechanism. RSG increases PPARγ level, which in turn regulates AQP3 expression and glycerol shuttling. PNPLA3 I148M activates JNK pathway, which suppresses PPARγ and consequently AQP3 expression. healthy liver AQP3 is also widely present, however its abundance decreased with fibrosis progression, and was also dependent on PNPLA3 mutation. This was in line with another previous report 16 in which AQPs expression decreased in liver of ethanol/lipopolysaccharide fibrotic animal model.
A study from our group recently demonstrated an up-regulation of the JNK/AP-1 pathway in PNPLA3 I148M cells, which could be partially counteracted by rosiglitazone treatment 8 . Notably, PPARγ itself suppresses JNK activation via directly binding Jun D and downregulating AP-1 12 . Since others showed a down-regulation of AQP3 in keratinocytes after UV radiation, another JNK activator 40,41 and that PPARγ blocks JNK pathway through direct binding to Jun D 12 , we used a JNK inhibitor to explore the impact of this pathway on AQP3 regulation in HSC. Indeed, although others showed the involvement of PI3K/Akt/mTor pathway in AQPs regulation in adipocytes and hepatocytes 42 , in HSC this appears to occur through JNK. Moreover, JNK inhibition also increased the amounts of vitamin A and lipids in the cells, indicating that reduction of AQP3 may disrupt glycerol metabolism in HSC.
PPARs were shown to be promising targets for treatment of fibrosis; in details, a PPARγ agonist like pioglitazone has been shown to increase AQPs expression 21 , ameliorate insulin resistance inducing adipogenesis, and a local effect in HSC should not be underestimated as it renders them quiescent. Moreover, it was recently demonstrated pioglitazone antifibrotic effects in a cohort of human diabetics with NASH 43 . A long term treatment, up to 3 years, was valued safe, well tolerated with no significant drug-related side effects; pioglitazone treatment showed to improve liver steatosis, inflammation and ballooning scores, also ameliorating metabolic and histological parameters in NASH, pre-diabetic and type 2 diabetic patients 43 .
In conclusion, AQP3 is regulated in PNPLA3 I148M via the JNK pathway through a PPARγ mediated activation and associates to different degrees of fibrosis. Future studies are required to decipher whether AQPs are targetable within HSCs activation, in order to develop new treatments for liver fibrosis in PNPLA3 I148M patients.

Materials and Methods
Isolation and culture of primary human HSCs and LX-2. HSCs were isolated from surgical liver resections unsuitable for transplantation, all experimental protocols were approved by the Ethics Committee of Medical University of Florence 44 after informed consent from all subjects, and carried out in accordance with relevant guidelines and regulations. After mechanical digestion, a multi-step enzyme digestion with a collagenase/ pronase/DNAase 44 solution was performed. Hepatic cells were centrifuged and separated using a density gradient (Percoll, GE Amersham, Arlington Heights, IL). HSCs were seeded on uncoated plastic dishes and cultivated with Iscove's Modified Medium (Dulbecco's medium, EuroClone, Italy) supplemented with 20% non-heat inactivated fetal bovine serum, 0.2 mol/L glutamine, sodium pyruvate 0.1 mol/L, non-essential amino acid solution 100x, antibiotic, antimycotic solution 100x (Life Technologies, Carlsbad). After 5 days of culture the purity was ~95% as estimated by retinoid auto fluorescence. All experiments with primary HSCs were performed on cells from passage 1 to 8. Primary HSCs at the same passage were used for comparison experiments of WT and I148M PNPLA3. The genotype of each isolated HSC line has been analyzed by real-time PCR for the I148M SNP, as done routinely in our lab 8  Western Blotting. Approximately, 500.000 cells per condition were collected in RIPA buffer (Radio immune precipitation assay buffer, 0.01 mol/L Tris-Cl (pH 8.0), 0.001 mol/L EDTA, 5 × 10 −4 mol/L EGTA, 0.1% sodium deoxycholate, 0.1% SDS, 1% NP-40) and protein concentration was measured using 660 nm protein assay kit. Twenty micrograms per sample were loaded on a SDS-PAGE using 10% polyacrylamide gels. α-SMA was identified using a 1:2000 dilution of the monoclonal mouse anti-human α-SMA (Sigma-Aldrich), AQP3 was detected using a 1:500 dilution of monoclonal rabbit anti-AQP3 as performed before 47 . PNPLA3 was detected with a 1:1000 dilution of the rabbit polyclonal anti-PNPLA3 (Abcam, Cambridge, UK). Band intensity achieved from these antibodies was normalized to band intensity of Calnexin using mouse anti-calnexin 1:3000 (Santa Cruz Biotechnology Inc, Dallas, TX, USA). Flow cytometry. HSC (approximately 70.000 cells/well) were incubated with the primary antibody (Rabbit anti-AQP3, Sigma Aldrich), diluted 1:500 in blocking solution, for 45 mins at RT, followed by washing and incubation with a fluorochrome-labelled secondary antibody (Alexa Fluor 594 -goat anti rabbit, Thermo Scientific) diluted 1:500 in the blocking solution for 1 h in the dark. Cells were washed twice in PBS and prepared for flow cytometric analysis. Retinol amount contained in LX2 cells was analysed exploiting its intrinsic autofluorescence with an 351 nanometers excitation as seen 48 . Flow cytometry was performed with BD FACSCanto ™ II and BD FACSDiva ™ software (Becton Dickinson New Jersey, USA).
Immunohistochemistry. Healthy specimens were collected after liver resection of colorectal metastasis (n = 4), whereas fibrosis samples (n = 10) were obtained by percutaneous liver biopsy (informed consent was obtained from all subjects); the study and all experimental protocols were approved by the local ethics committee (EK747/2011) and carried out in accordance with relevant guidelines and regulations, samples were evaluated by a board certified pathologist. Formalin-fixed human liver slides were de-paraffinized and prepared for hematoxylin & eosin staining. For the immunofluorescence, slides were blocked for 1 h in blocking buffer (1XPBS, 5% goat serum and 0.3% Triton ™ X-100) as seen previously 49 . Blocking buffer was discarded and sections incubated overnight with polyclonal rabbit anti-human AQP3 antibody (Sigma-Aldrich) diluted 1:250 in PBS with 5% goat serum (Dako, Glostrup Municipality, Denmark) at 4 °C. Slides were washed three times in PBS and incubated for 1 h at room temperature with a 1:500 dilution of monoclonal mouse anti-human α-SMA antibody (Sigma-Aldrich). Slides were then washed and incubated for 1 h in the darkness with the secondary antibody Alexa Fluor 594 goat anti rabbit IgG (1:500, Thermo Scientific) for AQP3. Slides were washed three times in PBS and the process was repeated with Alexa Fluor 488 goat anti mouse IgG (1:500, Thermo Scientific) for α-SMA.
Thereafter, nuclei were counterstained with DAPI (Sigma) for 10 min, washed and mounted (VECTASHIELD ® Mounting medium) for microscope analysis (Olympus BX51). The relative amount of AQP3 + cells was quantified and divided by the amount of nuclei per field. A number of 5 pictures were taken for each liver slide, and the average was representative for one patient as already performed in 49 . Statistics. Data are presented as mean ± standard deviation of 3 independent experiments performed in duplicates. Kruskal-Wallis test was employed for non-parametric multi-group comparisons, Mann-Whitney U test for comparisons between two groups only. A P-value < 0.05 was considered statistically significant. All statistics were calculated using SPSS 22.0 software (Chicago, IL, USA). Data availability. All data generated or analysed during this study are included in this published article (see supplemmatry material for original Western blots).