Potential conflict of interest: Nothing to report.
Hepatocellular carcinoma (HCC) growth severely affects prognosis. Ki-67, a known marker of cell proliferation, is a negative prognostic factor in HCC. Growth factors such as the epidermal growth factor (EGF) induce HCC cell proliferation but do not explain the great heterogeneity of HCC growth. Laminin-5 (Ln-5) is an extracellular matrix protein (ECM) present in the tissue microenvironment of HCC. The two main receptors for Ln-5, integrins α3β1 and α6β4, are expressed on the cell surface of HCC cells. The aim of this study is to investigate an alternative mechanism of HCC growth whereby Ln-5 promotes HCC cell proliferation through α3β1 and α6β4. HCC tissues containing Ln-5 display a larger diameter and higher number of positive cells for Ki-67, a well known proliferative index, as determined by double immunofluorescence staining and real-time PCR on microdissected tissues. In vitro, Ln-5, but not collagen I, collagen IV or fibronectin, induces proliferation as much as EGF does, via Erk phosphorylation as a consequence of β4 integrin phosphorylation. However, the two HCC cell lines do not proliferate in presence of Ln-5 despite β4 integrin and Erk1/2 activation. After transfection with α3 integrin, in the presence of Ln-5 one of these HCC cell lines acquires a proliferative activity whereas one of the proliferative HCC cell lines, knocked-down for α3 integrin, loses its proliferative activity. Conclusions: Our study suggests a new mechanism of HCC growth whereby Ln-5 stimulates proliferation via a different function of α6β4 and α3β1. (HEPATOLOGY 2007.)
Hepatocellular carcinoma (HCC) is a highly lethal cancer with an increasing frequency worldwide.1, 2 All the available therapies are mainly oriented to physically removing or destroying the tumor, whereas none so far block HCC progression, mainly because the molecular mechanisms underlying cancer cell proliferation are poorly understood. In fact, HCC growth is very different in one patient from another, as demonstrated by the different doubling time ranging from 1 to 20 months.3 Therefore, Ki-67, one of the best characterized proliferation markers, is considered a negative prognostic factor in HCC.4
Why some HCC grow faster than others is still unknown, but growing evidence suggests that various different growth factors, including the epidermal growth factor (EGF), stimulate HCC cell proliferation by phosphorylating different tyrosine-kinase (TK) receptors triggering signaling cascades, as recently reviewed.5, 6 However, HCC progression cannot be entirely explained by the action of growth factors or expression of the TK receptor.7
The tissue microenvironment is reported to affect the biological behavior of HCC because cancer cells grow completely embedded in a tissue enriched by extracellular matrix (ECM) proteins, growth factors, and proteolytic enzymes, deposited as a consequence of cirrhosis.8 Two different components of the tissue microenvironment, laminin-5 (Ln-5) and transforming growth factor-β1 (TGF-β1), are reported to induce the epithelial to mesenchymal transition of HCC cells though a multi-step cascade.9
The presence of Ln-5, an ECM protein belonging to the laminin family mainly expressed in the extra-cellular space at the basal membrane level,10 has been reported in HCC as well as other cancer tissues, significantly correlating with cancer metastasis and a worse prognosis.11–15
Ln-5 is the ligand of the α3β1 and α6β4 integrins, a family of transmembrane receptors formed by an α and β chain assembled with disulfide bonds.16 Integrins α3β1 and α6β4 mediate adhesion, scattering and migration Ln-5 dependent,17 while α3β1 expression has been shown at the cellular surface of HCC cells in tissues and in HCC cell lines displaying a motile and invasive activity.18
Aim of this study is to investigate an alternative mechanism of HCC growth, whereby Ln-5 stimulates HCC cell proliferation through a different function of α3β1 and α6β4.
The human HCC cell lines Hep3B and Alexander were obtained and cultured as previously described.18 Anti-Ln-5 antibodies were prepared as described.19 Anti-AKT, anti-Phospho AKT, anti-p44/42 MAP Kinase, anti-Phospho p44/42 MAP Kinase were purchased from Cell Signaling Technology (Danvers, MA). Anti-human Laminin-5 γ2 chain was kindly provided by S. Salo (Oulu, Finland). Anti-human Ki-67 was purchased from DAKO Cytomation (Glostrup, Denmark) and Abcam (Cambridge, UK). Rat Ln-5 was prepared as previously described.20 Collagen (Coll)-IV was purchased from Sigma (Saint Louis, MO), Coll I and Fibronectin (Fn) from Becton Dickinson (Bedford, MA). Anti-α3 integrin subunit (AB-10185) was purchased from Immunological Sciences (Rome, Italy), anti-β-Actin (clone AC-15) from Sigma, recombinant human epidermal growth factor (EGF) from PeproTech EC (London, UK), MAP kinase kinase inhibitor (PD98059) and PI3-kinase inhibitor (LY294002) from Calbiochem (Darmstadt, Germany). 3-[4,5-dimethylthiazole-2-yl]-2,5-diphenyltetrazoliumbromide (MTT), DAPI, Protease Inhibitor Cocktail and Crystal Violet from Sigma. Anti-β4 integrin antibodies were kindly provided by R. Falcioni (Rome, Italy) and V. Quaranta (Nashville, TN).
Cell Proliferation and Adhesion Assays
Cell proliferation was determined by MTT assay, crystal violet dissolution assay and manual counting in accordance with other studies.21 All experiments were repeated at least in quadruplicate, and each condition at least in triplicate. Briefly, 2 × 105 cells were incubated overnight; the next day, serum-free media containing Ln-5 (1μg/mL) or EGF (100ng/mL), and in some experiments LY294002 (10νM), PD98059 (50μM), MIG1, CM6, TR1 antibodies (100μg/mL) were used. Finally, cells were fixed, stained with crystal violet, solubilized and quantified in a spectrophotometer at 570nm. Alternatively, MTT was added to cell media. In manual counting, the cell number was determined by Trypan blue exclusion in a Burker chamber. Adhesion assay was performed as previously reported.18
Immunofluorescence was used to co-localize Ln-5 with Ki-67, and α3 or β4 with Ki-67 in tumoral and paired peritumoral tissues of the same HCC patients as previously described.9 Ten randomly chosen 400× microscopic fields of each section were captured, Ki-67 positive stained nuclei were counted and the mean value was calculated.
Immunoblotting and Immunoprecipitation
Cells (1.5 × 106) were allowed to adhere, serum-starved for 48 hours, incubated with Ln-5 and EGF for 15 minutes. Proteins were extracted in modified RIPA buffer supplemented with protease inhibitor cocktail. For immunoprecipitation, 2μL of anti-βL integrin subunit were added to 300μg of each protein sample and incubated for 2 hours. Immunocomplexes were then captured and processed following the manufacturer's instructions.
Laser Capture Microdissection and Real-Time RT-PCR Analysis
Frozen sections were microdissected and processed by quantitative Real-Time PCR analysis as described.9 The sequence of oligonucleotides is as follows: GAPDH, 5′-GGAGTCAACGGATTTGGT-3′ and 5′-GTGATGGGATTTCCATTGAT-3′; Ln5 α2 chain, 5′-AAAGCCACGTTGAGTCAGCC-3 and 5′-TCTTCCACCTGA- AAGGACTG-3′; Ki-67, 5′-GAACCTTCGCACTCTTCTGC-3′ and 5′-CAAGACTCGGTCCCTGAAAA-3′.
Plasmids, siRNA and Transfection
Human α3 integrin subunit full-length cDNA was kindly provided by Yoshikazu Takada (Davis, CA). α3 integrin cDNA was inserted as a XbaI/XbaI (Promega, WI) fragment into the NheI (Sigma) digested pIRES2-EGFP vector (BD clontech). For transfection, sub-confluent Hep3B cells were treated with 4μg of pIRES2-α3integrin-EGFP or pIRES2-EGFP plasmids and Lipofectamine 2000 (Invitrogen, Carlsbad, CA) according to the manufacturer's protocol. Stably transfected cells were selected with G418 (Sigma) and subcloned to select the highest expression level α3 integrin clones. Short-interfering RNAs (siRNAs) were selected according to Maurice Ho's Rational siRNA Design (available at: http://boz094.ust.hk/RNAi/siRNA).
Three α3 integrin cDNA target and one scrambled nonsense negative control RNA duplexes were purchased from Sigma-Proligo with the following sequences: 5′-CCAAGGAAACCUCUAUAUU-3′ (position 805 on α3 integrin mRNA), 5′-UCAUCAACAUCGUCCACAA-3′(position 1467), 5′-ACACUGAGGUCCAGUUCCA-3′ (position 1890) and 5′-UCACUACAAUCGCUCAACA-3′ (scrambled). 100 picomoles of each siRNA and 5μL of Lipofectamine2000 were used to transfect 5 × 105 HLF cells.
Quantitative data were summarized as median and range if not Gaussian distributed, otherwise data were normally distributed and mean and standard deviation are shown. Qualitative data were summarized as count and percentage. Independent comparison was performed with Student t test for normally distributed data or Wilcoxon sum rank test if not normally distributed. Correlations between quantitative variables were analyzed by mean Spearman coefficient.
Ln-5 Positive HCC Have Higher Ki-67 Expression and a Larger Tumor Diameter.
Ki-67 and Ln-5 were investigated in the tumoral and corresponding peritumoral tissue of 49 patients (demographic characteristics reported in Table 1). Ln-5 was considered positive when staining was present in the parenchymal tissue but not in the blood vessels, as previously described.18
Table 1. Clinical Characteristics of HCC Patients
No. of patients
Age (years, mean value)
69.7 ± 9.6
Multiple viral infection
Tumor size (cm, mean value)
5.1 ± 2.2
HCC histological grading
Ki-67 is expressed by proliferating cells at the nuclear level. A significantly (P < 0.0001) higher number of HCC Ki-67 positive cells was observed in tumoral as compared to the paired peritumoral tissue in all 49 patients. Ln-5 was absent in all peritumoral tissues and present in 26/49 tumoral HCC tissues. In these tissues, by double immunofluorescence staining, Ln-5 appeared expressed in the extracellular space and Ki-67 positive cells were mainly distributed near to or in some cases in contact with Ln-5, Fig. 1A. In serial sections, in Ln-5 positive tissues α3 and β4 were distributed around the HCC cellular surface, while in Ln-5 negative tissues both integrins were localized at the epithelial cells present around blood vessels, as previously described,18 Fig. 1B.
In the tumoral tissues, the number of positively stained Ki-67 HCC cells was significantly correlated (r = 0.4, P = 0.002) with the tumor diameter. Ln-5 positive HCC had a significantly larger diameter (P = 0.013) than Ln-5 negative HCC, and the number of Ki-67 positive cells was also significantly higher (P < 0.001) in Ln-5 positive than Ln-5 negative HCC. This difference was quantified by counting the number of Ki-67 positive cells in ten different randomly chosen microscopic fields of each sample. To ensure the reliability of the results, Ki-67 quantification was performed by three blinded independent operators; reproducible results were obtained for all sections. In Fig. 2A, the box plots show a different distribution of patients in the two groups (Ln-5 positive and negative); the Ln-5 negative group has a close interquartile range, almost all patients being positioned near the lower quartile. No correlation was observed among Ki-67, Ln-5 and the histological grade of differentiation.
To further confirm this difference at the molecular level, the same areas of serial sections previously studied by immunofluorescence were analyzed by real-time PCR after microdissection with the LCM technique. Consistently, Ki-67 mRNA levels were significantly increased (P < 0.005) in Ln-5 positive as compared with Ln-5 negative tissues, Fig. 2B.
In conclusion, HCC containing Ln-5 had a higher number of Ki-67 positive cells and a larger diameter.
Ln-5 Stimulates Proliferation of HLE and HLF, But Not Alexander and Hep3B HCC Cells.
To gain insight into the molecular mechanisms underlying the in vivo data, four different HCC cell lines, previously characterized,18 were challenged to proliferate, in serum-free conditions in the presence of Ln-5 and EGF, a known proliferative growth factor. Proliferation was assessed by means of three different methods, namely the MTT protocol, manual counting and crystal violet; consistent results were obtained for the different techniques as shown in Table 2.
Table 2. Comparison Between Different Proliferation Assays
Percentage of Proliferation
30.3 ± 3.2
30.6 ± 5.2
32.5 ± 6.7
29.4 ± 4.2
34.0 ± 5.9
30.5 ± 6.3
35.4 ± 4.5
35.7 ± 4.4
30.4 ± 5.2
30.5 ± 5.6
32.4 ± 5.1
28.0 ± 5.9
The optimal concentration of EGF (100 ng/mL) was determined by dose (from 0.1 up to 100 ng/mL) response experiments (data not shown). To determine the optimum timing for investigating cell proliferation, a kinetic scheme of proliferation ranging from 24 to 72 hours was performed. As reported in Fig. 3, in both HLE and HLF cell lines, Ln-5 and EGF significantly (P < 0.001) stimulated proliferation already after 24 hours, reaching the peak after 48, whereas Ln-5 failed to stimulate proliferation in either Alexander and Hep3B, while EGF still induced a significant effect (29% and 34%, respectively; P < 0.001).
Then the proliferative effect of other ECM proteins such as Collagen (Coll) I, Coll IV, Fibronectin (Fn), was investigated. As shown in Fig. 4, Ln-5, but not Coll IV, Coll I, and Fn, stimulated proliferation as much as EGF in HLF and in HLE cells compared to controls; crystal violet dissolution assay showed a statistically significant difference (P < 0.001).
To investigate the specificity of these results, HCC cell proliferation was performed in presence of functional antibodies against Ln-5.19, 22 As reported in Fig. 5, CM6 completely blocked (P < 0.001) the proliferative effect induced by Ln-5 while MIG1 and TR1 did not, in both HLE and HLF cells.
In conclusion, Ln-5 stimulated proliferation with a similar efficiency to EGF, and its effect was specific since it was blocked by CM6 antibody.
Ln-5 Phosphorylated Intracellular Pathways Are Responsible for HCC Cell Proliferation.
To investigate why Ln-5 induced proliferation on HLE and HLF but failed on Alexander and Hep3B cells, we explored the possibility that this different effect was due to different intracellular signaling pathways among the four cell lines, by incubating all the HCC cell lines in serum-free conditions using Ln-5 or EGF as control. Fig. 6A shows that Akt and Erk1/2 were very weakly phosphorylated at the baseline levels, but became strongly phosphorylated after exposure to Ln-5, as observed with EGF in HLF, and a similar effect was observed in HLE (data not shown). Surprisingly, a similar pattern was also observed in Hep3B and Alexander cells (data not shown) that, instead, did not proliferate in the presence of Ln-5. To make sure these patterns were truly responsible for Ln-5 stimulated proliferation, HLE and HLF were challenged to proliferate also in the presence of LY294002 and PD98059, inhibitors of PI3K/Akt and MEK/Erk, respectively. As shown in Fig. 6B, Ln-5 stimulated proliferation was completely blocked by PD98059 (P < 0.001) but not by LY294002 in both HLE and HLF cells. A similar effect was observed with EGF induced proliferation. As control, neither PD98059 and LY294002 showed any effect on unstimulated proliferation. Finally, PD98059 and LY294002 blocked Ln-5-induced phosphorylation of Akt and Erk1/2, as shown by western blotting in HLF (Fig. 6C) and HLE (data not shown).
In conclusion, Ln-5 activated Erk1/2 and consistently promoted HLE and HLF proliferation, as did EGF. However, the lack of proliferative activity was not caused by altered proliferative molecular pathways.
α3β1 Is Essential for Ln-5 Induced Proliferation.
To investigate the reason why Alexander and Hep3B cells did not proliferate in presence of Ln-5 despite activation of the Erk1/2 pathway, but did proliferate in presence of EGF activation of the same pathway, we tested the hypothesis that the two main receptors of Ln-5, α6β4 and α3β1 integrins, play an essential role in Ln-5 induced proliferation.
HLE, HLF, Alexander and Hep3B cells were serum starved, incubated with Ln-5 or EGF and immunoprecipitated with the anti-β4 monoclonal antibody 450 11-A followed by immunoblotting with anti-p-Tyr antibody. Both Ln-5 and EGF phosphorylated the β4 subunit in the cells that did and did not proliferate in the presence of Ln-5. These findings suggest that β4 is involved in the proliferation pathway because it is activated by Ln-5, but also that this subunit does not distinguish the different ability to proliferate in response to Ln-5.
Because we have previously shown that both Hep3B and Alexander lack α3 subunit expression,18 Hep3B cells were transfected to encode the full length of α3, as shown by western blot in Fig. 7A. To prove that such a transfection was functional, wild type Hep3B, EGFP and α3 transfected cells adhesion on Ln-5 was tested in a rapid adhesion assay. In this experiment HLF, that we have previously shown to strongly adhere onto Ln-5,18 was used as positive internal control. As shown in Fig. 7B, Hep3B/α3 displayed a strong adhesion (P < 0.001) on Ln-5 but wild type or vector transfected cells did not. At this point wild type, vector and α3 transfected Hep3B cells were challenged to proliferate in the presence of Ln5 or EGF. Ln-5 did not stimulate proliferation of the wild type or vector transfected cells, whereas the α3 transfected cells strongly (42%; P < 0.001) proliferated in the presence of Ln-5, as measured by both crystal violet and manual counting assays, Fig. 7C–D. To confirm the role of the α3 subunit, HLF proliferative cells were knocked-down for α3, as shown by western blotting in Fig. 8A. In functional assays, the three different α3 integrin siRNA transfected cells weakly adhere (P < 0.001) onto Ln-5 in a rapid adhesion assay while scrambled siRNA cells adhere as well as the wild type (Fig. 8B). Consistently Ln-5 failed to promote proliferation in the different α3 integrin siRNA transfected cells, whereas proliferation (29%; P < 0.001) was still stimulated on scrambled siRNA, that proliferate at similar levels to wild-type cells. In conclusion, we propose a model whereby HCC cells can proliferate in presence of either EGF or Ln-5. In this case, integrin α6β4 is involved because the intracytoplasmic tail of the β4 subunit is phosphorylated by Ln-5, that is engaged with integrin α3β1 to yield the proliferative effect.
This study may suggest a new, different mechanism whereby an ECM molecule, Ln-5, present in the tissue microenvironment, stimulates HCC growth through a different function of α6β4 and α3β1 integrin receptors. What physiological relevance have our data since HCC cells grow embedded in a tissue enriched with growth factors shown to induce HCC proliferation? The activation of different TK receptors mediated by growth factors is not sufficient to explain the high heterogeneity of tumor growth observed among HCC patients with comparable stages of disease and demographic characteristics, suggesting that other molecular pathways are implicated.6, 23 Therefore, the knowledge of an alternative HCC proliferation mechanism could help to make a better patient selection of candidates for biological therapies using TK receptors as molecular targets, as recently suggested.24, 25
As reported in the literature, we observed a highly variable number of Ki-67 positive cells among the different HCC tissues, correlating with tumor size.4 Such variability could be explained by the presence of Ln-5 in those tissues with a higher number of Ki-67 positive cells. The peculiar distribution of HCC proliferative cells near to Ln-5, together with the statistical association between Ki-67 and Ln-5, suggest that this molecule is implicated in HCC growth. This hypothesis was confirmed by “in vitro” data whereby Ln-5 stimulates HCC cell proliferation by means of three different methods. Although in our experimental condition this stimulation seems to be moderate, it is noteworthy that EGF, used at concentrations inducing the strongest proliferative effect in our experimental conditions, displays a similar effect. Ln-5 is widely reported to be involved in the metastatic activity of different malignancies including HCC, where we recently reported that it triggers the epithelial-mesenchymal transition.9 On the contrary, a proliferative activity of Ln-5 has recently been proposed in a developmental disorder and confirmed in HCC.26 These results explain why Iressa, an EGFR inhibitor currently used in clinical trials, lacks an antiproliferative effect on the same HCC cell lines used here, in presence of Ln-5.27 The possibility that Ln-5 phosphorylates EGFR thanks to the DIII domain generated by proteolytic cleavage by matrix metalloproteinases has been proposed, but does not seem to be the case in our experimental conditions, since EGF phosphorylates EGFR but Ln-5 does not.28
In our experimental model, the proliferative effect is mediated by Erk, as also reported in the case of polycystic kidney cells.26 However, in their case Erk was phosphorylated as a consequence of β4 integrin subunit activation, as previously suggested.26, 29 The role of β4 in cell proliferation is also demonstrated by the fact that β4 mutated keratinocytes do not proliferate in the presence of Ln-5, and Erk remains confined to the cytoplasm rather than being translocated into the nucleus.30 Consistent with these data, also in our model Ln-5 phosphorylates the β4 integrin subunit, but we took advantage of the fact that all the cell lines used expressed high levels of α6β4 while the non proliferative cells lacked the α3 integrin subunit, as previously reported.18 Therefore, one of the non proliferative HCC cell lines, Hep3B, transfected with the α3 integrin subunit, showed a proliferative activity in the presence of Ln-5 whereas the vector control did not. Furthermore, one of the proliferative HCC cell lines, HLF, knocked-down for the α3 integrin subunit, no longer proliferated in the presence of Ln-5 whereas the scrambled transfected lines still did so.
In conclusion, we propose a different mechanism of HCC growth as compared to that of growth factors such as EGF. In this scenario, Ln-5 phosphorylates α6β4 integrin, activating the Erk cascade, but cells proliferate only in the presence of α3β1, either because the Ln-5 G domain ligands for the two receptors are GIII for α3β1 and in between GIV-GV for α6ββγρ;4, or because of the existence of a sort of “inside out” pathway, as shown for other integrin receptors.31, 32 Our results further emphasize the role of tissue microenvironment factors in modulating the biological and hence clinical outcome of patients with HCC, helping to understand the heterogeneity of HCC growth and to better tailor potential future biological therapies.
The Authors are grateful to Mary V. Pragnell, B.A. for language revision.