Nutritional status modifies pregnane X receptor regulated transcriptome

Pregnane X receptor (PXR) regulates glucose and lipid metabolism, but little is known of the nutritional regulation of PXR function. We investigated the genome wide effects of the nutritional status on the PXR mediated gene regulation in the liver. Mice were treated with a PXR ligand pregnenolone 16α-carbonitrile (PCN) for 4 days and subsequently either fasted for 5 hours or after 4-hour fast treated with intragastric glucose 1 hour before sample collection. Gene expression microarray study indicated that PCN both induced and repressed much higher number of genes in the glucose fed mice and the induction of multiple well-established PXR target genes was potentiated by glucose. A subset of genes, including bile acid synthesis gene Cyp8b1, responded in an opposite direction during fasting and after glucose feeding. PXR knockout abolished these effects. In agreement with the Cyp8b1 regulation, PCN also modified the bile acid composition in the glucose fed mice. Contribution of glucose, insulin and glucagon on the observed nutritional effects was investigated in primary hepatocytes. However, only mild impact on PXR function was observed. These results show that nutritional status modifies the PXR regulated transcriptome both qualitatively and quantitatively and reveal a complex crosstalk between PXR and energy homeostasis.

genes (genes upregulated at least 1.5-fold), only 67 were common between the fasting and the glucose treatment conditions indicating significant influence of the nutritional status on PXR target gene preference ( Fig. 1a and Supplementary Table S1). PCN also repressed more genes in the livers of the glucose fed mice than in the fasting mice. During fasting only 50 genes were repressed at least 1.5-fold, while after glucose treatment 368 genes were repressed to a similar degree ( Fig. 1b and Supplementary Table S2). 18 genes were repressed at least 1.5-fold both in the fasting and after the glucose treatment ( Fig. 1b and Supplementary Table S2).
To explore the differences in the PCN mediated gene regulation during fasting and after glucose feeding we performed Ingenuity pathway analysis (IPA). In the fasted mice the top canonical pathways affected (based on P-value < 0.05 (i.e., −log10 ≥ 1.3), Fisher's exact test) were mainly related to metabolism of xenobiotics as well as several metabolic pathways catalyzing synthesis or catabolism of hormones or other endobiotics (Fig. 1c). Typical for these pathways is the involvement of cytochrome P450 (CYP) enzymes. In addition, some inflammation related pathways such as "LPS/IL-1 mediated inhibition of RXR function" and "Acute phase response signaling" are among the top 20 canonical pathways (Fig. 1c).
The top 20 pathways were quite similar also in the glucose fed mice (Fig. 1d). However, the top pathways included also lipid and bile acid related pathways such as "FXR/RXR Activation" and "LXR/RXR Activation" and a pathway related to energy homeostasis: "Sirtuin signaling Pathway". Furthermore, "NRF2-mediated Oxidative Stress Response" pathway was significantly affected suggesting increase in the cellular oxidative stress.
To further evaluate the effect of the nutritional status on PCN prompted canonical pathways we compared the activation z-scores between fasting and glucose fed mice (Fig. 2). This analysis revealed clearly that PCN activates many more functional pathways in the glucose fed mice livers than in the fasting mice. Furthermore, several of these additional pathways ("Glycolysis I", "Gluconeogenesis I", "Stearate biosynthesis" and "Cholesterol biosynthesis" (several different)) are related to glucose and lipid metabolism (Fig. 2, Supplementary Table S3). Although the gene expression data suggest activation of cholesterol synthesis, the plasma cholesterol level was not affected by PCN either in the fasting or the glucose fed mice (Supplementary Fig. S1).

Nutritional status modifies PCN mediated gene regulation in a target gene specific manner.
Analysis of the top PCN-upregulated genes in the fasting mice indicated the expected profile of many genes coding for drug metabolizing enzymes (Fig. 3a). Interestingly, many of the top upregulated genes in the fasting mice were induced much more potently by the PCN treatment in the mice treated with glucose (Fig. 3a). The top four upregulated genes were all induced 1.5 to 2 times more potently in the livers of the glucose treated mice compared to the fasting mice. We also analyzed the top PCN-upregulated genes in the glucose treated mice. The list of the top 10 upregulated genes is rather similar to the fasting animals and include phase 1 and phase 2 drug metabolizing enzymes, however, the fold changes are all higher in the glucose fed mice (Fig. 3b).
Since the glucose feeding appeared to modify the PCN response of individual genes, we next filtered the gene list to find the genes, for which the PCN response was the most strongly affected by the glucose feeding. The top 20 genes with the highest ratio of the PCN induction fold in the glucose fed and the fasting mice are presented in Table 1. The genes with the highest ratio contained a few genes that were clearly induced by PCN in both conditions such as Cyp3a11 and Gstm2; however, many genes in this list are either nonresponding or repressed during fasting but induced after glucose treatment. The gene with the highest ratio was Cyp8b1, one of the key genes in the bile acid synthesis. Cyp8b1 was repressed about 50% in the fasting mice but induced almost 5-fold in the glucose fed mice (Table 1).
To determinate the effect of glucose feeding on PXR function more precisely, selected genes with different response types in the gene expression profiling analysis were further investigated with QPCR using larger animal groups. Three categories were identified: (1) genes, for which the PCN induction was potentiated by PCN, (2) genes, which were induced similarly by PCN after fasting and glucose treatment, (3) genes regulated in opposite directions by PCN after fasting and glucose feeding.
According to the microarray analysis, Cyp3a11, Cyp2b10 and Cyp2a5 were all induced by PCN during fasting, but glucose potentiated the PCN effect (Fig. 2.). Indeed, in the QPCR analysis, Cyp3a11 and Cyp2b10 were both   www.nature.com/scientificreports www.nature.com/scientificreports/ induced by PCN in the fasting condition by 3-and 25-fold, respectively. However, both genes were induced much more potently after glucose feeding. Cyp3a11 was induced 13-fold after glucose treatment, i.e. glucose potentiated the PCN mediated induction about four-fold (P < 0.01) (Fig. 4a). Also, for Cyp2b10, the glucose treatment quadrupled the PCN response; however, the difference in the PCN induction between the glucose fed mice and the fasting mice was not statistically significant (Fig. 4a). In contrast with the gene expression array result, the Cyp2a5 expression was not affected by PCN in the fasting condition when measured with QPCR in larger animal groups. However, after glucose feeding PCN induced Cyp2a5 expression 2.7-fold (Fig. 4a).
According to the gene expression profiling analysis, the PCN mediated induction of Cyp3a13 and 5'-aminolevulinate synthase 1 (Alas1) was not affected by the nutritional state (Supplementary Table S1). This was supported also by the QPCR analysis of the bigger sample size. Both genes were induced by PCN about 2-fold both during fasting and after glucose feeding. However, the PCN response was statistically significant only after glucose, apparently because of the smaller variation in the mice treated with glucose ( Fig. 4a).
Cyp8b1, Cyp4a12a, Cyp4a12b and oxidative stress induced growth inhibitor 1 (Osgin1) were among a third group of genes. In the gene expression array, they were all repressed in the fasted mice (Supplementary Table S2). However, they were induced in the livers of the glucose treated mice and were among the genes for which the PCN response was the most affected by the glucose administration (Table 1, Supplementary Table S1). In the QPCR analysis, they appeared to be repressed by PCN during fasting, although this decrease was not statistically significant. During glucose feeding Cyp8b1 was significantly induced by PCN (P < 0.05), while for the other genes there was a trend for upregulation, but no statistically significant effect. Nevertheless, for all these genes, glucose feeding significantly changed the PCN response from downregulation to upregulation (Fig. 4a).
As a control, we measured expression of two glucose metabolism related genes previously reported to be repressed by PXR activation 5,6 . As expected, both glucose-6-phosphatase (G6pc) and phosphoenolpyruvate carboxykinase (Pck1) tended to be repressed by PCN in the fasting mice (Fig. 4a). Glucose treatment had significant effect on Pck1 response to PCN, but not on that of G6pc (Fig. 4a). Altogether, the QPCR results confirmed the overall gene expression patterns observed in the gene expression profiling analysis and indicate that nutritional status modifies PCN mediated gene regulation in a target gene specific manner.

PXR is indispensable for the PCN induced gene regulation and the nutritional status dependent modifications.
To establish the involvement of PXR in the observed gene regulation phenomena by PCN and glucose feeding we utilized PXR knockout mice. The Pxr −/− mice were subjected to identical experiment with four-day PCN treatment and subsequent fasting or glucose feeding as describe above. The PXR knockout abolished all the regulatory effects observed in the wildtype mice indicating that PXR is crucial for PCN mediated gene regulation as well as for the glucose induced alteration of the responses (Fig. 4b).
Effect of glucose, insulin and glucagon on PXR mediated target gene response in mouse primary hepatocytes. In the current animal experiment, we used glucose feeding as the nutritional model. We therefore hypothesized that the increased postprandial glucose level could have a direct effect on PXR mediated gene www.nature.com/scientificreports www.nature.com/scientificreports/ regulation. To test this hypothesis, we utilized cultured mouse primary hepatocytes and investigated the effect of glucose concentration on PXR mediated gene regulation.
Cyp3a11 was induced efficiently by PCN in the cultured hepatocytes, and the induction magnitude was comparable to the one in the liver in vivo. In contrast, some other genes like Cyp2b10 were induced poorly by PCN in the cultured hepatocytes compared with the liver in vivo (Fig. 5). High glucose level tended to increase Cyp3a11 induction by PCN, from 13-fold induction with in the cells with no glucose to 16-fold induction in the cells with 25 and 50 mM glucose in the media; however, the effect was rather modest and was not statistically significant (Fig. 5). The glucose itself did not influence Cyp3a11 expression (Fig. 5). We also tested the effect of glucose feeding alone on Cyp3a11 expression in the liver in a separate experiment in mice. However, there was no effect ( Supplementary Fig. S3). For the other genes studied, glucose level did not have any consistent, dose responsive effect on the PCN mediated regulation in the mouse primary hepatocytes. These results suggest that glucose level itself does not play a major role in the observed feeding effect on PXR mediated gene regulation in the liver. www.nature.com/scientificreports www.nature.com/scientificreports/ Feeding induces multiple systemic effects that could modify PXR function. Fasting to feeding transition greatly changes balance of glucagon and insulin, two hormones with major effects on liver functions. We therefore studied the effect of these pancreatic hormones on PCN mediated gene regulation in primary hepatocytes. Insulin treatment increased Cyp2a5 and Cyp2b10 induction dose dependently (Fig. 6). Cyp3a11 induction was increased 2.4-fold with 50 nM insulin compared to no insulin in the culture media (Fig. 6). However, further upregulation of insulin had negative effect on induction.
Glucagon had the clearest effect on Cyp2b10 induction that was dose dependently potentiated until 25 nM concentration (Fig. 6). Also, Cyp2a5 induction was increased by 25 nM glucagon (Fig. 6). Otherwise, glucagon had no effect on PCN mediated induction of the studied genes.
Effect of PCN on bile acid profile in glucose treated mice. Glucose feeding had dramatic effect on regulation of Cyp8b1 by PXR (Fig. 4a, Table 1). While PCN repressed Cyp8b1 in the fasting mice, the effect was transformed into induction after glucose feeding. Regulation of other genes coding for the key enzymes and transporters involved in liver bile acid homeostasis was explored based on the gene expression array data. Cyp7a1 and Cyp7b1 catalyzing key steps in the initiation of the classical and alternative bile acid synthesis pathways were systematically repressed both during fasting and after glucose treatment, although the Cyp7a1 repression during the fasting and Cyp7b1 repression after glucose did not reach the 1.5-fold threshold (Supplementary Table S4. Cyp27a1 was induced only after glucose treatment (Supplementary Table S4). Among the bile acid transporters, ATP binding cassette subfamily C member 3 (Abcc3) was induced both during fasting and after glucose, however, the induction was about 2 times more powerful after glucose treatment. Abcc2 was induced only after glucose treatment (Supplementary Table S4). Altogether these data indicate that nutritional status modifies the response of the bile acid homeostasis related genes to PCN. Nevertheless, the most prominent change is the reversed regulation of the Cyp8b1 during fasting and after glucose feeding. CYP8B1 enzyme plays a major role in determination of the bile acid composition and functions as the branch enzyme directing synthesis towards cholic acid (CA) (Fig. 7a). Therefore, increased Cyp8b1 expression could be expected to increase synthesis of CA and decrease production of chenodeoxycholic acid (CDCA), which mice, unlike humans, further metabolize to αand β-muricholic acid (MCA) (Fig. 7a).
Plasma and gallbladder total bile acid levels were measured from mice treated with PCN for four days, fasted overnight and then were given glucose (2 g/kg) by oral gavage 1 h before sacrificing. There was a tendency for lower plasma levels of the total bile acids, however, the difference was not statistically significant. Instead, the gallbladder bile acids were significantly reduced by PCN supporting the finding that PCN treatment suppresses bile acid synthesis through repression of the Cyp7 enzymes (Fig. 7b). Furthermore, individual bile acids were www.nature.com/scientificreports www.nature.com/scientificreports/ measured from the plasma. CA plasma level was increased, however, there were large interindividual differences in the CA levels and the difference was not statistically significant. CDCA level was reduced significantly by PCN treatment (Fig. 7c). Tauro-α/β-MCA, the further metabolites of the CDCA, and the secondary bile acids deoxycholic acid (DCA), metabolized by the intestinal bacteria from CA, were also lower, however not significantly. Altogether, these results suggest that PCN treatment combined with glucose feeding suppresses the total bile acid production and directs bigger proportion of the synthesis to the CA pathway instead of the CDCA pathway.

Discussion
Since the early reports 15 years ago associating PXR with the regulation of glucose and lipid metabolism 6,16 , mounting evidence has been connecting PXR to different aspects of hepatic energy homeostasis 3,17 . However, in many ways the mechanisms are still unclear and discrepant results have been published by different groups using various models 3 . One potential confounding factor is the influence of the nutritional status on the PXR function.
In the present study we show that nutritional status has dramatic effects on PXR mediated gene regulation. Treatment of mice with oral glucose both qualitatively and quantitatively modified the PCN mediated transcriptome in the liver in comparison to the fasting condition. In the glucose treated mice, PCN induced six times and repressed seven times more genes than in the fasting mice. Less than half of the genes induced by PCN during fasting were affected after glucose treatment indicating significant diversity in the PCN responsive transcriptome www.nature.com/scientificreports www.nature.com/scientificreports/ in the different nutritional conditions. The liver samples were collected 1 hour after glucose treatment with an intention to investigate the fasting to feeding transition. Although the time point is rather early, it is expected to be sufficient for transcriptional changes 18 . The early time point decreases the chance to detect indirect PXR target genes. Nevertheless, the observed changes in the transcriptome most probably reflect changes both in transcription and in transcriptome stability and dynamics.
Both in the fasting and in the glucose treated mice livers, the top upregulated genes were mainly related to the metabolism of xenobiotics. Indeed, this represents the classical and the most well-established function under the PXR control 19 . In addition, the most strongly and significantly affected canonical pathways were mostly related www.nature.com/scientificreports www.nature.com/scientificreports/ to xenobiotic metabolism or other cellular pathways involving cytochrome P450 enzymes. In accordance with the higher number of genes regulated by PCN after glucose treatment, in the glucose fed mice a much higher number of canonical pathways were activated by PCN. Interestingly, many of these additional pathways were predominantly related to the glucose and lipid metabolism. Indeed, PXR is known to affect the glucose and lipid metabolism in liver 3 . These findings suggest that PXR function is adapting in the postprandial phase to the needs of the nutritional status.
In addition to the differences in the number and the type of genes affected by PCN in the different nutritional conditions, glucose treatment also modified the response of individual PXR target genes to the PCN treatment. At least three different categories of PXR target genes could be identified. Some genes, such as Cyp3a13 and Alas1, were induced similarly both in the fasted and the glucose fed animals. The second group, including the prototypic PXR target gene Cyp3a11, was induced in both conditions, however, the glucose feeding potentiated the induction. Third, for some PCN regulated genes the glucose treatment appeared to change the direction of the response. This response type included Cyp8b1 coding for one of the key bile acid synthesis genes. Experiment in PXR knockout mice confirmed that all these changes require PXR.
In agreement with the previous results, PXR activation by PCN decreased biliary bile acid level 20 . This is probably related to the repression of Cyp7a1 controlling the major bile acid synthesis pathway and, to lesser extent, repression of Cyp7b1 controlling the alternative bile acid synthesis pathway 21 . Furthermore, the sinusoidal efflux transporter Abcc3 was also induced which may facilitate bile acid transfer to plasma. The repression of Cyp7a1 and the induction of Abcc3 by PXR activation are in agreement with the previous studies 20,22,23 .
Curiously, the glucose feeding had a very strong effect on regulation of Cyp8b1. In the fasting mice, PCN tended to repress Cyp8b1 expression in agreement with previous studies 16,24 . However, in the glucose fed animals Cyp8b1 was induced. CYP8B1 is a branch enzyme directing bile acids synthesis towards cholic acid (CA) and therefore the higher CYP8B1 enzyme expression is expected to have opposite effects on CA and CDCA synthesis (Fig. 7a). Indeed, PCN treatment of the glucose fed mice decreased the CDCA level significantly. Furthermore, the CA level tended to be higher. These results suggest that, in addition to the total bile acid level, PXR activation in the postprandial state may affect the bile acid composition.
Bile acids function as part of the digestive system facilitating intestinal absorption of dietary fats and oils. Moreover, bile acids are important signaling molecules activating several cellular receptors, especially the nuclear farnesoid X receptor (FXR) and the takeda G protein-coupled receptor 5 (TGR5) and regulate many important, physiological functions such as carbohydrate and lipid metabolism and immune system 25 . Individual bile acids have different potencies to activate bile acid receptors. For example, FXR is activated by CDCA > DCA > CA, while T-α-and T-β-MCA are antagonists 25 . Thus, modulation of bile acid composition may significantly affect the bile acid regulated functions. Indeed, PXR is known to be involved in regulation of glucose and lipid homeostasis and inflammation, all functions affected by bile acids 3,4 .
Our current study indicates that nutritional status has a major modifying effect on PXR mediated gene regulation. However, the exact mechanisms involved remain to be determined in the future studies. A recent study reported that high-glucose induces CYP3A4 mRNA level and PXR regulated reporter gene activity in HepG2 cells 26 . CYP3A4 mRNA was induced by high-glucose also in the human intestinal LS180 cells, but very modestly and not significantly in the human primary hepatocytes 26 . Furthermore, glucose level has been reported to regulate PXR phosphorylation status 15 . In the present study, we did not observe any effect of glucose alone on Cyp3a11 expression in the mouse primary hepatocytes with concentrations up to 50 mM. Furthermore, glucose feeding alone did not influence Cyp3a11 expression in the mouse liver. 25-50 mM glucose potentiated slightly, but no significantly, PCN mediated induction of Cyp3a11 in cultured hepatocytes. For the other genes studied, glucose did not have any consistent effect on PCN response. These results suggest that postprandial increase in plasma glucose level is not sufficient to mediate directly the observed changes in the PXR function by glucose feeding.
Moreover, we studied the effects of insulin and glucagon on PCN induced gene regulation in primary hepatocytes. The most consistent effect was the ability of both insulin and glucagon to stimulate the induction of Cyp2a5 and Cyp2b10 by PCN. The mechanisms of these effects are currently unknown. However, insulin and glucagon regulate many hepatic transcription factors, such as FOXO1, SREPBP1 and PGC-1α, that are known to interact with PXR protein 6,12,16,27 .
In summary, our study provides a genome wide view to the effect of nutritional status on PXR mediated gene regulation. In response to glucose feeding the PXR regulates, directly or indirectly, much higher number and wider range of genes than during fasting. Many well-established PXR target genes are induced more powerfully by a PXR ligand after glucose feeding that during fasting, while some are regulated in a totally opposite manner during fasting and glucose feeding. These results indicate that nutritional status should be carefully controlled and reported in studies investigating PXR mediated gene regulation. Furthermore, these results provide further insight to the crosstalk of PXR and energy homeostasis.

Methods
Animal experiments. Eight-week-old wildtype C57BL/6 mice or Pxr −/− mice in C57BL/6 background were used in this study. The Pxr −/− mouse line was kindly provided by Dr. Wen Xie (University of Pittsburgh). The mice were treated i.p. with pregnenolone-16-α-carbonitrile (PCN, 50 mg/kg dissolved in DMSO-corn oil 1:3) or vehicle (DMSO-corn oil 1:3) once daily for four days. The mice were fasted for five hours before sacrificing or they were first fasted for four hours and then administered glucose (2 g/kg) by oral gavage one hour before sacrificing. The livers were collected and frozen in liquid nitrogen and used for total RNA extraction. The number of wildtype mice was 8 in the vehicle treated fasting group and 7 in all the other groups. Three randomly selected samples from each group was used for gene expression profiling with microarray, while all the samples were used for the QPCR measurements. The number of Pxr −/− mice was 4 in each group. Furthermore, the same experimental Gene expression profiling. Total RNA from mouse liver tissue was isolated by using hybrid RNAzol ® RT (Sigma Aldrich, St. Louis, MO, USA)/RNeasy kit (Qiagen, The Netherlands) protocol. Briefly, the tissue was homogenized into RNAzol RT solution and after mixing with RNase free water, the RNA containing supernatant was transferred to RNeasy column and the RNeasy protocol was then followed according to the manufacturer's instructions. The quality and integrity of the isolated RNA was monitored by The work was carried out at the Core Facility of the Estonian Genome Center, University of Tartu, Estonia. Probe intensity and detection data were obtained using Illumina BeadStudio software, and further processed with GeneSpring GX 14.5 software (Agilent Technologies, Santa Clara, CA, USA). The raw expression data was quality assessed, log2-transformed, and subjected to normalization to the 75 th percentile and baseline transformation to median of all samples. Genes were defined as differentially expressed if the fold change was at least 1.5-fold and statistically significant (P < 0.05, One-way ANOVA and Tukey's post hoc -test). The complete data sets are available from the NCBI's Gene Expression Omnibus (GEO) database and gene expression profiling data comply with the MIAME (Minimum Information About a Microarray Experiment) standard. The data can be obtained from the GEO database with the accession numbers GSE125695.

IPA pathway analysis of differentially expressed genes. Differentially expressed genes and their
corresponding expression values were uploaded into the Ingenuity Pathway Analysis (IPA) software (Qiagen, Redwood City, CA, USA), and a core analysis was performed separately for every nutritional treatment by using default parameters. The following parameters were used: core analysis, reference set user-defined (i.e., only the set of differentially expressed genes by GeneSpring-software mapped to the IPA database), direct and indirect relationships included, endogenous chemicals included, confidence = experimentally observed. Significant associations of well-documented canonical pathways in the Ingenuity Pathway Knowledge database and the differentially expressed genes were analyzed by IPA-software (Fisher's exact test, P-value < 0.05 (i.e., −log10 ≥ 1.3)). In addition, an activation z-score was generated for each functional category. The entries that had a P-value < 0.05 and an absolute z-score value 2 were considered significant. Furthermore, we identified upstream transcriptional regulators with p-value of overlap < 0.05 and the absolute activation z-score > 2.
Cell culture. Mouse primary hepatocytes were isolated from untreated C57BL/6 mice as described previously 7 . The cells were seeded with 300 000 cells/well in 12-well plates in plating medium (William's E (Sigma Aldrich) with 10% FBS, supplemented with dexamethasone 10 ng/ml, gentamycin 10 µg/ml and insulin-transferrin medium supplement [containing 5 mg/L insulin, 5 mg/L transferrin and 5 µg/L sodium selenite] (Sigma Aldrich) for 3 hours. After the cells were attached, the medium was changed to serum-free plating medium. The cells were first treated with pregnenolone-16-ɑ-carbonitrile (PCN, 10 µM dissolved in DMSO) or vehicle (DMSO) for 21 h. Next the cells were then treated with different concentrations of either glucose, insulin or glucagon for 24 h while the 10 µM PCN or vehicle treatment was also continued. Glucose treatment was done in serum free DMEM with 4 replicates/group. Insulin and glucagon treatments were done in plain William´s E in triplicates. The cells were harvested for RNA isolation.
Quantitative real-time PCR. Total RNA was isolated from liver samples, or from primary hepatocytes using RNAzol ® RT followed by cDNA synthesis with p(dN) 6 Table 2. Primer concentrations were optimized to achieve 100% efficiency for amplification. The fluorescence values of the QPCR products were corrected with the fluorescence signals of the passive reference dye (ROX). The RNA levels of target genes were normalized against the 18 S or GAPDH control levels using the comparative CT (ΔΔCT) method.

Bile acid measurements. Total bile acids from plasma and gallbladder bile were measured with Mouse
Total Bile Acids Assay Kit (Crystal Chem, Zaandam, Netherlands) according to the manufacturer's protocol.

Cholesterol measurement. Plasma total cholesterol was measured with Cholesterol Quantitation Kit
(Sigma Aldrich) according to the manufacturer's protocol.
Statistical analysis. The statistical data analysis was performed using GraphPad Prism Software (La Jolla, CA, USA). The comparison of means of two groups was done by Student's two-tailed t-test, whereas multiple groups were compared by One-way ANOVA followed by Dunnett's post hoc test. Differences were considered significant at P < 0.05.

Data availability
The datasets generated during the current study are available in the Gene Expression Omnibus (GEO) repository with the Accession Number GSE125695.