Celecoxib reduces hepatic vascular resistance in portal hypertension by amelioration of endothelial oxidative stress

Abstract The balance between endothelial nitric oxide (NO) synthase (eNOS) activation and production of reactive oxygen species (ROS) is very important for NO homeostasis in liver sinusoidal endothelial cells (LSECs). Overexpression of cyclooxygenase‐2 (COX‐2), a major intravascular source of ROS production, has been observed in LSECs of cirrhotic liver. However, the links between low NO bioavailability and COX‐2 overexpression in LSECs are unknown. This study has confirmed the link between low NO bioavailability and COX‐2 overexpression by COX‐2‐dependent PGE2‐EP2‐ERK1/2‐NOX1/NOX4 signalling pathway in LSECs in vivo and in vitro. In addition, the regulation of COX‐2‐independent LKB1‐AMPK‐NRF2‐HO‐1 signalling pathway on NO homeostasis in LSECs was also elucidated. The combinative effects of celecoxib on diminishment of ROS via COX‐2‐dependent and COX‐2‐independent signalling pathways greatly decreased NO scavenging. As a result, LSECs capillarisation was reduced, and endothelial dysfunction was corrected. Furthermore, portal hypertension of cirrhotic liver was ameliorated with substantial decreasing hepatic vascular resistance and great increase of portal blood flow. With the advance understanding of the mechanisms of LSECs protection, celecoxib may serve as a potential therapeutic candidate for patients with cirrhotic portal hypertension.

Nitric oxide (NO) is the natural ligand for soluble guanylate cyclase and is responsible for an increase in cyclic guanosine monophosphate (cGMP), the final agent responsible for the relaxation of the vascular wall through the extrusion of cytosolic Ca 2+ .
Endothelial-derived NO, synthesised from L-arginine by endothelial NO synthase (eNOS), plays a crucial role in maintaining endothelial homeostasis and vascular tone under physiological conditions. 7 In the cirrhotic liver, there is a reduced NO bioavailability that plays a major role in the activation of peri-sinusoidal hepatic stellate cells (HSCs), LSECs capillarisation, increasing intra-hepatic vascular resistance and thereby worsening PHT. 8,9 Dysfunction of LSECs not only increases intra-hepatic resistance and portal pressure (PP) but also remodels liver sinusoid. The balance between eNOS activation and the production of reactive oxygen species (ROS) is very important for NO homeostasis in LSECs. 10  Cyclooxygenase (COX) is a key rate-limiting enzyme that can catalyse the conversion of arachidonic acid (AA) to prostanoids. 12 Among its isoforms, the constitutive COX-1 is involved primarily in housekeeping functions, while the inducible COX-2 modulates various inflammatory responses. Besides low NO bioavailability in LSECs, COX-2 overexpression in cirrhotic livers has been observed in our previous studies. 13,14 However, there is less study to link them.
Interestingly, COX-2 is considered as a major intravascular source of ROS production. 15 As a vasoconstrictor, COX-derived prostanoids may be involved in HVR of cirrhosis through the ROS-NO signalling pathway. 16,17 COX-2 inhibitors exhibit anti-oxidative effects in hypertension, cardiac hypertrophy, 18 hepatic ischaemia-reperfusion injury, 19 and sepsis. 20 Furthermore, our previous studies have shown that celecoxib, a selective COX-2 inhibitor, could prevent PHT by inhibiting intra-hepatic angiogenesis, 13,21 suggesting its potential protective effects on endothelial function. On the basis of that fragmentary information, we hypothesised that there may be some links between low NO bioavailability and COX-2 overexpression in LSECs.
This study was aimed to elucidate the unclear molecular pathways related to low NO bioavailability, ROS production and COX-2 overexpression in LSECs of cirrhotic livers and determine the effects of celecoxib on the regulation of oxidative and anti-oxidative processes of LSECs in PHT.

| Animals and experimental designs
Male Sprague-Dawley rats, weighing 180-220 g, were obtained from the Experimental Animal Center of Sichuan University. They were randomly assigned to the control, thioacetamide (TAA), and TAA +celecoxib groups with 12 animals in each group. The fibrosiscirrhosis rat model was established by intraperitoneal injection of TAA (Sigma-Aldrich) 200 mg/kg in normal saline every 3 days for 16 weeks ( Figure S1). The rats in the control group were injected with an equal volume of normal saline. During the last 8 weeks, the rats in TAA +celecoxib group were fed celecoxib (20 mg/kg/day, Pfizer) by gastric gavage. The rats in TAA or control group were fed their vehicle (distilled water) in the same way. All of the rats were anaesthetised with pentobarbital (50 mg/kg intraperitoneally, Sigma-Aldrich) 72 hours after the last TAA injection. After hemodynamic evaluation, they were sacrificed via anaesthesia overdose.
Then, the liver samples were harvested. The animal procedures were approved by the Animal Use and Care Committee of West China Hospital, Sichuan University (No. 2017005A).

| Histological examination and hydroxyproline measurement
Liver tissues fixed in 4% paraformaldehyde (PFA) were embedded in paraffin, sectioned (5 μm thickness), and stained with hematoxylin and eosin (H&E) and Sirius Red. Five randomised visual fields per section (at ×100 magnification) were selected and assessed by two experienced pathologists independently blinded to the experimental grouping. The degrees of fibrosis were evaluated by the Ishak scoring system 22 and the percentage of fibrotic area. Additionally, hepatic hydroxyproline, a major component of collagen, was measured in 400-500 mg liver samples using a commercial kit (Nanjing Jiancheng Bioengineering Institute), according to the manufacturer's instructions.

| In vivo hemodynamic evaluation
Hemodynamic evaluation was performed under anaesthesia after overnight fasting, as described previously. 23 Briefly, the femoral artery and portal vein were cannulated with 24G catheters to measure mean arterial pressure (MAP) and heart rate (HR), and PP, respectively. Simultaneously, a perivascular ultrasonic transit-time flow probe connected to a flow metre (Transonic Systems) was placed around the portal vein as close as possible to the liver to measure portal blood flow (PBF). Blood pressures and flows were recorded on a multichannel computer-based recorder (Chengdu Techman Software Co., Ltd). HVR was calculated as PP/PBF.

| Scanning electron microscope
After hemodynamic measurement, the livers were perfused in situ with normal saline, followed by a fixation solution of 4% PFA containing 2.5% glutaraldehyde. Then the fixed tissues were cut into short columns, snap-frozen in liquid nitrogen, and drubbed from the middle portion. After drying in a critical point apparatus, the samples were coated with platinum and observed under a scanning electron microscope (JEOL).

| Measurement of cGMP levels
cGMP, a marker of NO bioavailability, was determined in liver homogenates using an enzyme immunoassay (NewEast Biosciences) according to the manufacturer's instructions. Total protein concentration was quantified by the BCA assay (Beyotime Institute of Biotechnology), and cGMP levels were normalised to total protein concentration.

| Measurement of hepatic O − 2 contents
Fresh liver tissues were embedded in optimum cutting temperature compound (Sakura Finetek), frozen with liquid nitrogen, and sectioned into 10 μm thick slices. In situ O − 2 levels were evaluated with an oxidative fluorescent dye dihydroethidium (DHE, 10 μM, Nanjing KeyGen Biotech Co., Ltd). Five visual fields per section (at ×100 magnification) were randomly selected and quantitatively analysed using Image-Pro Plus 6.0.

| Targeted metabolomic analysis of AA
Major metabolites of AA through COX-2 pathway in liver homogenates were analysed by the Shanghai Institute of Nutrition and Health, Chinese Academy of Sciences, according to the published protocols using liquid chromatography-mass spectrometry (LC-MS). 24 The radioimmunoassay for serum prostaglandin E2 (PGE2) was performed by the Beijing Sino-UK Institute of Biological Technology.

| Immunohistochemistry staining
Paraffin-embedded liver sections (3 μm thickness) were deparaffinised in xylene and rehydrated with graded ethanol dilutions. Antigen retrieval was performed at high temperature under high pressure in sodium citrate buffer (10 mM, pH = 6.0) for 15 minutes. After blocking with H 2 O 2 and 10% goat serum, the sections were incubated with primary antibodies overnight at 4°C followed by incubation with biotin-streptavidin-horseradish peroxidase (HRP) detection system (ZSGB-BIO) at room temperature. Finally, the sections were stained with a solution of 3,3′-diaminobenzidine (DAB, ZSGB-BIO) and counterstained with hematoxylin. All primary antibodies used were listed in Table S1.

| Quantitative real-time PCR
Total RNA was extracted from the frozen liver tissues with RNAiso Plus (Takara), and cDNA was synthesised from total RNA (1 μg) using the RevertAid ™ First Strand cDNA Synthesis Kit (Thermo Fisher Scientific). Then, quantitative PCR was performed in triplicate using SYBR Green qPCR Master Mix (Bimake) on the CFX96 Real-Time PCR Detection System (Bio-Rad). The mRNA expression was normalised against glyceraldehyde-3-phosphate dehydrogenase (GAPDH) and was shown as fold changes relative to the control group. The primer sequences were listed in Table S2.

| Western blot analysis
Whole proteins from liver tissues or cultured cells were extracted using a protein extraction kit (Nanjing KeyGen Biotech Co., Ltd). Equal amounts of proteins for each sample were resolved by 10% or 12% SDS-PAGE, transferred to PVDF membrane (Merck Millipore), and blocked with 5% non-fat dry milk before being incubated with primary antibodies overnight at 4°C. Then, the blots were washed and incubated with HRP-conjugated secondary antibodies (1:20000, ZSGB-BIO) at room temperature for 2 hours. Protein bands were visualised by chemiluminescence using Western Blotting Luminol Reagent (Santa Cruz Biotechnology), and densitometric analyses were performed using Quantity One 4.6.2 (Bio-Rad). Protein levels were normalised against GAPDH and were shown as fold changes relative to the control group. All primary antibodies used were listed in Table S1.

| Cell culture and treatments
Freshly isolated LSECs were unstable in our preliminary experiments due to spontaneous capillarisation and short survival. SK-Hep1, an immortalised human LSECs line, can maintain endothelial phenotype persistently without reagent induction. 25
The serum-free medium was subsequently replaced with a complete fresh medium after 6 hours of transfection and cultured for additional 48 hours. Overexpression of COX-2 was verified by qRT-PCR and Western blot in mRNA and protein levels, respectively.

| Statistical analysis
All data were expressed as mean ±standard deviation (SD) and analysed using SPSS 19.0 software (SPSS). For multiple-group comparison, oneway ANOVA followed by Students-Newman-Keuls (SNK) post hoc test was performed. A p-value <0.05 was considered statistically significant.

| Celecoxib attenuated liver cirrhosis by inhibition of HSCs activation
As expected, cirrhotic livers with extensive collagen deposition and pseudolobule formation were harvested from TAA group ( Figure 1A). Impressively, celecoxib attenuated the cirrhotic lesions with significant lower Ishak score (3.73 ± 0.30 vs. 5.60 ± 0.57, p < 0.05) and fibrotic area Although body weight of each animal group was gradually increased during the experimental period. TAA-cirrhotic rats had lower body weight gain than control rats, whereas a significant improvement was observed after celecoxib treatment (Table S3). There was no significant difference in serum TB, ALT, AST, ALB, UREA, and CREA among the three groups (Table S4), suggesting that the dose of TAA or celecoxib used in present experiment had less adverse effects on liver and renal functions.

| Celecoxib significantly decreased PP and HVR
Cirrhotic PHT was developed in all of the rats exposed to TAA for 16 weeks ( Figure S1D). PP of rats treated with celecoxib was decreased by 22 (Table 1). In addition, the spleen index (spleen weight/rat weight) was also decreased with celecoxib treatment, further verifying the alleviation of PHT (Table S3).

| Celecoxib arrested LSECs capillarisation by increasing NO bioavailability
LSECs in TAA-cirrhotic rats underwent capillarisation with an obvious decline in the numbers of fenestrae (Figure 2A). Celecoxib attenuated the capillarisation induced by TAA (Figure 2A). Hepatic cGMP level, a marker of NO bioavailability, was significantly increased in TAA +celecoxib group compared with TAA group, p < 0.05 ( Figure 2B).   Figure 3A). After treatment with celecoxib, DHE staining was significantly diminished ( Figure 3A). Quantitatively, fluorescence intensity in TAA +celecoxib group was decreased by 65% compared with that in TAA group (p < 0.05, Figure 3B). Consistently, celecoxib administration resulted in a significant reduction in hydroxyl radical and hydrogen peroxide levels (p < 0.05, Figure S2A

| Celecoxib restored hepatic antioxidative capacity via activation of the NRF2-HO-1 signalling pathway
Given that NRF2-KEAP1 complex is the key regulator of antioxidative signalling, we determined whether celecoxib diminished superoxide radicals generation through activation of the NRF2-KEAP1 pathway ( Figure 4A-F). HO-1, a downstream effector molecule of NRF2, was also determined in mRNA and protein levels.
As expected, the down-regulation of NRF2 and HO-1 expression in TAA group was significantly inhibited by celecoxib treatment, p < 0.05 ( Figure 4A-D, F). The data from Westen blot showed that there was no significant difference of KEAP1 expression among the three groups ( Figure 4A Figure S2C). Collectively, these findings illustrated that celecoxib

| Celecoxib inhibited the COX-2-PGE2-EP2 signalling pathway
COX-2 expression in TAA-cirrhotic livers was significantly increased compared to control rats, p < 0.05 ( Figure 5A, B). Celecoxib treatment significantly decreased COX-2 expression in the liver, p < 0.05 ( Figure 5A, B). Meanwhile, COX-2-derived targeted metabolomic analysis showed that PGE2 levels in both liver and serum were remarkably decreased after celecoxib treatment, p < 0.05 ( Figure 5C, D), while other prostaglandins levels were kept normal (data not shown). Consistently, EP2 expression was down-regulated in the livers of TAA +celecoxib group compared to that in TAA group, p < 0.05 ( Figure 5E), indicating a direct inhibition of the COX-2-PGE2-EP2 signalling pathway.

| COX-2 promoted ROS production through the PGE2-EP2-ERK1/2 signalling pathway involving upregulation of NOX1 and NOX4
To further validate the link between COX-2 and NOX-derived ROS,

| Celecoxib induced endothelial HO-1 via COX-2-independent activation of the LKB1-AMPK-NRF2 signalling pathway
Both AMPK and NRF2 pathways are involved in the defence against oxidative stress. Because AMPKα and NRF2-HO-1 could be restored by celecoxib treatment in vivo, the effect of celecoxib on this pathway was also determined with SK-Hep1 cells in vitro. HO-1 protein levels were elevated by celecoxib in a concentration-dependent manner, p < 0.05 ( Figure 7A). When celecoxib was instead by DMC, With regarding to AMPK signalling pathway involved in the regulation of HO-1 induction, it was supported by the following results.
First, AMPKα phosphorylation and NRF2 expression in SK-Hep1 cells were increased with celecoxib treatment in a concentrationdependent manner, p < 0.05 ( Figure 7C). Next, AMPKα activation was enhanced 3-5 folds by celecoxib treatment for 60 minutes or 120 minutes, whereas NFR2 and HO-1 protein levels were not altered ( Figure S4). Moreover, LKB1, the upstream activator of AMPK, was also activated by celecoxib in a concentration-and time-dependent manner, as evidenced by phosphorylation at Ser 428 , p < 0.05 ( Figure 7C and Figure S4). These findings suggested that LKB1-AMPK pathway may modulate NRF2-HO-1 expression. To verify this hypothesis, SK-Hep1 cells were transfected with AMPKα-siRNA or NRF2-siRNA for 48 hours, followed by induction with celecoxib (10 μM) for an additional 24 hours. Up-regulation of HO-1 or NRF2 expression was suppressed by silencing AMPKα before celecoxib treatment, p < 0.05 ( Figure 7D). Similarly, the increased HO-1 expression was significantly inhibited by siRNA-mediated knockdown of NRF2, p < 0.05 ( Figure 7E). Consistent with this, we also observed similar results in DMC-induced HO-1 expression ( Figure S5), confirming the COX-2 independency of celecoxib on the LKB1-AMPK-NRF2-HO-1 signalling pathway.

| DISCUSS ION
Over the past 30 years, pharmacological therapy for PHT is consisting mainly of the continued oral administration of non-selective beta-blockers (propranolol, nadolol, carvedilol) for the prevention of first or recurrent variceal bleeding, and on the short-term intravenous infusion of terlipressin, somatostatin, or somatostatin analogues for acute variceal bleeding. 6 These treatments are aimed at correcting the increased splanchnic blood flow. It is only recently that this strategy has been changed. Progress of the modulation of intra-hepatic endothelial dysfunction has opened entirely new perspectives for developing more effective treatment. 2 NO-cGMP is crucial regulators for LSECs to maintain the fenestration, HSCs quiescence and portal pressure. 27 Both insufficient syntheses of NO and increased levels of superoxide and peroxynitrite have been well documented. 8 The present study firstly established the elaborate links between low NO bioavailability and COX-2 overexpression in LSECs. The combinative effects of celecoxib on diminishment of ROS decreased NO scavenging. As a result, LSECs capillarisation was reduced, losing fenestrae was protected, and endothelial dysfunction was corrected. Furthermore, PHT of cirrhotic liver was ameliorated with substantial decreasing HVR and great increase of portal blood flow. Figure 7F  In accordance with NO pathophysiological abnormalities, a drug that would be able to increase NO bioavailability in the liver would be hopeful to arrest liver fibrosis and decrease PP. Several efforts to improve NO bioavailability within the liver have been attempted including exogenous administration of NO donors, 28,29 transfecting the liver with adenovirus encoding eNOS, 30 enhancing eNOS activity by simvastatin, 31 and decreasing NO scavenging by means of antioxidants, such as metformin. 32 Actually, livertargeted NO donor drugs that mimic the cellular effects of endogenous NO and specifically release in the liver have been evaluated in the treatment of PHT, 28,29 whereas they failed to reduce PP in clinical trials. 33 Simvastatin administration significantly decreased hepatic venous pressure gradient but did not reduce HVR. 34 Therefore, addition of simvastatin to carvedilol for 3 months for primary prophylaxis of variceal bleeding did not improve hemodynamic response over carvedilol monotherapy. 35 Metformin caused a significant reduction in liver fibrosis and PP but no enhancement of portal blood flow. 32 The ideal drug for PHT is pictured as one that should reduce PP by decreasing intra-hepatic vascular resistance, while maintaining or enhancing hepatic blood flow. 36 By this picture, celecoxib should be quite satisfactory compared with other drugs. It significantly reduced HVR and increased portal blood flow in PTH by increasing eNOS-derived NO synthesis and reducing ROS-mediated NO scavenging in this study. The dual function of celecoxib was obviously superior to simply NO supplement or scavenging. the design of a study that continuous administration of celecoxib through a subcutaneous osmotic pump kept hepatic PGE2 levels below the physiological conditions. 42 As PGE2 has been proposed to mediate both profibrotic and antifibrotic effects of COX-2 inhibition. 43,44 In contrast, once-daily oral gavage of celecoxib in our study merely suppressed the elevated COX-2-derived PGE2 synthesis in cirrhotic livers.
COX-2 selective inhibitors (coxibs) have been widely used in the treatment of chronic inflammatory diseases. The long-term safety of celecoxib is still of concern. This experiment indicated that administration of celecoxib at dose of 20 mg/kg/d for 8 weeks did not show adverse effects on liver and renal functions. With regarding to the risk of cardiovascular events, increasing data indicate the clinical safety of celecoxib. The PRECISION study including 24,081 patients revealed that celecoxib at moderate doses was non-inferior to ibuprofen or naproxen in cardiovascular safety. 45 Furthermore, celecoxib therapy has generated significant clinical benefits for endothelial function improvement in patients with coronary artery disease and hypertension. 46,47 In addition, preliminary clinical evidence has shown that celecoxib does not increase the risk of variceal and non-variceal upper gastrointestinal events, 48 nor impair renal function in decompensated cirrhosis. 49 Although celecoxib has favourable ameliorating effects against liver fibrosis and PHT, further clinical trials should be performed to evaluate its efficacy and safety.
In conclusion, celecoxib ameliorated PHT of cirrhotic liver by substantial decreasing HVR on the basis of a great increase of portal blood flow. Protection of LSECs from capillarisation and endothelial dysfunction by celecoxib was contributed to the reduction of HVR. The function of LSECs was improved by enhancing their NO bioavailability, which was the counterbalance of pro-oxidative and anti-oxidative machinery through both COX-2-dependent and COX-2-independent pathways. With the advance understanding of the mechanisms of LSECs protection, celecoxib may be a potential therapeutic candidate for patients with cirrhotic PHT.

CO N FLI C T O F I NTE R E S T
All authors declare no conflict of interest.

DATA AVA I L A B I L I T Y S TAT E M E N T
All the data are present in the manuscript. All the data are available from the corresponding author Jinhang Gao (gjh731@foxmail.com) under reasonable request.