HCC is a heterogeneous disease in terms of etiology, biologic and clinical behavior. Meanwhile, hepatocarcinogenesis is a long-term, multistep process associated with changes in gene expression profiles. In the last several years, there have been important gains in our understanding of the pathogenesis of HCC and our appreciation of the critical oncogenic and tumor suppressor pathways involved in hepatocarcinogenesis [1-5]. Despite this, current knowledge about the molecular pathogenesis of HCC is a result of investigations of fully developed HCC. Very little is known about how many genes concur at the molecular level of tumor development, progression and aggressiveness.

Molecular profiling has been successfully used to identify candidate genes for HCC in human and animal model systems[3]. Although many approaches (including genome-scale studies) provide insights into some of the stages in human tumorigenesis, a sequential analysis of the development of tumors in humans is very difficult. Most of them have not given us the gene expression profiles that could point to those genes that play key roles during the whole carcinogenetic process from initiation to metastasis. Animal models of carcinogenesis have permitted the examination of the stages of neoplastic development in considerable detail.

In this study, we established the rat model of liver cancer induced by DEN to explore the processes of initiation and progression of HCC[6]. HCC develops commonly, but not exclusively, in a setting of chronic liver cell injury, which leads to inflammation, hepatocyte regeneration, liver matrix remodeling, fibrosis, and ultimately, cirrhosis[7,8]. The histological changes in DEN-induced liver cancer in rats are similar to those seen in human HCC. We think the similar phenotype might be based on similar gene expression profiles. Affymetrix GeneChip Rat 230 2.0 arrays were used to analyze gene expression profiles of liver tissues from control and DEN-treated rats during the process from cirrhosis to metastasis. This allowed us to obtain an almost complete picture of the early genetic alterations that are directly or indirectly involved in the development of HCC. We supposed that the genes expression profiles deregulated during the process from liver cirrhosis to carcinoma and metastasis play key roles in the hepatocarcinogenesis. The data obtained from the gene expression profiles will allow us to acquire insights into the molecular mechanisms of hepatocarcinogenesis and identify specific genes (or gene products) that can be used for early molecular diagnosis, risk analysis, prognosis prediction, and development of new therapies.

Some major drawbacks limit the usefulness of studies regarding molecular pathogenesis in HCC. First, few studies analyze the genetic and genomic alterations that emerge at different time points during the entire progressive process of the disease. Second, the limited size of the studies is often a factor that undermines the capability to provide consistent genomic data[9]. Animal models of hepatocarcinogenesis summarize the primal biology of liver tumorigenesis and have provided reliable data for understanding the cellular development of HCC in humans[1,10,11]. In the present study, the pathologic changes of livers in rats treated by DEN included non-specific injuries, regeneration and repair, fibrosis, and cirrhosis, dysplastic nodules, early tumorous nodules, advanced tumorous nodules and metastasis foci, resembling the process of human hepatocarcinogenesis. DEGs obtained by compare normal rats with DEN-treated animals at stages from cirrhosis to metastasis allowed us to screen for upregulated and downregulated gene expressional profiles. The number of DEGs at each stage was large and the information obtained was powerful. We were thus able to visualize the complicated process of hepatocarcinogenesis at the genomic level.

The annotated information of the DEGs show that extensive and diverse biological processes and molecular functions are involved in hepatocaricnogenesis. Most of the DEGs are involved in metabolism and transport, indicating that significant alterations occurred in the process of metabolism and transport during the developmnet of HCC. For example, tumor cells always perform anaerobic glycolysis, even when there is an adequate oxygen supply[12,13], partly a result of alterations in the profile of enzymes associated with glycolysis. In this study, the gene expression level of lactate dehydrogenase B increased from the cirrhosis phase to the metastasis phase. Evidence shows that some genetic changes promoting tumor growth influences glucose energy metabolism directly[14,15]. Many intermediate products from glycolysis are used to synthesize proteins, nucleic acids and lipids by tumor cells, providing the essential materials for the growth and hyperplasia of tumor cells. For aggressive tumors, increased glycolysis and metabolism alterations often occurred. The microenvironment acidosis provided by the conversion of pyruvic acid to lactic acid promotes invasion and metastasis of tumor cells [16-18]. Therapy targeting glycolysis could preferentially kill tumor cells with metabolic alterations and there is experimental evidence that some types of tumors can be treated by inhibiting the glycolytic and phosphopentose pathways[16]. In this study, some DEGs associated with metabolisms of glucose were shown in Figure 6A.

Fat metabolism have significant changes in the process of tumorigenesis, e.g. a high fat diet was related to the development of many tumors [19]. Enhanced fat synthesis in tumor cells could not only support the increased membrane synthesis and energy metabolism, but also higher level of fatty acid synthetase provides the base for interpretation the relation between the fat metabolism and the capacity of hyperplasia and metastasis of tumor cells[20]. Stearoyl-CoA desaturase (SCD), which have four known isomers, takes part in regulating lipid synthesis. SCD2 plays key roles in the early development and survival of embryos in mice, whose expressional levels in the livers of wild mice embryos and newborn mice were higher than that of adult mice[21]. Inhibition of lipid synthesis caused by the depletion of SCD2 was related to the decreased expression level of peroxisome proliferator-activated receptor gamma (PPAR-γ)[22]. Fatty acid binding proteins (FABPs) are proteins that could bind to fatty acid and other lipids reversibly. Researchers found expression of FABP5, coding epidermal fatty acid binding protein (E-FABP-GenBank Accession), upregulated in primary tongue carcinomas[23]. FABP4, as a bridge between the inflammation and other metabolism syndromes[24], could not only transport the nuclear receptor PPAR-γ from cytoplasm to nucleus but also cause increased transcript activation of it[25]. In this study, the expressional levels of SCD2, FABP4 and FABP5 increased during the process from cirrhosis to metastasis in rat model, suggesting that an alteration of the fat metabolism occurred in hepatocarcinogenesis of rat model. Other DEGs associated with fatty metabolisms were shown in Figure 6A.

In the present study, some enzymes related to the glutathione (GSH) metabolism were found to be significantly altered. For example, the expressional level of Gstm3 (glutathione S-transferase, mu type 3) decreased in all stages of hepatocarcinogenesis, while the expression levels of of enzymes increased, which including Glul (Glutamate-ammonia ligase), Gclc (Glutamate-cysteine ligase, catalytic subunit), GPX2 (Glutathione peroxidase 2), GPX3 (Glutathioneperoxidase 3), GSR (Glutathione reductase), Yc2 (Glutathione S-transferase Yc2 subunit), Gstm5 (Glutathione S-transferase, mu 5), Gstp1 (Glutathione-S-transferase, pi 1) and GSS (Glutathione synthetase). Some studies reported that GSH and the associated enzymes were considered to promot the tumor transformation from dysplastic nodules and take part in the development and progression of hepatocarcinomas[26,27]. In gastroenteric tumors, the increase of GSH and the associated enzymes exhibit activation, indicating the existence of ROS and anti-oxidation defense related to GSH metabolism[28]. There were five binding sites for β-catenin/TCF at the promoter region of GPX2, indicating that GPX2 might take part in the corresponding signal pathways[29]. Thus previous research and our data indicate that genes related to oxidative stress and GSH metabolism play important roles in the process of progression from dysplastic nodules to tumor. The expressional level of GSH increased in tissue of HCC and the active hyperplasia liver cells[30,31]. Research has shown that DNA oxidative injury is increased in human HCC [32,33].

Many other enzymes associated with metabolism are involved in the defense and stress reaction, such as oxidative stress. For example, AKR1B7 (aldo-keto reductase family 1, member 7) takes part in the detoxification of oxides, such as aldehyde. During the detoxification of aldehyde, the expressional level of AKR1B7 mRNA increased. There are five binding sites with NF-κB at the 5' upstream region of the AKR1B7 gene, and oxidative stress upregulates the expression of AKR1B7 mediated by NF-κB[34,35]. The expression level of aldehyde dehydrogenase ALDH3A1 (Aldehyde dehydrogenase 3A1) also increased after oxidative stress. In the present study, the expression levels of AKR1B7, AKR1B8 and ALDH3A1 were up-regulated at all stages of hepatocarcinogenesis. In the tumor cells, reactive oxigen species (ROS) was produced through the oxidative stress. ROS as signal molecules mediate various reactions relating to growth, such as angiogenesis. ROS in endothelial cells is mainly from NADPH oxidation enzymes, consisting of Nox1, Nox2, Nox4, Nox5, p22 (phox), p47 (phox) and Rac1 (small G-protein Rac1). NADPH oxidative enzymes were activated by different factors including VEGF, angiopoietin-1, hypoxia and ischemia. Furthermore, ROS has been shown to be involved in spontaneous phosphorylation[36]. Nox4 mediated growth factors, such as anti-apoptosis of IGF, is partly due to the ROS produced by NADPH oxidative enzymes inhibiting the key protein tyrosine phosphatases(PTPs), then continually causing JAK2 kinase phosphorylation which resists the apoptosis reaction[37]. In this study, However, the gene expression level of NADPH oxidative enzymes decreased in the livers of our rat model at all stages of hepatocarcinogenesis. The mechanism is unclear.

Cytochrome P450s (CYPs) are key enzymes in tumorigenesis, taking part in the activation and inactivation of chemotherapeutic agents in tumor tissues[38]. The expression level of CYP1B1 in breast carcinomas was up-regulated significantly, providing a new therapy target and phenotype biomarker. The significant increase in CYP2E1 correlates with invasiveness and is a potential prognosis factor[39,40]. Other studies have shown that the expression of CYP could influence the synthesis of arachidonic acid derivatives, thus altering the various downstream signal pathways, which was thought to be the prelude of carcinogenesis[41]. At present, it is not possible to always predict the efficacy of a chemotherapeutic agent or its degree of toxicity. Analysis of CYPs expressional levels in tumor cells may allow prognosis decisions and therapy predictions. In this study, only the expression level of CYP2C40 increased at all stages of hepatocarcinogenesis in rat models, while the remaining CYPs decreased (Figure 6C). Clearly, further investigation is needed to determine the role(s) of CYPs in hepatocarcinogenesis.

In addition to the deregulated expression of metabolism associated genes, we noticed that among the DEGs in the hepatocarcinogenesis of rat models, some known tumor-associated genes, such as Rb1 and Myc, showed deregulated expression occurring at all the stages of hepatocarcinogenesis. Their persisting activation or deactivation could induce the tumor phenotype, as well as play roles at the later stage of progression and metastasis. Meanwhile, some known metastasis-associated genes are found deregulated at the promotion stage of tumor development. For example, the expression level of Ndrg2 and Hrasls3 (HRAS like suppressor 3) decreased at all stages compared to the normal livers, while the expression level of Nme1 (expressed in non-metastatic cells 1) increased. Generally, it was thought that genes involved in the development of carcinoma activation participated at the early stage, while genes participating in the metastasis were activated at the latter stage of tumor progression[42]. In opposition to the traditional model, Bernards and Weinberg proposed that the metastatic ability of tumor cells occurred at the early stage of tumor development[43]. Some oncogenes such as Ras and Src assigned the tumor cells with the metastatic phenotype [44-46].

As we known, the important characteristic of malignant tumor cells is the capability of invading the vicinity, forming metastasis foci at the remote organ, overcoming the host's control over the microenvironment[47,48]. The malignant transformation of liver cells occurred on the basis of chronic injury, regeneration and cirrhosis. The liver cancer cells could synthesize ECM components and the ECM surrounding liver cancer cells was found to be different from that of stroma in the normal organ [49-51]. Integrin and laminin are the major components of ECM. The interaction between integrin and laminin is closely related to the signal transduction, providing survival signals for the cells, mediating the liver cancer cells formation of pseudopodia, and adherence with laminin, which are imperative if a liver cancer cell is to migrate and invade [52-55]. In the process of hepatocarcinogenesis in this rat model, the deregulated expression of many ECM associated genes plays important roles in the hepatocarcinogenesis, e.g. Itga6, Lamc1, Col1a1 and Spp1, etc. (Table 2, 3 and additonal file 2). The differential expression profile of ECM associated genes in time course and space is very important to cellular proliferation and migration.

It is worthy of note that the expression of the annexin family members and calcium-binding proteins are deregulated in this rat model of liver cancer. In our study, the expressional level of Annexin A1, A2, A3, A5 and A7 increased compared with the normal liver tissue. Annexins consist of a conserved protein family. Annexin A2 is closely associated with cell division regulation and tumor growth, and is deregulated in many tumors[56,57]. Two Annexin A2 molecules bind to the long chains of p11/S100A10 dimers through its N-terminals, form the isotetramer, regulating the reactions of Annexin A2 and membranes and actin in cortical areas, and the distribution of recirculating endosomes[58]. In addition, S100A10 and Annexin A2 form isodimers, prompting the invasion and metastasis of the tumor by activating plasminogen[59]. In the present study, the expression level of S100a10, S100a11, S100a6, S100a8 and S100a9 increased from cirrhosis to metastatic process when compared with the normal liver. S100A8/A9 form the compounds that play a role in inducing apoptosis in tumor cells. S100A8/A9 at low concentrations prompts growth activity, the phosphorylation of MAPK pathway and NF-κB is activated in cells after S100A8/A9 treatment.

The majority of HCCs slowly unfold against a background of chronic hepatitis and cirrhosis, which can be considered as preneoplastic conditions of the liver. Chronic hepatitis is characterized by persistent inflammation, cytokine and oxidative stress-mediated hepatocyte death and active proliferation of residual hepatocytes to replace the lost parenchyma[1,60]. During the process of hepatocarcinogenesis in rat models, chronic inflammation precedes cirrhosis. Epidemiology studies showed that chronic inflammation increased the risk of tumors, and the microenvironment of tumorigenesis resembles the reaction of inflammation to injury in many ways[61]. In the tumor microenvironment, the chemotactic factors and receptors mediated angiogenesis, recruited cells, prompting cellular survival and proliferation. On the other hand, oxidative stress occurred in inflammatory processes. The inflammatory cells and tumor cells both produce free radicals and soluble factors such as arachidonic acid, cytokines and chemotactic factors, seubsequently producing reactive oxygen. All these factors strongly recruit the inflammatory cells to produce cytokines, which promotes a vicious cycle. The intermediate products of active oxygen oxidize DNA directly or interfere with DNA repair. These oxides activate protein, carbohydrate and lipids quickly, the derived products interfere with inter- and intracellular homeostasis, favoring DNA mutation. Thus, the chronic inflammation prompts the malignant transformation of cells[62]. Chronic inflammation also favors angiogenesis[63]. In the present study, many DEGs are related to inflammation reaction, immune reaction and stress. (Tables 4 and 5)

In conclusion, analysis of gene expression profile is a useful tool to provide new clues and produce new research targets in the hepatocarcinogenesis field. In the present study, a rat model of liver cancer was established. We have listed the deregulated expression genes in the process from cirrhosis to liver cancer in the DEN-treated rat model. As indicated in the literature, this model shows that cirrhosis is a precancerous lesion of the liver. Although we only discuss some parts of the great quantity of information in this study, much unknown information remains. Functional analysis of these genes revealed discrete expression clusters, including cell proliferation, protein synthesis, and hepatocyte-specific functions. We still need to discern the key genes playing core roles in the promotion and progression of liver cancer. In this article, we focused our attention on the global molecular events that occurred in DEN-treated rats (and probably represent the earliest ones that start the multistep process of hepatocarcinogenesis). Additional information may be mined from this and similar studies to provide clues to many areas including the very important search for diagnostic markers, therapy targets and prognosis prediction markers.

HCC: Hepatocellular carcinoma; DEN: diethylnitrosamine; DEGs: differential expression genes.

The authors declare that they have no competing interests.

YFL wrote the manuscript. BSZ performed the validation of genes. HLZ and XJZ established the animal model. YHL prepared the tissue slides. JZ helped write the manuscript. JPZ, ZQF and XHG participated in the design of the study and helped to draft the manuscript. All authors read and approved the final manuscript.