• Open Access

Bone morphogenetic protein-9 induces epithelial to mesenchymal transition in hepatocellular carcinoma cells


  • Qi Li,

    1. Department of Medicine II, Section Molecular Hepatology – Alcohol Associated Diseases, Medical Faculty Mannheim at Heidelberg University, Mannheim, Germany
    2. Department of Gastroenterology and Hepatology, Beijing You'an Hospital, Capital Medical University, Beijing, China
    Search for more papers by this author
    • These authors contributed equally to this manuscript.
  • Xing Gu,

    1. Department of Laboratory Medicine, Eastern Hepatobiliary Surgery Hospital, Second Military Medical University, Shanghai, China
    Search for more papers by this author
    • These authors contributed equally to this manuscript.
  • Honglei Weng,

    1. Department of Medicine II, Section Molecular Hepatology – Alcohol Associated Diseases, Medical Faculty Mannheim at Heidelberg University, Mannheim, Germany
    Search for more papers by this author
  • Shahrouz Ghafoory,

    1. Institute of Pharmacy and Molecular Biotechnology, Heidelberg University, Heidelberg, Germany
    Search for more papers by this author
  • Yan Liu,

    1. Department of Medicine II, Section Molecular Hepatology – Alcohol Associated Diseases, Medical Faculty Mannheim at Heidelberg University, Mannheim, Germany
    2. Department of Molecular Cell Biology and Centre for Biomedical Genetics, Leiden University Medical Center, Leiden, the Netherlands
    Search for more papers by this author
  • Teng Feng,

    1. Department of Medicine II, Section Molecular Hepatology – Alcohol Associated Diseases, Medical Faculty Mannheim at Heidelberg University, Mannheim, Germany
    Search for more papers by this author
  • Johanna Dzieran,

    1. Department of Medicine II, Section Molecular Hepatology – Alcohol Associated Diseases, Medical Faculty Mannheim at Heidelberg University, Mannheim, Germany
    Search for more papers by this author
  • Li Li,

    1. Medical Research Center, Medical Faculty Mannheim, University of Heidelberg, Heidelberg, Germany
    Search for more papers by this author
  • Iryna Ilkavets,

    1. Department of Medicine II, Section Molecular Hepatology – Alcohol Associated Diseases, Medical Faculty Mannheim at Heidelberg University, Mannheim, Germany
    Search for more papers by this author
  • Marianna Kruithof- de Julio,

    1. Department of Molecular Cell Biology and Centre for Biomedical Genetics, Leiden University Medical Center, Leiden, the Netherlands
    Search for more papers by this author
  • Stefan Munker,

    1. Department of Medicine II, Section Molecular Hepatology – Alcohol Associated Diseases, Medical Faculty Mannheim at Heidelberg University, Mannheim, Germany
    Search for more papers by this author
  • Alexander Marx,

    1. Institute of Pathology, University Medical Centre Mannheim, University of Heidelberg, Heidelberg, Germany
    Search for more papers by this author
  • Albrecht Piiper,

    1. Department of Medicine I, Johann Wolfgang Goethe University, Frankfurt, Germany
    Search for more papers by this author
  • Eduardo Augusto Alonso,

    1. Department of Medicine I, Johann Wolfgang Goethe University, Frankfurt, Germany
    Search for more papers by this author
  • Norbert Gretz,

    1. Medical Research Center, Medical Faculty Mannheim, University of Heidelberg, Heidelberg, Germany
    Search for more papers by this author
  • Chunfang Gao,

    1. Department of Laboratory Medicine, Eastern Hepatobiliary Surgery Hospital, Second Military Medical University, Shanghai, China
    Search for more papers by this author
  • Stefan Wölfl,

    1. Institute of Pharmacy and Molecular Biotechnology, Heidelberg University, Heidelberg, Germany
    Search for more papers by this author
  • Steven Dooley,

    1. Department of Medicine II, Section Molecular Hepatology – Alcohol Associated Diseases, Medical Faculty Mannheim at Heidelberg University, Mannheim, Germany
    Search for more papers by this author
  • Katja Breitkopf-Heinlein

    Corresponding author
    • Department of Medicine II, Section Molecular Hepatology – Alcohol Associated Diseases, Medical Faculty Mannheim at Heidelberg University, Mannheim, Germany
    Search for more papers by this author

To whom correspondence should be addressed.

E-mail: katja.breitkopf@medma.uni-heidelberg.de


Epithelial-mesenchymal transition (EMT) is an important mechanism to initiate cancer invasion and metastasis. Bone morphogenetic protein (BMP)-9 is a member of the transforming growth factor (TGF)-β superfamily. It has been suggested to play a role in cancer development in some non-hepatic tumors. In the present study, two hepatocellular carcinoma (HCC) lines, HLE and HepG2, were treated with BMP-9 in vitro, and phenotypic changes and cell motility were analyzed. In situ hybridization (ISH) and immunohistochemical analyses were performed with human HCC tissue samples in order to assess expression levels of BMP-9. In vivo, BMP-9 protein and mRNA were expressed in all the tested patients to diverse degrees. At the protein level, mildly positive (1 + ) BMP-9 staining could be observed in 25/41 (61%), and moderately to strongly positive (2 + ) in 16/41 (39%) of the patients. In 27/41 (65%) patients, the BMP-9 protein expression level was consistent with the mRNA expression level as measured by ISH. In those patients with 2 +  protein level, nuclear pSmad1 expression in cancer cells was also significantly increased. Expression of BMP-9 was positively related to nuclear Snail expression and reversely correlated to cell surface E-cadherin expression, although this did not reach statistical significance. Expression levels of BMP-9 were significantly associated with the T stages of the investigated tumors and high levels of BMP-9 were detected by immunofluorescence especially at the tumor borders in samples from an HCC mouse model. In vitro, BMP-9 treatment caused a reduction of E-cadherin and ZO-1 and an induction of Vimentin and Snail expression. Furthermore, cell migration was enhanced by BMP-9 in both HCC cell lines. These results imply that EMT induced by BMP-9 is related to invasiveness of HCC.

Eighty to ninety percent of primary liver cancer cases originate from hepatocellular carcinoma (HCC), which has been regarded as the fifth most common cancer and the third most cancer-related death in the world.[1] Current therapeutic options including surgical resection, liver transplantation and chemoembolization are only used at early stages of HCC with limited efficacy.[2] Cancer recurrence occurs in around 50% of patients.[3, 4] Despite extensive scientific efforts, the prognosis of HCC is nowadays still poor since HCC is inclined to tumor invasiveness and formation of intra- and extra-hepatic metastases.

Epithelial-mesenchymal transition (EMT) is one major mechanism participating in malignant progression of cancer cells. Epithelial-mesenchymal transition is characterized by loss of differentiated traits in epithelial cells, for example, cell–cell contacts and cell polarity, as well as acquisition of mesenchymal appearances such as higher motility, invasiveness and resistance to apoptosis.[5] Properties typical for EMT comprise downregulation of epithelial markers like E-cadherin, ZO-1, nuclear translocation of β-catenin and upregulation of mesenchymal markers like Vimentin, N-cadherin and α-smooth muscle actin (SMA).[6, 7] The zinc-finger transcription factors Snail and Slug, and the basic helix-loop-helix transcription factor Twist play prominent roles as master regulators of EMT, as they repress expression of E-cadherin.[8-10] In liver carcinogenesis, there is growing evidence for a central role of EMT in different stages of disease progression. E-cadherin, β-catenin, Twist, Snail, Slug, SOX4 and others have been identified as markers for EMT in hepatocytes and HCC.[11]

Hepatocyte growth factor (HGF) has been reported to induce EMT in HCC via c-met signaling and cross-talk with Akt- and COX-2- pathways.[12, 13] Furthermore, transforming growth factor-β (TGF-β) signaling has been described as a critical inducer of a complete EMT phenotype in malignant hepatocytes. It is reported to display dual roles towards epithelial cells including hepatocytes. In healthy liver and early stages of hepatocarcinogenesis, TGF-β mediates cell cycle arrest and apoptosis, whereas at later stages, it causes cell dedifferentiation, EMT and metastasis.[14] Morphologically, TGF-β mediates a gain of plasticity in neoplastic hepatocytes which can be documented as cell spreading.[11] Transforming growth factor-β additionally activates other factors and signaling pathways such as COX-2, platelet derived growth factor (PDGF) or the PI3K/Akt pathway in malignant hepatocytes, which facilitate tumor cell survival.[15] Transforming growth factor-β may provide tumor promoting effects towards hepatocytes via Smad and non-Smad signaling pathways.[16] Bone morphogenetic proteins (BMPs) (4, 6, 7, 8, 9, 10, 11, 13 and 15) are upregulated in HCC cell lines and there is evidence for their contribution to migration and invasion of HCC cells, predominantly via Smad signaling.[17] Further, BMP-2 induces angiogenesis in HCC xenografted nude mice[18] and BMP-4 was shown to promote HCC progression and has potential as prognostic marker in HCC.[19, 20]

In the present report, we have investigated BMP-9, a member of the TGF-β superfamily, also named as Growth and Differentiation Factor (GDF)-2, which was initially cloned from a fetal mouse liver cDNA library.[21] Bone morphogenetic protein-9 binds to the type I receptors ALK1 (activin receptor-like kinase 1) and ALK2 to induce osteogenic signaling in mesenchymal stem cells.[22] After binding BMP-9, type II receptors phosphorylate the type I receptor and subsequently downstream signaling via activation of the R-Smads -1, -5, -8 is initiated. The complex formation of these Smads with Smad4 enables nuclear translocation of this transcriptionally active complex to regulate BMP-9-target genes.[23] Bone morphogenetic protein-9 participates in various physiologic processes including bone formation, hematopoiesis, glucose homeostasis, iron homeostasis and angiogenesis.[24-27] Bone morphogenetic protein-9 was recently suggested to be involved in cancer development in a small number of studies. But its effects on different cancer cells are controversial. It was reported that about 25% of epithelial ovarian cancers express BMP-9, whereas normal human ovarian surface epithelial specimens do not. Furthermore, some ovarian cancer cell lines acquire autocrine BMP-9 signaling that facilitates cancer cell proliferation.[28] On the contrary, in prostate cancer, BMP-9 suppresses cancer cell growth via inducing cell apoptosis.[29]

The biological role of BMP-9 in HCC, especially if BMP-9 can promote cancer invasion via EMT, has not been elucidated so far. In the present study, we investigated human samples from HCC patients for BMP-9 expression and HCC cell lines as well as samples from an HCC mouse model to investigate its impact on EMT marker expression and cell migration. In HCC patient samples, we found that BMP-9 expression was positively associated with T stage and that it enhanced cell migration and induced EMT in HCC cells in vitro. Therefore, we identified BMP-9 as a new EMT inducing factor in HCC.

Materials and Methods


Polyclonal anti-human BMP-9 antibody from AbD Serotec (Düsseldorf, Germany) for IHC on human sections, polyclonal rabbit anti BMP-9 (ab35088) from Abcam (Cambridge, UK) for IF on mouse sections, Phospho-Smad1 (Ser463/465)/Smad5 (Ser463/465)/Smad8 (Ser426/428) polyclonal antibody and anti-rabbit E-cadherin from Cell Signaling Technology (New England Biolabs GmbH), rabbit anti ZO1 and rabbit anti Snail from Abcam were used in immunohistochemistry. The following antibodies were used for Western blot: Smad1 rabbit monoclonal antibody and pSmad3 rabbit monoclonal antibody from Epitomics (Biomol GmbH), polyclonal anti-rabbit GAPDH antibody from Santa Cruz Biotechnology (Heidelberg, Germany) and anti-mouse β-actin and rabbit anti snail from Sigma (Munich, Germany). Anti-mouse E-cadherin from BD Biosciences (Heidelberg, Germany), and anti-mouse Vimentin from Abcam were used for both Western blot and immunofluorescence staining.


Recombinant human BMP-9 was purchased from R&D Systems (Wiesbaden-Nordenstadt, Germany). CellTiter-Glo Luminescent Cell Viability Assay was from Promega (Mannheim, Germany).

Tissue specimens

A total of 41 cases diagnosed as primary HCC from the surgical files in the Eastern Hepatobiliary Surgery Hospital, Second Military Medical University, Shanghai, China were enrolled in this study. Thirty-nine of them (95.12%) were hepatitis B surface antigen (HBsAg)-positive. The diagnosis was determined according to the International Union Against Cancer TNM classification of primary liver cancer, 6th edition.[30] The study protocol conformed to the ethical guidelines of the Declaration of Helsinki (1975) and was approved by the ethics committee of the Second Military Medical University, Shanghai, China. All patients provided an informed consent before the study.

Cell culture

Hepatocellular carcinoma cell lines HepG2 and human hepatoma cell line (HLE) cells were cultured in DMEM (Lonza, Basel, Switzerland) supplemented with 10% FCS (Invitrogen, Karlsruhe, Germany), 4 mM L-glutamine (Lonza, Basel, Switzerland) and 100 U/mL penicillin/streptomicin (Biochrom KG, Berlin, Germany). Cells were cultured at 37° in 5% CO2 atmosphere.

Adenovirus infection

AddnALK1/2, AdcaALK1 and AdcaALK2 were kindly provided by Professor Dr Peter ten Dijke (Leiden, Netherland). AdSmad1 and AdLacZ was a gift from Professor Dr Carl-Henrik Heldin (Uppsala, Sweden). The virus stocks were amplified in HEK293A cells. Virus titers were measured with Adeno-x rapid titer kit (Takara Bio Europe/Clontech, Germany). HLE and HepG2 cells were plated in 6-well plates at a density of 0.25 × 105/well and 4 × 105/well, respectively. In the afternoon of the second day, medium was changed and different adenoviruses of m.o.i 10 were added into each well. On the third day, medium (10% FCS) was changed. Serum starvation was performed in the morning of the fourth day so that cells had time to overcome the stress induced by the viruses. In the afternoon, cells were stimulated with BMP-9 (50 ng/mL). Proteins were harvested on the seventh day. In total, cells were infected by adenoviruses for 5 days and stimulated with BMP-9 for 72 h.

Transwell assay

The migration of HepG2 and HLE cells was detected in a modified 24-well transwell chamber (BD Biosciences). In the upper chamber, 25 000 cells in 0.25 mL of serum-free culture medium were treated with BMP-9 (50 ng/mL). Medium with 10% FCS and with or without BMP-9 (50 ng/mL) was loaded in the lower wells serving as chemotactic stimulus. After 6 (HLE cells) or 24 (HepG2 cells) h at 37°C, the cells on the upper surface of the filter and the underside were trypsinized, washed with serum-free DMEM and transferred to a 96-well flat bottomed plate, respectively. After centrifugation at 129g for 10 min, the supernatant was discarded and 40 μL buffer of the CellTiter-Glo Luminescent Cell Viability Assay (Promega) was added to each well. Luminescence was measured by the iControl1.6 program. The percentage of migrated cells was calculated. Each experiment was conducted in triplicate, and the mean ± SD was calculated.

Preparation of cell lysates and immunoblotting

Total cell protein was extracted on ice with RIPA lysis buffer (1 ×  Tris-buffer saline, 1% Nonidet P40, 0.5% sodium deoxycholate, and 0.1% sodium dodecylsulfate) in the presence of freshly added protease and phosphatase inhibitors (Roche, Mannheim, Germany). Protein concentration was determined using Bradford method with a Bio-Rad protein assay (Biorad, Munich, Germany). Thirty micrograms of protein extract was separated by 4–12% SDS-PAGE (4–12% Bis-Tris Gel, NuPAGE, Invitrogen) and transferred to nitrocellulose membranes (Pierce, Rockford, IL, USA). Nonspecific binding was blocked with 5% non-fat milk in tris buffered saline with 0.05% Tween 20 for 1 h. The membrane was incubated with the primary antibodies at 4°C overnight. Horseradish peroxidase-linked goat anti-mouse and anti-rabbit antibodies (Santa Cruz) were used as secondary antibodies. The membranes were developed with Supersignal Ultra (Pierce, Hamburg, Germany) and chemiluminescence was detected with a Fujifilms LAS 1 000 image detection system.


Total RNA was extracted from cells with an RNA purification kit (Roche) according to the manufactures' instructions. RNA was reverse transcribed into cDNA with the Transcriptor First Strand cDNA synthesis kit (Roche). Polymerase chain reaction conditions were 95°C for 5 min, followed by 28 or 38 cycles of 95°C for 30 s, 60°C for 30 s, and 72°C for 75 s. The final extension period consisted of 7 min at 72°C. Polymerase chain reaction products were separated on 1.5% agarose gels stained with ethdium bromide and visualized under UV light. Forward and reverse primers for the indicated gene amplification are described in Table 1.

Table 1. Primer sequences used in this study
Gene nameForwardReverseSize of PCR products (bp)

In situ hybridization

Bone morphogenetic protein-9 specific cRNA in situ hybridization probes were prepared using double stranded cDNA templates with flanking SP6 and T7-RNA-polymerase promoters, prepared using gene specific PCR-primers as described.[31] In short: total RNA was isolated from Jurkat human T lymphocyte cells; first-strand cDNA was synthesized with 3 μg total RNA using random hexamer primers and AMV Reverse Transcriptase (Promega, Madison, WI, USA); BMP-9 specific PCR primers included SP6-RNA-polymerase promoter flanking a short gene specific 5′ sequence and a T7-RNA-polymerase promoter flanking a short gene specific 3′sequence (amplified fragment: 998nt-1820nt of BMP-9 mRNA, GeneBank: NM_016204.1; Primer sequences used; H. BMP-9. SP6: 5′-CAGTGAATTGATTTAGGTGACACTATAGAAGTGGAACAAGAGAGCGTGCTCAAGAAGC-3′ and H. BMP-9. T7: 5′-CAGTGAATTGTAATACGACTCACTATAGGGAGACTCCTCCACCTCTCTAACTTCCATC-3′). Then anti-sense cRNA probes were synthesized using T7-RNA-polymerase, and sense cRNA probes were obtained with SP6-RNA-polymerase transcription. In situ hybridization was performed on 4 μm tissue slices as described.[31] Positive staining is visible as purple color from nitro-blue tetrazolium/5-bromo-4 chloro-3′-indolyl phosphate precipitate.

For the semi-quantitative assessment of ISH staining, staining scores were calculated with the following method: positive cell number was graded as 0–4 (0, no positive cells; 1, less than 25% positive cells; 2, 25–50% positive cells; 3, 50–75% positive cells; 4, more than 75% positive cells). The intensity of positivity was graded as 1–3 (1, weak purple staining; 2, strong and purple staining; 3, very strong and deep purple staining). The score was calculated according to this formula: number × intensity. According to the calculated score, the staining level was classified into 3 levels: 0, no positive staining; 1 + , score of 1–4; 2 + , 5 and more.

Immunofluorescent staining

Cells were plated on a glass chamber slide and every condition was done in duplicate. Following serum starvation over-night, the cells were simulated with rh-BMP-9 (50 ng/mL) for 72 h. Cells were fixed with ice-cold acetone and permeabilized with 0.1% Triton for 5 min in TRIS-buffered saline. After blocking with 1% BSA for 60 min, immunofluorescent staining was performed using primary antibodies against E-cadherin and vimentin with a dilution of 1:200 and second antibodies, the Alexa 488 labeled anti-rabbit or mouse IgG with a dilution of 1:200. The nucleus was stained with DRAQ5 (1:5 000) or DAPI (1:10000). Then the slide was mounted using DakoCytomation Fluorescent Mounting Medium (DakoCytomation, Hamburg, Germany) and visualized by confocal microscopy. Confocal images were obtained by using a Leica laser scanning spectral confocal microscope, model DM IRE2 (Leica Microsystems, Wetzlar, Germany). Excitation was performed with an argon laser emitting at 488 nm, a krypton laser emitting at 568 nm, and a helium/neon laser emitting at 633 nm. Images were acquired with a TCS SP2 scanner and Leica Confocal software, version 2.5 (Leica Microsystems, Wetzlar, Germany).

Immunofluorescent staining of cryosections of TGFα/cmyc stage 3 HCC mice were performed as follows: sections were washed in PBS and antigen retrieval was performed (10 min high pressure antimasking solution vectashield). Sections were then treated with 3% H2O2 for 10 min followed by PBS washes and blocked in PBST (tween 20, 0.1%) with 10% serum. Antibodies at appropriate dilution were added. BMP9 antibody was amplified with tyramide amplification kit from Invitrogen according to the manufacturer's protocol. Secondary antibodies (Alexa-555 for E-Cadherin) were used at a dilution of 1:250. TOPRO3 was used as a nuclear dye and sections were mounted in Prolog G antifading reagent (Invitrogen). Sections were scanned on a Leica SP5 confocal microscope.

Immunohistochemical staining

Following formalin-fixation, paraffin-embedding and sectioning (4 μm), human liver tissues were deparaffinized in xylene and rehydrated in graded ethanol before being washed with distilled water. Then antigen unmasking in EDTA buffer with microwave was performed. The section slides were immersed in 3% H2O2 for 30 min, and washed with PBS three times. Then DAKO peroxidase blocking reagent was used to block the endogenous peroxidase for 15 min. After washing with PBS three times, the slides were incubated with 1st antibody (BMP-9 1:400; pSmad1/5/8 1:100; E-cadherin 1:400; Snail 1:1000) at 4°C overnight. On the following day, the slides were re-warmed at room temperature for 1 h and washed with PBS three times before visualizing the immune-reactivity with Dako Envision + System-horseradish peroxidase kit according to the manufacturer's instructions. Slides were counterstained with hematoxylin and mounted.

Semi-quantitative assessment was performed for IHC staining of HCC tissue against BMP-9, pSmad1, E-cadherin and Snail. Staining scores were calculated with the following method: positive cell number was graded as 0–4 (0, no positive cells; 1, less than 25% positive cells; 2, 25–50% positive cells; 3, 50–75% positive cells; 4, more than 75% positive cells). The intensity of positivity was graded as 1–3 (1, weak yellow staining; 2, strong and brown staining; 3, very strong and deep brown staining). The score was calculated according to this formula: number × intensity. According to the calculated score, the staining level was classified into 3 levels: 0, no positive staining; 1 + , score of 1–4; 2 + , 5 and more.

Statistical analysis

The association between immunohistochemical staining levels and clinicopathological features was evaluated with kendall–tau rank correlation analysis in sas version 9.2 (SAS Inc., Cary, NC, USA). Student's t-test was used to analyze the result of the migration assay. Error bars were the standard error of the mean. < 0.05 was considered statistically significant.


BMP-9 is expressed in HCC tissue from patients

We analyzed BMP-9 mRNA and protein expression in 41 HCC tissues and 36 paired non-cancer liver tissues from liver cancer patients by immunohistochemical staining (IHC) and in situ hybridization (ISH). We found that BMP-9 protein and mRNA expression existed in all 41 patients, although the expression levels strongly varied. In 6/36 (17%) patients BMP-9 protein expression levels were equally high (2 + ) in the cancer area as compared to surrounding liver tissue. In another 6/36 (17%) the non-cancerous tissue showed highest expression (2 + ) although the corresponding HCC tissue showed only intermediate (1 + ) levels. In 15/36 (42%) the surrounding and the HCC tissues showed both intermediate levels (1 + ). Finally in one case each (3%) the tumor levels were either high (2 + ) or intermediate (1 + ) whereas in the surrounding tissue there was no BMP-9 detectable (Table 2). A detailed classification of all these “non-tumor” samples revealed that these tumor surrounding tissues were already highly fibrotic/cirrhotic (Table S1) and can therefore not be considered as “normal” liver tissue. Bone morphogenetic protein-9 staining was found in cancerous regions, and ISH data show that it is not only binding to the cell surface, but that BMP-9 mRNA is also produced by HCC cells themselves (Fig. 1a,b). According to an evaluation system as indicated in Materials and Methods, mildly positive (1 + ) BMP-9 protein staining was observed in 25/41 (61%), and moderately to strongly positive (2 + ) staining was present in 16/41 (39%) cancer samples of the patients (Fig. 1a and Table 3). In 27/41 (66%) patients, BMP-9 protein and mRNA expression levels were consistent (Fig. 1a,b). In 7/41 (17%) patients, there were 1 + mRNA levels correlated with 2 +  protein levels, whereas in another 7/41 (17%) patients, it was the other way around.

Table 2. Bone morphogenetic protein-9 expression in cancer area and paired non-cancer area
T  n Total
2+6 (17)7 (19)1 (3)14
1+6 (17)15 (42)1 (3)22
Total12 (33)22 (61)2 (6)36
Table 3. The association between immunohistochemical staining levels of all 41 hepato cellular carcinoma (HCC) samples and clinicopathological features of the patients was evaluated with kendall-tau rank correlation analysis
  BMP-9 protein expression  
Clinicopathological factor n 1 + (%)2 + (%) τ B P value
Total4125 (61)16 (39)  
<603218 (56)14 (44)0.0590.709
>6097 (78)2 (22)
Male3721 (57)16 (43)0.2630.096
Female44 (100)0 (0)
T stage
T1, 22016 (80)4 (20)0.3810.016
T3, 4219 (43)12 (57)
Size (cm)
<51812 (67)6 (33)0.1030.514
>52313 (57)10 (43)
Tumor grade
1,2108 (80)2 (20)−0.0360.822
3,43117 (55)14 (45)
Tumor number
Single3221 (66)11 (34)0.1800.256
Multiple94 (44)5 (56)
A2210 (45)12 (55)−0.3280.040
B,C1915 (79)4 (21)
TNM stage
l + ll1916 (84)3 (16)0.0420.793
lll + IV229 (41)13 (59)
Figure 1.

Bone morphogenetic protein-9 (BMP-9) is expressed in hepatocellular carcinoma (HCC) and expression of BMP-9 is associated with Smad1 phosphorylation, E-cadherin and Snail expression. Bone morphogenetic protein-9 expression analyses were performed in paraffin-embedded HCC tissues from liver cancer patients using immunohistochemical staining (IHC) and ISH (n = 41 for tumor area and n = 36 for non-tumor area). (a) Immunohistochemical staining for BMP-9 in tumor and paired non-tumor regions of livers from HCC patients (brown color) varied from case to case. The extent was determined as mild (1 + ) and moderate to strong (2 + ). Magnification, 40 ×. (b) Detection of BMP-9 mRNA expression by ISH (violet color). In situ hybridization with sense probe was used as a negative control. Magnification, 40 ×. (c) At 1 +  BMP-9 expression levels, pSmad1 staining was mildly positive, E-cadherin cell surface expression was present and positive nuclear Snail expression was lower compared with those samples having 2 +  BMP-9 expression levels. Representative images are shown. Magnification, 40 ×. (d) Immunohistochemical staining of pSmad1, Snail and E-cadherin in HCC cells was semi-quantitatively evaluated in three grades (–, 1 + , 2 + ) as described in Materials and Methods, and the association with BMP-9 expression was calculated, demonstrating that BMP-9 expression is significantly correlated with nuclear pSmad1, positively related to nuclear Snail, and reversely related to cell surface E-cadherin in HCC cells. (e) Immunofluorescent staining against BMP-9 (green) and E-Cadherin (red) of cryo sections from tumor necrosis factor (TNF)α/cmyc bitransgenic mice, which spontaneously develop HCC. The upper two pictures show typical examples of stage 3 HCC border areas, the second example is shown further enlarged. Black arrows point to cells with high expression of E-Cadherin but low expression of BMP-9 and white arrows point to cells with an opposite pattern.

A comparison of BMP-9 expression levels with clinicopathological features of the 41 investigated HCC patients indicates a significant association of the BMP-9 protein expression level with T stage (τB = 0.381, = 0.016) and Child–Pugh score (τB = −0.328, = 0.04) (Table 2), indicating that BMP-9 levels are positively correlated with invasion, but negatively with the severity of disease.

BMP-9 expression is associated with increased pSmad1 and Snail and decreased E-cadherin levels in HCC

We next examined BMP-9 downstream signaling and features of EMT using antibodies for phosphorylated Smad1 (pSmad1), Snail and E-cadherin and associated their occurrence/expression with BMP-9 protein expression. In HCC patients with 1 + BMP-9 levels, E-cadherin signals were stronger, while those of Snail and pSmad1 were weaker, as compared to liver cancer samples with 2 + BMP-9 levels (Fig. 1c). A semi-quantitative analysis of the staining results indicates a significant correlation of BMP-9 expression with increased activity of the Smad1 signal transduction. Further, expression of BMP-9 was positively related to Snail expression and negatively correlated to E-cadherin expression, although this trend did not reach statistical significance (Fig. 1d).

Because there were no patient samples available showing the invasive front of the tumor, we searched for an HCC animal model that could be used to investigate if BMP-9 is expressed preferentially at the tumor borders. To do so we used cryo sections from mouse livers of TGFα/c-myc bitransgenic mice, which develop HCC endogenously.[32] Immunofluorescent staining of such sections showed that indeed strong BMP-9 positive cells are located at the tumor borders and that cells with high expression of BMP-9 show rather low expression of the epithelial marker E-Cadherin and vice versa (Fig. 1e).

BMP-9 signaling is functional in HCC cell lines

In order to investigate BMP-9 signaling in liver cancer cells in more detail, we assessed whether BMP-9/Smad1 signaling components are expressed in the well-established HCC cell lines, HepG2 and HLE. These cells were originally derived from primary tumors.[33, 34] Histologically, HepG2 are classified as a well-differentiated liver cancer cell line and HLE as an invasive HCC cell line, but commonly representing epithelial phenotypes.

BMP-type I and II receptors ALK1, ALK2, BMPRII and ACVR2A as well as Smads-1 and -4 were expressed in both cell lines at different levels, as determined by PCR (Fig. 2a).

Figure 2.

Bone morphogenetic protein-9 (BMP-9) induces Smad1 phosphorylation in hepatocellular carcinoma (HCC) cells. (a) Components of the Smad1 pathway were slightly or strongly expressed in HepG2 and HLE cells as determined by polymerase chain reaction (PCR). (b) Cells were untreated or treated with different concentrations of BMP-9 for 1 h. pSmad1 and total Smad1 were analyzed by Western blot. (c) Cells were treated with 50 ng/mL BMP-9 for different times as indicated. Glyceraldehyde 3-phosphate dehydrogenase (GAPDH) or β-actin was used as a loading control.

To investigate whether expression of the receptors correlated with BMP-9 responsiveness of the cells, we treated them with rhBMP-9 (50 ng/mL) and examined Smad1 C-terminal phosphorylation. Both HCC cell lines were responsive to BMP-9 even at concentrations as low as 5 ng/mL. Smad1 phosphorylation occurred as early as 10 min after the stimulation, lasting at least 72 h (Fig. 2b,c). Taken together, these data indicate that the BMP-9/Smad1 pathway is functional in HCC cells in vitro with a prolonged instead of a short term response, which is in line with previous descriptions of Smad signaling participating in EMT.[35]

BMP-9 induces EMT in HCC cells in vitro

The immunohistochemical analyses of patient samples shown above point to a correlation between BMP-9 levels and EMT in HCC. Since increased migration is a typical feature of EMT, we performed Transwell assays to investigate if BMP-9 induces EMT in HCC cells in vitro. Treatment with BMP-9 (50 ng/mL) increased cell migration in HLE and HepG2 cells by 2.38- and 3.72-fold, respectively (Fig. 3a). We then investigated cell morphology and expression levels of EMT markers upon stimulation of the cells with BMP-9 for 72 h. Cell morphology showed that BMP-9 led to a conversion from epithelial to fibroblastic cell morphology (Fig. 3b). Although both cell lines are considered to be of epithelial phenotype, basal E-cadherin expression was rather diffuse and not strictly localized to the cell membranes and cell-cell contacts (Fig. 3c). However, BMP-9 treatment reduced the staining intensity for E-cadherin and enhanced that of the mesenchymal marker Vimentin (Fig. 3c). By Western blot analyses we confirmed downregulation of E-cadherin and induced expression of Vimentin (Fig. 3d). Snail, a key transcription factor responsible for the downregulation of E-cadherin was significantly induced by BMP-9 at the RNA level (Fig. 3e).

Figure 3.

Bone morphogenetic protein-9 (BMP-9) enhances cell motility and induces epithelial-mesenchymal transition (EMT) in hepatocellular carcinoma (HCC)cells. (a) Hepatocellular carcinoma cell lines were plated into transwell chambers and treated or not with BMP-9 (50 ng/mL). At the end point of the migration assay, cells in the upper and lower chambers were trypsinized and transferred to a fresh 96-well flat bottom plate. Adenosine triphosphate (ATP) content, which correlates to cell number, was evaluated for migrated and non-migrated cells. Error bars reflect standard error of the mean. *< 0.05, Student's t-test was done to analyze statistical difference. Experiments were performed in triplicate and repeated three times. The average ± standard error of the mean (SEM) of all experiments is shown. (b) Photographs of HCC cells taken after 72 h of stimulation with BMP-9 or left untreated. (c) HLE or HepG2 cells were untreated or treated with BMP-9 50 ng/mL for 72 h after serum-starvation overnight. Cells were immunostained with anti-E-cadherin or anti-Vimentin antibodies. DRAQ5 was used to show the nuclei (scale as indicated in the image). (d) Western blots against E-cadherin and Vimentin using lysates of HCC cells. Glyceraldehyde 3-phosphate dehydrogenase (GAPDH) served as a loading control. (e) Polymerase chain reaction (PCR) showing that BMP-9 induced Snail mRNA expression in both HCC cell lines.

Collectively, these results support the conclusion that BMP-9 mediates EMT and enhances migration in HCC cells, which might be an important feature during tumor invasion and formation of metastasis in vivo.

ALK1 and ALK2 are essential type I receptors for BMP-9-induced downregulation of E-cadherin in HCC cells

Some studies using other cell types have demonstrated that ALK1 and ALK2 are BMP-9 specific type I receptors.[22] In this study, we used adenoviral infection to functionally link BMP-9 mediated signal transduction with downregulation of E-cadherin in HCC cell lines (Fig. 4a). Overexpression of dominant negative forms of these receptors (AddnALK1/2) strongly reduced phosphorylation of Smad1 along with downregulation of E-cadherin upon stimulation with BMP-9 (Fig. 4b). Furthermore, overexpression of constitutive active mutants of both receptors (AdcaALK1 or AdcaALK2) could mimic the effects of BMP-9, leading to phosphorylation of Smad1 and downregulation of E-cadherin (Fig. 4c,d). Similarly BMP-9 mediated upregulation of Snail was mimicked by overexpression of constitutively active Alks and was inhibited by dominant negative forms (Fig. 4e). We further performed immunofluorescent stainings on adenovirally transduced cells. We stained for two markers, ZO-1 (epithelial marker for cell–cell contact sites) and vimentin (mesenchymal marker). In accordance with the above-described Western blot results we found that dominant negative Alks neutralized the BMP-9 effect on EMT while both constitutive active forms induced EMT (Fig. 5a,b). These results imply that BMP-9 induces an EMT response in HCC via the type I receptors ALK1 and ALK2 in HCC cells.

Figure 4.

ALK1 and/or 2 are required for bone morphogenetic protein-9 (BMP-9) mediated E-cadherin expression. HepG2 and HLE cells were infected or not with either a control virus (AdLacZ), a virus expressing a dominant negative mutant of the receptor (AddnALK1/2) or a virus expressing a constitutive active mutant of the receptors (AdcaALK1/2), and treated or not with BMP-9, as indicated. (a) Schema of the experimental setup for adenovirus infection. (b–e) Immunoblots showing Smad1 phosphorylation, total Smad1 expression, E-cadherin and Snail expression and β-actin as a loading control and presence of the HA-tag demonstrates successful overexpression of the constructs.

Figure 5.

Immunofluorescent stainings for ZO-1 (green), vimentin (red) and 4′6′-diamidino-2-phenylindole dihydrochloride (DAPI) (blue; staining the nuclei) were performed on hepatocellular carcinoma (HCC) cells that were treated as described for Fig. 4. (a) HLE cells and (b) HepG2 (original magnification: 400-fold).


Bone morphogenetic protein-9, which belongs to the TGF-β superfamily of cytokines is one of the most potent BMPs to induce bone formation and is predominantly synthesized in the liver.[36, 37] In healthy rat liver, non-parenchymal cells, like hepatic stellate cells, Kupffer cells and endothelial cells were reported as major sources of BMP-9. In another study, cholangiocytes and hepatocytes were identified as major producers of BMP-9.[37] These authors also detected BMP-9 expression in whole liver lysates as well as different liver cell types including cholangiocytes, hepatocytes, hepatic stellate cells and hepatic sinusoidal endothelial cells at the RNA and protein levels. These data indicate that basal expression of BMP-9 exists in healthy liver. Furthermore, the BMP-9 mRNA level in cholangiocytes was 6 times higher than in hepatocytes and nearly 100 times higher than in hepatic stellate cells and hepatic sinusoidal endothelial cells. These results are mainly consistent with ours. With ISH we found a positive signal for BMP-9 mRNA in normal mouse liver, which was located mainly in cholangiocytes, whereas in hepatocytes and other cell types, it was rather low or negative (data not shown). Most likely, ISH is not sensitive enough to detect basal BMP-9 levels in these other cell types under healthy conditions.

Regarding the physiological functions of BMP-9, it was found to circulate in plasma and keep endothelial cells in a resting state. Further, significant roles in glucose-[38] and iron homeostasis[39] were identified. In liver cells, BMP-9 has been reported to induce proliferation of rat hepatocytes[40] and participate in liver regeneration. Although these findings give some preliminary hints, BMP-9′s function, especially in diseased liver, is only poorly understood up to now.

The functions of TGF-β in carcinogenesis and tumor progression of various cancer entities have been intensely studied already. In premalignant cells, TGF-β serves as a tumor suppressor, whereas in later stages of cancer, it functions as a tumor promoter. It has not been clearly clarified how and when TGF-β is converted from a tumor suppressor to a tumor promoter during carcinogenesis. Bone morphogenetic proteins have also been demonstrated to participate in formation and progression of human cancer. For example, BMP-2 inhibits growth of gastric cancer cells through increasing the levels of p21/WAF1/CIP1, leading to cell cycle arrest in the G1-phase,[41] while it facilitates proliferation of lung cancer cells through Smad1/5 signaling and induction of Id-1 expression.[42] Nine different BMPs are upregulated in HCC cells and their described functions in disease progression include cell proliferation, migration and angiogenesis.[17] Bone morphogenetic protein-4 and BMP-7 are upregulated in HBx-induced HCC mouse models and in HBV-related HCC patients and were therefore included to what they defined “most common regulators” during the transition from normal liver to HCC. Furthermore, their ectopic overexpression increased cell viability and promoted migration in Hep3B cells.[43] Overexpression of BMP-4 is significantly associated with the number of tumor nodules, TNM stage and vascular invasion, and was proposed as a new marker to predict recurrence and prognosis of HCC patients.[20] Further, BMP-4 induces cell proliferation and migration in CC cells, HepG2 and Hep3B,[44] whereas BMP-2 induces angiogenesis in Bel7402 and SMMC7721 tumor xenografted nude mice. WSS25, an antagonist of BMP-2, inhibits angiogenesis via blocking BMP/Smad1/Id1 signaling in these mouse models.[18]

In human cancer, variant functions for BMP-9 have been described depending on the cancer type. In prostate and breast cancer, BMP-9 provides tumor suppressor activity, since it inhibits growth, migration and invasion,[29, 45] whereas in ovarian cancer, it acts as a proliferation promoter via activation of ALK2/Smad1 signaling.[28] Although BMP-9 was found upregulated in certain HCC cell lines,[17] so far, its contribution to HCC development and progression has not been explored.

In the present study, we demonstrate that BMP-9 mRNA and protein are expressed at different degrees in liver tissues from HCC patients. Slightly positive (1 + ) staining of BMP-9 protein was observed in 61% and moderately to strongly positive staining (2 + ) in 39% of HCC tissue. Bone morphogenetic protein-9 protein expression levels were higher in the cancer area than in its adjacent liver tissue in 25% of the patients and were similar in both areas in 58% of patients. In the remaining 16%, expression levels were even lower in the cancer area than in its adjacent liver tissue. Based on the finding that liver is the main organ producing BMP-9,[40] we can expect that low levels of BMP-9 expression should always be detectable. Furthermore, the “normal” liver tissues investigated in the present study came from HCC patients and should therefore not be considered healthy (see Table S1). Comparing these results with our ISH of healthy livers, which showed strong positivity only in cholangiocytes (data not shown), together with the 25% of patients showing increased presence of BMP-9 in HCC compared to adjacent tissue, we conclude that there is a tendency of increase in BMP-9 in diseased liver, especially in HCC. In order to further investigate if BMP-9 is especially expressed within the tumor, at the borders of the tumor or in the adjacent tissue, we used sections from livers of TGFα/c-myc bitransgenic mice, which develop HCC endogenously.[32] The resulting stainings show high presence of BMP-9 especially in cells at the tumor border. In addition, many cells in this area express either the epithelial marker E-Cadherin or are strongly positive for BMP-9, giving first hints that BMP-9 leads to reduced expression of E-Cadherin. We want to mention also that 95% of the investigated HCC patients were HBV-infected and it will be an interesting task to investigate in the future if HBV infection as such impacts BMP-9 expression and if there are differences in regard to disease etiology.

Expression of BMP-9 in HCC, as resolved with IHC staining, was significantly associated with the T stage, showing stronger expression in patients with T3, T4 than in those with T1, T2, indicating that high levels of BMP-9 correlate with the degree of tumor invasion into the surrounding tissue. When correlating BMP-9 expression with the Child–Pugh class of the patients, we found that Child–Pugh class B/C was stronger associated with the group with lower BMP-9 levels (1 + ; 79%), while in patients with Child–Pugh class A the distribution was almost 50% (1 + ) and 50% (2 + ). This inverse correlation shows that although BMP-9 correlates with invasiveness of the tumor (T stage), it does not seem to increase along with loss of liver function as determined by Child–Pugh classification.

Studying HCC cells in vitro, we found that BMP-9 promotes cell migration, a typical feature of mesenchymal cells. Together with the known pro-proliferative effect of BMP-9 in liver cells,[40] this observation prompted us to investigate if BMP-9 induces EMT in HCC cells. Epithelial-mesenchymal transition is required for epithelial cancer cell migration and metastasis and downregulated epithelial (E)-Cadherin and upregulated Snail are well described features of EMT. E-cadherin is a Ca2 + -dependent cell adhesion transmembrane glycoprotein, which links adjacent cells by hemophilic interactions,[46] thus representing a typical marker for epithelial polarity. Membranous E-cadherin as well as the cell–cell contact protein ZO-1 are downregulated during loss of the epithelial phenotype representing classical molecular hall marks of EMT. Expression of E-cadherin is negatively regulated by Snail at the transcriptional level. Appearance of mesenchymal markers like Vimentin or N-cadherin is another typical feature of EMT. Indeed, besides promoting cell migration, BMP-9 downregulated E-cadherin and ZO-1 and induced Snail and Vimentin expression, indicating that BMP-9 induces EMT in HCC cells. In accordance with these in vitro findings, in HCC patient samples, BMP-9 expression was positively correlated with Snail expression and was reversely correlated with E-cadherin expression, although these correlations did not reach statistical significance, which may be due to the relatively small number of samples. Hence, our results are supportive of the hypothesis that BMP-9 may promote migration and metastasis of HCC cells via induction of EMT.

We further investigated the BMP-9 downstream signaling pathway involved in EMT, and especially in BMP-9 induced downregulation of E-cadherin. In mesenchymal stem cells, BMP-9 was reported to signal via the BMP type I receptors ALK1 and ALK2, which both activate Smad1 signaling to mediate osteogenesis.[22] We here show that ALK2 is expressed in both HCC cell lines tested and by using adenovirally overexpressed receptor mutants, we could confirm that dominant negative or constitutive active ALKs 1 and 2 impact E-cadherin, ZO-1 and vimentin expression levels and EMT in HLE and HepG2 cells. We conclude that similar to other cell types, in HCC cells BMP-9 also signals via ALKs 1 and/or 2.

Downstream of ALK1/2 activation, Smad1 is phosphorylated and mediates the signal transduction to the nucleus. Bone morphogenetic protein/Smad1 signaling was described to participate in different tumor types with diverse outcomes, either as tumor promoter or as tumor suppressor. Activation of Smad1/5 is correlated with TGF-β-induced growth inhibition in B-cell lymphoma.[47] In human breast cancer cells, activation of BMP-2/Smad1 signaling inhibits cancer cell proliferation via upregulation of p21.[48] In breast cancer patients and a corresponding xenograft mouse model, enhanced phospho-Smad1 staining was found in bone metastasis as compared to the primary tumor or lymph node metastases, which suggests that Smad1 signaling contributes to bone metastasis of breast cancer.[49] In pancreatic cancer, BMP-2, 4 and 7 induce EMT and increase cancer cell invasiveness, and their common canonical downstream signaling component, Smad1 is indispensable for BMP-mediated invasiveness.[50] In liver cells, BMP-9 signaling has not been characterized in detail yet, but similar to HCC cells primary mouse hepatocytes respond to BMP-9 with strong activation of the Smad1 pathway (data not shown). In line with the presence of BMP-9, Smad1 phosphorylation was increased in HCC cells of liver samples from patients. These results further support the conclusion that the BMP-9/Smad1 pathway might be involved in the pathogenesis of HCC. However, we can at this stage not exclude that besides BMP-9, other members of the TGF-β superfamily (e.g. other BMPs or TGF-β) also take part in inducing Smad1 phosphorylation in vivo.

In summary, our results implicate the BMP-9/Smad1 pathway as a potential promoter of HCC via inducing EMT. Further studies with larger numbers of human patient samples and using animal models with a disturbed BMP-9 signaling pathway are planned to elucidate the functional implication of BMP-9 and its downstream signaling in liver cancer progression and to estimate its potential as new marker for HCC progression and prognosis as well as therapeutic target to treat HCC.


The study was supported by German Research Foundation programs “SFB TRR77 Liver Cancer” and “Do373/8-1″ and Federal Ministry of Education and Research grants “The Virtual Liver” and “cell therapy in Liver Regeneration” (S.D). Qi Li is a fellow supported by China Scholarship Council. We wish to thank Professor Dr Peter ten Dijke (Leiden, Netherland) for providing AddnALK1/2, AdcaALK1, AdcaALK2, Professor Dr Carl-Henrik Heldin (Uppsala, Sweden) for AdLacZ and AdSmad1 and Professor S. Thorgeirsson (National Cancer Institute, NIH, Bethesda, Maryland, USA) for giving permission to Professor Piiper to perform the TNFα/c-myc HCC mouse model.

Disclosure Statement

The authors have no conflict of interest.