Relationship analysis between gene expression profiles and rat liver cirrhosis occurrence

Liver cirrhosis (LC) is a kind of liver disease which is pathologically characterized by abnormality and necrosis of hepatic cells, proliferation of fibrous tissue, nodular regeneration and pseudolobule formation. To explore the mechanism of LC occurrence at the level of mRNA, Rat Genome 230 2.0 Array was used to detect gene expression profiles of rat fibrotic livers in 3, 6 and 9 weeks after CCl4 treatment in this study. It was found that a total of 305 genes including 153 up-regulated, 150 down-regulated and 2 up/down-regulated genes, were related to LC occurrence. Then, k-means clustering was employed to classify above 305 genes into 5 clusters based on gene expression similarity, and EASE analysis further indicated that the above genes were mainly associated with metabolic process, stress reaction, cell growth and apoptosis/death. Thereafter, ingenuity pathway analysis (IPA) software was used to analyze potential effects of the above-mentioned 305 genes, and the results suggested that lipid metabolism and cell growth were inhibited while cell apoptosis/death was activated, but immune/inflammatory response was first activated and then inhibited. Furthermore, IPA also predicted that several signal pathways “ERK/MAPK Signaling”, “p38 MAPK”, “Endothelin-1 Signaling”, “Growth Hormone Signaling”, “LPS/IL-1-mediated inhibition of RXR function” and “IL-6 Signaling” were involved in regulating the occurrence of liver cirrhosis. It was concluded that 305 genes and 3 kinds of physiological activities were closely related to LC occurrence.

then liver cirrhosis occurs.
Many cytokines have been reported to be involved in the initiation of liver cirrhosis.Tumor necrosis factor alpha (TNFα) and transforming growth factor beta (TGFβ) could stimulate activation of HSCs, which are known to be activated as the predominant cells responsible for liver fibrosis.Furthermore, interleukins (ILs) were shown to have complicated effects on immune response, inflammation, and liver fibrogenesis (Zhou et al., 2014).Chou et al. (2006) demonstrated that IL-10 inhibited fibrogenic and pro-inflammatory gene responses in CCl 4induced mice, and Zhang et al. (2007) showed that IL-10 presented an anti-fibrogenic effect by down-regulating the activity of HSCs.CCl 4 is widely used for inducing liver fibrosis and cirrhosis in animal models, leading to the production of CCl 3 and CCl 3 OO• free radicals (Chavez et al., 2012).Previously, Chan et al. (2016) examined and compared gene expression profiles of cirrhotic livers and noncirrhotic livers from 40 patients who underwent liver resection or liver transplantation with Affymetrix HuGene 2.0 Chip, and found that a total of 213 genes were significantly differentially expressed for more than twofold change in cirrhotic livers.However, the small sample size and heterogeneous patient characteristics may limit the conclusions, thus more gene symbols need further investigation.
To further compare the gene expression profiles of normal liver tissue and that of LC at broader transcriptional level, this study established a model of rat LC induced by CCl 4 to analyze the gene expression changes during liver cirrhosis and then to explore the processes of occurrence and progression of liver cirrhosis.The data obtained from the gene expression profiles could provide more useful information on the global gene expression changes due to CCl 4 administration and bring important insights into the mechanisms of LC.

Preparation of rat model of liver cirrhosis
A total of 30 healthy male Sprague-Dawley rats, each weighting 180±20 g, were supplied by the Experimental Animal Center of Henan Normal University, and were housed at 21±2°C; relative humidity, 60±10%; illumination time, 12 h/day (8:00-20:00).They were randomly divided into model group (LC) with 24 rats and control group with 6 rats.Rats in LC were fed with normal food and 0.35 g/L phenobarbital sodium solution in the first week.A dose of 0.5 ml/100 g of CCl4 diluted 2:3 with colza oil was injected into their abdominal cavity twice per week in 2 to 4 weeks, and their single drinking water was 10% alcohol.Their drinking water was exchanged with 30% alcohol in 5 to 9 weeks.At the same time, the same amount of physiological saline was injected into those in the control group at the corresponding time.At 3, 6 and 9 weeks after CCl4 administration, the liver tissues from the middle part of right lobe were collected in RNase-free tubes and stored at -80°C until use.All operations and handling procedures were carried out in accordance with the current Animal Protection Law of China.

Histopathological detection of liver tissues in LC
Small cuboids of about 5 × 5 × (2-3) mm from the right lobe of the liver were fixed with 10% neutral-buffered formalin for 24 h and washed for 24 h.Then, they were routinely dehydrated with a graded series of ethanol, cleared in xylene, embedded in paraffin, sectioned at 5 μm thickness.Afterwards, the slices were stained with haematoxylin for 3 min, immerged in ammonia water (pH 8.0) for 30 s, and counterstained with 0.5% eosin for 5 min.Finally, they were dehydrated by gradient ethanol, cleared in xylene and sealed with neutral gum.Histopathologic examinations of the liver sections were conducted and peer-reviewed.

Rat Genome 230 2.0 microarray detection
Total RNA was extracted from the frozen mixed middle parts of right lobes of all the rats at each time point according to the Trizol mini kit (Invitrogen Corporation, Carlsbad, California, USA) and purified according to the RNeasy mini protocol (Qiagen, Inc, Valencia, California, USA).Then, the total RNA was regarded as qualified sample when 28S to 18S RNA was equal to 2:1.After that, total RNA from liver tissues at 0, 3, 6 and 9 weeks of rat model of LC were applied for microarray analysis using Rat Genome 230 2.0 Array (Affymetrix Inc., Santa Clara, CA, USA).To minimize technical errors in the array analyses, the liver sample was detected repeatedly for three times at each time point by Rat Genome 230 2.0 Array, totaling 3 arrays × 4 time points (Xu et al., 2011).

Identification of liver cirrhosis-related genes
Affymetrix GCOS 2.0 was used to convert the images showing gene expression abundance into signal values, signal detection values (P, A, M) and experiment/control values (Ri).To normalize the data of each array, all signals were scaled to a target intensity of 200.When P value is < 0.05, it means that the gene is present (P), when < 0.065, marginal (M), and when > 0.065, means marked absent (A).On the other hand, the normalized signal values in LC to that in control were used to calculate the relative value, that is, ratio value of gene expression abundance.When ratio value is ≥ 3, it means that gene expression was significantly up-regulated, when ≤ 0.33, it means significantly down-regulated, and when 0.33-2.99, it means biologically insignificant.To minimize the technical errors from microarray analysis, each sample was tested at least three times using Rat Genome 230 2.0 microarray, and the average value was considered as a reliable value.The genes in LC which has expressed significantly were considered as LC-related genes (supplementary Table 1).

Quantitative real-time PCR
To verifying the chip data, four genes were selected for RT-PCR analysis.The primers were designed with Primer Express 2.0 software according to the sequences of target genes LCN2 (NM_130741), MYC (NM_012603), CCND1 (NM_171992) and SPINK3 (NM_012674) and synthesized by Shanghai Generay Biotech Co., Ltd.The gene-specific primers were the following: forward primer 5'-CACCCTGTACGGAAGAACC-3' and reverse primer 5'-CACATCCCAGTCAGCCAC-3' for LCN2, forward primer 5'-GAGGAGAAACGAGCTGAAGCG-3' and reverse primer 5'-TGAACGGACAGGATGTAGGC-3' for MYC, forward primer 5'-CCTGACTGCCGAGAAGTTGTGC-3' and reverse primer 5'-TGGAGGGTGGGTTGGAAATGAA-3' for CCND1, and forward primer 5'-CACCCTGCACAGTTCGTC-3' and reverse primer 5'-AGGGCAATTAGGCGTTTT-3' for SPINK3.Then, 2 µg total RNA from each of sample was reverse-transcribed using random primers and reverse transcription kit (Promega, A3500, USA).First-strand cDNA samples were subject to quantitative PCR amplification using SYBR ® Green I on the Rotor-Gene 3000A (Corbett Research, Australia).Every sample was analyzed in triplicate.Standard curves were generated from five repeated ten-fold serial dilutions of cDNA, and the copy numbers of target genes in every milliliter of the sample were calculated according to standard curves (Wang and Xu, 2010).

Ethical approval
All applicable international, national, and/or institutional guidelines for the care and use of animals were followed.

Bioinformatics analysis
In order to characterize the expression patterns of the genes involved in LC, k-means clustering was applied to classify the differentially expressed genes, then the genes mapped in each cluster during the entire time course of LC were assigned to DAVID functional analysis, after that, Expression Analysis Systematic Explorer (EASE) software was utilized to determine whether the Gene Ontology (GO) categories are over-represented or not according to a modified Fisher's exact test (Otu et al., 2007).Physiological processes were selected according to EASE score (P-value), and only these processes assigned with P-value < 0.05 were considered to be significantly over-represented during liver cirrhosis occurrence.In addition, to detect whether such physiological processes were activated or inhibited based on the distinct up-and down-regulation pattern of the expressed genes, and to predict the potential predominant pathways, differentially expressed genes were analyzed by ingenuity pathway analysis (IPA) version 9.0 software (Kramer et al., 2014).First, a dataset containing these differentially expressed genes was uploaded into "Dataset Files" of the IPA.Then IPA core analysis was performed, and physiological processes and signal transduction pathways could be obtained from the modules of "Diseases and Biofunctions" and "Heat-map" in "Canonical Pathways" through comparison analyses.

Pathological changes of liver tissues during the occurrence and development of rat liver cirrhosis
For rat normal livers, the structure of hepatic lobes and liver sinusoid were clear.Hepatocytes were orderly arranged and shaped in polygon (Figure 1A).In the third week of LC group induced by CCl 4 , central venous was congested obviously, and spotty necrosis was observed with increased number of cells undergoing hydropic degeneration (Figure 1B).At the sixth week, steatosis was increasingly severe, pseudolobules began to form partly, and a few fibrous hyperplasia scattered (Figure 1C).At 9 weeks, infiltration of inflammatory cell was further enhanced, and fibrous collagens were becoming thick gradually.Furthermore, fibrous septa, pseudolobule and regenerative nodules of liver emerged, and fasciculus appeared in pseudolobule (Figure 1D).

Comparison analysis of the detection results of realtime RT-PCR and microarray
To evaluate the validity of the chip data obtained in this study, several genes were selected for real time RT-PCR assay to detect their expression changes in liver tissues of CCl 4 -induced liver cirrhosis.The genes surveyed were composed of four up-regulated genes: LCN2, MYC, CCND1 and SPINK3.The comparative results indicated that, although there were somewhat differences in the relative degree of up-or down-regulation measured by the above two methods, expression profiles of these four genes detected by real-time RT-PCR were almost in accordance with those obtained by chip analysis in 3, 6 and 9 weeks of rat liver cirrhosis, suggesting that array results were reliable (Figure 2).

Expression profiles of significantly expressed genes during rat liver cirrhosis
Rat Genome 230 2.0 Array was used to detect gene expression profiles of LC in 3, 6 and 9 weeks, and the obtained microarray data were submitted to the Gene Expression Omnibus database with the accession number of GSE73499.As a result, 305 known genes were found to be significantly changed in expression, including 153 up-regulated, 150 down-regulated genes and 2 up/down-regulated genes (Table S1).Subsequently, the above-mentioned 305 genes were divided into upand down-regulation groups and classified into 5 clusters by k-means with log2 fold change according to gene expression similarity (Figure 3A): Cluster 1 (C1) included 85 genes, which exhibited down-regulated expression pattern in 6 and 9 weeks.Though, 64 genes in C2 were down-regulated at each time point during LC, no significant expression change was observed among different time points.C3 involved 13 genes which were strikingly up-regulated at 9 weeks.121 genes were contained in C4, and strengthened in expression level in both 6 and 9 weeks.There were 22 genes classified into C5, which displayed increasing trend at 3 and 9 weeks during the occurrence of LC (Figure 3B).

Functional enrichment analysis of significantly expressed genes during rat liver cirrhosis
After the expression patterns of liver cirrhosis-related genes were characterized, DAVID analysis was carried out for the gene sets of C1-5 in Figure 3, and then the over-represented GO categories with EASE scores (P values) of < 0.05 were selected.C1 was enriched with categories of metabolic processes and regulation of cell growth.Obviously, lipid metabolic processes were involved in C2 and inflammation response-related genes were mainly observed in C3 and C5.Meanwhile, stress response, cell apoptosis/death and cell growth were predominant in C4.In brief, a total of 305 genes were divided into three categories with metabolic processes scattering in C1 and C2, stress reaction enriched in C3, C4 and C5, and cell growth and apoptosis/death involved in C1 and C4 (Table 1).

Activity prediction of bio-processes and signal pathways during rat liver cirrhosis
A total of 305 genes were uploaded to IPA software for core analysis and comparison analysis, and bio-functions of LC at three time points were obtained (Figure 4).The results showed that inflammation response was slightly activated at 3 weeks but obviously inhibited at 6 and 9 weeks, which displayed the same trend as that of immune response of cells.Meanwhile, among several categories of lipid metabolism, only fatty acid metabolism was activated at 3 weeks and inhibited at 6 weeks while conversion of fatty acid, concentration of cholesterol and sterol all showed the trend of suppression.Proliferation of liver cells and hepatocytes were both inhibited, and cell death and apoptosis of hepatocytes were accordingly activated.
To display the relationships between expression changes of genes and activities of physiological processes during the occurrence of LC, the gene regulatory networks regulating three main categories of bio-process were predicted and respectively shown in Figure 5.The 12 genes regulating inflammation were all Represents EASE score (a modified Fisher's exact test).
all up-regulated at 3 weeks, and 5 of them participated in activating inflammation response.At 6 weeks, 14 genes including 6 up-regulated and 8 down-regulated might be involved in inhibiting inflammation, and 13 genes with 7 upregulated and 6 down-regulated genes led to inflammation attenuation at 9 weeks.Expression changes of genes involved in fatty acid metabolism showed that 6 up-regulated genes from a total of 11 genes strikingly activated the process at 3 weeks.However, 8 up-regulated and 1 down-regulated genes might play vital roles in inhibiting fatty acid metabolism at 6 weeks.At 9 weeks, fatty acid metabolism was slightly boosted by 9 genes, which included 6 up-regulated and 3 down-regulated genes.As for the proliferation and death of liver cells during LC, this study found that cell death of hepatocytes was induced at 9 weeks by 3 down-regulated and 1 up-regulated genes, but 4 genes with 3 up-regulated suppressed proliferation of liver cells at 6 weeks and 4 genes with 2 up-regulated inhibited the process at 9 weeks.In addition, to clarify which signaling pathway might play important role in liver cirrhosis, pathway analysis was performed to associate the differentially expressed genes with canonical signaling pathways by IPA software.The enriched "Heat-map" in the "Canonical Pathways" was amputated in Figure 6.Obviously, "ERK/MAPK Signaling" and "p38 MAPK" involved in regulation of cell proliferation were both activated at 6 and 9 weeks, "Endothelin-1 Signaling" was at 6 weeks, while "Growth Hormone Signaling" was strikingly inhibited at 6 weeks.Moreover, "LPS/IL-1-mediated inhibition of RXR function" and "IL-6 Signaling" were similar, which were activated at 6 weeks, but inhibited at 9 weeks.

DISCUSSION
Though several studies have detected gene expression profiles associated with toxin-induced liver fibrosis or cirrhosis in rats and mice, the number of tested genes remains to be improved (Gant et al., 2003;Liu et al., 2009).Furthermore, the studies in humans always investigated the patients with liver fibrosis resulting from HBV and HCV infection or liver cancer, and the small patient sample size and heterogeneous patient characteristics may always lead to the limitations of conclusions.To further systematically study the toxin-induced liver fibrosis at broader transcriptional level, this study used Rat Genome 230 2.0 Array including 24619 genes to detect the gene expression profiles of liver cells in rats following 3, 6 and 9 weeks after CCl 4 administration.It was found that 305 genes were significantly changed, including 153 up-regulated, 150 downregulated and 2 up/down-regulated genes.DAVID functional analysis categorized the differentially expressed genes into 3 groups including metabolic process, stress reaction, cell growth and apoptosis/death.IPA further predicted activities of enriched bio-processes and potential signaling pathways during rat liver cirrhosis, which could provide information for the understanding of the pathophysiological changes of cirrhotic liver.
Hepatic fibrosis was always concomitant with hepatic inflammation (Wu et al., 2015).In this study, stress reaction-associated genes were enriched in C3, C4 and C5, they displayed increased expression at mRNA level.However, IPA software predicted that inflammation response and immune response were significantly activated only at 3 weeks in bio-function heat map (Figure 4).Similarly, several previous studies demonstrated that liver injury could stimulate immune system, and then induce immune response (Zhang et al., 2009;Dienes and Drebber, 2010).Therefore, CCl 4induced hepatic injury may lead to the activation of inflammation and immune response during the early stage of liver cirrhosis.On the contrary, inflammation and immune responses were inhibited at 6 and 9 weeks in Figure 4. Accordingly, Ji et al. (2012) found that the top network of down-regulated proteins expressed by HSCs  was involved in immune response and speculated that the immune response of HSCs was impaired upon activation.However, it deserves further study to identify whether the decreased inflammation and immune responses were due to the activation of HSCs after 6 and 9 weeks of CCl 4 administration.Among the genes participating in inflammation, A2M was always used as a biological indicator to assess hepatic fibrosis and cirrhosis, and was found to be up-regulated at 3 weeks in this study.Gangadharan et al. (2007) pointed out that the expression of A2M was strengthened with the development of hepatic fibrosis, which was consistent with the current study result.Wald et al. (2004) found that up-regulation of CXCL12 in liver during chronic HCV and HBV infection may support the establishment of a chronic inflammatory state and a progressive fibrotic process, thus the down-regulation of CXCL12 at 6 and 9 weeks lead to inflammation response inhibition, as shown in Figure 5A and indicated an accordingly result.The upregulation of SPP1 might be a primary pathway of HSC activation (Erkan et al., 2010).Consistently, its expression was increased in 6 weeks of LC in this study.One of the IL-1 family members, IL-33 was downregulated at the mRNA level in 6 and 9 weeks, and it has been identified as an important factor contributing to inflammatory response and liver injury (Wang et al., 2012), which together suggested its important role in inhibiting inflammation during the occurrence of liver cirrhosis.IL-6 showed both pro-inflammatory and antiinflammatory effects on hepatic system through ERK/MAPK signaling (Hassan et al., 2014), and IL-6 signaling and ERK/MAPK signaling pathways were both enriched in "canonical pathway" by IPA (Figure 6), which showed the important roles of above two pathways in regulating inflammation during liver cirrhosis occurrence.
The genes involved in metabolic processes were decreased at the mRNA level and enriched in C1 and 2, which exhibited the same trend as the activities of metabolic processes, such as conversion of fatty acid, concentration of cholesterol and sterol.Previous studies proved that lipid metabolism was severely impaired in liver cirrhosis originating from hepatitis C and B viruses (Vere et al., 2012), implying that the process of lipid metabolism was attenuated during liver cirrhosis.Surprisingly, fatty acid metabolism was activated at 3 weeks and slightly inhibited at 6 weeks.Moreover, the regulatory network of fatty acid metabolism showed that the up-regulation of FABP4 at three time points participated in activating fatty acid metabolism (Figure 5B).FABP4 is one of fatty acid binding proteins, and the expression of FABP4 and FABP5 was enhanced during cirrhosis in hepatocarcinogenesis of rat model (Liu et al., 2009), which was not contradictory with our result and indicated an alteration of fat metabolism.One previous study showed that cytochrome P450 (CYP) 1A1 could protect lung from oxidative injury by decreasing levels of lipid hydroperoxides (Lingappan et al., 2014), and the Wang et al. 155 Present study demonstrated that CYP1A1 was upregulated at three time points and activated fatty acid metabolism at 3 and 9 weeks, suggesting that CYP1A1 may also play a protective role by activating fatty acid metabolism in liver cirrhosis.According to the IPA analysis, LPS/IL-1-mediated inhibition of RXR function ranked first in the identified canonical pathways, and the study performed by Raghu et al. (2012) indicated that this pathway was involved in ethanol metabolism in alcoholic fatty liver.On the other hand, different concentration of alcohol were supplied for the model rats in this study, thus LPS/IL-1-mediated inhibition of RXR function might participate in rat LC occurrence by regulating lipid metabolism.In addition, one previous study proved that IL-6 had positive effects on hepatic lipid metabolism (Hassan et al., 2014).Therefore, IL-6 signaling pathway may be involved in modulating lipid metabolism during liver cirrhosis.
Cell proliferation and apoptosis/death related biological processes were enriched in C1 and 4 with different gene expression profiles.The study performed by Jiang et al. (2015) showed that the levels of early, late apoptosis and cell death were significantly higher in the cirrhosis model.Accordingly, this study showed that cell death and apoptosis of hepatocytes were activated in "Diseases and Bio-functions" analysis by IPA.However, proliferation of liver cells and hepatocytes were both found to be inhibited.On the other hand, Chen et al. (1999) found that proliferation continued in spite of increasing apoptosis of hepatocytes apoptosis at 72 h after subcutaneous injection of 50% CCl 4 .Thus, it is speculated that proliferation of liver cells might be inhibited due to the intense stimulation of CCl 4 and alcohol in this study.The oncogene MET played an important part in preventing Fas-mediated apoptosis of hepatocytes (Zou et al., 2007), and proliferation and survival of hepatocytes were impaired in mice with Met mutants after partial hepatectomy (Borowiak et al., 2004).Consistent with the above results, this study showed that the down-regulation of MET at 9 weeks may activate cell death of hepatocytes and its up-regulation at 9 weeks may suppress proliferation of liver cells.It is well known that CDKN1A inhibits the activity of cyclin-dependent kinase 2 (CDK2) and arrests cell cycle at G1. Accordingly, its up-regulation may account for the activated cell death of hepatocytes at 9 weeks and inhibitory proliferation of liver cells at 6 and 9 weeks.Proliferation and activation of HSCs could be suppressed via ERK/MAPK in hepatic fibrosis (Peng et al., 2014), and p38 MAPK was also associated with cell apoptosis in liver fibrosis (Wang et al., 2015), so it could be concluded that the above two pathways, which were both enriched in Figure 6, play a vital role in cell growth and apoptosis/death during liver cirrhosis occurrence.
In conclusion, Rat Genome 230 2.0 Array detection and gene synergy analysis identified a total of 305 genes, which were differentially expressed at 3, 6 and 9 weeks and associated with liver cirrhosis occurrence.These genes were involved in three primary physiological activities based on GO annotation and several potential related pathways according to IPA software.In the future, the result above will be confirmed using gene addition and RNA interference, etc. and provide more information for therapy and prevention of liver cirrhosis.

Figure 2 .
Figure 2. Verification of gene expression in CCl4-induced liver cirrhosis by real-time PCR.The results of RT-PCR and Rat Genome 230 2.0 Array are presented as real line and dotted line, respectively.

Figure 3 .
Figure 3. Global comparison of gene expression patterns in liver cirrhosis.A. K-means clustering of a total of 305 genes.Red and green colors denote the expression level higher and lower than the control, respectively.B. Five clusters (C1 to C5) display different gene expression profiles over time in the development of LC.

Figure 4 .
Figure 4. Biofunction heatmap of 305 significantly expressed genes predicted by IPA software.The color of heatmap square represents the activity of biofunctions.Jacinth represents activation of biofunction, and blue represents inhibition of biofunction.

Figure 6 .
Figure 6.Signaling pathway heatmap of 305 significantly expressed genes predicted by IPA software.The color of heatmap square represents the activity of biofunctions.Jacinth represents activation of biofunction, and blue represents inhibition of biofunction.

Table 1 .
Over-represented functional categories in five clusters during rat liver cirrhosis occurrence. #