Potential conflict of interest: Nothing to report.
This research was supported by the Intramural Research Program of the National Institutes of Health, National Cancer Institute.
Hepatocyte growth factor (HGF)/c-Met supports a pleiotrophic signal transduction pathway that controls stem cell homeostasis. Here, we directly addressed the role of c-Met in stem-cell–mediated liver regeneration by utilizing mice harboring c-met floxed alleles and Alb-Cre or Mx1-Cre transgenes. To activate oval cells, the hepatic stem cell (HSC) progeny, we used a model of liver injury induced by diet containing the porphyrinogenic agent, 3,5-diethocarbonyl-1,4-dihydrocollidine (DDC). Deletion of c-met in oval cells was confirmed in both models by polymerase chain reaction analysis of fluorescence-activated cell-sorted epithelial cell adhesion molecule (EpCam)-positive cells. Loss of c-Met receptor decreased the sphere-forming capacity of oval cells in vitro as well as reduced oval cell pool, impaired migration, and decreased hepatocytic differentiation in vivo, as demonstrated by double immunofluorescence using oval- (A6 and EpCam) and hepatocyte-specific (i.e. hepatocyte nuclear factor 4-alpha) antibodies. Furthermore, lack of c-Met had a profound effect on tissue remodeling and overall composition of HSC niche, which was associated with greatly reduced matrix metalloproteinase (MMP)9 activity and decreased expression of stromal-cell–derived factor 1. Using a combination of double immunofluorescence of cell-type–specific markers with MMP9 and gelatin zymography on the isolated cell populations, we identified macrophages as a major source of MMP9 in DDC-treated livers. The Mx1-Cre-driven c-met deletion caused the greatest phenotypic impact on HSCs response, as compared to the selective inactivation in the epithelial cell lineages achieved in c-Metfl/fl; Alb-Cre+/− mice. However, in both models, genetic loss of c-met triggered a similar cascade of events, leading to the failure of HSC mobilization and death of the mice. Conclusion: These results establish a direct contribution of c-Met in the regulation of HSC response and support a unique role for HGF/c-Met as an essential growth-factor–signaling pathway for regeneration of diseased liver. (HEPATOLOGY 2012)
It is now well recognized that the adult liver contains a stem cell compartment that can be activated under conditions of severe liver injury to give rise to both hepatocytic and biliary epithelial cell (BEC) lineages.1–4 Hepatic stem cells (HSCs) are thought to reside within the terminal bile ductules (Hering canals) located at the interface between parenchyma and biliary tracts. Upon activation, HSCs give rise to oval cells, which form a network of proliferating branching ducts that migrate into parenchyma, where they finally differentiate into hepatocytes.5-7 Numerous molecular factors and cell types contribute to HSC activation either directly or indirectly.8-10 We and others have established that oval cell expansion requires a close cooperation with accompanying stellate cells, which provide hepatocyte growth factor (HGF) and also promote pericellular collagen deposition, thus creating a microenvironment supporting the growth of expanding progenitor cells.11-16
HGF was originally characterized as a potent mitogen for mature hepatocytes.17 All biological effects of HGF are mediated by a single tyrosine kinase receptor (c-Met).18, 19 Gene-knockout studies have shown that both HGF and c-Met are absolutely required for survival, including liver development.20, 21 The unique property of c-Met signaling is the activation of a complex biological program supporting morphogenesis, mitogenesis, and motogenesis (also referred to as “invasive growth”).22, 23 The program operates during embryogenesis, tissue regeneration, and cancer metastasis and controls proliferation, branching morphogenesis, survival, migration through extracellular matrix (ECM), and differentiation. Previously, we have shown that HGF/c-Met signaling promotes the activation and early expansion of oval cells after severe liver injury in an acetylaminofluorene/partial hepatectomy rat model.12 However, the molecular mechanisms supporting adult stem cell activation are not well understood, and knowledge about the role of the HGF/c-Met pathway in this process is still limited.
Recently, we and others have provided direct genetic evidence for the essential role of HGF/c-Met in hepatocyte-mediated liver regeneration.24-26 Here, we analyzed the contribution of the c-Met-signaling pathway in stem-cell–mediated liver regeneration by utilizing liver-specific c-Met conditional knockout mice. To gain insight in the intricate nature of epithelial-mesenchymal cross-talk that defines stem cell behavior, inactivation of c-Met was achieved either in epithelial cell lineages (c-Metfl/fl; Alb-Cre+/−) or in various subsets of liver cells, including stromal cells (c-Metfl/fl; Mx1-Cre+/−), by crossing c-Metfl/fl mice with transgenic mice expressing Cre-recombinase under the control of a constitutively active albumin promoter or a ubiquitous interferon-inducible Mx1 promoter. To activate oval cells, we used a model of chronic liver injury induced by diet containing the porphyrinogenic agent, 3,5-diethoxycarbonyl-1,4-dihydrocollidine (DDC), which has been described previously.27, 28
Our results show that the absence of c-Met caused severe damage both to hepatocytes and biliary epithelium, disrupted the balance between ECM production and degradation, and prevented stem-cell–mediated liver regeneration. Consequently, our study establishes the HGF/c-Met-signaling pathway as an essential component of hepatic regenerative capability.
Male 8-10-week-old Metfl/fl; Mx1-Cre+/− and c-Metfl/fl; Alb-Cre+/− mice were generated and genotyped as previously described.25, 26 Metfl/fl and c-Metwt/wt; Alb-Cre+/− mice were used as corresponding controls. For Mx1-Cre-mediated c-Met inactivation, Metfl/fl and Metfl/fl; Mx1-Cre+/− mice received three intraperitoneal injections of 300 μg of pIpC in saline at 2-day intervals, which, in the liver, was shown to result in a complete deletion of gene flanked by LoxP recombinase recognition sites.29 To induce oval cells, mice were given a diet containing 0.1% DDC (Bio-Serv, Frenchtown, NJ). Metfl/fl; Mx1-Cre+/− and Metfl/fl mice received DDC diet 3 days after the last pIpC injection. Mice were housed in an Association for Assessment and Accreditation of Laboratory Animal Care facility and cared for in accord with the guidelines from the Animal Care and Use Committee at the National Cancer Institute, National Institutes of Health (NIH; Bethesda, MD).
Gelatin and In Situ Zymography.
The activity of MMPs in tissue extracts was examined by electrophoresis on 10% sodium dodecyl sulfate polymerase acrylamide gel electrophoresis containing gelatin (Invitrogen, Carlsbad, CA) without previous heating or reduction. Gels were stained with SimplyBlue SafeStain (Invitrogen). Densitometry was performed on inverted black-and-white gel images. In situ zymography was performed on 7-μm liver cryosections as previously described.30
The statistical differences for two-group comparison were determined by the Bootstrap t test, with 10,000 repetitions for small sample sizes (n < 4), and by the two-sample Student's t test or Mann-Whitney U test for a larger sample size. The Kolmogorov-Smirnov test and Leven's test were used to verify the normality assumption and equality of variances, respectively. For three-group comparison, a one-way analysis of variance test was applied, if the samples satisfied normality assumption, and the Kruskal-Wallis rank-sum test, if the samples failed normality assumption. For a discrete random variable, the statistical differences were determined using the Poisson generalized linear model. We used R statistical software (version 2.8.0) and considered P values ≤0.05 (*), ≤0.01 (**), and ≤0.001 (***) as significant.
Lack of c-Met Induces Severe Liver Dysfunction, Fibrosis, and Cholestasis.
The phenotype of both c-Met mutant mice was very similar, albeit more severe, in mice with total (c-Metfl/fl; Mx1-Cre+/−), than selective (c-Metfl/fl; Alb-Cre+/−), c-Met inactivation (Fig. 1; Supporting Fig. 1). In both cases, Met-deficient mice did not show compensatory regeneration and developed severe liver atrophy resulting from significant reduction in hepatocyte proliferation and a parallel increase in hepatocyte apoptosis (Fig. 1A-C; Supporting Fig.1A-C). Consistent with more extensive liver damage, both conditional knockout models displayed a considerable decrease in serum albumin levels (Fig. 1D; Supporting Fig. 1D), whereas the levels of aspartate aminotransferase (AST), alkaline phosphatase, and direct bilirubin were progressively increased (Fig. 1E; Supporting Fig. 1E-I).
At the molecular level, c-Met mutant livers were unable to activate the major downstream signaling pathways involved in cell proliferation, motility regulation, and apoptosis protection, such as extracellular signal-regulated kinases (i.e. Erk1/2), Akt, and Stat3 (Fig. 1F). Histologically, the most striking difference was a considerable reduction in oval cell proliferation. Control livers developed an extensive network of branching oval cell ducts, with small lumens radiating from the periportal areas toward the parenchyma. In contrast, the mutant biliary epithelium displayed a dramatic accumulation of protoporphyrin plugs and showed only a rudimentary outgrowth, which was more reminiscent of a classical bile duct proliferation restricted by a more severe periportal fibrosis (Supporting Fig. 2). By 8 weeks of DDC treatment, all c-Metfl/fl; Mx1-Cre+/− and c-Metfl/fl; Alb-Cre+/− mice (n = 5 each genotype) died from liver failure, whereas all control mice survived (n = 10). Together, the data show that the absence of c-Met function caused severe damage to both hepatocytes and biliary epithelium, impaired oval cell expansion, and thus blocked liver regeneration.
Lack of c-Met Affects Sphere-Forming Capacity of Oval Cells.
Sphere-forming assays are widely used in stem cell biology to determine the dynamics of stem cells in vivo.31 To address the sphere-forming potential of c-Met deleted oval cells, we first isolated the bulk nonparenchymal cell fraction and fluorescence-activated cell-sorting (FACS)-sorted single oval cells using an oval-cell–specific marker, epithelial cell adhesion molecule (EpCam),32 in combination with lineage cocktail antibodies. The latter are designed to react with five major hematopoietic lineages and were used to ensure the purity of the FACS-sorted epithelial cells. We confirmed that c-met was deleted in the EpCam+/Lineage− cells in both models, as shown by polymerase chain reaction (PCR) analysis (Fig. 2A,B).
To generate spheres, we then cultured the sorted EpCam+/Lineage− cells in Matrigel in the presence of HGF, epidermal growth factor (EGF), or both growth factors. Quantification and morphological assessment of cultures showed that the number of primary spheres generated from the c-Met deleted oval cells was reduced by 80%. In addition, the mutant spheres were considerably smaller (Fig. 2C,D). As expected, c-Met-deficient cells were responsive only to mitogenic EGF, but not HGF. In c-Met-expressing cells, HGF alone was more effective in increasing both the number and the sphere size, as compared to EGF. These experiments demonstrate that c-met deletion altered functional properties of oval cells.
Lack of c-Met Affects Oval Cell Proliferation.
To corroborate these findings invivo, we used Ki-67 immunohistochemistry (IHC). A quantitative time-course analysis of Ki-67 staining showed a drastic, progressive decline in the frequency of proliferating oval in c-Met-deficient livers (Fig. 3A). Reduction in proliferation was found in both c-Met models, as shown by a similar decrease in oval cell density, as determined by quantification of the number of A6-positive cells (Fig. 3B). Immunostaining with an additional marker of cell-cycle progression confirmed a significant decrease in size of the oval cell pool (Fig. 3C). Interestingly, loss of Met appeared to be more compatible with BEC proliferation (Fig. 3C, bottom images), implying a failure of oval cell outgrowth.
Lack of c-Met Affects Oval Cell Differentiation.
To test whether the differentiation potency of oval cells was impaired, we performed dual-label experiments using two oval-cell–specific antibodies: A6 and EpCam. Our previous work has established that A6, in addition to being a specific marker of oval cells and BEC, can also recognize a fraction of newly formed hepatocytes,33, 34 whereas EpCam is expressed exclusively by oval and bile duct cells.32, 35 Quantitative analysis revealed a progressive accumulation of A6+/EpCam−-positive cell clusters with a hepatocyte-like morphology, which were located in close proximity to oval cells only in the Metfl/fl control livers, but not in c-Metfl/fl; Mx1-Cre+/− or c-Metfl/fl; Alb-Cre+/− livers (Fig. 4A,B; and data not shown). Significantly, only A6+ hepatocyte-like cells expressed hepatocyte nuclear factor 4-alpha (HNF-4α) transcription factor, a well-known marker of hepatocytic differentiation,36 whereas ductular oval cells were HNF-4α negative (Fig. 4C). These data demonstrate that loss of c-Met impaired the ability of oval cells to differentiate into hepatocytic lineage.
Lack of c-Met Affects Oval Cell Migration.
Next, we examined the changes in distribution of oval cells migrating inside the parenchyma. For this, we divided the hepatic lobule into three zones—periportal (0-97 μm), middle (97-194 μm), and central (194-290 μm)—and measured the distance between the portal tract and migrating oval cells visualized by A6 staining. In control livers, oval cells formed small ducts expanding toward the central zone (Fig. 5). The average distance between the portal veins and endpoint of A6-positive small branching ducts with poorly defined lumen increased from 92.6 μm at 1 week to 132.7 μm at 4 weeks. In contrast, in c-Met-deficient livers, A6-positive cells lined larger ducts with round lumen, which were confined to portal tracts and did not spread into parenchyma (the average distance from portal tracts was 78.2 and 79.0 μm at 1 and 4 weeks, respectively) (Fig. 5A-C). Thus, the absence of c-Met altered the pattern of ductular reaction and impaired its distribution in the parenchyma.
Loss of c-Met in HSCs Alters Functional Interactions With Stem Cell Niche Microenvironment.
Next, we assessed whether the absence of c-Met signaling altered the stem cell/oval cell microenvironment. Consistent with the protective role of HGF/c-Met against fibrosis,37 both c-Met mutant models developed a more extensive periportal fibrosis, as judged by the quantification of Sirius red staining, which was more pronounced in c-Metfl/fl; Mx1-Cre+/− livers (Fig. 6A,B). By 4 weeks after the initiation of the DDC diet, the Sirius red–positive areas were significantly larger, both in c-Metfl/fl; Mx1-Cre+/− and in c-Metfl/fl; Alb-Cre+/− livers, as compared to the respective DDC-treated control mice (Fig. 6C). Monitoring liver fibrosis, using second harmonic generation confocal imaging, confirmed the presence of a much more dense and altered collagen matrix structure in c-Met-deficient mice maintained on the DDC diet (Fig. 6A). In contrast with straight and well-organized collagen fibers in DDC-treated control livers, mutant livers displayed irregular, wavy, and significantly less aligned collagen fibers or bundles.
This was paralleled by a diminished macrophage mobilization, as measured by IHC and FACS analysis using Kupffer-cell–specific F4/80 antibody (Fig. 6A, D, E). Remarkably, both total and selective c-Met inactivation in liver cells decreased the recruitment of stromal/inflammatory cells to the site of oval cell reaction, suggesting a leading role for epithelial compartment in maintaining functional HSC microenvironment.
Decreased Fibrolytic Activity in c-Met-Deficient Mice.
To address the mechanisms underlying increased fibrosis and aberrant tissue remodeling in c-Met deleted livers, we examined the levels of MMPs, the primary proteolytic enzymes involved in the breakdown of ECM. A time course of MMP9 activation showed that in control mice, the proteolytic activity of MMP9 was progressively increasing, along with the expansion of oval cells, whereas c-Met-deficient mice displayed a decrease in MMP9 (Fig. 7A,B). This was consistent with the results of in situ zymography combined with A6 staining, which showed a close proximity of MMP9 activity to oval cell reaction (Fig. 7C; Supporting Fig. 3). Levels of MMP9 were reduced in both models of liver- and epithelial-specific c-Met deletion (Fig. 7D,E). There was no difference in MMP2 activity, regardless of genotype. These data link the aberrant tissue remodeling in c-Met-deficient livers with a reduction in stem cell niche component MMP9.
Cell Source of MMP9.
Finally, we determined the cell source of MMP9 in DDC-treated livers. For this, we carried out gelatin zymography on isolated hepatocytes, nonparenchymal cell (NPC) fraction, and FACS-sorted F4/80-positive macrophages. Quantification of the intensity of active MMP9 band showed that the main source of active MMP9 was NPC cells, and that monocytes/macrophages accounted for approximately 80% of this activity (Fig. 8A,B). Confirming the zymography results, dual immunofluorescence staining for MMP9 and markers for oval (A6), Kupffer (F4/80), and stellate (alpha smooth muscle actin; αSMA) cells revealed colocalization at the interface with Kupffer and oval cells, but not with stellate cells (Fig. 8C). These data show that macrophage is the primary cell source of active MMP9 in this model. To provide additional evidence that the absence of c-Met creates a defective stem cell microenvironment, we examined the expression of chemokine stromal-cell–derived factor 1 (SDF1), known as a powerful chemoattractant for bone-marrow–derived monocytes. SDF1 protein levels were considerably decreased in both Met mutant models as well as the number of A6+/SDF1+ oval cells (Supporting Fig. 4).
The aim of this study was to define the role of c-Met-signaling pathway in different phases of adult hepatic stem cell activation by utilizing mice harboring c-met floxed alleles and Alb-Cre or Mx1-Cre transgenes. Using conditional mouse genetics and a DDC toxic liver injury model, we demonstrate that the lack of c-Met signals impaired both hepatocyte- and stem-cell–mediated liver regeneration, leading to the death of mice. Genetic loss of c-Met function has profound effects on tissue remodeling and overall composition of the HSC niche microenvironment concomitant with a failure of HSCs to expand and differentiate into hepatocytes. The Mx1-Cre-driven deletion of c-met in liver cells resulted in a more severe effect on stem cell activation, as compared to the selective inactivation of c-met in the epithelial cell lineages achieved in c-Metfl/fl; Alb-Cre+/− mice. However, in both models, loss of c-Met function caused a similar cascade of events disrupting HSC response.
Consistent with the c-Met involvement in diverse cellular functions, c-met deletion had a broad, profound impact on HSC properties. Assessment of proliferation and the frequency of oval cells (i.e. HSC progeny), using a combination of oval-cell–specific and proliferative markers, revealed a striking decrease in the size of the oval cell pool (Fig. 3). We also found a marked reduction in the number of A6+/HNF4α+ cells, reflecting a reduced capacity of c-Met-deficient oval cells to differentiate into hepatocytes (Fig. 4), as well as an almost complete lack of their migration into parenchyma (Fig. 5).
Concurring with fewer oval cells being present in c-Met mutant livers, the frequency of primary spheres generated from EpCam+/Lineage− cells isolated from both c-Metfl/fl; Mx1-Cre+/− and c-Metfl/fl; Alb-Cre+/− mutant livers was significantly reduced, as compared to the sphere-forming activity of c-Met-expressing oval cells (Fig. 2). c-Met-deficient spheres were smaller in size and failed to attach and expand in the two-dimensional monolayer, whereas control spheres could be further subcultured as adherent clones. These data provide a strong indication that c-Met signals play a prominent role not only in hepatocyte proliferation,24-26 but also affect the dynamics of hepatic progenitor cells in vivo. Significantly, EGF supplementation was capable of increasing the sphere-forming ability of c-Met-deficient oval cells to the levels found in the similarly treated control cells. These in vitro experiments suggest that signaling molecules shared by tyrosine kinase receptors could compensate, at least in part, for the lack of c-Met. This was in striking contrast to the situation in vivo where hepatic deletion of a single c-met gene caused far-reaching alterations in hepatic homeostasis and created a microenvironment that compromised normal stem cell functions via direct and indirect mechanisms.
Significantly, c-Met deficiency promoted development of the periportal fibrosis consistent with the antifibrotic role of HGF/c-Met signaling.37 High-resolution imaging of primary fibrillar collagens in unfixed livers, using second harmonic generation microscopy,38 corroborated a denser collagen distribution and also revealed prominent differences in the orientation and length of collagen fibers. Furthermore, c-Met deficiency caused an aberrant tissue distribution of the collagen-producing stellate cells (Fig. 6). In contrast with control livers, in both models of c-Met deletion, stellate cells did not accompany the migrating oval cells, but accumulated in the areas of periportal fibrosis. Our previous results, as well as others, implicated HSCs as an integral component of HSC niche by providing a variety of growth factors as well as a mechanical support for expanding the population of oval cells.11-16
Another striking consequence of c-Met deficiency was defective mobilization of F4/80-positive Kupffer cells and greatly reduced secretion of proteolytic enzyme MMP9 at the margins of adjacent oval cells disrupting the balance between ECM production and degradation (Fig. 7; Supporting Fig. 3). The structural abnormalities caused by the lack of c-Met function could compromise the movement of the expanding oval cell ducts into parenchyma and thereby interrupt cross-talks with the components of HSC niche. This phenomenon occurred regardless of total or selective c-Met inactivation in liver cells, although it was more prominent in Metfl/fl; Mx1-Cre+/− mice, suggesting that loss of c-Met function in the epithelial compartment was a common denominator responsible for the striking similarities in phenotypes.
MMP9 is a matrix-degrading enzyme involved in the resolution of fibrotic matrix and basement membrane degradation39 critical for oval cell migration into parenchyma and subsequent differentiation into hepatocytes.6, 40, 41 Using a combination of double immunofluorescence of MMP9 with cell-type–specific markers as well as gelatin zymography on the isolated cell populations, we identified macrophages as a major source of MMP9 in DDC-treated livers (Fig. 8). These data are in line with the early reports describing macrophages as the primary source of gelatinases in liver fibrosis.42 The invading macrophages have also been referred to as major determinants of liver progenitor cell expansion in the models of diet- and immune-mediated liver injury by providing promitogenic cytokines (e.g., tumor necrosis factor alpha and tumor necrosis factor–like weak inducer of apoptosis)15, 43-45 and MMPs, including MMP9.46
In addition to matrix-degrading potential, MMP9 is also known for its ability to recruit bone-marrow–derived cells to the injured liver to facilitate the resolution of fibrotic matrix.47, 48 As a part of a general impairment of tissue remodeling caused by the c-Met absence, we also found reduced levels of SDF1 (Supporting Fig. 4), another stem cell niche mediator that can attract and retain hematopoietic cells within fibrotic livers.47
Cre-mediated recombination of Metfl/fl was achieved both in hepatocytes and ductular oval cells and BECs, regardless of using a Mx1-Cre or Alb-Cre promoter, similar to the findings published previously.49, 50 Accordingly, these two epithelial cell types sustained a considerable structural and functional damage, as shown by reduced albumin secretion and a marked increase in serum AST levels (Fig. 1; Supporting Fig. 1). Consistent with a protective role of HGF/c-Met signaling against accumulating toxic bile acids and developing intrahepatic cholesterol,51 both c-Met conditional knockout mouse models also displayed a marked increase in serum bile acid and bilirubin concentrations, rendering c-Met mice more susceptible to DDC toxic injury. Thus, the combined impact of increased bile acid production and a defective hepatobiliary transport capacity appear to contribute to increased cholestasis and liver injury promoted by the lack of c-Met signaling. The latter underscores the fundamental role of the HGF/c-Met-signaling pathway for regeneration of the diseased liver.
In summary, using a DDC toxic liver injury model, we have shown that c-Met is a major determinant of adult HSC and HSC niche homeostasis. Lack of c-Met affected the proliferative potential of oval cells, capacity to migrate, pattern of differentiation, and dynamic interaction with the microenvironment. Future studies aiming at isolating and characterizing oval cells induced by other models of liver injury relevant to human studies (e.g., viral injury, acetaminophen toxicity, and bile duct ligation) will provide a further understanding of the role of c-Met signaling in the regulation of adult liver stem cells.
The authors thank Dr. Joe Grisham for valuable discussions, Susan Garfield for her help with confocal microscopy, and Tanya Hoang and Anita Ton for their assistance with PCR analysis, IHC, and animal care.