Extracellular matrix modulates sensitivity of hepatocytes to fibroblastoid dedifferentiation and transforming growth factor β–induced apoptosis


  • Patricio Godoy,

    1. Molecular Alcohol Research in Gastroenterology, Department of Medicine II, Faculty of Medicine at Mannheim, University of Heidelberg, Germany
    2. IfADo-Leibniz Research Centre for Working Environment and Human Factors at the Technical University Dortmund, Dortmund, Germany
    Search for more papers by this author
  • Jan G. Hengstler,

    1. IfADo-Leibniz Research Centre for Working Environment and Human Factors at the Technical University Dortmund, Dortmund, Germany
    Search for more papers by this author
  • Iryna Ilkavets,

    1. Molecular Alcohol Research in Gastroenterology, Department of Medicine II, Faculty of Medicine at Mannheim, University of Heidelberg, Germany
    Search for more papers by this author
  • Christoph Meyer,

    1. Molecular Alcohol Research in Gastroenterology, Department of Medicine II, Faculty of Medicine at Mannheim, University of Heidelberg, Germany
    Search for more papers by this author
  • Anastasia Bachmann,

    1. Molecular Alcohol Research in Gastroenterology, Department of Medicine II, Faculty of Medicine at Mannheim, University of Heidelberg, Germany
    Search for more papers by this author
  • Alexandra Müller,

    1. Molecular Alcohol Research in Gastroenterology, Department of Medicine II, Faculty of Medicine at Mannheim, University of Heidelberg, Germany
    Search for more papers by this author
  • Gregor Tuschl,

    1. Merck KGaA, Merck Serono, Early and Explanatory Toxicology, Darmstadt, Germany
    Search for more papers by this author
  • Stefan O. Mueller,

    1. Merck KGaA, Merck Serono, Early and Explanatory Toxicology, Darmstadt, Germany
    Search for more papers by this author
  • Steven Dooley

    Corresponding author
    1. Molecular Alcohol Research in Gastroenterology, Department of Medicine II, Faculty of Medicine at Mannheim, University of Heidelberg, Germany
    • Department of Medicine II, Gastroenterology and Hepatology, University Hospital, Mannheim, Germany, Theodor-Kutzer Ufer 1-3, 68135 Mannheim, Germany
    Search for more papers by this author

  • Potential conflict of interest: Nothing to report.


Hepatocytes in culture are a valuable tool to investigate mechanisms involved in the response of the liver to cytokines. However, it is well established that hepatocytes cultured on monolayers of dried stiff collagen dedifferentiate, losing specialized liver functions. In this study, we show that hepatocyte dedifferentiation is a reversible consequence of a specific signaling network constellation triggered by the extracellular matrix. A dried stiff collagen activates focal adhesion kinase (FAK) via Src, leading to activation of the Akt and extracellular signal-regulated kinase (ERK) 1/2 pathways. Akt causes resistance to transforming growth factor β (TGF-β)–induced apoptosis by antagonizing p38, whereas ERK1/2 signaling opens the route to epithelial–mesenchymal transition (EMT). Apoptosis resistance is reversible by inhibiting Akt or Src, and EMT can be abrogated by blocking the ERK1/2 pathway. In contrast to stiff collagen, a softer collagen gel does not activate FAK, keeping the hepatocytes in a state where they remain sensitive to TGF-β–induced apoptosis and do not undergo EMT. In this culture system, inhibition of p38 as well as overexpression of constitutively active Akt causes apoptosis resistance, whereas constitutively active Ras induces EMT. Finally, we show that matrix-induced EMT is reversible by replating cells from dried stiff to soft gel collagen. Our results demonstrate that hepatocyte dedifferentiation in vitro is an active process driven by FAK-mediated Akt and ERK1/2 signaling. This leads to similar functional and morphological alterations as observed for regenerating hepatocytes in vivo and is reversible when Akt and/or ERK1/2 signaling pathways are antagonized. Conclusion: Hepatocytes can exist in a differentiated and a dedifferentiated state that are reversible and can be switched by manipulating the responsible key factors of the signaling network. (HEPATOLOGY 2009.)

In most tissues, cells are surrounded by extracellular matrix (ECM) that determines processes such as differentiation, proliferation, and apoptosis.1, 2 Recognition of biochemical and biophysical properties of ECM and cell adhesion are mediated by integrins.3 By anchoring the cytoskeleton to the plasma membrane, integrins connect physical properties of the ECM to the cell.4-6 Integrin–ECM interaction leads to activation of signal transduction pathways through integrin-associated signaling modules such as focal adhesion kinase (FAK) and Src family tyrosine kinases. Collagen type-I, the most common component of the ECM,7 has been used extensively as a substrate for basic cell–matrix interactions in vitro.8-11 The biophysical properties of collagen have profound effects on cell behavior. For instance, contact to a stiff layer of dried collagen stimulates spreading and promotes cell proliferation, whereas adhesion to a malleable collagen gel matrix (as normally found in vivo) inhibits cell cycle entry.8, 12

Primary cultured hepatocytes are a valuable tool to investigate molecular mechanisms involved in the response of the liver to cytokines and growth factors and to study drug metabolism.13-15In vitro, hepatocytes are remarkably sensitive to the ECM in which they are cultured. Essential hepatocyte features such as polarity, bile canalicular transport, enzymatic activities, and metabolic functions are progressively lost when plated on plates coated with a stiff dried collagen. In turn, when cultured between two layers of collagen gel (referred to as a collagen sandwich), most of these features are preserved for extended culture periods.13, 16-18 This indicates that biochemical and physical features of collagen strongly influence hepatocyte morphology and physiological behavior. After a few days on stiff collagen–coated plates, hepatocytes start to spread and acquire a fibroblast-like shape, whereas on a collagen sandwich (CS) they remain in a distinctive cuboidal shape for long periods.19 In line with this, hepatocytes on dried collagen are responsive to growth factor–induced cell cycle progression, whereas in contact to collagen cell cycle entry is blunted.8 Obviously, stiff versus soft collagen induces different cell states of hepatocytes. In this study, we demonstrate that matrix-induced dedifferentiation results in epithelial–mesenchymal transition (EMT) and apoptosis resistance, and that both cell states are reversible consequences of specific signaling network constellations, whereby extracellular signal-regulated kinase (ERK) signaling is responsible for EMT and Akt signaling mediates resistance to apoptosis.


CG, collagen gel; CM, collagen monolayer; CS, collagen sandwich; ECM, extracellular matrix; EMT, epithelial–mesenchymal transition; ERK, extracellular signal-regulated kinase; GAPDH, glyceraldehyde 3-phosphate dehydrogenase; FAK, focal adhesion kinase; JNK, c-Jun N-terminal kinase; mRNA, messenger RNA; PARP, poly(ADP-ribose) polymerase; PCR, polymerase chain reaction; TGF-β, transforming growth factor β.

Materials and Methods


A complete list of chemicals and antibodies can be found in the Supporting Materials and Methods (Supporting Tables 1 and 2).

Mouse Hepatocyte Isolation and Culture.

Primary hepatocytes were isolated from livers of male Black6 mice (8 to 12 weeks of age) using collagenase perfusion as described.20 Cells were seeded on plates precoated with a solution of 250 μg/mL of rat tail collagen I for monolayer cultures (CM), or in a CS as described.19 A detailed protocol can be found in the Supporting Materials and Methods. In some experiments, hepatocytes were seeded onto CM followed by addition of an upper layer of collagen gel (CM+CG), or simply plated onto a single layer of collagen gel (CG). Collagen preparation and medium changes were performed as described for CS. In some experiments, hepatocytes plated on CM for 3 days were trypsinized for 5 minutes and replated onto either CM or CS for an additional 72 hours.

Immunofluorescence and Confocal Microscopy.

Immunostainings were performed as described.21 Nuclei were stained with SYTOX Green for AML12 cells or DRAQ5 for primary hepatocytes. Confocal microscopy and image acquisition were performed as described.21

Messenger RNA Isolation, Complementary DNA Synthesis, and Polymerase Chain Reaction.

Total messenger RNA (mRNA) was extracted from primary mouse hepatocytes using the RNAeasy Mini Kit (Qiagen, Hilden, Germany). Complementary DNA was synthesized from 1 μg RNA with the Quantitech Reverse Transcription kit (Qiagen, Hilden, Germany). Polymerase chain reaction (PCR) was performed with the Taq DNA Polymerase and dNTPack system (Roche, Mannheim, Germany). Real-time PCR for vimentin, claudin-1, and snail was performed with Applied Biosystems TaqMan kits as described.21 A complete list of primers used can be found in the Supporting Materials and Methods (Supporting Table 3). All real-time PCR quantifications are representative of at least two independent experiments.

Oligonuceotide Microarray Analysis.

A total of three independent hepatocyte preparations were cultured in collagen monolayer for 1 and 2 days. Total RNA was collected from untreated or transforming growth factor β (TGF-β)–stimulated cells at each time point and purified with the RNeasy Mini kit (Qiagen, Hilden, Germany), and the integrity was verified by denaturating agarose electrophoresis. Reverse transcription, RNA labeling and hybridization to arrays were performed as described.21 Differential expression analysis was performed with analysis of variance in BioConductor (http://www.bioconductor.org) using R V2.4.1.22 Genes with mean expression changes greater than two-fold (log base 2 greater than 1 or −1) and the P value < 0.05 were selected as significantly changed. Gene ontology analysis was performed with DAVID (Database for Annotation, Visualization, and Interpreted Discovery [http://apps1.niaid.nih.gov/david/]).23 Additional information of array analysis can be found in the Supporting Materials and Methods (Supporting Table 4).

Western Blot Analysis.

Immunoblots were performed as described.24 A description for densitometric anaylsis of band intensities can be found in the Supporting Materials and Methods. To validate western blots for signal transduction analysis on both CM and CS, we compared protein lysates obtained as described in the sodium dodecyl sulfate–polyacrylamide gel electrophoresis silver staining section (Supporting Materials and Methods) from hepatocytes in CS, untreated or stimulated for 1 hour with 5 ng/mL TGF-β, to a protein lysate obtained from CM-cultured hepatocytes stimulated with the same dose of TGF-β. pSmad2 and total Smad2 levels were comparable in all conditions tested (Supporting Fig. 1A,B). All western blots are representative of at least two independent experiments.


FAK was immunoprecipitated from CM or CS protein lysates in RIPA buffer as described.20 Briefly, 10 μg of rabbit-anti FAK were added to 500 μg of protein lysates, plus 20 μL of 50% Protein-G beads slurry (Santa Cruz, Heidelberg, Germany), rotated overnight at 4°C, washed three times with ice cold RIPA, boiled for 5 minutes, resolved on 10% sodium dodecyl sulfate–polyacrylamide gel electrophoresis, and transferred to nitrocellulose membranes, followed by immunoblotting of phospho-tyrosine and FAK. Immunoglobulin G isotype control immunoprecipitation showed no unspecific signal for FAK (Supporting Fig. 2).

Adenovirus Preparation, Quantification, and Infection.

Adenovirus vectors were expanded, purified, and quantified as described.24 Infections were performed in CM and CS hepatocytes 4 hours after plating, for 1 hour with 100 i.f.u of adenovirus vectors expressing constitutively active H-Ras-61L,25 constitutively active HA-myr-Akt,26 green fluorescent protein,27 or β-galactosidase.28 The medium was then removed and the cells were washed three times with Hank's balanced salt solution. For CS cultures, a second layer of collagen gel was added after infections, followed by addition of serum-free William's E medium plus 100 nM dexamethasone, on both CM and CS cultures. Infection rate was always higher than 90%, as verified with adenovirus/green fluorescent protein–expressing vector (Supporting Fig. 3).

Apoptosis Analysis via Genomic DNA Fragmentation.

DNA fragmentation was analyzed via DNA laddering electrophoresis or chromatin condensation. Additional information can be found in the Supporting Materials and Methods.

Hepatocyte Viability and Bile Canaliculi Function.

A detailed protocol can be found in the Supporting Materials and Methods.


Dedifferentiation of Primary Hepatocytes in Culture Involves EMT and Resistance to TGF-β–Induced Apoptosis.

Hepatocytes cultured on dishes coated with collagen (CM) spread and dedifferentiate during a 3-day period.19, 21 Spreading and dedifferentiation were strongly reduced by culturing hepatocytes between a CS (Fig. 1A). The dedifferentiation process in CM has features of EMT, namely the acquisition of a fibroblast-like morphology (Fig. 1A), stress fiber formation (Fig. 1B), loss of membrane localization of E-cadherin (Fig. 1C), increased vimentin (Fig. 1D) and snail (Supporting Fig. 4A), and decreased claudin-1 (Supporting Fig. 4B) expression. Conversely, in CS the cells remained with epithelial phenotype for several days (Fig. 1A-D; Supporting Fig. 4B,C). Hepatocytes in CS form active bile canaliculi that were observed to a much lower degree in CM (Fig. 1A; Supporting Fig. 5A-C). TGF-β can potently induce EMT in hepatocytes.21, 29 This effect was observed in both CM and CS cultures (Fig. 1B-D; Supporting Fig. 4A-C). However, due to the spontaneously occurring EMT, TGF-β stimulation only enhanced this process in CM. A well-documented effect of TGF-β in hepatocytes is apoptosis.30-32 Interestingly, hepatocytes strongly differed in their susceptibility to TGF-β–induced apoptosis depending on the culture system. In CS, TGF-β caused massive induction of cell death, evidenced by formation of apoptotic bodies and chromatin condensation (Fig. 2A,B; Supporting Fig. 5), activation of caspase-3, and degradation of poly(ADP-ribose) polymerase (PARP) (Fig. 2C). In contrast, CM-cultured cells showed resistance to TGF-β induced apoptosis (Fig. 2A-C; Supporting Fig. 7). In conclusion, hepatocytes cultured on CM undergo EMT and become resistant to apoptosis.

Figure 1.

Spontaneous and TGF-β–induced EMT in primary hepatocytes. (A) Phase contrast microscopy of primary hepatocytes cultured in CM or in CS. (B) Confocal scan of F-actin staining (red) in hepatocytes. Nuclei are pseudocolored blue. (C) Confocal scanning of E-cadherin (red) in primary hepatocytes cultured in CM. Nuclei are pseudocolored blue. (D) Real-time PCR analysis of mesenchymal marker vimentin in primary hepatocytes.

Figure 2.

ECM modulates TGF-β–induced apoptosis in hepatocytes. Primary hepatocytes were stimulated for 48 hours with 5 ng/mL TGF-β. (A) Top row: Phase contrast microscopy of hepatocytes in CM or CS. Apoptosis is evidenced by cellular condensation in apoptotic bodies (arrows). Bottom row: Hoechst nuclear staining indicates chromatin condensation as a late marker for apoptosis (arrows). Normal nuclei are shown by arrowheads. (B) Quantification of apoptotic bodies and condensed nuclei from control and TGF-β–treated hepatocytes in CM or CS. (C) Western blot of full-length and cleaved PARP, active caspase-3, and pSmad2. β-Actin was used as a loading control. The results are representative of three independent experiments.

Comparative Transcriptome Analysis in CS- versus CM-Cultured Mouse Hepatocytes.

It is known that hepatocytes lose many differentiated functions in monolayer culture,33 which has been associated to transcriptional down-regulation of the respective genes,34 thus giving the impression that hepatocytes in CM suffer a detrimental process. However, our whole genome analysis with Affymetrix arrays indicates that a complex transcriptional response is induced in CM, where more genes are up-regulated than down-regulated (Supporting Table 4). Among the down-regulated genes were glucose 6-phosphatase, glycogen synthase, and several cytochromes and UDP-glucuronosyltransferases (Supporting Table 4), indicating a general loss of liver metabolic functions. Gene ontology analysis of up-regulated genes included groups involved in signal transduction, developmental process, cell communication, regulation of transcription, and, surprisingly, metabolic functions (Supporting Fig. 6A), which suggested that the dedifferentiation in CM is mediated by an active transcriptional response. Indeed, in the presence of protein or RNA synthesis or the inhibitors cycloheximide or actinomycin-D, hepatocytes remained in a honeycomb shape, and spontaneous actin stress fiber formation was strongly diminished (Supporting Fig. 6B). These inhibitors did not induce cell death, because CMFDA hydrolysis to its fluorescent metabolite (a feature of live cells) was not compromised by Act-D or CHX (Supporting Fig. 6B).

From the gene array data we selected three up-regulated genes (vimentin, ZEB1, and snail-1) and one down-regulated gene (claudin-1) that had already been published to represent EMT markers.29, 35 The genes were confirmed as up- or down-regulated via quantitative real-time PCR (Fig. 1D; Supporting Fig. 4A-D). In subsequent experiments, EMT was characterized by the expression of these markers at mRNA level and by morphological markers shown in Supporting Table 5.

Dedifferentiation and Apoptosis Sensitivity Depend on Structure and Composition of the Matrix.

Two features of the ECM may be relevant to explain the differences between CM and CS, namely the spatial localization and the stiffness of the matrix. First, in CM the ECM is a two-dimensional support that only contacts hepatocytes on their ventral side, whereas in CS the presence of two layers of ECM on it provides a three-dimensional support similar to the in vivo configuration in the liver. The second is the stiffness generated by the suprafibrillar configuration of collagen itself. In CM, the fibrils are cross-linked to the plastic surface, whereas in CG and CS, they form a configuration that results in a relatively softer matrix.36-38 Therefore, we asked whether the spatial localization (two-dimensional versus three-dimensional), or its suprafibrillar configuration are critical for hepatocyte dedifferentiation or sensitivity to TGF-β–induced apoptosis. We stimulated hepatocytes cultured in matrix combinations that included seeding onto conventional CM, between CM and a top layer of collagen gel (CM+CG), onto a single layer of collagen gel (CG), or in conventional CS. Interestingly, only CS prevented EMT. Contact to a stiff collagen matrix caused EMT, independent from the presence of an upper layer of soft collagen gel (Supporting Fig. 7B). However, addition of the upper layer led to bile canaliculi formation. In contrast, cells plated onto a single layer of collagen gel form bile canaliculi to a similar extent as cells in CM (Supporting Fig. 5A-C).

A second major observation was that contact to CM caused resistance to TGF-β–induced apoptosis. In contrast, the strongest apoptotic response was observed in cells cultured on CG and CS, judged by extensive induction of apoptotic bodies, degradation of PARP, and activation of caspase-3 (Supporting Fig. 7A-D). These results indicate that bile canaliculi formation depends on a three-dimensional ECM regardless of the nature of the matrix, and that contact to a stiff collagen matrix is sufficient to induce resistance to TGF-β–induced apoptosis.

p38 Activation Mediates TGF-β–Induced Apoptosis in CS.

The enhanced sensitivity to TGF-β–induced apoptosis in CS-cultured hepatocytes could result from a differential response to TGF-β, for instance, mediated via the core ALK5/Smad2-3 complex. In order to compare the kinetics of initiation and termination of Smad signaling, we performed time course stimulation with a low dose of TGF-β over 24 hours. Interestingly, we observed a similar kinetic response to TGF-β induced phosphorylation of Smad1, Smad2, and Smad3 in hepatocytes cultured in CM or CS (Fig 3a; Supporting Fig. 8A).

Figure 3.

Extracellular matrix influences the TGF-β–induced p-38 pathway but not Smad pathways. (A) Western blot analysis of Smad2 and Smad1/3 phosphorylation in CM- and CS-cultured hepatocyes. Cells were stimulated with 1 ng/mL TGF-β for the indicated times. β-Actin was used as a loading control. (B) Time course of CS- and CM-cultured hepatocytes stimulated with 5 ng/mL TGF-β. Western blot analysis of p-p38, pJNK, and pSmad2. β-Actin was used as a loading control. (C) Role of p38 in TGF-β–induced apoptosis. Hepatocytes cultured on CS were stimulated with 5 ng/mL TGF-β in the presence or absence of 5 μM SB43143 or 10 μM SB203580 for 48 hours. Dimethyl sulfoxide was used as a vehicle contol (0.1%). Phase contrast images show apoptotic bodies (arrows), and EMT by spindle shape of hepatocytes. (D) TGF-β–induced apoptosis analyzed via western blotting of PARP (full-length and cleaved). pSmad2 was used as indicator of active TGF-β signaling. Glyceraldehyde 3-phosphate dehydrogenase (GAPDH) was used as a loading control.

TGF-β can also activate non-Smad signal transduction pathways.39 Stress-induced mitogen-activated protein kinase p38 and c-Jun N-terminal kinase (JNK) mediate TGF-β–dependent apoptosis in primary hepatocytes and in the nondifferentiated hepatoma cell line AML12.31, 32, 40 Therefore, we compared TGF-β induction of pJNK and p-p38 in both culture systems. JNK activation was constitutively stronger in CM than in CS, and we did not observe any induction upon TGF-β stimulation (Fig. 3B; Supporting Fig. 8B). Conversely, p38 activation was strongly induced by TGF-β in hepatocytes on CS, whereas no activation was seen in those on CM (Fig. 3B; Supporting Fig. 8C). Consistent with previous reports,32 we found that TGF-β activation of p38 is a relatively late event in hepatocytes, occurring only after 4 hours in AML12 cells (Supporting Fig. 9A) and after 8 hours in primary cells (Fig. 3B). Compared with CM, hepatoyctes on CG and CS showed the strongest sensitivity to p38 activation (Supporting Fig. 9B,C).

Apoptosis is modulated by Bcl-family proteins.41 In hepatocytes, the p38 pathway leads to apoptosis by inducing the expression of the proapoptotic Bcl-family member Bim.42 Consistently, we observed a stronger induction of Bim mRNA in CS-cultured cells than in CM (Supporting Fig. 9D).

Another important effect of TGF-β is down-regulation of antiapoptotic Bcl-2, which depends on the ALK5–Smad3 pathway.43 Regardless of the culture system, TGF-β induced strong down-regulation of Bcl-2 after 48 hours (Supporting Fig. 7C).

To validate the role of ALK5 and p38 in TGF-β–induced apoptosis, we stimulated hepatocytes on CS in the presence of either the ALK5 inhibitor SB431542 or the p38 inhibitor SB203580 (Fig. 3C,D). The ALK5 inhibitor completely abrogated TGF-β–induced Smad2 phosphorylation and apoptosis, while the p38 inhibitor clearly reduced apoptosis (Fig. 3C,D), without altering the level of pSmad2 (Fig. 3D). In marked contrast to its key role in apoptosis, p38 activity does not influence TGF-β–induced EMT (Fig. 3D). This effect was also observed in AML12 cells (Supporting Fig. 10A,B), which indicates that TGF-β–induced EMT and apoptosis are generally controlled by separate pathways in hepatocytes.

Culture in CM Induces Constitutive Activation of the Survival Pathways ERK and Akt in Hepatocytes.

ERK and Akt pathways are involved in hepatocyte proliferation and survival.44-46 Therefore, we compared expression and activation of ERK and Akt in mouse liver tissue in freshly isolated and cultured hepatocytes. In CM, high basal levels of ERK and Akt phosphorylation were observed, which increased during culture (Fig. 4A; Supporting Fig. 11A,B). In contrast, hepatocytes in CS displayed similarly low levels of ERK and Akt phosphorylation as fresh liver tissue (Fig. 4A; Supporting Fig. 11A,B).

Figure 4.

ECM influence on culture-dependent and TGF-β–induced non-Smad pathways. (A) Comparison of ERK and Akt expression and activation in mouse liver tissue, freshly isolated and CM- or CS-cultured hepatocytes. Protein lysates were taken as described in the Materials and Methods, and analyzed via western blotting. GAPDH was used as a loading control. (B) Western blot analysis of total and phosphorylated Akt and ERK in CM- and CS-cultured hepatocyes, unstimulated or treated with 5 ng/mL TGF-β for 1 hour on day 1 and day 2. GAPDH was used as a loading control.

TGF-β may induce ERK and Akt signaling.39, 47 In our study, TGF-β induced phosphorylation of Akt in CM-cultured cells at days 1 and 2, whereas in CS only a very weak effect occured at day 1 (Fig. 4B; Supporting Fig. 11C). ERK phosphorylation was not induced by TGF-β in CM or CS (Fig. 4B; Supporting 11D).

Focal Adhesion Kinase Expression and Activation Are Enhanced in CM, Leading to Activation of ERK and Akt Pathways.

The activation of ERK and Akt in CM-cultured cells led us to investigate potential upstream mediators for this effect. Integrin clustering upon binding to the ECM results in recruitment and activation of FAK, which in turn activates downstream signals that lead to survival and differentiation.48 It has been shown that FAK expression increased depending on whether the cells are cultured on hard substrates compared with collagen gel.49 In agreement with this, FAK expression increased from 24 hours to 48 hours in CM cultures but not in CS at the mRNA and protein level (Fig. 5A,B). Upon recruitment to focal adhesion sites, FAK is activated by phosphorylation at tyrosine 397 (Y397), which results in interaction with SH-2–containing proteins, including Src.48 Src in turn phosphorylates FAK on several tyrosine residues, leading to full activation of FAK kinase activity and increased affinity to adaptor proteins that ultimately induce activation of downstream pathways mediating proliferation and survival.50 Thus, tyrosine phosphorylation of FAK is crucial for its activation and downstream signaling. We observed FAK phosphorylation in both culture systems (Fig. 5C). However, due to the stronger expression of FAK in CM-cultured cells, the total amount of tyrosine-phosphorylated FAK was higher in CM than in CS (Fig. 5C). TGF-β did not enhance phosphorylation of FAK (Fig. 5C). Thus, CM induces enhanced expression of FAK in primary hepatocytes.

Figure 5.

Focal adhesion kinase expression and activation in hepatocytes. (A) Western blot analysis of FAK in hepatocytes cultured on CM or CS, untreated or stimulated with 5 ng/mL TGF-β for the indicated periods. (B) FAK mRNA expression on hepatocytes as described in panel A. rS6 was used as a loading control. (C) Immunoprecipitation and western blot analysis of tyrosine phosphorylation of FAK and total FAK on hepatocytes cultured in CM or CS, treated as indicated in panel A.

Activation of FAK results in recruitment and activation of Src family members,48 which is crucial for its downstream signaling, including ERK and Akt pathways.48, 51 Consistent with this, pharmacological inhibition of Src with PP2 strongly reduced both ERK and Akt phosphorylation in CM (Fig. 6A; Supporting Fig. 12A,B). Inhibition of Akt phosphorylation was stronger than that of ERK.

Figure 6.

ERK and Akt pathways mediate antiapoptotic effects on CM. (A) Western blot analysis of matrix and TGF-β–induced signal transduction in the presence of MEK1/2, PI-3K, and PP2 inhibitors. U0126 (50 μM), LY294002 (25 μM) and PP2 (25 μM) were added for 30 minutes before stimulation for 1 hour with 5 ng/mL TGF-β. GAPDH was used as a loading control. Dimethyl sulfoxide was used as a vehicle control (0.1%). (B) Phase contrast microscopy of hepatocytes on CM, 48 hours after stimulation with 5 ng/mL TGF-β in the presence of the indicated inhibitors, used as described in panel A. Arrows indicate apoptotic cells. (C) Western blot analysis of apoptosis markers (PARP degradation, caspase-3 activation, Bcl-2) and p38 activation in CM-cultured hepatocytes, treated as described in panel B. GAPDH was used as a loading control. (D) Densitometric quantification of p38 phosphorylation relative to total p38, from hepatocytes treated as described in panel B.

Akt Is the Major Antiapoptotic Pathway in CM.

Growing hepatocytes in CM results in activation of ERK and Akt pathways, both known to promote cell survival.52 To determine the relevance of each pathway for apoptosis resistance, we used inhibitors for MEK1/2 (U0126) and PI-3K (LY294006). The PI-3K kinase inhibitor completely abrogated both endogenous and TGF-β–induced p-Akt without altering ERK phosphorylation. The MEK1/2 inhibitor strongly reduced ERK phosphorylation; however, it also had a partial inhibitory effect on p-Akt (Fig. 6A; Supporting 12A,B), indicating that there is a cross-talk between ERK and Akt pathways in CM-cultured hepatocytes, in which Akt activation is partially dependent on MEK/ERK activity. None of these inhibitors influenced TGF-β effects on Smad2 phosphorylation (Fig. 6A; Supporting Fig. 12C). To determine if these pathways inhibit TGF-β–induced apoptosis in CM, we stimulated hepatocytes with TGF-β for 48 hours in the presence of the respective inhibitors. Blocking of either ERK or Akt leads to enhanced sensitivity to TGF-β–induced apoptosis, as observed by the increase in apoptotic bodies (Fig. 6B), caspase-3 activation, and PARP degradation (Fig. 6C). This indicates that in CM both ERK and Akt induce survival signals that antagonize the proapoptotic effect of TGF-β. Interestingly, inhibition of Src induced the strongest increase in TGF-β–induced apoptosis (Fig. 6B,C). Because Src inhibition reduces both ERK and Akt activation, it is likely that the FAK/Src complex participates in induction of antiapoptotic signals in CM.

Because p38 activation is required for TGF-β–induced apoptosis, we analyzed whether inhibition of the antiapoptotic pathways has an effect on p38 activation. Indeed, all inhibitors enhanced TGF-β–induced activation of p-38 (Fig. 6C,D). The strongest impact was observed with PI-3K and Src inhibitors. We conclude that activation of survival pathways Akt and ERK inhibits TGF-β–induced activation of p38.

Constitutively Active Akt Protects Against TGF-β–Induced Apoptosis in CS.

The levels of Akt and ERK phosphorylation were much lower in hepatocytes cultured on CS compared with CM. Because both pathways are important in mediating antiapoptotic signals on CM, we overexpressed constitutively active forms of Akt26 or Ras25 in CS using adenovirus vectors. Expression of ca-Akt strongly inhibited TGF-β–induced apoptosis in CS (Fig. 7A,B), which is consistent with an antiapoptotic role of Akt in CM. Interestingly, expression of caAkt inhibited TGF-β–dependent down-regulation of Bcl-2 (Fig. 7B). Constitutively active Ras did not exert an antiapoptotic effect despite inducing strong activation of ERK (Fig. 7A,B). This is in agreement with the observations made with Akt and ERK inhibitors (Fig. 6) and identifies Akt as main antiapoptotic pathway in CM.

Figure 7.

Constitutive active Ras in CS-cultured hepatocytes leads to EMT, whereas constitutive active Akt results in apoptosis resistance. (A) Phase contrast images of CS-cultured hepatocytes, transduced with the indicated adenovirus constructs. Cells were either unstimulated or treated for 48 hours with 5 ng/mL TGF-β. Apoptotic bodies and fibroblastoid shape are observed as features of apoptosis and EMT. (B) Western blot analysis of ERK and Akt phosphorylation, and apoptosis induction by PARP degradation in CS-cultured hepatocytes treated as described in panel A. GAPDH was used as loading control. (C) Confocal microscopy of stress fiber formation by F-actin staining (red) as a maker of EMT in CS-cultured hepatocytes, treated as indicated in panel A. Nuclei were pseudo-colored in blue. (D) PCR analysis of EMT markers in CS-cultured hepatocytes, treated as described in panel A. rS6 was used as loading control.

Constitutively Active Ras Signaling Induces EMT Independent of TGF-β, Whereas Akt Signaling Inhibits EMT.

Overexpression of constitutively active Ras dramatically induced dedifferentiation of CS-cultured hepatocytes into a fibroblastoid phenotype (Fig. 7A). In addition, Ras signaling up-regulated expression of several EMT markers, independent of TGF-β stimulation (Fig. 7C,D). In contrast, ca-Akt did not induce EMT on its own, and conversely, protected against TGF-β–induced EMT (Fig. 7A,C,D).

CM-Induced EMT and Apoptosis Resistance Are Reversed upon Replating in CS.

To determine if EMT and resistance to apoptosis in CM are irreversible, we trypsinized hepatocytes cultured on CM for 3 days, and replated the cells on either CM or CS. Twenty-four hours after replating, hepatocytes on CS formed bile canaliculi to the same extent as in direct cultures on CS (Fig. 8A; Supporting Fig. 13A). Remarkably, CM-induced ERK phosphorylation was completely abrogated upon replating in CS (Fig. 8B). Conversely, replating onto CM did not induce bile canaliculi formation (Fig. 8A; Supporting Fig. 13A), nor did it reduce ERK activation (Fig. 8B). Consistent with this, EMT was reverted by replating in CS, as observed by a decrease in vimentin (Fig. 8C) and increased claudin-1 expression (Supporting Fig. 13B). In line with this, hepatocytes replated in CS were sensitive to TGF-β–induced apoptosis, whereas cells replated on CM were still resistant to it (Fig. 8D,E).

Figure 8.

CM-induced EMT and apoptosis resistance represents a reversible cell state. (A) Hepatocytes cultured on CM for 3 days were trypsinized and replated onto either CM or CS. Phase contrast microscopy indicates re-epithelization of hepatocytes in CS. (B) Western blot analysis of ERK expression and activation in hepatocytes cultured in CM or CS, or replated in CM or CS as described in panel A. GAPDH was used as a loading control. (C) Real-time PCR analysis of vimentin expression in hepatocytes cultured in CM or CS, and in cells replated on either CM or CS. (D) Western blot analysis of apoptosis by active caspase-3. GAPDH was used as a loading control. (E) Densitometric quantification of active caspase-3 relative to GAPDH, from control and TGF-treated hepatocytes that were replated as described in panel A. Hepatocytes were treated as described in panel B. (F) Integration of signaling pathways induced in CM lead to EMT and modulate TGF-β effects in hepatocytes. The core ALK5/Smad2-3 signaling is activated by TGF-β independent of the ECM (in blue). However, contact to a stiff monomeric collagen matrix is transduced by integrins into FAK/Src activation and further downstream signaling through ERK and Akt. ERK signaling leads to EMT, while Akt exerts antiapoptotic effects over TGF-β–induced p38, thus inhibiting TGF-β–induced apoptosis (in red). The dedifferentiated cell state characterized by EMT and apoptosis resistance is associated with high activity of the signaling factors indicated by red characters. In contrast, the differentiated, apoptosis-sensitive cell state is associated with low activity of the ERK and AKT pathways and by a high sensitivity to TGF-β induction of p38.


In this study, we showed that culturing hepatocytes in CM or CS led to different cell states that can be characterized by specific signal transduction networks (Fig. 8F). Contact to a stiff collagen matrix causes activation of PI-3K/Akt and MEK/ERK pathways. Akt leads to apoptosis resistance, whereas ERK signaling causes EMT. However, hepatocytes in CS can also be forced into EMT or into apoptosis resistance by expression of either constitutively active Ras or Akt. Conversely, resistance to apoptosis in CM can be overcome by PI-3K/Akt inhibition.

Both ERK and Akt activation in CM seem to depend on FAK/Src signaling. Src inhibition led to stronger inhibition on Akt than on ERK (Fig. 6A). This could be due to the fact that activation of ERK is dependent on Y925 phosphorylation of FAK,48, 51 whereas FAK/Src-mediated Akt activation depends on Y397 phosphorylation of FAK.53, 54 Thus, it is possible that inhibition of Src leads to differential effects on tyrosine phosphorylation at these FAK residues, which results in partial inhibition of ERK and complete inhibition of Akt activation.

PP2 also abrogated TGF-β–induced Akt activation. Participation of FAK/Src on TGF-β–induced p-Akt might depend on interactions between TGF-β receptors and integrin complexes at focal adhesion clusters. Indeed, TβR-II interaction with αvβ3 integrins was shown to participate in TGF-β–induced proliferation of lung fibroblasts.55 Similarly, TGF-β–induced p38 activation depends on integrin-β1 activity in NMuMG cells.56

Abrogation of ERK and Akt signaling resulted in enhanced TGF-β–induced p38 activation and enhanced apoptosis. Akt can phosphorylate and thus inhibit the activity of ASK1 and MEKK3, upstream activators of p38.57, 58 Therefore, it is likely that PI-3K/Akt inhibition releases a molecular constrain on upstream kinases needed for p38 activation. In addition to its effect on ERK phosphorylation, U0126 also decreased Akt phosphorylation. Considering that inhibition of the Akt pathway had a major influence on p38 phosphorylation, the proapoptotic effect of U0126 could be mediated by indirect inhibition of Akt. PP2 led to the strongest proapoptotic effect and also caused increased sensitivity for TGF-β–dependent phosphorylation of p38 (Fig. 6C). This may be explained by the fact that PP2 completely inhibited Akt phosphorylation. Although we demonstrated an important role for Src in transducing integrin signaling toward dedifferentiation and survival, other components of integrin signaling clusters might also be involved in these effects. Recently, it was shown that integrin linked kinase is fundamental for hepatocyte differentiation and viability.4, 5 Therefore, it is likely that many components of focal adhesion complexes contribute to the overall effect in hepatocytes.

The CM-induced stress situation might depend on the physical properties of the ECM. Collagen monomers in acidic solutions can precipitate when neutralized at room temperature and form striated fibers similar to those observed in natural connective tissue matrixes,18 which was the case for our collagen gels. Conversely, the monomers deposited on the tissue culture plates simply aggregate into densely packed structures different from striated fibrils.59 However, fibrils may also form in CM, because short UV irradiation induces cross-links on collagen monomers.60 Nevertheless, the fibers in the gel form a three-dimensional network of fibrils,61 whereas the ones generated in the monolayer system are fixed to the plastic substrate, which results in a relatively higher stiffness.37, 49 ECM stiffness has been reported to strongly influence cell behavior.62, 63 For example, contact to a stiff collagen matrix enhances focal adhesion complex formation leading to spreading and proliferation of smooth muscle cells. In contrast, contact to collagen gel inhibits cell cycle progression.12, 38 However, the plasticity of the collagen matrix is not the only factor affecting hepatocyte behavior in vitro, because bile canaliculi formation was strongly dependent on the presence of two layers of ECM. This implies that ECM stiffness and spatial localization (one versus two layers) are fundamental in modulating hepatocyte behavior in vitro.

Remarkably, the EMT induced by CM is a reversible state. A similar plasticity was reported for skin fibroblasts and bone marrow mesenchymal stem cells, which under certain stimuli can dedifferentiate into hepatocyte-like cells.64, 65 However, upon removal of the stimulus, they return to their basic mesenchymal phenotype.64 It is possible that the behavior of hepatocytes in CM represents a response to injury in terms of regeneration capacity and resistance to apoptosis. However, further studies are needed to clarify these issues.

In conclusion, ERK is a key factor for the switch between the differentiated, apoptosis sensitive to the dedifferentiated, apoptosis resistant cell state. The most important observation of this study is that both cell states are reversible and can be converted into each other by manipulation of the responsible factors of the signaling network.