Overactivation of the TGF-β pathway confers a mesenchymal-like phenotype and CXCR4-dependent migratory properties to liver tumor cells


  • Potential conflict of interest: Nothing to report.


Transforming growth factor-beta (TGF-β) is an important regulatory suppressor factor in hepatocytes. However, liver tumor cells develop mechanisms to overcome its suppressor effects and respond to this cytokine by inducing other processes, such as the epithelial-mesenchymal transition (EMT), which contributes to tumor progression and dissemination. Recent studies have placed chemokines and their receptors at the center not only of physiological cell migration but also of pathological processes, such as metastasis in cancer. In particular, CXCR4 and its ligand, stromal cell-derived factor 1α (SDF-1α) / chemokine (C-X-C motif) ligand 12 (CXCL12) have been revealed as regulatory molecules involved in the spreading and progression of a variety of tumors. Here we show that autocrine stimulation of TGF-β in human liver tumor cells correlates with a mesenchymal-like phenotype, resistance to TGF-β-induced suppressor effects, and high expression of CXCR4, which is required for TGF-β-induced cell migration. Silencing of the TGF-β receptor1 (TGFBR1), or its specific inhibition, recovered the epithelial phenotype and attenuated CXCR4 expression, inhibiting cell migratory capacity. In an experimental mouse model of hepatocarcinogenesis (diethylnitrosamine-induced), tumors showed increased activation of the TGF-β pathway and enhanced CXCR4 levels. In human hepatocellular carcinoma tumors, high levels of CXCR4 always correlated with activation of the TGF-β pathway, a less differentiated phenotype, and a cirrhotic background. CXCR4 concentrated at the tumor border and perivascular areas, suggesting its potential involvement in tumor cell dissemination. Conclusion: A crosstalk exists among the TGF-β and CXCR4 pathways in liver tumors, reflecting a novel molecular mechanism that explains the protumorigenic effects of TGF-β and opens new perspectives for tumor therapy. (Hepatology 2013; 58:2032–2044)






stromal cell-derived factor 1α




the epidermal growth factor receptor


epithelial-mesenchymal transition


hepatocellular carcinoma


transforming growth factor-beta


transforming growth factor-beta receptor-1

Transforming growth factor-beta (TGF-β) is an important regulatory suppressor factor; however, paradoxically, it also modulates other processes that contribute to tumorigenesis, such as fibrosis, immune regulation, microenvironment modification, and cell invasion.[1] Indeed, in addition to its suppressor effects, TGF-β induces antiapoptotic signals in fetal hepatocytes and hepatoma cells,[2, 3] through activation of the epidermal growth factor receptor (EGFR) pathway.[4] Cells that survive to TGF-β-induced apoptotic signals undergo epithelial-mesenchymal transition (EMT).[3, 5, 6] Upon progression of liver cancer, EMT is considered a key process that may drive intrahepatic metastasis.[7] TGF-β levels are increased in hepatocellular carcinoma (HCC) tissue, plasma, and urine and decreased in patients who underwent effective therapy for HCC.[8] Liver tumors expressing late TGF-β-responsive genes (antiapoptotic and EMT-related genes) display a higher invasive phenotype and increased tumor recurrence when compared to those that show an early TGF-β signature (suppressor genes).[9] Interestingly, blocking TGF-β up-regulates E-cadherin and reduces migration and invasion of HCC cells.[10]

Recent studies place chemokines and their receptors at the center not only of physiological cell migration, but also of pathological processes, such as metastasis in cancer.[11] In particular, CXCR4 and its ligand, stromal cell-derived factor 1α (SDF-1α) / chemokine (C-X-C motif) ligand 12 (CXCL12), have been revealed as important molecules involved in the spreading and progression of a variety of tumors.[12] Different data suggest that molecular strategies to inhibit the CXCR4/CXCL12 pathway could be of therapeutic use for the treatment of HCC.[13] CXCR4 is up-regulated in human HCC,[14] correlating with progression of the disease.[15] Its ligand CXCL12 stimulates human hepatoma cell growth, migration, and invasion.[14] We have recently described that TGF-β up-regulates CXCR4 in rat hepatoma cells[16] and sensitizes cells to respond to CXCL12, which mediates cell scattering and survival. These results suggest a crosstalk between the increased protumorigenic response to TGF-β and the establishment of a functional CXCR4/CXCL12 axis. Nothing is known about whether a similar situation occurs in human liver tumorigenesis.

The aim of this work was to analyze whether autocrine stimulation of TGF-β in human liver tumors may induce up-regulation, and/or intracellular reorganization, of CXCR4, which, concomitant with the EMT process induced by this factor, would contribute to the enhancement of cell migration and invasion.

Materials and Methods

Ethics Statement

Approval for experiments related to the study of liver carcinogenesis in experimental animal models was obtained from the General Direction of Environment and Biodiversity, Government of Catalonia, #4589, 2011. All animals received humane care and study protocols comply with the institution's guidelines. Human tissues were collected with the required approvals from the Institutional Review Board (Comité Ético de Investigación Clínica del Hospital Universitario de Bellvitge) and patient's written consent conformed to the ethical guidelines of the 1975 Declaration of Helsinki.

Cell Culture

Cell lines used in this study were from commercial sources. Hep3B, HepG2, and PLC/PRF/5 were obtained from the European Collection of Cell Cultures (ECACC). SNU449 were obtained from the American Tissue Culture Collection (ATCC). Huh7 and HLF cells were from the Japanese Collection of Research Bioresources (JCRB Cell Bank) and were kindly provided by Dr. Perales (University of Barcelona, Spain) and Dr. Giannelli (University of Bari, Italy), respectively. Cell lines were never used in the laboratory for longer than 4 months after receipt or resuscitation.

HepG2 and Hep3B were maintained in modified Eagle's medium (MEM) medium, PLC/PRF/5 and Huh7 in Dulbecco's modified Eagle's medium (DMEM) medium, SNU449 and HLF in RPMI medium. Neonatal mice hepatocytes were immortalized as described[17] and cultured in DMEM. All media (Lonza, Basel, Switzerland) were supplemented with 10% fetal bovine serum (FBS; Sera Laboratories International, Cinder Hill, UK) and cells maintained in a humidified atmosphere of 37°C, 5% CO2. Analysis of cell viability was performed by Crystal violet staining.[3]

Immunofluorescence Staining

Fluorescence microscopy studies were performed as described[3] (further details in the Supporting Materials and Methods). Cells were visualized with a Nikon eclipse 80i microscope with the appropriate filters. Representative images were taken with a Nikon DS-Ri1 digital camera. ImageJ software (National Institutes of Health [NIH], Bethesda, MD) was used to analyze fluorescence from TIFF images captured using the same exposure conditions.


Human HCC tissues were obtained from the Pathological Anatomy Service, University Hospital of Bellvitge, Barcelona. Paraffin-embedded tissues were cut into 4-μm-thick sections, incubated with the specific primary antibody overnight at 4°C, and binding developed with the Vectastain ABC kit (Vector Laboratories, Burlingame, CA). Further information is supplied in the Supporting Materials and Methods.

Western Blot Analysis

Total protein extracts and western blotting procedures were carried out as described.[3] Source of antibodies are detailed in the Supporting Materials and Methods.

Analysis of Gene Expression

RNeasy Mini Kit (Qiagen, Valencia, CA) was used for total RNA isolation. Reverse transcription (RT) was carried out using the High Capacity Reverse Transcriptase kit (Applied Biosystems, Foster City, CA), and 500 ng of total RNA from each sample for complementary DNA synthesis. For details about semiquantitative and real-time polymerase chain reaction (PCR) reactions, see the Supporting Materials and Methods.

RNA Interference Assays

Cells at 70% confluence were transiently transfected with 50 nM small interfering RNA (siRNA) for 8 hours using TransIT-siQuest following the manufacturer's instructions (Mirus, Madison, WI). For stable transfection of short hairpin RNA (shRNA), cells at 50%-60% confluence were transfected with MATra-A reagent (IBA, Germany) according to the manufacturer's recommendation (15 minutes on the magnet plate, 2 μg/mL of shRNA plasmid). Four different plasmids of TGFBRI shRNA were transfected separately or combined, as well as a control shRNA. Protocols used were as described.[18] For siRNA sequences and further experimental details, see the Supporting Materials and Methods.

Migration Assays

Cell motility was examined by two different methods: (1) a wound-healing assay[16] and (2) real-time migration assay through the xCELLigence system (Roche Applied Science). For the wound-healing assay, cells were grown at basal conditions to 95% confluence and monolayers were scratched with a pipette tip (0 hours). Cell migration was recorded by phase contrast microscopy (Olympus IX-70) at 48 hours after wound scratch. For real-time monitoring of cell migration, the xCELLigence system was used; 4 × 104 cells/well were seeded onto the top chamber of a CIM plate, which features microelectronic sensors integrated on the underside of the microporous membrane of a Boyden-like chamber. CIM plates were placed onto the Real-Time Cell Analyzer (RTCA) station (xCELLigence System, Roche, Mannheim, Germany). Cell migration was continuously monitored by measuring changes in the electrical impedance at the electrode/cell interface, as a population of cells migrated from the top to the bottom chamber. Continuous values are represented as cell index (CI), a dimensionless parameter that reflects a relative change in measured electrical impedance, and quantified as a slope (h−1) of the first 5 hours.

Diethylnitrosamine (DEN)-Induced Hepatocarcinogenesis in Mice

Male mice at day 15 of age received intraperitoneal injections of DEN (5 mg/kg) diluted in saline buffer, control animals were injected with saline buffer intraperitoneally. At 6, 9, and 12 months of age, mice were sacrificed and their livers removed. For histological studies, liver lobes were fixed in 4% paraformaldehyde overnight and paraffin-embedded for immunohistochemistry staining. Total RNA was isolated from frozen tissues to analyze gene expression by real-time quantitative PCR. Three to four animals/condition and two different tissue pieces/animal were processed for RNA extraction.


All data represent at least three experiments and are expressed as the mean ± SEM. Differences between groups were compared using either Student t test or one-way analysis of variance (ANOVA) associated with Dunnett's test. Statistical significance was assumed when P < 0.05. The analysis was performed using GraphPad Prism software (Graph-Pad for Science, San Diego, CA). For data from human samples, statistical significance between means was determined by the nonparametric Mann-Whitney U test. Correlation between TGF-β and CXCR4 mRNA levels was determined by the Pearson correlation coefficient.


Mesenchymal-Like Phenotype in HCC Cells Correlates With SMADs Activation and Resistance to TGF-β-Induced Suppressor Effects

In order to evaluate the relevance of the autocrine stimulation of TGF-β pathway in the acquisition of mesenchymal-like features, we analyzed the phenotype of six different human liver tumor cell lines whose characteristics are detailed in Supporting Table 1. A correlation between the decrease in E-cadherin and cytokeratin-18 (CK-18) expression, characteristics of an epithelial phenotype, and the appearance of cells expressing vimentin (a mesenchymal intermediate filament) was observed (Fig. 1A). The acquisition of a mesenchymal-like phenotype occurred concomitantly with an increase in the expression of TGFB1 (Fig. 1B) and with nuclear localization of both SMAD2 and SMAD3 (Supporting Fig. 1). Analysis of TGF-β in the culture medium revealed increased amounts of this cytokine in mesenchymal-like versus epithelial cell lines. Furthermore, conditioned medium from mesenchymal-like HCC cells induced higher Smad2 phosphorylation in immortalized mice hepatocytes (Supporting Fig. 1). With the exception of the HepG2 cells that show mutations in NRAS and are resistant to TGF-β-induced suppressor effects,[19] the epithelial phenotype correlated with response to TGF-β as a cytostatic factor, whereas cells with a mesenchymal-like phenotype did not arrest proliferation in the presence of TGF-β (Fig. 1C). This behavior confirms a previous classification of these cell lines according to the TGF-β signature[9] (early for PLC/PRF/5 and Huh7; late for SNU449, HLF). Results in Hep3B indicate that these cells represent a transition from an epithelial to a mesenchymal-like phenotype, since they showed decreased expression of E-cadherin and simultaneous expression of epithelial (CK-18) and mesenchymal (vimentin) intermediate filaments (Fig. 1A). Interestingly, this mixed phenotype correlated with a high activation of the TGF-β pathway (Supporting Fig. 1) and lower suppressor response to this cytokine (Fig. 1C). In summary, mesenchymal-like phenotype in HCC cell lines correlates with autocrine stimulation of the TGF-β pathway and resistance to TGF-β-induced suppressor effects.

Figure 1.

Autocrine stimulation of the TGF-β pathway in HCC cells correlate with a mesenchymal-like phenotype and resistance to TGF-β-induced suppressor effects. HepG2, PLC/PRF/5, Huh7, Hep3B, SNU449, HLF were cultured under standard conditions in 10% FBS. (A) Immunofluorescence of E-cadherin (green), CK-18 (green), vimentin (green), and DAPI (blue). (B) TGFB expression levels determined by real-time PCR. Mean ± SEM (n = 5). (C) Effect of 48 hours of treatment with 2 ng/mL TGF-β on cell viability, analyzed by Crystal violet staining: data were calculated relative to zero time and represent the mean ± SEM of at least six independent experiments. Student t test was calculated versus zero time for each cell type: ***P < 0.001.

Mesenchymal-Like Phenotype in HCC Cells Correlates With CXCR4 Up-regulation and Asymmetric Distribution, Which Is Required for Cell Migration

The analysis of the cytoskeleton organization reflected that cells with more mesenchymal phenotype presented F-actin located in stress fibers, whereas the more epithelial ones showed more pericellular distribution (Fig. 2A, left panels). Cells with mesenchymal characteristics showed CXCR4 in an asymmetric distribution in a great percentage of them (Fig. 2A, right panels). HepG2 cells showed homogeneous distribution of CXCR4 with no apparent polarization, whereas in the epithelial Huh7 and PLC/PRF/5 localization of CXCR4 was variable, with some cells showing polarized areas, but a great percentage containing homogeneous intracellular localization (Fig. 2A, right panels, quantification of the protrusions in Fig. 2B). Furthermore, analysis of CXCR4 expression at the messenger RNA (mRNA) levels revealed that cells with mesenchymal-like characteristics presented a higher expression of CXCR4, when compared with the more epithelial ones (such as HepG2) (Fig. 2C). Levels of TGFB1 mRNA showed correlation not only with the mesenchymal-like phenotype, but also with CXCR4 levels (Fig. 2D). In agreement with their mesenchymal characteristics and F-actin distribution, the migratory capacity of Hep3B and SNU449 was much higher than that observed in HepG2, analyzed through the xCELLigence technology or in a wound-healing assay (Fig. 2E,F). Interestingly, in mesenchymal-like cells, such as Hep3B (Fig. 2G) or SNU449 (results not shown), the cells in the migration front showed a strong polarization of CXCR4. The presence of AMD3100, a well-known inhibitor of the CXCR4 receptor, inhibited migration of both Hep3B and SNU449 (Fig. 2F). Furthermore, only cells that showed CXCR4 elevated expression and asymmetrical distribution, such as SNU449, responded to CXCL12 inducing migration, whereas HepG2 cells did not (Supporting Fig. 2). All these results together indicate that autocrine stimulation of the TGF-β pathway in HCC cell lines correlates with activation of the CXCR4/CXCL12 axis, which mediates cell migration.

Figure 2.

Overactivation of the TGF-β pathway correlates with high expression and asymmetric distribution of CXCR4, which is required for cell migration. (A) Immunofluorescence of F-ACTIN (red), CXCR4 (green), and DAPI (blue). (B) Percentage of CXCR4 protrusions/cell number. (C) CXCR4 mRNA levels by real-time PCR relative to HepG2 levels. (B,C) Mean ± SEM (n = 3). Student t test versus HepG2 cells: **P < 0.01, ***P < 0.001. (D) Correlation among TGFB and CXCR4 mRNA levels in the individual analyses performed in the different cell lines. (E) Real-time migration assay (xCELLigence system, Roche). (F) Cells were untreated or treated with 1 μg/mL AMD3100. Wound-healing assay (48 hours after wound scratch). (G) Immunofluorescence of CXCR4 (green) and DAPI (blue) of Hep3B, 10 minutes after wound scratch. (E-G) Representative experiments (n = 3).

Targeting TGF-β Receptor I Attenuates the Mesenchymal Phenotype, Decreases CXCR4 Expression, and Impairs Cell Migration in HCC Cells

To analyze whether the autocrine stimulation of the TGF-β pathway induces CXCR4 expression and/or its asymmetric distribution, we stably silenced TGFBR1 expression with specific shRNA in Hep3B (Fig. 3A) and PLC-PRF5 cells (Supporting Fig. 3). Increase in E-cadherin, which presented a pericellular distribution, and decrease in vimentin expression were observed in TGFBR1-silenced Hep3B cells (Fig. 3B,C, left). Cytoskeleton organization changed in the absence of TGFBR1 expression, showing a more pericellular distribution and fewer stress fibers (Fig. 3C,D). CXCR4 expression was inhibited in these cells (Fig. 3B,C, right, and D), which correlated with a significantly lower capacity to migrate (Fig. 3E). Silencing of TGFBR1 also correlated with reorganization of cytoskeleton and attenuation of CXCR4 expression and asymmetric distribution in PLC/PRF/5 cells (Supporting Fig. 3). A pharmacological inhibitor of the kinase activity of TGFBR1, LY36497, which attenuated SMAD2 phosphorylation in HCC cells both in the absence or presence of TGF-β (Supporting Fig. 4), decreased CXCR4 levels (Fig. 4A), increased E-cadherin (CDH1) mRNA levels (although changes were more moderate and less significant than the TGFBR1 silencing: Supporting Fig. 4), reorganized the cytoskeleton and decreased the percentage of cells with an asymmetric distribution of CXCR4 (Fig. 4B). Interestingly, treatment with LY36497 inhibited the capacity of cells to close the wound in migration experiments (Fig. 4C). In summary, TGF-β signaling is responsible for up-regulation and asymmetric distribution of CXCR4 in HCC cells.

Figure 3.

Stable silencing of TGFBR1 recovers the epithelial phenotype and attenuates CXCR4 expression, inhibiting cell migratory capacity. Hep3B cells were stably transfected with an unspecific shRNA (Hep3B-shUns) or a pool of four different shRNAs against TGFBRI, as detailed in the Supporting Materials and Methods, (Hep3B-shTGFBR1) and were comparatively studied: (A) western blot (left) and real-time PCR of TGFBRI (right). (B) E-cadherin (CDH1) and CXCR4 expression levels by real-time PCR. (C) Immunofluorescence of: left: E-cadherin (green), vimentin (green), and DAPI (blue); right: F-ACTIN (red), CXCR4 (green), and DAPI (blue). (D) Quantitative analysis of the F-ACTIN stress fibers area (top) and the number of CXCR4 protrusions relative to the number of cells (bottom). (E) Analysis of cell migration. Top: wound-healing experiment (48 hours after wound scratch). Bottom: real-time migration assay (xCELLigence system). (A, left, C, E, top) Representative experiments (n = 3). (A, right, B, D, E, bottom) Mean ± SEM (n = 3). Student t test *P < 0.05, **P < 0.01, ***P < 0.001.

Figure 4.

Pharmacological inhibition of TGFBR1 kinase activity prevents CXCR4 expression and polarization in the mesenchymal HCC cell lines. Hep3B, SNU449, and HLF cells were incubated in the presence or absence of 3 μM LY36497 for 48 hours. (A) Western blot analysis of CXCR4. A representative experiment of three is shown. (B) Immunofluorescence of F-ACTIN (red), CXCR4 (green), and DAPI (blue). Percentage of CXCR4 protrusions/cell number is represented on the right of each picture as the mean ± SEM of three independent experiments. Student t test versus untreated cells: **P < 0.01. (C) Wound-healing experiment in Hep3B cells with or without 3 μM LY36497 at 48 hours after wound scratch.

Mice Tumors From DEN-Induced Liver Carcinogenesis Show High Expression of Both TGF-β1 and CXCR4

In order to know whether the TGF-β/CXCR4 crosstalk shows significance during in vivo hepatocarcinogenesis, we started with the analysis of tumors in a model of DEN-induced experimental liver tumorigenesis in mice. At different times after a single dose of DEN in 15-day-old animals, liver was collected and analyzed. The appearance of tumors was observed microscopically in all male mice at 9 months of age, but clear macroscopic observation of relevant tumor masses was not observed until 12 months (Supporting Fig. 5). Real-time PCR analysis revealed a progressive increase in the expression of TGFB1, TGFBR1, and CXCR4 in livers from mice of 9 to 12 months of age (Fig. 5A). Increased expression of TGFB1 correlated with a higher percentage of cells showing nuclear localization of phospho-SMAD2 and phospho-SMAD3 in immunohistochemical studies (Fig. 5B). Cells in the border of the tumor presented the maximal level of CXCR4 expression (Fig. 5C). Importantly, it was possible to observe some CXCR4-positive cells invading the stroma. The expression of CXCL12/SDF-1α was concentrated in perivascular or ductal cells, which could induce the stimulus for cells to migrate toward these areas. Furthermore, we found that immortalized mice hepatocytes in culture were able to respond to TGF-β by inducing CXCR4 expression, a process that was SMAD2/3-dependent (Fig. 5D). In summary, tumor cells in the DEN-induced mice model of liver tumorigenesis show increased activation of the TGF-β pathway, which correlates with enhanced CXCR4 levels that concentrates particularly in the cells of the tumor border line.

Figure 5.

Tumorigenesis in mice following DEN treatment is associated with increased TGF-β and CXCR4 signaling. (A) Tgfb, TgfbrI, and Cxcr4 transcript levels analyzed by real-time PCR in control liver (PBS treatment) versus tumoral tissues in animals at 6, 9, and 12 months of age. Data represent mean ± SEM (n = 4 in control animals; 3 in DEN-treated animals; analysis in two different pieces of tissue/animal). Student t test *P < 0.05, **P < 0.01, ***P < 0.001. (B,C) Immunohistochemistry analysis of serial sections of liver from mice after DEN or PBS treatment (12 months). (D) Effect of transient knockdown of SMAD2 or SMAD3 on Cxcr4 mRNA levels analyzed by RT-PCR in immortalized neonatal hepatocytes treated during 24 hours with or without 2 ng/mL TGF-β. A representative experiment (n = 3).

Expression of CXCR4 in Human HCC Tissues Correlates With an Active TGF-β Pathway and Is Concentrated in Areas of Cell Spreading

Finally, we wanted to know whether TGF-β1 signaling and CXCR4 expression correlated in human HCC tissues. We analyzed tissues from 17 patients with HCC from different etiologies (Table 1). Heterogeneity among HCC tumors, with variable expression of TGFB1 and its receptor TGFBR1, was observed. Nevertheless, when calculated as the mean among the patients, the expression was significantly increased in tumor tissues versus their surrounding nontumoral tissues. Analysis of CXCR4 was also variable, but again the tendency was to an increased expression in the tumor tissues (Fig. 6A). However, the most interesting way to dissect the results was individually (Fig. 6B), considering each patient independently. In all the patients showing increased expression of CXCR4, TGFB1 expression was also enhanced, with the exception of patient 8, who presented CXCR4 expression mainly in areas of infiltration (results not shown). This patient suffered from an autoimmune disease. This direct correlation was not necessarily true the other way around, since some patients with increased expression of TGFB1 did not show higher expression of CXCR4 (patients 9, 10, 13, 17). Of note, the increased expression of TGFB1 at the mRNA level correlated with higher levels of TGFB1 protein in the tissues from these patients, not only in the tumoral cells but also in the surrounding stroma and perivascular areas (Fig. 6B,C). Nuclear location of phospho-SMAD2 confirmed the activation of the TGF-β signaling. In a similar way to that observed in the mice model, CXCR4-positive cells were mainly located in the border of the tumor or in the perivascular area (Fig. 6B,C) and CXCL12 expression was found in the stroma, infiltration areas, and in ductal and perivascular cells. It is worth noting that it was possible to observe CXCR4-positive cells trying to invade the vasculature and infiltrating the peritumoral capsule (Fig. 6D). Interestingly, CXCR4-positive tumor cells surrounding vascular areas showed disorganization of E-cadherin, which reflects a less differentiated, more mesenchymal, and migratory phenotype (Fig. 6C). In fact, the highest expression of both TGF-β and CXCR4 significantly correlated with the lowest stages of differentiation in the HCC patients analyzed (Supporting Fig. 6A). Furthermore, patients with a cirrhotic background showed the highest levels of CXCR4 and, interestingly, the tumor surrounding (cirrhotic) tissue from these patients contained significantly higher levels of both TGF-β and CXCR4 when compared with the surrounding tissue from noncirrhosis patients (Supporting Fig. 6B). Immunohistochemical analysis of CXCR4 in tissues from patients with different grades of fibrosis (no tumors yet) revealed progressive increase in the expression of this protein, which correlated with higher activation of the TGF-β pathway, analyzed as SMAD2 phosphorylation (Supporting Fig. 6C).

Table 1. Patient and Tumor Characteristics
CaseAge/SexEtiologyBackgroundSize (cm)Tumoral Focus/ Satellite NodulesHistological GradeMicroscopic Vascular InvasionMacroscopic Vascular InvasionpT / Stage
  1. Gender: F (female); M (male). Histological grade according to the criteria of Edmondson and Steiner: 1, well differentiated; 2, moderately differentiated; 3, poorly differentiated; 4, undifferentiated. Etiology: HBV (hepatitis B virus); HCV (hepatitis C virus); NBNC (HBV(-), HCV(-)); alcohol (heavy alcohol use); U (unknown etiology); Aut (autoimmune); background: NL (normal liver); LC (liver cirrhosis).

146/MAlcLC31/03NoNo1 / I
250/MAlc, HCVLC2,52/02-3NoNo2 / II
375/MAlcLC31/03NoNo1 / I
449/FUNL271/03NoNo1 / I
578/FUNL7,51/13YesNo2 / II
650/MAlcLC29/02YesNo2 / II
769/MHCVLC3-3NoYes1 / I
865/FAutLC65/02YesNo3 / IIIa
961/MHCVLC2,5-2NoNo2 / II
1062/MHCVLC4,51/03NoNo1 / I
1174/MUNL61/02NoNo1 / I
1252/MHCVLC3,81/02NoNo1 / I
1382/MUNL71/01-2NoNo1 / I
1446/MHCVLC8,51/multi3YesNo2a / II
1571/MUNL201/multi2YesNo2 / II
1664/MHCVLC3,31/03NoNo1 / I
1771/MHCVLC4,51/12-3NoNo1 / I
Figure 6.

High expression of CXCR4 correlates with activation of the TGF-β pathway and a less differentiated phenotype in HCC tumor tissues. (A) TGFB, TGFBRI, and CXCR4 transcript levels analyzed by real-time PCR, comparing tumor versus surrounding tissue in 17 HCC patients (up). Relative expression of TGFB and CXCR4 of each individual tumor versus its respective surrounding tissue, represented in a logarithmic scale for a better understanding of changes observed (down). (B) Immunohistochemistry analysis of serial sections in two representative HCC patients (1 and 5), compared with healthy tissue. (C) Immunohistochemistry analysis of tumor arteries. (D) Magnification of different CXCR4 localizations within the tumor.

In summary, a great percentage of HCC tumors express high levels of CXCR4 that is always coincident with activation of the TGF-β pathway and correlates with a dedifferentiation stage and a cirrhotic background. CXCR4 concentrates particularly in the cells of the tumor border and in the perivascular areas, a fact that may suggest its potential involvement in tumor cell migration.


In addition to the clear evidence for TGF-β signaling as a liver tumor suppressor, different studies have identified overexpression of TGF-β1 in HCC, which correlates with tumor progression and a bad prognosis.[9, 10] The ability of TGF-β to contribute to tumor progression depends on the capacity of the cells to overcome its growth inhibitory and proapoptotic effects. Different mechanisms could account for this resistance, among others: (1) alteration of oncogenic pathways, such as Ras/Erks or p53[19, 20]; (2) alterations in the TGF-β suppressor arm, such as dysregulation of embryonic liver fodrin (ELF, a crucial SMAD3/4 adaptor)[21] or up-regulation of SMAD7[22, 23]; or (3) interaction with hepatitis B virus X (HBx) protein.[24] Tumor cells that overcome TGF-β suppressor effects become susceptible to respond to these cytokine-inducing other effects, such as EMT processes that contribute to either fibrosis and/or tumor dissemination.[25] Furthermore, TGF-β may exert multiple effects on the microenvironment, as well as on vasculogenesis.[26] For all these reasons, the TGF-β signaling pathway is starting to be considered as a pharmaceutical target in HCC.[8] However, whereas interference with TGF-β signaling in various short-term animal models has provided promising results, liver disease progression in humans is a process of decades with different phases where targeting of TGF-β might have both beneficial and/or adverse effects.[27] Indeed, dissecting the downstream signals that govern the protumorigenic effects of the TGF-β pathway in liver tumor cells may help in the design of more specific targeted therapies for downstream TGF-β receptors and/or to select patients in whom a potential positive response to TGF-β inhibitors is predicted.

In this work, we show that some human HCC cells display a mesenchymal-like phenotype and migratory capacity under basal conditions, which is coincident with overactivation of the TGF-β pathway. An inverse correlation between the mesenchymal-like phenotype and the response to TGF-β as a tumor suppressor is observed. In liver cancer cells EMT, through Snail1 up-regulation, overcomes TGF-β-induced tumor-suppressor effects, switching its response to tumor progression, making cells resistant to cell death and prone to acquire invasive properties.[28] Furthermore, correlating with the autocrine stimulation of TGF-β, HCC cells express high levels of CXCR4, which is asymmetrically distributed and concentrated at the presumptive cell migratory front and mediates cell migration. Interestingly, both mesenchymal-like features and expression/polarization of CXCR4 are attenuated in cells where TGFBR1 expression is decreased with a specific shRNA, which correlates with the impairment of their migratory capacity. Although previous reports had reported the overexpression of TGF-β in HCC[9, 10] and the correlation of CXCR4 expression with invasive potential of HCC cells,[13, 15, 29, 30] this is the first study demonstrating that the tumor-promoting function of TGF-β signaling involves CXCR4/CXCL12, which results in enhanced migration in human liver tumor cells. Furthermore, activation of CXCR4 would affect several major signaling pathways related not only to cell migration, but also to proliferation and survival,[16] which may have relevant consequences in tumor progression.[31]

The results presented here also indicate that in the animal model of DEN-induced liver carcinogenesis, expression of TGF-β1 and CXCR4 is progressively increased, reaching maximum levels at late stages where tumors are macroscopically observed. We have also proven that in cultures of immortalized hepatocytes, TGF-β induces CXCR4 expression, a process that requires activation of both SMAD2 and SMAD3. In fact, an integrative genomic analysis of CXCR4 transcriptional regulation had previously suggested that TGF-β, Nodal, and Activin signals may induce CXCR4 upregulation based on SMAD2/3 and FOX family members.[32] The study in humans also indicates that a relevant percentage of liver tissues from HCC patients show a higher expression of CXCR4, which is always coincident with overactivation of the TGF-β pathway and correlates with a less differentiated phenotype and cirrhotic background. Cells that present a higher amount and polarized localization of CXCR4 are located in the borders of the tumor, in the migratory fronts, or in the perivascular zone, coincident with high expression of TGF-β in these areas. Interestingly, expression of CXCL12 is higher in the peritumoral cells, which suggest a paracrine regulation of the CXCR4 pathway. Indeed, overactivation of the TGF-β pathway sensitizes tumor cells to respond to CXCL12 produced by tumoral surrounding tissue. All these results together support the existence of crosstalk among TGF-β and CXCR4 pathways in HCC human tumors, which may contribute to tumor progression and dissemination.

The inhibition of the TGF-β pathway is emerging as a new therapeutic tool in cancer.[33] Since it regulates several steps in tumor progression, blocking this mediator should have multiple beneficial effects.[8] However, based on the results presented here, from both in vitro and in vivo experiments, the heterogeneity of the tumors might condition the response to these inhibitors. Indeed, overactivation of the TGF-β pathway differs among the different cell lines tested, as well as among the different tissues from patients. Interestingly, a strong correlation between TGF-β overactivation and mesenchymal-like and migratory phenotypes is observed, locating CXCR4 as a target of TGF-β both in cell lines and in HCC patients. From these results, CXCR4 localization in the migratory fronts of tumor tissues, coincident with high expression of TGF-β and/or high nuclear localization of p-SMAD2, may be used as biomarkers to predict the beneficial response to therapeutic agents that act on the TGF-β pathway. Increasing evidence demonstrates that activation of the CXCR4/CXCL12 pathway is a potential mechanism of tumor resistance to both conventional therapies and biological agents by way of complementary actions.[34] The use of TGF-β inhibitors, or inhibitors of the CXCR4/CXCL12 pathway, might increase the response to other therapeutic drugs when used in combination.

In conclusion, overactivation of the TGF-β pathway in HCC cells confers on them a mesenchymal-like phenotype and migratory properties through activation of the CXCR4/CXCL12 axis, a mechanism that would contribute to tumor progression in HCC patients. CXCR4 localization in the migratory fronts of tumor tissues, coincident with overactivation of the TGF-β signaling, may be considered in the future as a prognostic factor to predict patient response to drugs that target the TGF-β pathway.


The authors thank Greta Ripoll for technical support and participation in the analysis of the DEN model of hepatocarcinogenesis by Dr. Joana Visa (and the IDIBELL animal core facility) and graduate student Miguel Reina. We thank Drs. Perales and Giannelli for providing cells.