Transforming growth factor β (TGF-β) is a potent inhibitor of hepatocyte proliferation in vitro and is suggested to be a key negative regulator of liver growth. To directly address the role of TGF-β signaling in liver regeneration in vivo, the TGF-β type II receptor gene (Tgfbr2) was selectively deleted in hepatocytes by crossing “floxed” Tgfbr2 conditional knockout mice with transgenic mice expressing Cre under control of the albumin promoter. Hepatocytes isolated from liver-specific Tgfbr2 knockout (R2LivKO) mice were refractory to the growth inhibitory effects of TGF-β1. The peak of DNA synthesis after 70% partial hepatectomy occurred earlier (36 vs. 48 hours) and was 1.7-fold higher in R2LivKO mice compared with controls. Accelerated S-phase entry by proliferating R2LivKO hepatocytes coincided with the hyperphosphorylation of Rb protein and the early upregulation of cyclin D1 and cyclin E. However, by 120 hours after partial hepatectomy, hepatocyte proliferation was back to baseline in both control and R2LivKO liver. Regenerating R2LivKO liver showed evidence of increased signaling by activin A and persistent activity of the Smad pathway. Blockage of activin A signaling by the specific inhibitor follistatin resulted in increased hepatocyte proliferation at 120 hours, particularly in R2LivKO livers. In conclusion, TGF-β regulates G1 to S phase transition of hepatocytes, but intact signaling by TGF-β is not required for termination of liver regeneration. Increased signaling by activin A may compensate to regulate liver regeneration when signaling through the TGF-β pathway is abolished, and may be a principal factor in the termination of liver regeneration. (HEPATOLOGY 2004.)
Transforming growth factor β (TGF-β) is a multifunctional cytokine with diverse effects on development, growth, and homeostasis in most tissues.1 Mammals have three forms of TGF-β (TGF-β1-3), all of which reside on different chromosomes but are approximately 80% identical at the level of amino acid sequence.2 Signaling by TGF-β occurs through type I and type II transmembrane serine/threonine kinase receptors and intracellular Smad transduction molecules (see review by Shi and Massague3). The TGF-β ligand binds to its type II receptor, either directly or via coreceptors. Once activated by TGF-β, type II receptors recruit, bind, and transphosphorylate type I receptors, thereby stimulating their protein kinase activity. The activated type I receptors phosphorylate Smad2 or Smad3, which then bind to Smad4. The resulting Smad complex then moves into the nucleus, where it interacts in a cell-specific manner with various transcription factors to regulate the transcription of target genes.3 In most epithelial, endothelial, and hematopoietic cells, TGF-β is a potent inhibitor of cell proliferation.1 Signaling by TGF-β arrests the cell cycle in the G1 phase by stimulating production of the cyclin-dependent protein kinase inhibitors p15 and by inhibiting the function or production of essential cell cycle regulators, especially the cyclin-dependent protein kinases 2 and 4, and cyclins D1 and E.4 These changes result in decreased phosphorylation of the retinoblastoma gene product Rb, allowing it to bind to and sequester members of the E2F family of transcription factors and arrest cell cycle progression at the G1/S checkpoint.
A unique feature of the adult mammalian liver is its ability to accurately regenerate lost mass, such as occurs following surgical resection or diffuse liver injury. The molecular events that trigger liver regeneration are now beginning to unfold (see review by Michalopoulos and DeFrances5), but little is known about the mechanisms that restrict proliferation and return hepatocytes to quiescence after liver regeneration is complete. The role of TGF-β signaling in the control of liver regeneration has been studied mainly with a partial hepatectomy (PH) model in rats. After surgical resection of liver lobes, the remaining cells proliferate in a synchronous fashion to restore the liver mass.6 DNA synthesis is mostly complete by 72 hours, but the whole process may last up to 2-3 weeks.7 TGF-β is a logical candidate to limit and/or stop liver regeneration following PH.2, 7 However, studies to test this hypothesis have yielded conflicting results.2, 8 TGF-β has potent antiproliferative effects on hepatocytes in vitro.9 Expression of liver TGF-β1 messenger RNA increases very soon after PH, but it is not clear whether peak levels are reached before or after hepatocyte proliferation.10, 11 Nevertheless, these results suggest that mitoinhibitory effects of endogenous TGF-β may contribute to the termination of hepatocyte proliferation observed following the wave of DNA synthesis in the regenerating liver. Infusion of TGF-β into partially hepatectomized rats significantly delays the onset of DNA synthesis, although the inhibitory effects are transient.12 Interestingly, during liver regeneration, hepatocytes acquire temporary resistance to TGF-β,13 allowing them to proliferate despite rising levels of this cytokine. The mechanism is unknown, but it may be a result of downregulation of TGF-β receptors14 or upregulation of inhibitors of the TGF-β signaling pathway.15 It has been suggested that return of TGF-β sensitivity at later stages may limit hepatocyte proliferation and terminate liver regeneration.16
Because embryos lacking the TGF-β type II receptor (TβIIR) die in utero,17 we performed hepatocyte-specific disruption of TβIIR in conditional knockout mice to directly examine the role of TGF-β signaling in the control of liver regeneration in vivo. We show that disruption of TGF-β signaling in the liver enhances the proliferative response of hepatocytes after PH but does not affect the termination of liver regeneration. Increased signaling by activin A may act as a backup mechanism to regulate liver regeneration when signaling through the TGF-β pathway is blocked and may be involved in termination of liver regeneration.
TGF-β, transforming growth factor β; Tgfbr2, gene for TGF-β type II receptor; R2LivKO mice, liver-specific Tgfbr2 knockout mice; PH, partial hepatectomy; TβIIR, TGF-β type II receptor; Cre-Ctrl mice, Cre-expressing littermate control mice; PCR, polymerase chain reaction; BrdU, 5-bromo-2′-deoxyuridine.
Materials and Methods
Tgfbr2 conditional knockout mice were generated using the Cre-loxP approach as described previously.18 Exon 4 of mouse Tgfbr2 was selected as a target for mutagenesis, because this exon encodes the entire transmembrane domain as well as a majority of the kinase domain of the receptor, both of which are essential for receptor function.19 To obtain liver-specific disruption of Tgfbr2 in the liver, Tgfbr2 mice homozygous for the floxed allele were crossed with mice homozygous for an Alb/Cre transgene (The Jackson Laboratory, Bar Harbor, ME). The (Tgfbr2fl/+ × Alb/Cre+/−) mice so generated were interbred with littermates of the same genotype. Subsequently, (Tgfbr2fl/+ × Alb/Cre+/−) mice were crossed with (Tgfbr2fl/+ × Alb/Cre-/−) offspring to generate (Tgfbr2fl/fl × Alb/Cre+/−) liver-specific Tgfbr2 knockout (R2LivKO) mice and (Tgfbr2+/+ × Alb/Cre+/−) Cre-expressing littermate control (Cre-Ctrl) mice. Liver-specific gene deletion approaches 100% at 6 weeks postnatally using the albumin promoter to drive Cre expression.20 Therefore, mice aged 7-8 weeks were used for all experiments in this study.
Genotyping and Screening for Gene Deletion.
Genotyping was performed via polymerase chain reaction (PCR) on DNA isolated from tail clippings as described previously.21 Primers P1 (5′-TAT GGA CTG GCT GCT TTT GTA TTC-3′) and P2 (5′-TGG GGA TAG AGG TAG AAA GAC ATA-3′), which amplify the region around the 5′ loxP site (Fig. 1A), were used to detect the wild-type (422-bp) and floxed (575-bp) alleles of Tgfbr2 (Fig. 1B) as described previously.18 The Cre transgene was detected using primers Cre-F (5′-AGG TGT AGA GAA GGC ACT TAG C-3′) and Cre-R (5′-CTA ATC GCC ATC TTC CAG CAG G-3′) to yield a 411-bp product (Fig. 1B) as described previously.22 Detection of liver-specific disruption of Tgfbr2 was performed using primers P1 (5′-TAT GGA CTG GCT GCT TTT GTA TTC-3′) and P3 (5′-TAT TGG GTG TGG TTG T-3′) as described previously.18 PCR amplification yields a 692-bp product from Tgfbr2 null (−/−) liver only and no product from wild-type (wt/wt) or floxed (fl/fl) liver (Fig. 1C).
Growth Inhibition Assay.
Hepatocytes were isolated by a two-step collagenase perfusion of the liver.23 Viable hepatocytes were plated onto uncoated 96-well plates at a density of 5 × 103 cells per well and cultured in a defined medium as described previously.24 After overnight incubation in serum-free medium, human TGF-β1 (R&D ystems, Minneapolis, MN) was added to the cultures in concentrations of 0, 1.0, and 5.0 ng/mL, and the cells were cultured for another 72 hours. Hepatocyte viability was measured using a colorimetric assay based on the cleavage of the tetrazolium salt WST-1 by mitochondrial dehydrogenase in viable cells (Roche Diagnostics, Indianapolis, IN). The colorimetric reaction was quantified using a SPECTRAmax PLUS microplate spectrophotometer and SOFTmax PRO software (both from Molecular Devices, Sunnyvale, CA). The effects of exogenous TGF-β1 on viability of isolated hepatocytes were confirmed by trypan blue staining of hepatocytes cultured under the conditions detailed above. The medium was removed 72 hours after treatment and the cultures washed twice with phosphate-buffered saline and trypsinized. Cell viability was determined with a hemocytometer and trypan blue exclusion using a 0.4% solution of trypan blue in phosphate-buffered saline.
All experiments were performed in accordance with the National Institutes of Health guidelines for the humane use of laboratory animals. Mice (20-25 g) of matched sexes were subjected to a standard 70% PH under isoflurane anesthesia, and tissue was collected between 24 and 336 hours postoperatively. The livers were removed, weighed, and normalized to body weight. Samples of tissue were fixed in 70% alcohol/formalin (10:1) for 5-bromo-2′-deoxyuridine (BrdU) immunohistochemistry.
To monitor the kinetics of liver DNA synthesis, the animals were given intraperitoneal injections of BrdU (Boehringer Mannheim, Indianapolis, IN) at 150 mg/kg body weight 1 hour before sacrifice. S-phase nuclei were immunostained with a BrdU mononclonal antibody (Becton Dickinson, San Jose, CA), the reaction was developed with a Vector ABC Elite Kit (Vector Laboratories, Burlingame, CA), and the tissue was counterstained with hematoxylin as described previously.25 DNA synthesis was determined by counting BrdU-positive nuclei, and the labeling index was expressed as the percentage of the total cells counted. Approximately 2,000-3,000 hepatocyte nuclei were counted per animal at each time point.
Staining for phosphorylated Smad2 was performed using the EnVision+ System (Dako Cytomation, Carpinteria, CA) according to the manufacturer's instructions. Following deparaffinization and blocking of endogenous peroxidase, target antigen retrieval was performed by heating to 95°C for 25 minutes in 1 mmol/L ethylenediaminetetraacetic acid. The sections were incubated with rabbit anti–phospho-Smad2 (Chemicon International, Temecula, CA) at 1:100 dilution or negative control reagent. Thereafter, peroxidase-labeled polymer conjugated to goat anti-rabbit immunoglobulin was added, and the antigen-antibody reaction was detected using 3,3′-diaminobenzidine chromogen solution. The positively stained nuclei were revealed under light microscopy after counterstaining with hematoxylin.
The apoptotic index was scored on hematoxylin-eosin-stained livers from 3 to 4 animals per time point. Briefly, 1,000 hepatocytes/mouse were randomly evaluated with a light microscope (Nikon Microphot FXA). The index was represented as a percent (mean ± SE) of the total cells counted.
Administration of Follistatin.
Recombinant human follistatin, a specific activin-binding protein that blocks the action of activin A,26 was generously provided by Dr. Yuzuru Eto of the Central Research Laboratory, Ajinomoto Co., Ltd. (Kawasaki, Japan). Mice were subjected to 70% PH, immediately followed by administration of follistatin at 1 μg dissolved in 200 μL of phosphate-buffered saline into the inferior vena cava with a 30-gauge needle. A booster dose at 1 μg was given via tail vein injection 48 hours after PH to sustain elevated intrahepatic levels of follistatin up to 120 hours.27 Control animals received intravenous injections of phosphate-buffered saline at the same volume. Animals were sacrificed 120 hours after PH, and the remnant liver was removed and processed as described in the Partial Hepatectomy section.
Western Blot Analysis.
Whole cell lysates were prepared from liver samples using T-Per tissue protein extraction reagent (Pierce, Rockford, IL), and protein concentrations were determined using the Bradford assay (Bio-Rad, Hercules, CA) with bovine serum albumin as the standard. Protein samples (100 μg per lane) were separated via electrophoresis in a 10% to 20% NOVEX Tris-glycine precast gel (Invitrogen, Carlsbad, CA) under reducing conditions, and transferred to a polyvinylidene fluoride membrane (Invitrogen) by electroblotting. Standard immunoblotting procedures were performed, and the immunoreaction was detected via enhanced chemiluminescence using SuperSignal West Pico (Pierce) as the substrate. The following antibodies were used: cyclin D1, cyclin E, Smad2, Smad3, Smad4, and phospho-Smad2/3 (Santa Cruz Biotechnology, Santa Cruz, CA); pRb (Oncogene, San Diego, CA); and β-actin (Chemicon International).
RNA Isolation and Reverse-Transcriptase PCR.
Total RNA was isolated from approximately 100 μg of mouse liver using the guaniduim isothiocyanate extraction method. One microgram of RNA was converted to complementary DNA using the Superscript First-Strand Synthesis System (Invitrogen). Complementary DNA was then amplified via PCR using the following specific primers: 5′-TGC TGC ACT TGA AGA AGA GAC CC-3′ (sense) and 5′-TGG TCC TGG TTC TGT TAG CCT TG-3′ (antisense) for activin A; 5′-GGG AAA GAG ACA GAA CCA ACC AGA-3′ (sense) and 5′-TGG GCT GTG TGA CTT CCA TCT C-3′ (antisense) for activin receptor R2A; and 5′-CCA TGT TTG TGA TGG GTG TGA ACC-3′ (sense) and 5′-TGT GAG GGA GAT GCT CAG TGT TGG-3′ (antisense) for glyceraldehyde-3-phosphate dehydrogenase control. The PCR products were separated via electrophoresis through a 2% agarose gel containing ethidium bromide, and bands were visualized on a UV light box.
Differences between the groups were assessed using Student t test. A P value of .05 was considered statistically significant.
Hepatocytes From R2LivKO Liver Are Refractory to the Growth Inhibitory Effects of TGF-β1.
We performed hepatocyte-specific disruption of Tgfbr2 using conditional knockout mice with the Cre-loxP system and an albumin-expressing transgene (Fig. 1A-C). To confirm dysregulation of TGF-β signaling in hepatocytes, we examined the ability of cultured hepatocytes from R2LivKO mice and floxed littermate controls to escape the growth inhibitory effect of TGF-β1 using a 96-well proliferation assay. As shown in Fig. 1D, TGF-β1 at concentrations of 1.0 and 5.0 ng/mL resulted in a more than 60% reduction in viability of control hepatocytes. On the other hand, TGF-β1 had no effect on the viability of hepatocytes from R2LivKO mice, indicating that these cells were insensitive to the effects of TGF-β1. Similarly, trypan blue staining showed that exogenous TGF-β1 (both 1.0 and 5.0 ng/mL) resulted in an approximately 50% to 60% reduction in viablity of control hepatocytes, while R2LivKO hepatocytes remained unaffected (data not shown).
Enhanced DNA Synthesis, Faster Recovery but Normal Termination of Regeneration in Liver-specific Tgfbr2 Knockout Mice.
We next performed PH experiments on R2LivKO and Cre-Ctrl mice to address the specific role of TGF-β signaling in the control of liver regeneration. Mortality after 70% PH in R2LivKO and Cre-Ctrl mice was 4.4% and 5.6%, respectively, which was higher than in the background strain (B6/129 hybrid), suggesting a toxic effect of Cre expression in the liver.28 The BrdU labeling index curve for R2LivKO mice was shifted to the left compared with that of Cre-Ctrl mice, indicating a shortened interval to the onset of DNA synthesis (36 hours vs. 48 hours for Cre-Ctrl mice), and peak DNA synthesis was increased by 1.7-fold in R2LivKO mice (Fig. 2A). However, by 120 hours after PH, DNA synthesis was virtually back to baseline in both R2LivKO and Cre-Ctrl mice (Fig. 2A). The observation that G1 to S cell cycle transition was accelerated during liver regeneration in the absence TGF-β signaling was confirmed by Western blot evidence of the early induction of several cell cycle genes. As shown in Fig. 2C, regenerating liver from R2LivKO mice showed early phosphorylation of the Rb protein. TGF-β is known to cause growth arrest of hepatocytes by blocking the phosphorylation Rb protein by cyclin-activating complexes at the G1/S checkpoint of the cell cycle.3, 7 Similarly, there was early and prominent expression of cyclin D1 and, particularly, cyclin E in regenerating R2LivKO liver compared with Cre-Ctrl liver (Fig. 2C). In keeping with the enhanced cell cycle kinetics in regenerating R2LivKO liver, conditional knockout mice showed more rapid recovery of liver mass, with a significant increase in liver/body weight ratio at 96 and 120 hours after PH (Fig. 2B). Although there appeared to be a trend toward increased liver mass in R2LivKO mice 14 days after PH (Fig. 2B), the difference was not statistically significant. These results indicate that termination of liver regeneration occurs normally in R2LivKO mice, despite the absence of TGF-β signaling in the livers of these animals.
Persistent Smad Activation in Regenerating Livers from Liver-Specific Tgfbr2 Knockout Mice.
Smad proteins are the critical downstream targets of signaling by TGF-β.3 To assess the activity of the Smad pathway during liver regeneration in the presence of disrupted TGF-β signaling, we performed Western blot analysis on timed liver specimens of key downstream Smad proteins and their phosphorylated forms. As shown in Fig. 3A, the induction of Smad proteins and their phosphorylation in response to PH was dampened in R2LivKO liver compared with Cre-Ctrl, and the effects on Smad2/3 expression and phosphorylation declined after 48 hours in knockout liver. Nevertheless, there was persistent Smad activation in R2LivKO liver, despite elimination of TGF-β signaling. We thus directed our attention to activin A, whose type II receptor signals through the same Smads as the TβIIR. As shown in Fig. 3B, regenerating R2LivKO liver showed increased expression of messenger RNA for both activin A and its type II receptor (activin R2A) compared with Cre-Ctrl, and this was particularly evident after the 24-hour time point. These findings suggest that increased signaling through the activin pathway acts as a compensatory mechanism to regulate liver regeneration in R2LivKO liver.
Blockage of Signaling by Activin A in Liver-Specific Tgfbr2 Knockout Mice Delays Termination of Liver Regeneration.
Because the activin A pathway, like TGF-β, has been shown to inhibit hepatocyte proliferation,29 it was necessary to directly address the contribution of the activin pathway in arresting liver regeneration in the presence of disrupted TGF-β signaling. Signaling through the activin pathway can be blocked by follistatin, a physiological inhibitor of activin A action.26 We thus administered follistatin to mice at the time of PH, with a booster dose at 48 hours, to provide sustained blockage of the activin pathway. We specifically wanted to determine whether disruption of signaling by both the activin and TGF-β pathways affected the termination of liver regeneration. Liver tissue was thus analyzed 120 hours after PH, a time point at which hepatocyte proliferation should have virtually ceased. As shown in Fig. 4, phosphorylated Smad2 at 120 hours after PH was reduced in the livers of mice treated with follistatin (Fig. 4C, D), indicating efficient interruption of Smad pathway activation. Follistatin treatment resulted in a 2.2-fold increase in BrdU labeling in Cre-Ctrl liver (P < .05) and a 2.9-fold increase in R2LivKO liver (P < .007) at 120 hours (Fig. 5A). Furthermore, there was a significant increase in BrdU labeling of liver from R2LivKO plus follistatin mice compared with Cre-Ctrl plus follistatin (P < .02) at that time point (Fig. 5A). These results were confirmed by BrdU immunohistochemistry (Fig. 5D-G), which showed increased nuclear staining of hepatocytes at 120 hours in mice that received follistatin, particularly in R2LivKO mice (Fig. 5G). However, there was no significant difference in liver body mass between the two genotypes (Fig. 5B), most likely because of the higher rate of apoptosis observed in the R2LivKO mice (Fig. 5C). Taken together, these results suggest that signaling by activin A, not TGF-β, contributes to termination of hepatocyte proliferation after PH.
TGF-β has been proposed as a major candidate to limit the proliferation of hepatocytes during liver regeneration and to stop regeneration once functional liver mass has been regained.7 To evaluate the role of TGF-β signaling in the regulation of liver regeneration in vivo, we used a floxed TβIIR allele in conditional knockout mice to eliminate signaling through any of the three TGF-β isoforms in the liver. We postulated that disruption of TGF-β signaling in hepatocytes might result in prolonged proliferation of hepatocytes during liver regeneration induced by a 70% PH and could possibly delay the termination of the regenerative response. Intriguingly, despite some small alterations, liver regeneration appeared largely unperturbed by the lack of TGF-β signaling in hepatocyte-specific Tgfbr2 knockout mice. Proliferating R2LivKO hepatocytes showed acceleration of S-phase entry and augmentation of DNA synthesis, which resulted in more rapid recovery of liver mass; however, termination of liver regeneration occurred normally around 120 hours after PH, and no excess mass was gained. In support of these findings, we previously found that transgenic mice overexpressing TGF-β1 in the liver showed suppression of early DNA synthesis after PH, but termination of liver regeneration was unaffected.30 Similarly, Russell et al.12 showed that infusion of TGF-β1 and TGF-β2 into partially hepatectomized rats resulted in a marked but transient suppression of DNA synthesis during liver regeneration. Taken together, these results indicate that the rising levels of TGF-β detected after PH10, 11, 31 act to limit the proliferative response of hepatocytes during liver regeneration through restriction of G1 to S phase transition. However, intact TGF-β signaling is clearly not required to stop hepatocyte proliferation once the deficit in liver mass has been replaced. Our findings in conditional knockout mice thus argue against a return of sensitivity to TGF-β signaling due to increased expression of TGF-β receptors by hepatocytes16 as being a physiologically important mechanism underlying termination of liver regeneration.
The data obtained from PH experiments in our R2LivKO mice have identified the activin pathway as a potentially important source of inhibitory signals during liver regeneration. Activin A, whose type II receptor signals through the same Smad proteins as TβIIR, is known to be induced during liver regeneration.29 In partially hepatectomized rats, blockage of activin A action by recombinant follistatin increases remnant liver weight and liver regeneration rate at 120 hours, suggesting an effect of activin signaling on the termination of liver regeneration.32 Similarly, we found that follistatin treatment increased the proliferation rate at 120 hours, particularly in the presence of disrupted TGF-β signaling. It is important to note, however, that unlike targeted disruption of TβIIR, which is complete and permanent, the effects of follistatin are transient and wear off rapidly after 72 hours.27 We addressed this problem by administering a second dose of follistatin, which has been shown to markedly elevate the levels in the liver at 120 hours.27 Nevertheless, we found that follistatin treatment resulted in a decrease in Smad2 phosphorylation in the R2LivKO livers, suggesting that activin, not TGF-β, contributes to termination of hepatocyte proliferation after PH. In addition, the combined effects of TGF-β and activin A signaling on terminating the regenerative response appear to be relatively small. Therefore, there remains the possibility that neither TGF-β nor activin is a major factor in hepatic regeneration. Accordingly, the observation that apoptosis may compensate for increased DNA synthesis and maintain liver mass in follistatin-treated R2LivKO mice would further argue that other molecules are involved in terminating liver growth. G-protein-coupled endothelial differentiation gene and sphingosine 1-phosphate have recently been shown to have antiproliferative effects during liver regeneration after PH.33
In conclusion, TGF-β appears to limit the proliferative response of regenerating hepatocytes through inhibition of G1 to S phase cell cycle transition. However, intact signaling by TGF-β is not required for normal termination of liver regeneration. Signaling through the activin pathway may compensate when signaling by TGF-β is blocked, establishing activin as a principal factor in the termination of liver regeneration.
We thank Tyjen L. Tsai for technical assistance with the screening of the mice and Tanya Hoang for help with the immunohistochemical staining.