E2F transcription factors are key regulators of the cell cycle although the relative contribution of each E2F member in regulating cellular proliferation is still poorly defined. Present evidence suggests that E2F2 may act both as a suppressor and promoter of proliferation, depending on the cellular context. We used a loss-of-function mutant mouse model to investigate the function of E2F2 in liver regeneration after partial hepatectomy, a paradigm of cell-cycle progression. Liver mass recovery and histology were examined over 9 days in 70% hepatectomized E2F2−/− and wild-type animals. Transcriptome analysis was performed in quiescent and 48-h regenerating liver samples. TIGR MultiExperiment Viewer was used for the statistical analysis of microarray data, significance was determined by Fischer, and P values were adjusted applying Benjamini-Hochberg multiple-testing correction. We show that E2F2 is required for adult hepatocyte proliferation and for timely liver regeneration, as disruption of the E2F2 gene in hepatocytes leads to a reduced rate of S-phase entry and to delayed liver regeneration. Transcriptome analysis followed by ontological classification of differentially expressed genes and gene-interaction network analysis indicated that the majority of genes involved in normal liver regeneration were related to biosynthetic and catabolic processes of all major biomolecules as well as cellular location and intracellular transport, confirming the complex nature of the regeneration process. Remarkably, transcripts of genes included in functional categories that are crucial for cell cycle, apoptosis and wound-healing response, and fibrosis were absent in the transcriptome of posthepatectomized E2F2−/− mice. Our results indicate that the transcriptional activity of E2F2 contributes to promote adult hepatocyte proliferation and liver regeneration.
- cell cycle
- liver repair
- partial hepatectomy
the e2f transcription factors are key regulators of cell proliferation and rate-limiting factors in the S-phase entry, as they regulate the expression of many genes involved in the G1/S transition. According to the most accepted model of cell-cycle control, unphosphorylated retinoblastoma tumor suppressor protein (pRb) binds to E2F in G0/G1, forming a complex that actively represses E2F-responsive genes. In late G1 pRb is inactivated because of phosphorylation by cyclin-dependent kinases, releasing free E2Fs that activate the expression of their target genes (4, 12, 42).
The E2F family is presently composed of nine members encoded by eight different genes (E2F1–8) (4, 12, 20, 42). Members of E2F family can be classified into “activators” (E2F1–3a) and “repressors” (E2F3b-8). Activator E2Fs are considered potent activators of transcription and positive regulators of the cell cycle. Overexpression of any of these activator E2Fs is sufficient to promote G1/S transition and DNA replication in immortalized, quiescent rodent fibroblasts in the absence of growth factors (23). However, there is increasing evidence that they may also act as transcriptional repressors (2, 7, 18, 28). E2F1 seems to act both as an oncogene and as a tumor suppressor in the liver (6) but was found not to be essential for normal liver regeneration (22). The E2F2 transcription factor has not been studied as extensively as E2F1 and E2F3, and its biological function remains unclear. On one hand, E2F2 promotes cell division in mouse embryonic fibroblasts (45) and in hematopoietic progenitor cells (21). Transgenic mice overexpressing E2F2 in thymic epithelial cells develop thymomas (34), whereas ectopic expression of E2F2 in cultured cells stimulates their entry into S-phase (9), indicating that E2F2 promotes cell-cycle progression. Surprisingly, E2F2−/− lymphoid and pancreatic cells exhibit an increased capacity to replicate their DNA (17, 18, 26), and E2F2−/− mice develop autoimmune disorders (17) and tumors (48), which are further enhanced by the expression of a Myc transgene (29), suggesting that E2F2 is also a negative regulator of proliferation and tumor suppressor. Thus the roles of E2F2 in either suppressing or promoting cell growth may be tissue specific. Indeed, E2F2 has been shown to have a tissue-restricted role in erythropoiesis and neuronal differentiation (13, 31), and several studies have suggested that E2F2 may also be involved in apoptosis (43) although this still needs to be confirmed.
One of the most paradigmatic models of cell-cycle progression is liver regeneration. Following partial hepatectomy (PH), 90–95% of the remaining parenchymal cells synchronously reenter the cell cycle to begin regeneration (39). A complete restoration of normal tissue-specific mass and function takes place in a few days, and maximum DNA synthesis occurs within the initial 40–48 h after PH in mice (24, 41).
Taking into account the tissue-specific role attributed to E2F2, this work investigated whether E2F2 regulates liver growth. We report here that E2F2 is required for normal hepatocyte proliferation and timely liver regeneration after PH. Mice with targeted deletion of the E2F2 gene exhibited impaired liver regeneration, and hepatocytes showed delayed cell-cycle entry from quiescence. Transcriptomic analyses of regenerating liver tissues revealed that many genes critical to liver regeneration, such as cell-cycle regulation, apoptosis, and response to wounding, are regulated by E2F2, implying a crucial role for this transcription factor in liver growth after PH.
MATERIALS AND METHODS
Animals, PH, and sampling.
Eight- to ten-week-old wild-type (WT) and E2F2 knockout (E2F2−/−) (17) female mice of the mixed C57Bl6:129Sv strain were maintained on a 12-h:12-h light/dark cycle with free access to water and food. Seventy percent PH was performed on 2-h-fasted animals under isofluorane anesthesia. The resected, quiescent liver was weighed and used as 0-h control sample. Mice were euthanized at 6, 12, 24, 48, 72, 96, 120, 144, 168, and 216 h after PH. Tissue at 0 h was harvested identically to the subsequently harvested regenerating liver. At the time of euthanasia, blood glucose levels were measured and the mice and the remnant liver tissue were weighed. Pieces (100 mg) of quiescent and regenerating livers were used for RNA extraction, and the remaining tissue was formalin fixed. Animal care and surgical procedures were approved by the University of the Basque Country Animal Care and Use Committee. Postoperative mortality was low and similar in the two genotypes.
BrdU labeling and immunohistochemistry.
Liver hematoxylin and eosin (H&E) staining and hepatocellular bromodeoxyuridine (BrdU) incorporation were assessed as described in the supplemental materials (supplemental material for this article is available online at the American Journal of Physiology Gastrointestinal and Liver Physiology website).
Gene expression analysis by quantitative real-time RT-PCR.
RNA isolation from liver tissue and gene expression analysis by qRT-PCR was performed using established protocols as described in the supplemental material.
Affymetrix chips and microarray data analysis.
Three RNA pools of the 0-h time point and three RNA pools of the 48-h time point were prepared of WT and E2F2−/− mice genotypes. Each pool contained liver samples from eight animals, pairwise in the 0- and 48-h pools. This way, three paired comparisons between the regenerating (48 h) and the quiescent (0 h) liver tissues were performed within each genotype.
Hybridization procedures, scanning, and normalization were performed according to Affymetrix's protocol at Progenika Biopharma S.A. (Derio, Spain) using GeneChip Mouse Genome 430 2.0 microarray platform (Affymetrix, Santa Clara, CA). The quality of the hybridization mixture was checked with spike controls and by the 3′/5′ test. Global scaling was applied to the quantification data to adjust the average recorded data to a target intensity of 100 using GeneChip Operating Software from Affymetrix. Data normalization was made by GeneSpring analysis program (Silicon Genetics, Redwood City, CA).
Statistical analysis and clustering of microarray data.
TIGR MultiExperiment Viewer Mev version 4.1 (Institute of Genomic Research, Rockville, MD) was used for the statistical analysis of the microarray data (32). Hybridization intensities were log2 transformed, and arrays were normalized against each other. To identify the genes whose expression levels changed significantly during liver regeneration in the WT and E2F2−/− mice, we used Welch's t-test with a P < 0.01 significance level cutoff. Functional classification of the resultant genes was made using FatiGO+, a public domain web tool for finding significant associations of Gene Ontology terms within groups of genes (1). Statistical significance was determined by Fischer's exact test, and P values were adjusted applying Benjamini-Hochberg multiple-testing correction. STRING database (version 8.3) was used to build a gene interaction network, with a confidence score >0.9 (37).
Other statistical analyses.
Unless otherwise noted, numerical results are presented as means ± SD. Statistical significance was determined using an unpaired, two-tailed Student's t-test for comparisons between two individual data groups and two-way ANOVA followed by Bonferroni's Multiple-Comparisons Test for comparisons between cohorts. A P < 0.05 was considered statistically significant.
Delayed recovery of liver mass in E2F2−/− mice after PH.
The role of E2F2 in liver mass-recovery after PH was examined in age- and sex-matched adult WT and E2F2−/− mice over 9 days (Fig. 1A). The extent of regeneration was estimated by the measurement of the liver/body weight ratio (LW/BW), whereby day 0 (before PH) represents 100%. A strong regenerative response of the liver was seen in the early phases in the two genotypes, without significant differences between the regeneration profiles. On day 3 the LW/BW ratio was ∼75% of the initial in the two groups. Subsequently, however, regeneration was delayed in E2F2−/− mice compared with controls (P < 0.01, ANOVA). By day 4, the LW/BW ratio in WT mice was similar to that of nonresected controls; by contrast, this ratio was about 20% lower in E2F2−/− mice. Mutant mice did not recover the original hepatic ratio until day 6 after surgery.
Animals exhibited severe hypoglycemia after PH (Fig. 1B), with initial blood glucose levels of around 10 mg/dl for both genotypes. E2F2 deficiency was associated with delayed recovery to euglycemia; glucose levels around 80 mg/dl were reached 24 h later in E2F2−/− mice.
Delayed hepatocyte proliferation and steatosis resolution in E2F2−/− mice after PH.
DNA replication was examined during liver regeneration to determine whether the absence of E2F2 had an effect on S-phase entry in hepatocytes. DNA replication index was determined by measuring the in vivo 2-h BrdU incorporation into the cell nuclei (Fig. 1C). Differences between WT and E2F2−/− were apparent at 48 h of regeneration. At this time point, WT mice displayed a sharp peak in liver DNA synthesis, with ∼35% of BrdU-positive hepatocytic cells, confirming previous data (24, 41). Subsequently, replication index then reduced abruptly to ∼8% on day 3 in WT mice and remained at this level on day 4. In contrast, only ∼10% of liver cells were BrdU positive in E2F2-deficient mice at 48 h after PH, and this percentage remained similarly low during the following days. Thus our data suggest that the delay in liver regeneration in E2F2−/− mice may be attributable to an impaired entry of cells into S-phase.
Regenerating liver tissues develop hepatic steatosis within the first hour after surgery (10). H&E staining of liver sections confirmed transient steatosis in WT and E2F2−/− mice (Fig. 2). Both genotypes showed mild mixed micro- and macrovesicular at 0 h that evolved into massive microvesicular steatosis by days 1 to 3 after PH. Whereas resolution of hepatic steatosis occurred by day 4 in WT mice, moderate amounts of microvesicular steatosis were present in E2F2−/− mice and remained detectable 7 days after partial hepatectomy.
E2F1 and E2F3a mRNA expression changes after PH.
As liver regeneration was unimpeded in E2F2−/− mice, we evaluated whether compensatory mechanisms allowed cells to progress through the cell cycle. The existence of a certain degree of redundancy among E2F family members is widely accepted (8, 9), so the loss of E2F2 might be compensated by activity changes in the other “activator E2Fs”. Expression kinetics of E2F1 and E2F3a transcripts were almost superimposable after partial hepatectomy in WT mice (Fig. 3). They were present at low levels in quiescent liver and upregulated in a time-dependent manner during regeneration, peaking on day 3. E2F1 and E2F3a expression kinetics in E2F2−/− mice lagged WT mice, with a peak expression on day 6, suggesting that E2F2 could modulate E2F1 and E2F3a activation during normal liver regeneration.
However, the delayed proliferation exhibited by E2F2−/− liver cells cannot be attributed solely to differences in E2F1 or E2F3a expression. Although differences in DNA replication between WT and E2F2−/− mice were maximal at the 48-h time point, levels of E2F1 and E2F3a were comparable in both genotypes. In aggregate, the data suggest that E2F2 plays a nonredundant role in the G1- to S-phase transition during liver regeneration, which can be compensated, in part, by the activity of E2F1 and E2F3a.
Microarray analysis and data validation.
To identify genes that respond to E2F2 in liver regeneration, we applied microarray technology for liver-gene-expression analysis in WT and E2F2−/− mice before (0 h, quiescence) and 48 h after PH, the time of maximal DNA replication in WT and where the highest differences in the replication index between genotypes were observed. For each genotype, microarray data produced a list of Affymetrix codes (genes plus expressed sequence tags) with an increased or reduced expression at the 48-h time point compared with the 0-h time point. Affymetrix data are available in the ArrayExpress database (www.ebi.ac.uk/arrayexpress) under accession number E-MEXP-1413 (30). Finally, microarray data were validated by qRT-PCR by assessing the changes in expression of 15 genes, selected by quantitative criteria (Supplemental Fig. S1 and Table S1).
Applying hierarchical clustering to microarray data with the GeneSpring software (Fig. 4A), a clear-cut separation of two groups was originated, one corresponding to quiescent livers (0 h), where samples of either genotype were clustered together, and the other group corresponding to 48-h-regenerated livers, where E2F2−/− samples were clustered apart from WT samples. These results evidence an E2F2-dependent differential gene expression during liver regeneration, implying that E2F2 is required for the normal regulation of many genes linked to this proliferation process.
Many genes showed significant expression changes during liver regeneration (Fig. 4B). The expression of 1,478 genes was changed in WT mice; of these, 45.7% were upregulated and 54.3% were downregulated. In E2F2−/− mice, the expression of 1,801 genes was changed; of these, 57.9% were upregulated and 42.1% were downregulated. Surprisingly, only 377 genes were similarly regulated in both strains of mice (Supplemental Table S3), of which 181 genes were involved in metabolic pathways and 40 genes in metabolic regulation (Fig. 4C). The identity of the remaining genes differed between the WT (1,101 genes) and the E2F2−/− transcriptomes (1,424 genes).
Gene ontology analysis of microarray data.
To identify the most representative functional categories involved in liver regeneration, we carried out an enrichment analysis of the microarray data on the basis of the Gene Ontology (GO) consortium criteria for Biological Process (Table 1). Regenerating WT hepatic tissue exhibited 63 overrepresented functional categories at GO level 3. In the case of regenerating E2F2−/− samples, this number was reduced to 44. More than half of the categories were shared by the WT and the E2F2−/− transcriptomes. The processes with higher enrichment in the two genotypes include various core metabolic pathways (related mainly to proteins, nucleotides, amino acids, nitrogen compounds, lipids, carbohydrates, alcohols and cofactors), regulators of physiological processes, cellular localization, and intracellular transport (Table 2). The proportions of genes included in each category were similar in WT and E2F2−/− mice, even though their identities differed. For example, WT liver samples exhibited upregulation of Tpi1 and Ldha and downregulation of Aldob and Pygl, whose expression was unchanged in E2F2−/− mice. By contrast, disruption of E2F2 affected the expression of other carbohydrate metabolism genes, such as Aldoa, Pgm1, Pgd, Tkt, and Pck1, which were upregulated, and Pdhb and Dlat, which were downregulated. WT samples showed higher expression of many genes involved in essential mitochondrial functions, including tricarboxylic acid cycle-mediated energy production (Idh3a, Dlst), mtDNA replication (Polg), fatty acid oxidation (Echs1, Hadh), hydrogen peroxide neutralization (Sod2), and pyruvate metabolism (Pdhb, Dlat, and Pcx). Others, such as Mdh2, Fh1 (tricarboxylic acid cycle reversible reactions) and Cps1 (catalyzes the overall flow limiting step through the urea cycle), were higher in E2F2−/− mice.
Several genes upregulated by E2F2 loss (Agpat3, Pnpla2, and Fabp4) participate in the metabolism of triglycerides (TG) and phospholipids (PL). In addition to this, Dgapt2, responsible for the last step of TG biosynthesis, was downregulated in WT mice, whereas it showed no change in E2F2−/− mice. These findings suggest a higher potential for the E2F2−/− hepatocytes for TG and PL synthesis and metabolization. Also related to lipid metabolism, Acsl4, an acyl-CoA synthetic isoenzyme that has been demonstrated to have preference for arachidonic acid as substrate (5), was upregulated in E2F2−/− mice. The gene corresponding to heme oxygenase-1 (Hmox1), the substrate inducible isoform of the rate-limiting step in heme degradation, was also upregulated in the absence of E2F2.
Remarkably, functional categories that are crucial for liver regeneration, including cell cycle, apoptosis, and response to wounding, were only enriched in the WT samples. By contrast, the biological processes related to transport, secretion, and response to other organisms were only overrepresented in the E2F2−/− samples.
To get a better understanding of interconnections among the genes included in the overrepresented GO categories during liver regeneration, we performed a network analysis by using the STRING database (37). The analysis carried out with WT samples generated a complex map of gene interactions that could be arranged into functional clusters related to protein translation, proteasomal functions, glutathione metabolism, oxidative phosphorylation, amino acid metabolism, lipid metabolism, DNA replication/cell cycle, cytokine signaling/focal adhesion, and response to wounding/coagulation (Supplemental Fig. S3A), arguing that an integrated network of these molecular pathways is necessary for liver regeneration. Gene interaction maps generated with E2F2−/− samples showed many similarities with WT controls but also remarkable differences; most notably interaction networks related to DNA replication/cell cycle and response to wounding/coagulation were absent in E2F2-deficient samples, and an interaction network related to spliceosomal functions was only present in E2F2-deficient samples (Supplemental Fig. S3B).
Analysis of genes involved in cell cycle, apoptosis, and response to wounding.
We examined the genes involved in the regulation of cell cycle and apoptosis that showed E2F2-dependent modification in liver regeneration (Table 3). Most of them were only regulated in WT mice (Tgfb1, Cdh1, Cdkn1c, Pola1), suggesting that their expression is E2F2 dependent. By contrast, Cdkn1a, a known inhibitor of cdk2 and cdk4, was overexpressed in both genotypes. Cdk6 and genes coding for cyclins E1 (Ccne1) and D1 (Ccnd1) were also upregulated in both genotypes, but, whereas Ccne1 upregulation was higher in WT, Cdk6 and Ccnd1 upregulation was higher in E2F2−/− mice. Remarkably, upregulation of some members of the Mcm and Orc gene family, involved in DNA replication, was only observed in WT livers (Table 3 and Supplemental Fig. S2). However, an earlier (12–24 h) activation of some of these genes (Ccne1, Mcm3, Mcm4) was detected in E2F2−/− livers (Supplemental Fig. S2 and Table S2). Intriguingly, some genes such as the Myc oncogene and the gene coding for E2F dimerization protein Tddp2 were only upregulated in the E2F2−/− mice. Among the cell-cycle genes, Cdc20, whose product is required for the microtubule activation required before anaphase for chromosome segregation, exhibited the greatest regeneration-associated upregulation in WT mice. Notably, additional qRT-PCR experiments showed that overexpression of Cdc20 and of eleven more cell-cycle genes was drastically abolished by disruption of E2F2 (Table 4).
Most of the genes included in the category of apoptosis were only regulated in WT mice (Cradd, Bcl2l2, Casp2, Cycs), with some exceptions, such as the tumor suppressor Bcl2l11, strongly upregulated in E2F2−/− but not in WT mice. The expression changes of many genes involved in the NF-κB pathway, such as Nfkbia, Ikbkg, Tnfrsf21, or Ikbkb, were found to be mediated by E2F2. Similarly, almost all genes involved in the response to wounding (Table 5), including some coagulation factors (F2, F5, F7, and F13b) and complement components (C2, C3, and Cfi), showed an E2F2-mediated expression, as expression changes were observed only in WT mice.
E2F transcription factors are thought to be essential for cell-cycle progression, and deficiency of an E2F member would be expected to lead to defects in proliferation associated with liver regeneration. However, published data with E2F1−/− mice have shown that E2F1 is dispensable for this process (22), suggesting that other E2F family members could be mediating liver regeneration. In this work using E2F2 nullizygous mice, we have demonstrated that disruption of E2F2 leads to impaired liver regeneration, arguing that E2F2 transcription factor is essential for efficient regenerative and proliferative processes that take place after liver injury. The kinetics of liver regeneration was delayed, although not blocked, upon E2F2 loss, and the original liver mass was finally restored in E2F2−/− mice, suggesting the existence of compensatory mechanisms that allowed the progression of liver cells through the cell cycle. E2F1 or E2F3 could be contributing to compensation because their levels were maintained similar to WT controls in E2F2−/− samples undergoing liver regeneration.
Cellular and animal models have shown that E2F2 can play both an activator and a repressor role in cellular proliferation (9, 17, 21, 26, 29, 34, 48). We have previously demonstrated that loss of E2F2 leads to accelerated entry and progression of T cells into the cell cycle (26). By contrast, loss of E2F2 does not promote hepatocyte proliferation upon PH. Instead, it impairs timely DNA replication and cell-cycle progression of regenerating liver cells. These results suggest that the cellular context may determine the functional role played by E2F2, by mechanisms that remain to be defined.
In a recent letter, Dong and Xu reported that a large number of genes involved in cellular proliferation were altered following PH in WT rat (14). In agreement with this, a large array of genes involved in cellular proliferation appears to be regulated in WT mice liver cells following PH, and the functional category cell-cycle regulation is overrepresented in WT samples. In contrast, the majority of these genes did not change in regenerating E2F2−/− samples, indicating that they are under E2F2 control. Indeed, many of these genes, including those coding for proteins involved in the assembly of DNA replication complexes (Mcm3–7, Orc4l, or Orc6l), contain E2F response elements (15, 28, 36, 38, 44) and constitute a tight interaction network, as shown by STRING analysis (Supplemental Fig. S3, A and B). Therefore, it is likely that delayed S-phase entry exhibited by E2F2−/− hepatocytes is due, at least in part, to the absence or reduced expression of such assembly genes. In turn, this would lead to delayed cell-cycle progression (3, 11). Gene expression data provide support for such delay in mutant mice. For example, the expression of Cdc20 and Hells, genes involved in mitosis, are increased in WT but not in E2F2−/− cells. Intriguingly, Myc, an immediate early gene (33), is only upregulated in E2F2−/− mice.
It is also interesting to note that WT mice show greater expression of many genes involved in key mitochondrial processes, indicative of higher mitochondrial activities. The efficiency of mitochondrial oxidative phosphorylation is reduced during the prereplicative phase of liver regeneration, and then, subsequently recovers (16). Therefore, repressed mitochondrial process genes in the E2F2−/− mice suggest that E2F2 deletion retards liver regeneration by reducing mitochondrial activity.
Several studies have described a transient steatosis associated with the first stages of the liver regeneration, suggesting that it is essential for appropriate regeneration (35, 40). In contrast to Newberry et al. (27), who demonstrated that altered hepatic TG content had no effect on liver regeneration, others have shown that regeneration is impaired in genetic models with excess hepatic fat (19, 46, 47) or after disruption of hepatic adipogenesis (35). The present work confirms massive and reversible hepatic steatosis occurring during the first 3 days of the regeneration. This regeneration-associated hepatosteatosis was also evident in the absence of E2F2 but was prolonged and concomitant with delayed liver regeneration. Although we have not specifically examined whether E2F2 directly modulates regeneration-associated hepatosteatosis, others have suggested that they may be independent processes. We show here that, first, the degree of steatosis was similar in both genotypes, and its kinetics correlated with hepatic growth. Second, and most important, no functional category related to steatosis was deregulated upon E2F2 loss. Taken together, our results indicate that E2F2 contributes to liver regeneration by regulating the expression of genes that are essential for timely G1/S progression of the cell cycle, as well as the expression of genes that are specific to the regeneration process. Whether the differential expression of genes involved in response to wounding/coagulation in WT and KO mice is responsible for the faster recovery from steatosis and hepatectomy that WT mice exhibit is an open question. Future studies should elucidate the extent of functional compensation performed by other E2F members in liver homeostasis.
This study was supported by grants from Ministry of Education and Science (SAF2009/12037) and Gobierno Vasco (Saiotek calls and IT-336-10). I. Delgado and A. Iglesias were recipients of Basque Government fellowships for graduate students.
Authors declare that they have no conflict of interests or financial interests.
- Copyright © 2011 the American Physiological Society