|Home | About | Journals | Submit | Contact Us | Français|
Smad family proteins Smad2 and Smad3 are activated by transforming growth factor β (TGF-β)/activin/nodal receptors and mediate transcriptional regulation. Although differential functional roles of Smad2 and Smad3 are apparent in mammalian development, the relative functional roles of Smad2 and Smad3 in postnatal systems remain unclear. We used Cre/loxP-mediated gene targeting for hepatocyte-specific deletion of Smad2 (S2HeKO) in adult mice and generated hepatocyte-selective Smad2/Smad3 double knockouts by intercrossing AlbCre/Smad2f/f (S2HeKO) and Smad3-deficient Smad3ex8/ex8 (S3KO) mice. All strains were viable and had normal adult liver. However, necrogenic CCL4-induced hepatocyte proliferation was significantly increased in S2HeKO compared to Ctrl and S3KO livers, and transplanted S2HeKO hepatocytes repopulated recipient liver at dramatically increased rates compared to Ctrl hepatocytes in vivo. Using primary hepatocytes, we found that TGF-β-induced G1 arrest, apoptosis, and epithelial-to-mesenchymal transition in Ctrl and S2HeKO but not in S3KO hepatocytes. Interestingly, S2HeKO cells spontaneously acquired mesenchymal features characteristic of epithelial-to-mesenchymal transition (EMT). Collectively, these results demonstrate that Smad2 suppresses hepatocyte growth and dedifferentiation independent of TGF-β signaling. Smad2 is not required for TGF-β-stimulated apoptosis, EMT, and growth inhibition in hepatocytes.
Transforming growth factor β (TGF-β) family members regulate a wide range of cellular functions and have essential roles in embryonic development (23). The three TGF-β isoforms act on virtually every cell type in mammals by engaging a ubiquitous intracellular signaling cascade of Smad family proteins through ligand-induced activation of heteromeric transmembrane TGF-β receptor kinases (9, 23). Smad2 and Smad3 may be activated through carboxy-terminal serine phosphorylation by the TGF-β receptor type 1 (Tgfbr1) (25). Receptor-activated Smads (R-Smads) form oligomeric complexes of R-Smad(s) and a common Smad4 and translocate into the nucleus to participate in transcriptional regulation.
Although Smad2 and Smad3 are highly similar and share regulation and overlapping functions, several distinguishing features determine unique patterns of gene activation and signal transduction observed in vitro (49). Smad3-deficient mice are viable but succumb from defects in mucosal immunity between 1 and 8 months of age (7, 50). Numerous reports indicate a central role for Smad3 as a mediator of TGF-β-induced growth inhibition (7), epithelial-to-mesenchymal transition (EMT) (37, 53), and apoptosis (6, 34) and demonstrate its important roles in fibrosis, wound healing, and inflammation in vitro (reviewed in reference 13).
In contrast, various gene targeting approaches demonstrate that Smad2 is required for embryonic axis patterning and endoderm specification (16, 28, 45, 46). Interestingly, Smad2 and Smad3 are coexpressed in the early mouse embryo from the blastocyst stage onward, and both mediate autoregulatory activation of the Nodal gene (11). Unlike Smad3, full-length Smad2 cannot bind DNA due to the presence of a unique 30-amino-acid insert encoding exon 3 of the MH1 domain that is subject to alternative splicing (8, 47). Smad2/Smad3 double homozygous mutants entirely lack mesoderm and fail to gastrulate (11). Recent results demonstrate that the exon3-spliced short form of Smad2 is sufficient to mediate all transcriptional responses of TGF-β during normal development in the absence of full-length Smad2 and Smad3 (12), whereas exon3 containing the long form of Smad2 had no role in mediating essential transcriptional regulation by TGF-β/activin/nodal during development (12). Although these studies indicate that the exon3-less Smad2 isoform and Smad3 signal cooperatively to mediate cell fate decisions in the early mouse embryo, the absolute or cooperative postnatal roles of the long or the short forms of Smad2 and of Smad3 remain unclear. Although conditional knockout models for Smad2 have been generated recently (20, 44), functions of Smad2 in normal adult physiology and disease pathogenesis have not been defined.
In liver, TGF-β inhibits DNA synthesis in primary cultures of rat hepatocytes (5), induces EMT (42), and stimulates apoptosis (29). Liver regeneration induced by necrogenic doses of carbon tetrachloride (CCL4) is a well-studied model of cell cycle control and signal transduction due to the robust kinetic profile of hepatocyte necrosis and proliferation following CCL4-induced liver injury (24). TGF-β1 has been proposed as a major growth factor to arrest proliferation once functional mass has been regained. TGF-β1 mRNA and protein levels increased a few hours after partial hepatectomy or toxic injury (3). The level of phophorylated-Smad2 was dramatically elevated as early as 6 h after CCL4 treatment, suggesting that Smad2 is actively involved in opposing hepatocyte proliferation (22). The functional roles of Smad2 and of Smad3 in normal adult liver, as well as in liver regeneration, remain unclear.
Here we report the generation of a Smad2 conditional knockout mouse model by flanking the exon2 of the Smad2 gene with two LoxP sites. By crossing with Albumin-Cre transgenic mice, we successfully inactivated Smad2 in mouse liver hepatocytes and were able to generate hepatocyte-selective Smad2/Smad3 double knockouts by intercrossing AlbCre/Smad2f/f and Smad3-deficient Smad3ex8/ex8 mice. This genetic system provides the first in vivo model to study inactivation of either Smad2, Smad3, or both TGF-β-activatable Smads in postnatal liver. Our results demonstrate novel roles for Smad2 as cell-autonomous negative regulator of hepatocyte proliferation and migration, as well as mesenchymal transition. In contrast, TGF-β signaling in hepatocytes does not require Smad2 but is dependent on Smad3.
The murine Smad2 gene was isolated as described previously (16). The 2-kbp loxP-neomycin-loxP expression cassette was cloned into BglII site in intron 1-2, and the third loxP was inserted at the PstI site in intron 2-3. Targeting vector (40 μg) was linearized at the unique NotI site and was electroporated into 2 × 107 WW6 embryonic stem (ES) cells (17). Stable transfected ES cells were selected as described previously (16) and screened for homologous recombination events by PCR analysis. Correct targeting events were verified by Southern blot hybridization. Chimeric mice were generated with ES cells from clones 2C6 and 3D3. Chimeric mice were mated to wild-type (WT) C57BL/6 × 129V mice to establish mouse strains with germ line transmission. Heterozygous mice were maintained in a C57BL/6 genetic background. To eliminate the neomycin resistance cassette, heterozygous mice were crossed with cytomegalovirus (CMV)-Cre transgenic mice (kindly offered by W. Edelmann). Partial and complete deletions were detected by PCR analysis in mosaic mice. Mice carrying Smad2f allele were obtained by mating mosaic mice with C57BL/6 mice.
Tail biopsies from F1 animals were analyzed by PCR, and the results were confirmed by Southern blot analysis. F2 animals were analyzed by PCR. The presence of Smad2f allele was detected by primers 127 (5′-TTCCATCATCCTTCATGCAAT-3′) and 101 (5′-CTTGTGGCAAATGCCCTTAT-3′), resulting in a 451-bp PCR product. The wild-type allele gives a 271-bp product.
Albumin-Cre transgenic mouse line (Jackson Laboratory) genotyping followed the protocol provided by JAX lab. Eight-week-old mice were used unless otherwise indicated. By this age, the efficiency of DNA excision by the Albcre transgene is more than 90% (32).
A total of 40 μg of genomic DNA was isolated from total liver and digested by BamHI. 5′ 1-kb BamHI/BglII fragment was used to probe the digested genomic DNA. Probe was labeled by using a random primer labeling kit (RadPrime DNA Labeling System; Invitrogen). Hybridizations were performed by using standard procedures.
Immunoblotting was performed as previously described (53) with the following antibodies: anti-Smad2 monoclonal antibody (raised against a peptide sequence at the linker region of mouse Smad2) (BD Transduction Laboratories), affinity-purified polyclonal antibody (detecting the phosphorylated serine residues in the C terminus) of Smad2 (Cell Signaling), anti-c-Myc (Santa Cruz), antivimentin (BD Transduction Laboratories), anti-p21CIP1 (Chemicon), anti-cyclin D1 (Santa Cruz), anti-p15INK4b (Santa Cruz), anti-E-cadherin (Zymed Laboratories), anti-MK14 (Covance Research Products, Inc.), anti-Myc (Invitrogen), antihemagglutinin (anti-HA; Sigma), and anti-β-tubulin antibody (Sigma).
Wild-type (WT) and Smad2 knockout mouse embryo fibroblasts (MEFs) were seeded at a density of 2 × 104 cells/well per six-well dish. The next day, cells were transfected with reporter constructs using Effectene transfection reagent (QIAGEN) according to the manufacturer's protocol. Luciferase activity was analyzed by using a Dual-Luciferase Reporter Assay kit (Promega). In all transfections, the expression plasmid pRL-CMV served as an internal control to correct for transfection efficiency.
After attachment of hepatocytes, medium was switched to 0.2% serum medium containing all of the supplements. Cells were treated with TGF-β1 (5 ng/ml) for 48 and 72 h. DNA fragmentation was detected as previously described (38).
Cultured hepatocytes were harvested by cell scrapers after 24 h of TGF-β1 (5 ng/ml) treatment. Caspase 3 activity was analyzed according to the manufacturer's instructions (BD ApoAlert Caspase Assay Plates; BD Bioscience) by using a 1420 Multilabel counter (Perkin-Elmer).
Immunofluorescence staining was performed as described previously (53). Fluorescein isothiocyanate-phalloidin (Sigma) was used to detect F-actin. E-cadherin was detected by using mouse anti-E-cadherin (Zymed), followed by the addition of Cy3-conjugated donkey anti-mouse antibody (Jackson Immunoresearch). Cell nuclei were counterstained with DAPI (4′,6′-diamidino-2-phenylindole; Sigma). Fluorescent images were acquired by a Zeiss Axioskop fluorescence microscope at Mount Sinai Medical Center.
Primary cultured hepatocytes were “wounded” by scratching the cells with a 200-μl pipette tip in the presence or absence of TGF-β1 after attachment and then monitored after 24 h by phase-contrast microscope photography as described previously (4).
Mouse hepatocytes were isolated from livers of 8-week-old mice by a modified two-step collagenase perfusion protocol (27). The hepatocytes were plated on collagen 1-coated dishes in Dulbecco modified Eagle medium-F-12 (Gibco/BRL) with supplements as described previously (2). A total of 2 × 105 cells/well was seeded into 12-well plates for [3H]thymidine incorporation assay. After the cells were attached, the medium was changed to 0.2% fetal bovine serum medium containing supplements and 10 ng of epidermal growth factor (Upstate Biotechnology, Lake Placid, NY)/ml. DNA synthesis in primary hepatocytes was measured after incubation the cells with [3H]thymidine for 24, 48, and 72 h as previously described (2).
A single dose of diluted CCL4 solution (875 μl of Crisco oil was added to 125 μl of CCL4 and filtered by using 0.22-μm-pore-size filter) was administered by intraperitoneal injection at a dose of 3.3 ml/kg (body weight). Livers were excised for analysis 1, 2, and 3 days after injection.
Primary hepatocytes were transplanted into dipeptidyl peptidase IV−/− (DPPIV−/−)Rag2−/− mice by intrasplenic injection, as previously reported (52). The control transplantation group includes transplantations with donor hepatocytes isolated from WT (C57BL/6J) and Smad2f/f mice, respectively. Multiple sections from both the middle and left liver lobes were taken from each recipient at various times after cell transplantation. DPPIV staining was used to quantify donor cell engraftment and proliferation, as described previously (52). The slides were examined independently by two observers and scored in a blinded fashion (by A.O. and D.A.S.). All results were expressed as mean ± the standard deviation. Repopulation efficiencies in the various groups were compared by using the Student t test.
Total RNA was prepared by using TRIZOL reagent and quantitative reverse transcription-PCR (qRT-PCR) was performed as described previously (53). The sequences of primers used are available upon request.
Mice carrying conditional Smad2f alleles (Fig. (Fig.1A)1A) were generated as described in detail in Materials and Methods. To investigate postnatal functional roles of Smad2 in liver, we crossed Smad2f/f mice with Albcre transgenic mice, expressing Cre recombinase in hepatocytes under control of a rat albumin promoter/enhancer (31). Deletion of exon 2 by Cre-mediated recombination at the Smad2 locus was confirmed by Southern blot (Fig. (Fig.1B),1B), and deletion of Smad2 protein was confirmed by Western blot analysis in total liver lysates of Albcre/Smad2f/f mice (Fig. (Fig.1C)1C) and in primary hepatocytes isolated from Albcre/Smad2f/f mice (S2HeKO) (Fig. (Fig.1D).1D). Note that a reduced amount of wild-type Smad2 is detectable in whole liver lysates from S2HeKO mice due to the presence of nonparenchymal, Albcre-negative cells (Fig. (Fig.1C,1C, lanes 1 and 2). C-terminal serine phosphorylation of Smad2 was induced by TGF-β in Ctrl but not in S2HeKO cells (Fig. (Fig.1E,1E, upper panel). Recent studies in MEFs carrying homozygous deletion of exon 2 of Smad2 demonstrated a truncated phosphorylated Smad2 consisting of an MH2 domain and a small part of the linker region (33). A faint band consistent with the truncated Met241 phospo-Smad2 protein of ca. 26 kDa was detectable only after extensively prolonged film exposure (Fig. (Fig.1E,1E, lower panel). However, the abundance of truncated phospho-Smad2 (P-Smad2Δ1-240) in S2HeKO cells stimulated with TGF-β1, as determined by band densitometry, was ~80-fold lower compared to full-length phospho-Smad2 in Ctrl cells at 1 and 4 h of TGF-β stimulation.
Levels of wild-type and truncated Smad2 transcripts were not significantly different in WT, S2HeKO, Smad3-deficient Smad3ex8/ex8 (S3KO), and double-knockout (DKO) hepatocytes (Fig. (Fig.2B).2B). It is currently not possible to assess the abundance of unphosphorylated Smad2Δ1-240 (Fig. (Fig.2A)2A) because suitable antibodies are not available at the time of this writing. To compare the relative levels of phosphorylation of full-length Smad2 (Smad2FL) and truncated Smad2Δ1-240 directly, we transfected MEFs with Myc-tagged Smad2FL and Myc-tagged Smad2Δ1-240 expression plasmids, followed by TGF-β treatment (Fig. (Fig.2C).2C). TGF-β induced phosphorylation of a significant fraction of Myc-Smad2FL (Fig. (Fig.2C,2C, lanes 3 and 4). In contrast, only a very minor fraction of total Myc-Smad2Δ1-240 was phosphorylated in untreated and TGF-β treated transfected cells (Fig. (Fig.2C,2C, lanes 5 and 6). Next, we examined the effect of recombinant expression of Smad2Δ1-240 or Smad2FL in S2KO MEFs on Smad2-dependent transcriptional activation by using ARE-Lux reporter contransfection assays (Fig. (Fig.2D).2D). Reconstitution of Smad2FL in S2KO MEFs restored the induction of ARE-Lux reporter activity by TGF-β. In contrast, expression of Smad2Δ1-240 did not restore TGF-β inducibility of ARE-Lux. In addition, cotransfection of Smad2FL and Smad2Δ1-240 had no significant effect on induction of Smad3-dependent SBE4 Luc reporter activity induced by TGF-β in WT MEFs (Fig. (Fig.2E2E).
By crossing Smad2f/f mice with mice carrying Albcre transgene (31) and Smad3dex8 allele (50), we generated offspring with the following genotypes: Smad2f/f Smad3+/+ (normal control [Ctrl]); Smad2f/f Smad3dex8/dex8 (Smad3 knockout [S3KO]); Albcre/Smad2f/f Smad3+/+ (hepatocyte-specific Smad2 knockout [S2HeKO]); and Albcre/Smad2f/f Smad3dex8/dex8 (Smad2/Smad3 double knockout [DKO]). S2HeKO mice were viable and fertile and manifested normal postnatal liver and body growth and function for up to 1 year of observation. S3KO were viable and manifested normal liver growth and function but succumbed to a primary defect in mucosal immune function at between 1 and 8 months, as previously reported (50). Interestingly, postnatal liver morphology, growth, and function were normal in DKO mice within the life span afforded by S3KO, indicating that Smad2 and Smad3 were not required for liver development and homeostasis after a critical stage during embryonic development.
We generated primary hepatocytes from 8-week-old Ctrl, S2HeKO, S3KO, and DKO mice according to standard protocols (27). Interestingly, the viability of primary DKO hepatocytes immediately after isolation and prior to plating was consistently and dramatically reduced (lower than 70% viable cells) compared to Ctrl, S2HeKO, and S3KO cells (Fig. (Fig.3A).3A). After attachment of primary hepatocytes on standard collagen-coated culture dishes, S2HeKO cells were elongated with process formation and cell separation compared to Ctrl cells, which had cobblestone epithelial morphology (Fig. (Fig.3B).3B). S3KO cells resembled the epithelial morphology of Ctrl hepatocytes, but cell-cell contacts and intercellular space appeared to be wider compared to Ctrl cells (Fig. (Fig.3B).3B). In contrast, only few DKO cells were able to attach with atypical morphology after seeding (Fig. (Fig.3B).3B). Thus, Smad2 and Smad3 deficiency caused distinct hepatocyte phenotypes, respectively, and loss of Smad2 was characterized by a fibroblastoid/mesenchymal cell morphology. In addition, lack of both Smad2 and Smad3 was associated with atypical cell morphology and features of cell stress, a finding consistent with decreased viability of primary DKO cells prior to seeding.
Because S2HeKO cell morphology appeared mesenchymal compared to Ctrl cells, we reasoned that Smad2 may have a role in regulating EMTs of hepatocytes (15). We examined molecular markers of EMT in primary Ctrl and S2HeKO hepatocytes. Ctrl cells manifested characteristic cell-cell junction staining for adherens junction protein E-cadherin and cortical F-actin (Fig. (Fig.4).4). Treatment with TGF-β for 24 h was sufficient to induce EMT in Ctrl cells, as demonstrated by loss of E-cadherin staining and redistribution of F-actin in stress fibers (Fig. 4A to C). E-cadherin adherens junctions and cortical actin staining was reduced, while stress fiber labeling was increased in S2HeKO cells at baseline compared to Ctrl cells (Fig. 4A to B), indicating that S2HeKO cells were unable to maintain the characteristic epithelial hepatocyte phenotype. Importantly, transient or stable overexpression of truncated Smad2Δ1-240 in H2.35 hepatocytes had no effect on characteristic epithelial cobblestone morphology and on E-cadherin protein levels, while overexpression of Snail, a known inducer of EMT in hepatocytes, was able to induce mesenchymal changes and to repress E-cadherin levels in these cells (data not shown). These results suggest that the dedifferentiation of S2HeKO hepatocytes was due to Smad2 deficiency rather than the presence of low levels of truncated Smad2Δ1-240. In contrast, S3KO hepatocytes manifested an epithelial phenotype at baseline, and TGF-β treatment had no effect on E-cadherin and F-actin expression (data not shown), as previously demonstrated (53). Due to the sparse attachment and growth patterns and the atypical cell morphology of DKO hepatocytes (see Fig. Fig.3B),3B), similar analysis was not possible in DKO cells.
Next, we examined the effect of Smad2 or Smad3 deficiency on cell motility using standard scratch-wound assays as previously described (53). At 24 h after wounding, untreated Ctrl cells were unable to migrate into the wound area, whereas TGF-β induced a partial coverage of the wound area by Ctrl cells (Fig. (Fig.4D).4D). In contrast, S2HeKO cells were covering the entire width of the scratch wound area in the absence of TGF-β (Fig. (Fig.4D).4D). S3KO cells did not cover scratch wound at baseline, and TGF-β had no significant effect (Fig. (Fig.4D;4D; note that a larger pipette tip was used for S3KO wounding, resulting in a wider scratch wound area).
In addition, epithelial marker protein E-cadherin was reduced, and mesenchymal marker vimentin was increased, at baseline in S2HeKO cells (Fig. (Fig.4E).4E). TGF-β caused significant reductions of E-cadherin levels and de novo expression of vimentin in Ctrl cells within 24 h. S3KO cells manifested an intermediate molecular profile whereby baseline vimentin was elevated compared to Ctrl cells, but TGF-β treatment decreased vimentin protein in contrast with the opposite result observed in Ctrl cells (Fig. (Fig.4E).4E). Thus, untreated and TGF-β-regulated levels of vimentin are different between Ctrl and S3KO cells, indicating a role for Smad3 in control of vimentin expression. However, baseline and TGF-β-stimulated vimentin levels in S2HeKO significantly exceeded vimentin levels in Ctrl and S3KO, irrespective of treatment (Fig. (Fig.4E).4E). E-cadherin abundance was comparable at baseline between Ctrl and S3KO cells; however, TGF-β treatment suppressed E-cadherin in Ctrl cells but failed to suppress E-cadherin in S3KO cells (Fig. (Fig.4E).4E). Thus, Smad2 and Smad3 deficiency exert differential roles in the maintenance of epithelial hepatocyte phenotypes and TGF-β-inducible EMT. Our observations demonstrate a functional requirement for Smad2 in the maintenance of the differentiated epithelial hepatocyte phenotype and in suppression of promigratory mesenchymal phenotype transitions, a finding consistent with constitutive loss of E-cadherin expression and gain of vimentin expression. In contrast, Smad3 is required for TGF-β-inducible EMT and migration, a finding consistent with a critical role as mediator of TGF-β-induced suppression of E-cadherin. In addition, Smad3 deficieny interferes with constitutive and TGF-β-regulated expression of the mesenchymal marker vimentin.
TGF-β may induce apoptosis in hepatocytes in vitro and in vivo (29). Several proapoptotic signaling events have been described (reviewed in references 1 and 40); however, the precise roles of Smad2 and Smad3 in TGF-β-induced apoptosis in hepatocytes remain unknown. TGF-β treatment induced a characteristic DNA fragmentation (laddering) in Ctrl and S2HeKO but not in S3KO hepatocytes after 48 and 72 h (Fig. (Fig.5A).5A). These results were confirmed by a quantitative assay measuring activity of effector caspase 3 (Fig. (Fig.5B),5B), demonstrating that Smad3 is an essential mediator of apoptosis induced by TGF-β in hepatocytes. In contrast, absence of Smad2 did not alter proapoptotic signaling by TGF-β in these assays, indicating that Smad2 is not involved in apoptosis signaling.
We used established [3H]thymidine incorporation assays as previously described (2) to examine the effect of increasing concentrations of TGF-β on proliferation of primary Ctrl, S2HeKO, and S3KO hepatocytes after 24, 48, and 72 h. TGF-β1 reduced [3H]thymidine incorporation in a dose-dependent manner in Ctrl cells, irrespective of the time of treatment (Fig. (Fig.6A).6A). [3H]thymidine incorporation was significantly more reduced at low concentrations of TGF-β treatment in S2HeKO cells compared to Ctrl cells, indicating that S2HeKO may be more sensitive to growth inhibition by TGF-β. In contrast, TGF-β treatment had no significant effect on [3H]thymidine incorporation in S3KO cells (Fig. (Fig.6A).6A). Surprisingly, baseline [3H]thymidine incorporation in untreated S2HeKO hepatocytes was persistently close to threefold elevated compared to Ctrl and S3KO cells (Fig. (Fig.6B).6B). Together, these results suggest that Smad3, but not Smad2, is required for TGF-β-inducible inhibition of G1 cell cycle progression, whereas Smad2, but not Smad3, suppresses cell cycle progression constitutively and independently of TGF-β stimulation in hepatocytes.
The increased basal proliferation rate in S2HeKO cells was associated with strikingly increased cyclin D1 protein at baseline compared to Ctrl cells (Fig. (Fig.6C).6C). Cyclin D1 expression was also increased in S3KO cells, but to a much lesser extent compared to S2HeKO (Fig. (Fig.6C).6C). TGF-β suppressed cyclin D1 protein in S2HeKO cells and to a lesser degree in Ctrl and S3KO cells (Fig. (Fig.6C).6C). In contrast, p21CIP1 was strongly increased at baseline in S3KO hepatocytes compared to Ctrl and S2HeKO cells (Fig. (Fig.6C),6C), and TGF-β induced p21CIP1 at 8 h in Ctrl cells and after 24 h in S2HeKO (Fig. (Fig.6C)6C) but had no additional effect on p21CIP1 protein levels in S3KO cells. In addition, baseline expression levels of c-Myc and of p15INK4b proteins were comparable between Ctrl, S2HeKO, and S3KO, respectively (Fig. (Fig.6C).6C). TGF-β downregulated c-Myc protein and upregulated p15INK4b protein in Ctrl and, surprisingly, in S3KO cells (Fig. (Fig.6C).6C). Transient overexpression of Smad2Δ1-240 in H2.35 hepatocytes had no effect on cyclin D1 mRNA levels (data not shown), suggesting that the increased level of cyclin D1 in S2HeKO was due to Smad2 deficiency rather than the presence of low levels of Smad2Δ1-240. Interestingly, Smad2 deficiency was associated with upregulation of c-Myc and failure to induce p15INK4b in S2HeKO cells (Fig. (Fig.6C),6C), indicating a selective role for Smad2 in mediating suppression of c-Myc and the associated induction of p15INK4b by TGF-β.
Together, these findings suggest a novel functional role for Smad2 as a selective constitutive repressor of cyclin D1 and of baseline proliferation in primary hepatocytes, although Smad2 is not required for mediating antiproliferative signals of TGF-β, and Smad2 deficiency is associated with increased sensitivity to growth inhibition by TGF-β. In contrast, Smad3 deficiency was associated with a strong constitutive increase of p21CIP1 protein and a loss of antiproliferative function of TGF-β, but it was not required for repression of cyclin D1 and c-Myc and for induction of p15INK4b by TGF-β. Thus, Smad2 and Smad3 exert distinct roles in control of cell cycle regulators and in mediating antiproliferative TGF-β signaling.
To validate whether S2HeKO hepatocytes have a constitutive sustained growth advantage in vivo, we used a liver cell transplantation model as previously described (52). To control for both genetic background of ES cell lineages and the absence or presence of Smad2, we used donor hepatocytes generated from Smad2f/f and normal (WT) C57BL/6J mice as control groups (Ctrl) for comparison with S2HeKO donor hepatocytes. Histochemical labeling demonstrated that the number and size of DPPIV+ (red) donor cell clusters was increased in mice transplanted with S2HeKO donor cells (Fig. 7D to F) compared to Smad2f/f and C57BL/6J donor hepatocytes at 1, 2, and 3 months posttransplantation, as indicated (Fig. 7A to C). The total percent liver repopulation was significantly higher for S2HeKO donor cells compared to Smad2f/f and C57BL/6J donor hepatocytes (Fig. (Fig.7G).7G). The average number of cells per cluster was also increased significantly in S2HeKO compared to both controls (Fig. (Fig.7H),7H), indicating that S2HeKO donor cells had a persistent, long-term growth advantage compared to control cells. These results demonstrate that Smad2 deficiency confers a significant and sustained cell-autonomous growth advantage upon hepatocytes during liver repopulation in vivo.
Next, we determined whether Smad2 and/or Smad3 have functional roles in toxic liver injury and liver regeneration by using a model of toxic liver injury induced by a single injection of CCL4 (52). CCL4 exposure induced localized centrilobular necrosis/apoptosis/steatosis peaking at 48 h that was not significantly different between Ctrl, S2HeKO, and S3KO mice (Fig. 8A to C). In contrast, CCL4 induced advanced bridging central injury in DKO livers (Fig. (Fig.8D).8D). Together with the decreased viability of DKO cells after hepatocyte isolation (Fig. (Fig.3A),3A), these results demonstrate that deficiency of both Smad2 and Smad3 dramatically increased susceptibility to hepatotoxic injury in vivo, indicating that essential cell survival mechanisms are impaired in hepatocytes lacking both Smad2 and Smad3.
In addition, we analyzed rates of hepatocyte proliferation after CCL4 liver injury which induces a well-characterized proliferative response resulting in liver regeneration. At 24 h after CCL4 injection, fractions of Ki67-positive hepatocytes were less than 1% and not significantly different between Ctrl, S2HeKO, and S3KO (Fig. (Fig.9G).9G). By 48 h, Ki67-positive hepatocytes were comparable in Ctrl and S3KO livers, representing 25.41% ± 5.7% and 29.20% ± 2.2% of total hepatocytes, respectively (Fig. 9D and F, respectively). However, the fraction of Ki67 positive hepatocytes was significantly higher (50.03% ± 5.3%) in S2HeKO livers compared to Ctrl and S3KO (Fig. 9E and G). Similar results were obtained at 72 h after CCL4 injection; however, the differences between Ctrl, S3KO, and S2HeKO did not reach statistical significance by day 3 (Fig. (Fig.9G).9G). These results demonstrate that Smad2-deficient hepatocytes, but not Smad3-deficient cells, proliferate at increased rates during early liver regeneration after acute liver injury. These in vivo findings are consistent with the observed increased growth rates of Smad2-deficient primary hepatocytes in vitro and transplanted hepatocytes in vivo and establish a role for Smad2 as a negative regulator of hepatocyte proliferation in vivo.
To examine both the relative and the absolute roles of R-Smads in the regulation of target genes of TGF-β, we used qRT-PCR to determine steady-state transcript levels of extracellular matrix associated genes and genes involved in EMT in Ctrl, S2HeKO, S3KO, and DKO hepatocytes after exposure to TGF-β. All examined genes that were strongly upregulated by TGF-β in Ctrl cells showed similar responses in S2HeKO cells (Fig. 10A). In contrast, induction of these genes was significantly diminished but not completely eliminated in S3KO cells (Fig. 10A). Interestingly, TGF-β stimulation of these genes was completely eliminated, and baseline levels were considerably lower compared to the other genotypes specifically in DKO hepatocytes (Fig. 10A). The two examined genes that were strongly repressed by TGF-β in Ctrl hepatocytes, E-cadherin (Cdh1) and Snai2, displayed a similar response profile in S2HeKO cells. In contrast, their repression by TGF-β was reduced in S3KO and DKO hepatocytes compared to Ctrl and S2HeKO cells (Fig. 10B). Our findings demonstrate that, in the presence of Smad3, Smad2 is not required for induction or repression of target genes in hepatocytes. In contrast, in the presence of Smad2, Smad3 mediates partial gene induction and gene repression by TGF-β. The absence of both Smad2 and Smad3 resets baseline levels of TGF-β target genes and completely abolishes transcriptional regulation by TGF-β.
Our study defines for the first time distinct functional roles for Smad2 and Smad3 in postnatal liver. We report that postnatal development and homeostasis are not perturbed in mice with hepatocyte-specific deletion of Smad2 or a double knockout of Smad2 and Smad3. Thus, R-Smads are not required to maintain normal physiological conditions in hepatocytes. These findings are consistent with a recent report demonstrating normal development and appearance of liver in mice with Cre/loxP mediated hepatocyte-specific knockout of TgfbR2 (36), indicating that TGF-β signaling is dispensable in hepatocytes under unchallenged conditions in vivo. In addition, it is possible that Smad2 and Smad3 deficiency in hepatocytes may be compensated for under physiological conditions by unspecified stromal-epithelial interactions.
In contrast, our studies demonstrate distinct functional roles of Smad2 and Smad3 in hepatocytes when the physiological context is challenged in vivo. We identify a novel functional role for Smad2 as a cell-autonomous negative regulator of hepatocyte growth in vitro and in vivo. This conclusion is supported by significantly increased baseline rates of proliferation of primary Smad2-deficient hepatocytes compared to wild-type and Smad3-deficient hepatocytes in vitro and by significantly and persistently increased growth rates of transplanted Smad2-deficient hepatocytes for up to 3 months in a liver cell transplantation model in mice. These results suggest that basal proliferation of Smad2-deficient hepatocytes is reset to permit increased proliferative rates compared to wild-type control cells and that the reset is likely independent of TGF-β activity. The increased proliferation was associated with a strikingly increased abundance of cyclin D1. Cyclin D1 promotes G1/S progression of cell cycle (26), and its rapid increase in regenerating liver after partial hepatectomy was increased in hepatocyte-specific Tgfbr2 knockout livers compared to WT (30). Interestingly, TGF-β completely suppressed aberrant cyclin D1 expression in S2HeKO, indicating that the increased basal level of cyclin D1 is independent of extracellular TGF-β.
Our results further demonstrate that Smad2 mediates negative control of hepatocyte proliferation after toxic liver injury induced by single necrogenic dose of CCl4 (52). This conclusion is supported by significantly increased hepatocyte proliferation in livers of S2HeKO mice compared to control and Smad3KO mice. It remains to be determined whether this function is TGF-β dependent or independent. However, a recent report demonstrated that nuclear localization of Smad2 was reduced by 50% in Albcre/Tgfbr2f/f livers compared to controls after partial hepatectomy (36), indicating that activation of Smad2 may be dependent on TGF-β signaling in a different model of regeneration. Our results confirm an antimitogenic role for Smad2 in liver regeneration after acute toxic injury, whereas we were unable to demonstrate a role for Smad3. We also observed that both long Smad2FL and short Smad2Δexon3 transcripts are present in this model (data not shown). It will be important to determine which of these splice forms mediates the antimitogenic activity, since short Smad2Δexon3 has been shown to be capable of mediating all TGF-β signaling during embryonic development, whereas long Smad2FL was neither essential nor sufficient for TGF-β signaling or embryonic development (12).
In contrast to our in vivo observations, our in vitro studies demonstrate that Smad3 is required for TGF-β-stimulated growth inhibition of primary hepatocyte, whereas Smad2 was not required. However, the antimitogenic role of Smad3 appears to be highly cell type and/or context dependent. For example, Smad3 is not required for TGF-β growth inhibition in mammary gland epithelial cells (51) but is required in ras-transformed keratinocytes (43). Overexpression of Smad2, but not of Smad3, significantly reduced growth of Mv1Lu mink lung epithelial cells xenografted in SCID mice (39). Moreover, epidermal keratinocyte proliferation was reduced in 12-tetradecanoyl phorbol myristate acetate-induced skin of Smad3-deficient mice compared to controls (19). Together our in vivo and in vitro results and these published observations indicate that a role of Smad3 as a mediator of TGF-β-induced epithelial growth inhibition may not be universally applicable and appears to be highly context dependent.
Surprisingly, Smad2-deficient hepatocytes in primary culture were unable to maintain a differentiated epithelial phenotype and spontaneously manifested morphological and molecular features characteristic of EMT similar to those induced by TGF-β in control hepatocytes. In contrast, untreated S3KO cells had characteristic differentiated epithelial cell features, and TGF-β was unable to induce EMT in these cells, indicating that Smad3 was required for TGF-β induced EMT in vitro, as previously reported (37, 53). Although previous studies using overexpression or dominant-negative interference systems indicated that Smad2 and Smad3 mediate TGF-β-induced EMT (41), the results from our genetic approach presented here suggest that Smad2 is not involved in TGF-β-induced EMT in primary hepatocytes, whereas Smad3 is required. In fact, our results suggest that Smad2 function is required for stable epithelial phenotype of primary hepatocytes, possibly independent of TGF-β activity. However, the relevance of this surprising Smad2 function in vivo, for example, in liver injury or hepatocarcinogenesis, remains to be determined.
Our findings have implications for hepatocarcinogenesis. Genetic inactivation of Tgfbr2, Smad2, and Smad4 in hepatocellular carcinoma have been reported, albeit with various frequencies (14, 18, 48). In addition, there is substantial evidence for epigenetic inactivation of Smad2 or Smad3 function in various cancers, commonly mediated by direct interaction with oncoproteins (10, 21). Moreover, TGF-β may exert dual roles in carcinogenesis where TGF-β switches from a tumor suppressor in the premalignant stages of tumorigenesis to a proto-oncogene function at later stages of disease leading to metastasis (35). The molecular determinants that mediate this functional switch remain poorly understood. Our results lead us to speculate that the loss of Smad2 function (genetic or epigenetic) during carcinogenesis may increase the proliferative and metastatic potential, whereas loss of Smad3 function may decrease the metastatic potential of tumor cells. Thus, it is possible that the critical functional switch may be determined by changes of the relative balance of Smad2 (antimetastatic signaling) and Smad3 (prometastatic signaling) in malignant cells.
The present results confirm previous results obtained from mouse fibroblasts (49) indicating that Smad2 function is not required or may negatively modulate transcriptional regulation by TGF-β if Smad3 is present. Interestingly, Smad3-deficient hepatocytes showed various degrees of partial transcriptional responsiveness to TGF-β if Smad2 is present, depending on the individual target gene. In contrast, transcriptional activity of TGF-β was universally lost in hepatocytes lacking both Smad2 and Smad3. Based on genetic isoform replacement experiments in mice, expression of the Smad2 splice form Smad2Δexon3 or Smad3 cDNAs under control of the Smad2 locus was sufficient to mediate all transcriptional responses of TGF-β required for normal development in Smad2-deficient embryos (12). In contrast, expression of full-length Smad2 cDNA under control of the Smad2 locus was not sufficient to rescue phenotypes of Smad2 knockout embryos, indicating that the long form of Smad2 may exert functions unrelated to transcriptional regulation by TGF-β (12). Both Smad2 isoforms are present in hepatocytes (data not shown). Thus, our results suggest that the Smad2Δexon3 may compensate partially for transcriptional function of Smad3 in S3KO hepatocytes. Moreover, Smad2/Smad3 double knockout is also associated with a striking reset (reduced) of baseline transcript levels of target genes that are activated by TGF-β, indicating that deficiency of Smad2 and Smad3 profoundly alters the basal and inducible transcriptome of hepatocytes. Interestingly, although double deficiency of Smad2 and Smad3 is compatible with normal hepatocyte function under physiological conditions postnatally, DKO hepatocytes have dramatically increased susceptibility to undergo necrosis or apoptosis in response to toxic liver injury in vivo and to stress induced by standard hepatocyte isolation procedures.
We are grateful to Yutong Zhang for expert technical assistance and to Scott Friedman for helpful discussions. Hepatocyte isolation was performed with expert assistance by David Neufeld at the Cell Culture and Genetic Engineering Core of the Marion Bessin Liver Research Center (supported by 2P30DK041296-16).
Microscopy was performed at the MSSM-Microscopy Shared Resource Facility, supported, in part, with funding from NIH-NCI shared resources grant (R24 CA095823). W.J. was supported by the American Heart Association Research Fellowship. This study was supported by NIH grants RO1DK56077, RO1DK60043, and P5ODK064236-01003 to E.P.B.