Integrin signal masks growth-promotion activity of HB-EGF in monolayer cell cultures

Establishment of cell culture methods of mammalian cells stimulated revolutionary progress in biology and medicine, and has now become an essential technology in cell biology research. Mammalian cell culture is most commonly achieved by incubating cells with a defined nutrient medium supplemented with serum in tissue culture plates. Cells attach and adhere onto the flat surface of glass or plastic plates and grow two dimensionally, forming monolayer sheets. Although the standard cell culture system has provided us with fundamental knowledge of cell and gene functions, this system is not an accurate representation of the in vivo environment in which the cells originally exist (Lee et al., 2007; Schmeichel and Bissell, 2003; Yamada and Cukierman, 2007). Tissues and organs are three-dimensional (3D). Cells grown on flat tissue culture plates (TCPs) using standard cell culture methods differ from those growing in their natural environments in terms of their morphology, physicochemical properties of the substrate they attach to, and cell-cell and cell-matrix interactions. Thus, the behavior, growth and differentiation, as well as their response to internal and external signals, of cells grown in the standard cell culture environments are largely different to that of cells grown in vivo.

The EGFR-ligand system supports the proliferation, motility, differentiation and survival of various cell types, thereby contributing to the development, morphogenesis and maintenance of homeostasis in the body. The EGF-EGFR system also has a pivotal role in the progression and development of malignant tumor growth (Hynes and Lane, 2005; Normanno et al., 2006; Lynch et al., 2004; Paszek et al., 2005). Furthermore, the overexpression of EGFR ligands induces or enhances cell growth in nude mice (Miyamoto et al., 2004; Ongusaha et al., 2004; Wang et al., 2006; Normanno et al., 2006). However, in contrast to the marked effects of EGFR ligands on tumor cell growth in vivo, the growth-promoting properties of the EGFR ligands have not been well documented in most cell types using standard cell culture methods. Early studies indicated that EGF increased the lifetime, but not the growth rate, of primary keratinocytes in culture (Rheinwald and Green, 1977). Previous studies using normal keratinocytes, fibroblasts, breast cancer cells and neuroblastoma cells reported that EGFR ligands stimulated cell growth with a minimal increase in cell numbers (Hashimoto et al., 1994; Ho et al., 2005; Lembach, 1976; Osborne et al., 1980). Moreover, in some cases, EGFR ligands induced cell-cycle arrest or apoptosis (Cao et al., 2000; Fan et al., 1995). Even under serum-depleted conditions, the addition of growth factors to the cell culture medium resulted in only a slight growth promotion, if any. The lack of an appropriate in vitro culture system to demonstrate growth promotion by EGFR ligands posed a limitation in understanding the signaling mechanism and molecules responsible for the proliferating effect of these growth factors. The exceptions are the myeloid lineage cell lines 32D and Ba/F3. These cell lines, which normally grow in suspension in an interleukin-3-dependent manner, can proliferate specifically in response to EGFR ligands by expressing ectopic EGFR in the absence of interleukin-3, thus facilitating observation of its growth-promotion activity (Higashiyama et al., 1995; Iwamoto et al., 1999; Pierce et al., 1988; Yu et al., 2002). These findings suggest that cellular responses to EGFR ligands vary according to the cell culture conditions and cell systems used for the study. However, the reason for this low potency of EGFR ligands in most cell types has not been investigated in much detail.

HB-EGF is an essential member of the EGFR ligands in vivo, which is synthesized as a transmembrane precursor protein (proHB-EGF) (Higashiyama et al., 1991). Its extracellular domain is then cleaved by proteases, via a so-called ectodomain-shedding mechanism, which yields a soluble mature growth factor (sHB-EGF), which is similar to other EGFR ligands (Goishi et al., 1995; Massagué and Pandiella, 1993). In the course of the present study, we found that HB-EGF did not promote the growth of ovarian cancer cell lines under standard monolayer cell culture conditions, although it enhanced the cell growth rates when the same cell lines were injected into nude mice. Thus, we focused on the growth-stimulatory effect of HB-EGF under various culture conditions and found that this effect of HB-EGF was not well documented in the standard monolayer culture. However, it was particularly observed in culture systems in which the integrin signal was attenuated. We also show that such culture conditions enable the observation of cell proliferation by EGF, TGFα or ligands of other receptor tyrosine kinases (RTKs). The results of the present study indicate why the growth-promotion activity of growth factors has thus far not been well documented in vitro.

We previously reported that the tumorigenicity of the ovarian cancer cell lines SKOV3 and RMG-1 when injected into nude mice were strongly enhanced by exogenous expression of HB-EGF, whereas small-hairpin RNA-mediated knockdown of endogenous HB-EGF or administration of the HB-EGF-specific inhibitor CRM197 suppressed their tumorigenicities (Miyamoto et al., 2004). We confirmed the contribution of HB-EGF to SKOV3 cell growth in vivo by overexpression or knockdown experiments (Fig. 1A). Although the tumorigenic growth of xenografted cells might be affected by various host factors, the marked growth-promotion effect of HB-EGF in vivo led to the proposal that HB-EGF also contributes to SKOV3 cell growth in vitro. To test this hypothesis, we compared the growth rates of SKOV3 cells, HB-EGF-overexpressing SKOV3 (SKOV-HB) cells and HB-EGF-knockdown SKOV3 cells under monolayer culture conditions using standard polystyrene tissue culture plates (TCPs). Contrary to our expectation, no obvious differences in the growth rates between the parental and HB-EGF-overexpressing SKOV3 cells in TCPs were observed (Fig. 1A). Moreover, SKOV3 cells expressing small-hairpin RNAs specific for HB-EGF, which had lost their tumorigenicity, were able to grow almost as quickly as the parental cells in TCPs (Fig. 1A).

To further study the effects of HB-EGF on cell growth, we used BRL cells, which are originally non-tumorigenic and become tumorigenic upon HB-EGF overexpression (Wang et al., 2006). We established BRL cells overexpressing wild-type HB-EGF (BRL-HB) or an ectodomain-shedding deficient mutant HB-EGF (BRL-HBuc) (Miyamoto et al., 2004; Yamazaki et al., 2003), and compared their growth rates with that of mock virus-infected BRL cells (BRL-mock). Although BRL-HB and BRL-HBuc cells expressed comparable amounts of the HB-EGF precursor (proHB-EGF), the secretion of sHB-EGF by BRL-HBuc cells was significantly reduced compared with that of BRL-HB cells (supplementary material Fig. S1). When BRL-HB cells were subcutaneously injected into nude mice they formed tumors, as reported previously (Wang et al., 2006). However, neither BRL-HBuc cells nor BRL-mock cells formed tumors until 3 weeks after injection (Fig. 1B). Similarly to SKOV3 cells, these cells grew at comparable rates in TCPs (Fig. 1B). These results indicated that HB-EGF promotes the growth of non-tumorigenic cells in vivo in an ectodomain-shedding-dependent manner, although HB-EGF did not potentiate cell growth in cells grown using in vitro culture conditions.

All the cultures described above were performed in medium containing 10% FCS. We tested the effect of serum on the growth-stimulating effect of HB-EGF in TCPs. Even in medium containing 1% FCS, the cells grew as quickly as those cultured in medium containing 10% serum; HB-EGF did not potentiate their growth rate (Fig. 1C). To rule out any effects of serum, cells were cultured using a serum-free defined medium as described in the Materials and Methods. Overall, the growth rate of BRL and SKOV3 cells, regardless of HB-EGF expression, was largely reduced under serum-free conditions. Even in the serum-free medium, HB-EGF did not enhance BRL (Fig. 1D) and SKOV3 cell growth (supplementary material Fig. S2).

To test whether the inability of HB-EGF to enhance cell growth in TCPs was a common feature, we tested the effect of HB-EGF on cell growth in three other cell lines, A431, RMG-1 and U373 MG cells. Similarly to BRL and SKOV3 cells, the exogenous expression of HB-EGF did not increase the growth rates of A431, RMG-1 and U373 MG cells in TCPs, either in medium containing 10% FCS or in serum-free defined medium (supplementary material Fig. S3). These results suggest that the inability of HB-EGF to promote growth in TCPs is not attributable to the existence of FCS in the culture medium.

Tumorigenic cells can often grow in semisolid media, such as soft agar (Hanahan and Weinberg, 2000). HB-EGF induces oncogenic transformation in several cell lines and confers the ability to grow in soft agar (Fu et al., 1999; Harding et al., 1999; Ongusaha et al., 2004). Consistently with previous studies, HB-EGF strongly potentiated BRL cell growth in the presence of 10% FCS in soft agar (hereafter cell culture in soft agar is referred to as 3D-agar). BRL-HB cells grew to form spherical colonies in 3D-agar, but BRL-mock cells did not (Fig. 2A). One characteristic difference between cell culture conditions in 3D-agar and 2D plates is that cells embedded in gels grow three-dimensionally (in a multilayer). Therefore, we examined whether BRL cells grew in an HB-EGF-dependent manner under other 3D culture conditions. 3D cultures consisted of 3D matrices, such as collagen gels (hereafter referred to as 3D-COL) or a laminin-rich matrix (hereafter referred to as 3D-LM) (Paszek et al., 2005; Weaver et al., 1997). Although the morphology of BRL-HB cells growing in 3D-COL and 3D-LM was different to that in 3D-agar, HB-EGF markedly enhanced cell proliferation in these 3D cultures (Fig. 2A). In 3D-COL, BRL-mock cells proliferated to twice their initial number, whereas BRL-HB cells proliferated rapidly and increased their numbers around seven times faster than BRL-mock cells after 1 week of culture. BRL-HB cells proliferated even more rapidly in 3D-LM than in 3D-COL and multiplied around seven times faster than BRL-mock cells after 4 days of culture (Fig. 2A). Growth enhancement by HB-EGF in 3D culture conditions was observed for SKOV3, A431, RMG-1 and U373 MG cells (supplementary material Figs S2, S3).

In the above experiments, proHB-EGF was overexpressed in BRL cells and other cell lines to examine the effect of HB-EGF on cell growth. We also performed similar experiments by adding recombinant sHB-EGF to the medium as an alternative to proHB-EGF overexpression. sHB-EGF promoted the growth of BRL-mock cells in a dose-dependent manner in 3D-COL, but not in TCPs (supplementary material Fig. S4A). HB-EGF-dependent growth was also tested using MCF-10A, another non-transformed cell line. The addition of sHB-EGF induced a ∼3.5-fold increase in growth of MCF-10A cells in 3D-LM, but weakly induced (∼1.3-fold) growth of the cells in TCPs (supplementary material Fig. S4B). These results indicate that growth promotion by HB-EGF in 3D cultures is a general phenomenon observed for transformed and non-transformed cells, and that HB-EGF-dependent growth in 3D cultures is reflected by both the overexpression of proHB-EGF and by addition of sHB-EGF in the medium.

CRM197, a specific inhibitor of HB-EGF, suppresses HB-EGF-dependent growth of ovarian cancer cell lines in vivo (Miyamoto et al., 2004). CRM197 did not affect the growth of BRL-mock and BRL-HB cells in TCPs (supplementary material Fig. S5). In 3D-COL, however, CRM197 suppressed the growth of BRL-HB, but not BRL-mock cells, further confirming HB-EGF-dependent growth in 3D-COL (supplementary material Fig. S5).

As shown above, the growth enhancement by HB-EGF was clearly observed in 3D culture systems in which cells were embedded either into a soft agar, collagen or laminin-rich matrix. To determine whether the embedding of cells into the gel was critically required for cell growth enhancement by HB-EGF, we tested other cell culture conditions: (1) cells were cultured atop a soft agar surface (2D-agar); (2) cells were cultured on NanoCulture® plates (NCPs). We compared the growth rate of BRL-HB cells with those of BRL-mock cells grown under these culture conditions. When BRL cells were cultured in 2D-agar, cell adhesion and spreading were strongly suppressed. Under these conditions, BRL-mock cells scarcely grew, but BRL-HB cells proliferated and formed spheroidal cell masses (Fig. 2B). NCPs with a nanometer scale honeycomb-patterned structure on the surface are shown to reduce cell adhesion to the matrix (http://www.scivax.com/cell/english/index.html). In NCPs, cell adhesion, spreading and monolayer cell sheet formation were hampered in the case of BRL cells, and these cells showed HB-EGF-dependent growth (Fig. 2B).

Cells embedded in gels might differ from cells growing in TCPs largely because of their surrounding microenvironments. In 3D, for example, gas exchange, nutrient supply and diffusion of growth factors might be hampered, compared with TCPs. However, since growth promotion by HB-EGF was observed to be as high as that of 3D culture conditions using 2D-agar or NCP, the difference of these factors might not be critical in determining the requirement of HB-EGF for cell growth. It was interesting that growth enhancement by HB-EGF was observed in reduced cell-matrix adhesion conditions, as seen in 2D-agar and NCPs, whereas HB-EGF-dependent cell growth was not observed in TCPs, which allow cells to attach, spread and form monolayers.

To characterize the conditions that facilitate growth promotion by HB-EGF, we examined the growth rates of BRL cells under various culture conditions. For this experiment, methylcellulose gel was used instead of soft agar for the 3D-agar condition for the ease of cell counting (therefore referred to as 3D-MC). Since cell growth rate was different in each culture condition, cell number was counted at distinct culture periods. The growth curve of BRL-mock and BRL-HB cells in TCPs and 3D-COL are shown in Fig. 3A. Cell numbers of BRL-mock and BRL-HB cells cultured in NCP, 3D-LM, 2D-agar and 3D-MC for the days indicated are also shown in Fig. 3B. These results indicate that the growth rate of cells in TCPs is extremely high compared with that of other culture conditions in vitro. Based on Fig. 3A and 3B, we also examined the relationship of the `growth rate' of BRL cells without HB-EGF and the `upregulation ratio' by HB-EGF. The growth rate was estimated by comparison of the doubling time of BRL-mock cells in various culture conditions without HB-EGF, and the upregulation ratio was estimated by comparison of growth rate between BRL-mock and BRL-HB cells, as described in the Materials and Methods. Fig. 3C shows the relationship of the growth rate and upregulation ratio in each culture condition. The cells in culture conditions with lower growth rates showed a higher upregulation ratio, although this was not the case for 3D-MC. Thus, the enhancement of cell growth by HB-EGF was particularly observed in culture conditions in which overall cell growth was reduced.

We examined which EGFR downstream signaling molecules were responsible for cell growth in TCPs and 3D-COL upon HB-EGF stimulation. BRL-HB cell growth in 3D-COL was suppressed by inhibitors of EGFR (ZD1839), MEK (PD98059) and phosphatidylinositol-3 kinase (PI3K; LY294002), but not ErbB2 (AG825) or p38MAPK (SB203580; Fig. 4A). This suggested that EGFR and its downstream MEK-Erk and PI3K-Akt signals were required for BRL-HB cell growth in 3D-COL. The PI3K inhibitor LY294002 inhibited BRL-HB cells in TCPs as well as in 3D-COL. Kinetic studies with increasing concentrations of inhibitors indicated that BRL-HB cell growth was largely suppressed by the MEK inhibitor PD98059 in 3D-COL and partially suppressed in TCPs within a higher concentration range (Fig. 4B). These findings suggest that MEK and PI3K are required for cell growth in TCPs as well as 3D-COL, but cells grown in 3D-COL were more susceptible to the inhibitors than cells grown in TCPs.

HB-EGF-dependent cell growth in 3D-COL was suppressed by inhibitors of MEK or PI3K, suggesting that activation of the Raf-MEK-Erk and PI3K-Akt pathways is crucial for promoting cell growth. To test whether activation of these pathways could substitute for the activity of HB-EGF, wild-type Raf-1, MEK1 and Akt1 were overexpressed in BRL-mock cells. Raf-1 alone, MEK1 alone or Raf-1 plus MEK1 did not significantly promote cell growth, whereas Akt1 alone partially promoted cell growth (data not shown). However, coexpression of Raf-1, MEK1 and Akt1, induced BRL cells to change their morphology to that resembling BRL-HB cells and promoted cell growth at comparable levels to those induced by HB-EGF (Fig. 4C), suggested that the coordinated activation of the Raf-MEK-Erk and PI3K-Akt pathways was crucial for potentiating cell growth in 3D-COL. By contrast, cell growth was not significantly affected by coexpression of Raf-1, MEK1 and Akt1 in TCPs (Fig. 4C). These results indicate that HB-EGF enhanced cell growth in 3D culture by activating both the Raf-MEK-Erk and PI3K-Akt pathways.

Next, we examined the activation states of EGFR, Erk and Akt in cells grown in TCPs and 3D-COL before and after transient HB-EGF stimulation. We first compared the steady-state levels of EGFR, Erk and Akt proteins of BRL cells cultured for 16 hours without sHB-EGF. The levels of EGFR and Akt proteins were markedly reduced in cells grown in 3D-COL compared with those grown in TCPs (Fig. 5A, see lanes Time 0). The Erk2, but not the Erk1 levels, were also slightly reduced in 3D-COL. Consistent with the reduction in the level of EGFR, the phosphorylation of EGFR (Y845 and Y992), Akt and Erk before HB-EGF stimulation was reduced in 3D-COL. These results indicate that the activation states of EGFR and downstream signaling molecules in the absence of HB-EGF in 3D-COL were lower compared with those in TCPs. Upon stimulation by addition of recombinant sHB-EGF, the phosphorylation levels of EGFR (Y845, Y992 and Y1068), Akt and Erk in TCPs were significantly enhanced. Although the phosphorylation levels of EGFR, Akt and Erk in 3D-COL were also enhanced by HB-EGF, they were much lower than those in TCPs (Fig. 5A).

The reduction of a cell response to HB-EGF in 3D-COL upon transient stimulation by sHB-EGF might be attributable to the reduced diffusion rate of sHB-EGF in the collagen gel. To test whether attenuation of EGFR signaling was characteristic of HB-EGF-dependent cell growth or due to the reduced diffusion, the activation states of Erk and Akt in cells grown in NCPs or 2D-agar were studied. In this case sHB-EGF would diffuse freely as it does in TCPs. Although Akt (Fig. 5B) or EGFR (data not shown) was not reduced in NCPs or 2D-agar, steady states and sHB-EGF-induced phosphorylation levels of Erk and Akt in NCPs or 2D-agar were lower than those in TCPs, indicating that attenuation of EGFR signaling was not explained by the reduced diffusion rate of sHB-EGF (Fig. 5B).

Activation states of Erk and Akt were also examined under sustained cell culture conditions. For this assay, BRL-mock cells, BRL-HB cells and BRL cells expressing Raf-1, MEK1 and Akt1 were grown in TCPs or 3D-COL for 4 days in the presence of 10% FCS, and the phosphorylation levels of Erk and Akt were examined. The phosphorylation level of Erk in TCPs was much higher than that in 3D-COL regardless of the expression of HB-EGF or Raf-1, MEK1 and Akt1 (Fig. 5C). HB-EGF, or coexpression of Raf-1, MEK1 and Akt1, increased the phosphorylation level of Erk in 3D-COL, but the level was still lower than that of cells cultured in TCPs. This suggests a correlation between the phosphorylation level of Erk and the growth rate of BRL cells. The enhanced phosphorylation of Akt was also observed by expression of HB-EGF or coexpression of Raf-1, MEK1 and Akt1 in 3D-COL. This was consistent with results obtained using a transient assay (Fig. 5A), where expression of Erk and Akt in cells growing in 3D-COL was attenuated when compared with that in cells growing in TCPs. We observed that the phosphorylation level of Akt was reduced by expression of HB-EGF in TCPs, although the significance was not clarified.

We examined whether activation of the Raf-MEK-Erk and PI3K-Akt pathways is linked with cell cycle progression by evaluating cyclin D1 expression. BRL-mock cells showed reduced levels of cyclin D1 in 3D-COL compared with that in TCPs, suggesting that the cell cycle of BRL-mock cells in 3D-COL was arrested at G1 phase (Fig. 5C). Consistent with growth-promotion activity, HB-EGF or Raf, MEK and Akt increased cyclin D1 in 3D-COL to levels comparable with those in TCPs.

Finally, we examined EGFR downstream signaling in vivo. To determine the relationship of the activation state of EGFR signaling with HB-EGF dependency in cell growth, the phosphorylation levels of Erk and Akt of BRL-HB cells growing in tumors in mice were compared with those of BRL-HB cells growing in TCPs. The phosphorylation levels of Erk and Akt in tumors were lower than those of cells in TCPs (Fig. 5D), although the phosphorylation levels varied. Taken together, we conclude that the growth enhancement by HB-EGF was observed in culture conditions with lower levels of EGFR signaling compared with that seen in TCPs.

Cell adhesion to the ECM substrate is mediated predominantly by integrins. Growth factor receptors and integrins activate some common signaling pathways (Schwartz and Ginsberg, 2002), and both signals cooperate functionally in a variety of biological processes. A previous study indicated that the expression level of EGFR is positively linked to that of integrin β1 and that both were crossregulated in 3D-LM (Wang et al., 1998). BRL cells mainly express β1 and β3 integrin subunits. Thus, we examined the levels of integrin β1 and integrin β3 proteins in BRL-mock cells grown in TCPs and 3D-COL. The integrin β1 levels in BRL-mock cells were significantly reduced when they were grown in 3D-COL compared with TCPs (Fig. 6A), whereas the levels of integrin β3 remained constant (data not shown). When HB-EGF was expressed in BRL cells, the level of integrin β1, particularly its mature form, was increased in cells grown in TCPs or 3D-COL. The combination of Raf-1 with MEK1 and Akt1 also consistently increased the integrin β1 levels under these culture conditions. We examined the level of integrin β1 of the BRL-HB cells growing in vivo and found that it was reduced in vivo compared with cells growing in TCPs (Fig. 5D). These results suggest that the integrin levels are linked with growth factor signaling in cells growing in 3D-COL or in vivo, as well as in 3D-LM.

To determine whether a reduced integrin β1 level in reduced cell adhesion conditions causes the diminished growth of BRL cells and their HB-EGF dependency, we established BRL cells expressing exogenous integrin β1 (BRL-β1) and examined cell growth in various culture conditions. Exogenous expression of integrin β1 resulted in an increased EGFR level (Fig. 6B). BRL-β1 cells grew faster than BRL-mock cells in 3D-COL or 2D-agar, but not in TCPs (Fig. 6C). However, cell growth rate induced by integrin β1 expression in 3D-COL or 2D-agar was slower than that induced by HB-EGF expression. Furthermore, when BRL-mock cells and BRL-β1 cells were stimulated with various concentrations of recombinant sHB-EGF in 3D-COL, BRL-β1 cells grew faster than BRL-mock cells in the same concentrations of sHB-EGF (Fig. 6D). However, the high concentration of sHB-EGF in BRL-mock cells far overcame the growth advantage of the BRL-β1 cells. These results suggest that a reduced integrin β1 level could partly cause the diminished growth of BRL cells and their HB-EGF dependency.

Integrins form stable focal contacts when cells are attached and spread to hard substrates, such as TCPs, with a corresponding generation of strong signals. By contrast, when cells are cultured on soft substrates, the formation of focal contacts is hampered, and integrin signals are reduced (Cukierman et al., 2001; Paszek et al., 2005). Although it is difficult to accurately estimate the strength of integrin signals directly, the morphology of cells cultured under various conditions (Fig. 2) and the phosphorylation levels of Erk and Akt in the absence of HB-EGF (Fig. 5) represent the strength of integrin signals within a particular cell culture system. Therefore, we speculated that, in combination with the expression level of integrin β1, integrin signals are strengthened in TCPs, whereas they are considerably decreased in 3D or 2D adhesion-reduced culture conditions.

We hypothesized that the integrin signal might predominantly contribute to promote cell growth in TCPs, thereby allowing the cells to grow without HB-EGF. However, integrin signals were too low to grow in cells cultured in 3D and 2D adhesion-reduced conditions, resulting in enhanced cell proliferation by HB-EGF. To test this hypothesis, we investigated the effect of integrin inhibition on HB-EGF dependency for cell growth in TCPs. First, BRL cells were cultured in the presence or absence of integrin antibodies. Although a single blockade of integrin β1 or integrin β3 scarcely inhibited the growth of BRL cells (data not shown), simultaneous blockade of integrins β1 and β3 suppressed the growth of BRL cells in TCPs (Fig. 7A). When sHB-EGF was added to the culture with the integrin antibodies, sHB-EGF almost recovered the growth rate to the level of BRL cells without the integrin antibodies, indicating that BRL cells exhibit HB-EGF-dependent growth in the presence of integrin antibodies. Growth stimulation by HB-EGF in TCPs was also shown upon comparison of the growth rate of BRL-mock and BRL-HB cells in the presence of integrin antibodies (Fig. 7B).

To further examine the contribution of integrin to cell growth in TCPs, we examined the HB-EGF dependency of cells obtained from an integrin-β1-knockout mouse. Fig. 7C shows that growth of cells lacking integrin β1 was HB-EGF-dependent in TCPs. Growth stimulation of integrin-β1-knockout cells was observed by addition of sHB-EGF to the medium. Reintroduction of integrin β1 increased the growth of integrin-β1-knockout cells in TCPs and induced the cells to become HB-EGF independent. These results indicate that the integrin system largely contributes to the growth of cells cultured in TCPs. FAK is a critical downstream effector of integrin signaling (Mitra and Schlaepfer, 2006). To further strengthen our hypothesis, we examined whether fibroblasts obtained from FAK-knockout mice (MEF^FAK–/–) showed HB-EGF dependency for growth in TCPs. Similarly to cases of integrin blockade with integrin antibodies (Fig. 7A,B) and integrin-β1-knockout cells (Fig. 7C), MEF^FAK–/– cells exhibited HB-EGF-dependent growth in TCPs upon addition of sHB-EGF to the medium (Fig. 7D). Reintroduction of FAK increased the growth of FAK-knockout cells in TCPs, and the cells became less HB-EGF independent. Taken together, we conclude that integrin and HB-EGF share common signals for promoting cell growth, and that in TCPs integrins generate sufficient signals for growth, thereby contributing to the dispensability of HB-EGF for growth purposes. In 3D or 2D adhesion-reduced conditions the integrin signals are not sufficient to stimulate rapid cell growth, and therefore there is a cumulative contribution of integrins and growth factors toward cell growth.

We demonstrated here that growth stimulation by HB-EGF was facilitated in 3D culture systems or 2D cell adhesion-reduced conditions. To determine whether this feature was specific for HB-EGF or was a general feature of other growth factors, we tested the growth-stimulatory effects of various growth factors in TCPs and 3D-COL. None of the growth factors tested greatly enhanced cell growth of BRL-mock cells in TCPs. EGF and TGFα of the EGF family enhanced cell growth to the same extent as HB-EGF did in 3D-COL (Fig. 8). In addition, the growth of BRL cells in 3D-COL was enhanced by FGF-2 and IGF-1 and slightly enhanced by PDGF-BB (Fig. 8). These results indicate that culture conditions with reduced integrin signals also facilitated cell growth promotion by other members of the EGF family of growth factors or ligands of other RTKs.

In this study, SKOV3 and BRL cells injected into nude mice exhibited HB-EGF-dependent growth; however, when these cells were cultured in TCPs, they displayed no HB-EGF dependency. Based on this finding, we investigated the effects of HB-EGF on cell growth under different culture conditions. Our results indicated the following: first, the HB-EGF dependency of cell growth was emphasized in culture systems in which cells grew in reduced cell-substrate adhesion, and this phenomenon was observed for transformed and non-transformed cells; second, the phosphorylation of Erk and Akt was drastically diminished in reduced cell-substrate adhesion conditions compared with that of TCPs, and therefore activation of Raf-MEK-Erk and PI3K-Akt pathways by HB-EGF or other growth factors is required for cell growth; third, the required signals for cell growth appeared to be produced by integrin in TCPs; and finally, monolayer cell culture in TCPs masked the growth-promoting activity of EGF, TGFα and ligands of other RTKs. To our knowledge, the present study is the first to examine the effects of EGFR ligands on cell growth under various culture conditions, using the same cell lines.

Integrins are major players of cell adhesion to the ECM substrate. EGFR and integrins are thought to activate most common signaling pathways (Schwartz and Ginsberg, 2002). Here, we demonstrated that integrin signals largely affect EGFR-ligand-dependent cell growth. In the cases of cell-adhesion-reduced cultures (i.e. 3D-COL, 3D-LM, 3D-agar, 2D-agar and NCPs), integrin signals were reduced compared with that of TCPs, which was not sufficient to support cell growth. Under such conditions, the stimulation of EGFR by its ligands would be required for activation of downstream signaling molecules and promotion of cell growth. Coating of plastic surfaces of TCPs with poly-L-lysine reduces the integrin signal. HB-EGF stimulated SKOV3 cell growth in TCPs coated with poly-L-lysine (Yagi et al., 2005). Moreover, EGFR-expressing 32D and Ba/F3 cells show EGFR ligand-dependent growth in suspension culture. In addition, HB-EGF dependency was observed in the presence of the MEK inhibitor PD98059 in TCPs (Fig. 4B). These observations support our hypothesis that HB-EGF stimulates cell growth when integrin signals are reduced.

We have provided evidence that the activation states of EGFR and its downstream signaling molecules in TCPs differ from those in 3D-COL, NCPs and 2D-agar. The difference seems to be the level of signal strength rather than in signaling pathways. One critical difference is in the significant reduction in phosphorylation levels of EGFR, Erk and Akt in 3D-COL compared with TCPs, before and after HB-EGF stimulation. The reduced level of phosphorylation could explain why PD98059 prevented cell growth more effectively in 3D-COL than in TCPs (Fig. 4B). Decreased phosphorylation levels of EGFR, Erk and Akt might be partially due to the reduced level of EGFR and Akt protein levels. Since the protein levels of Akt (Fig. 5B) and EGFR (data not shown) were not reduced in NCPs or 2D-agar, this might not be the major cause of attenuation of the activation state of EGFR and downstream signaling molecules. Our results also suggest that robust stimulation of EGFR ligands is required for cells to maintain their growth potential in growth-suppressive conditions used for the present study compared with TCPs. HB-EGF dependency could thus be preferentially observed in cells in reduced cell-adhesion conditions.

We have provided evidence that the integrin β1 level was reduced in 3D-COL compared with that of TCPs, and that the expression of HB-EGF or the simultaneous expression of Raf-1, MEK1 and Akt1 increased integrin β1 expression (Fig. 6A). Consistently, previous studies showed that EGFR inhibition reduced integrin β1 levels, and vice versa, in mammary carcinoma cells cultured with 3D-LM (Wang et al., 1998). These results imply that reduction of integrin and EGFR in 3D culture systems might be partly responsible for HB-EGF dependency.

Integrins are enhanced and/or required for growth factor signaling (Streuli and Akhtar, 2009). In the present study, we showed that HB-EGF-dependent growth was observed under integrin-reduced conditions. This does not deny the requirement of integrins in growth-factor-dependent cell growth. In fact, we showed here that overexpression of integrin β1 in BRL cells resulted in enhanced cell growth in 3D-COL and 2D-agar (Fig. 6C,D). In the culture conditions tested in this study (i.e. anti-integrin antibodies, integrin-β1-knockout cells and FAK-knockout cells), integrin signals might have been partially inhibited. Complete inhibition of integrin signals might cause cell growth arrest or cell death, even in the presence of growth factors.

Colony formation in semisolid media, also known as anchorage-independent growth, is a hallmark of oncogenic transformation in vitro (Hanahan and Weinberg, 2000). Formation of `foci', which are aberrantly accumulated cell masses seen in TCPs, is another criterion for oncogenic transformation of cells. Although these criteria of oncogenic transformation are well reflected in the tumorigenicities of cancer cells in vivo, the reason why oncogenic cells exhibit such transformed characters in vitro is not completely understood. The present study sheds light on these characteristics of transformed cells. It will be possible to consider tumor formation of cancer cells in vivo as a type of 3D growth. Focus formation, as well as colony formation in semisolid media, requires the ability to grow in 3D. As shown here, HB-EGF or other growth factors are required for cell growth in 3D or 2D integrin-signal-impeded conditions, therefore normal cells do not grow under such culture conditions in the absence of exogenously added growth factors. By contrast, transformed cells could be activated by signaling via Erk and Akt pathways, or by mutations in or overexpression of EGFR, Ras or other signaling molecules. This could result in the acquisition of cell proliferation ability in integrin-signal-impeded conditions in the absence of exogenously added growth factors, although cell proliferation could be enhanced if growth factors, such as HB-EGF, are exogenously provided. In this context, the requirement of cell adhesion for growth in normal cells or `anchorage dependency' should be taken as meaning the requirement of growth-promoting signals stimulated by cell adhesion, rather than just adhesion of the cell to a substrate.

EGFR systems and their downstream signaling molecules are promising targets for cancer treatment. Actually, many EGFR inhibitors and EGFR-targeted antibodies are currently being used in the clinic or being assessed in clinical trials (Hynes and Lane, 2005; Mosesson and Yarden, 2004). As shown in the present study, the cell growth potential of EGFR ligands is not well documented in TCPs, whereas 3D or 2D integrin-signal-impeded conditions demonstrate this efficiently and reflect the tumorigenicity of nude mice. CRM197, an inhibitor of HB-EGF, strongly inhibited tumor formation by ovarian cancer cells in nude mice (Miyamoto et al., 2004) and HB-EGF-induced cell growth in 3D-COL, but this effect was not detected in cells grown in TCPs (supplementary material Fig. S4). Thus, 3D cultures or culture systems that impede integrin signals might be a better model for the evaluation of inhibitors that target the EGFR system and its downstream signaling molecules. The present study also suggested that EGFR-ligand-induced cells and oncogenic-transformed cells share a common feature in terms of their 3D growth potential. Therefore, the identification of molecules that are specifically required for growth in 3D environments could result in the development of novel targets for cancer therapy. Such molecules might not be required for the normal functioning of epithelial cells that are grown in monolayers. Therefore, therapeutics that target such molecules could inhibit tumor cells without adverse toxicity and side effects. Further studies on molecules that facilitate cell growth in 3D and 2D integrin-signal-impeded culture systems might provide novel molecular targets for cancer therapy.

The anti-HB-EGF extracellular domain antibody and recombinant sHB-EGF, EGF and TGFα were obtained from R&D Systems. The anti-HB-EGF C-terminal domain and anti-EGFR antibodies were from Santa Cruz Biotechnology. The anti-Erk, anti-phosphoErk (T202/Y204), anti-Akt, anti-phospho-Akt (S473) and anti-phospho-EGFR (Y845, Y992 and Y1068) antibodies were from Cell Signaling Technology. The anti-actin and anti-integrin-β1 cytoplasmic domain antibodies were from Chemicon. The function-blocking antibodies against integrins β1 (Ha2/5) and β3 (2C9.G2) were from PharMingen. The anti-cyclin-D1 antibody (DCS-6) was from Medical & Biological Laboratories. ZD1839 was a gift from AstraZeneca, and the other kinase inhibitors were from Calbiochem. Recombinant IGF-I, PDGF-BB and FGF-2 were from PeproTech. CRM197 was prepared as described previously (Miyamoto et al., 2004).

SKOV3 cells and SKOV3 cells expressing human HB-EGF or small-hairpin RNAs for human HB-EGF were established as described previously (Miyamoto et al., 2004). Buffalo rat liver (BRL) and ecotropic retrovirus packaging Plat-E (Morita et al., 2000) cells were gifts from Kaoru Miyazaki (Yokohama City University, Japan) and Toshio Kitamura (The University of Tokyo, Japan), respectively. FAK-null cells (Ilic et al., 1995) were obtained from Tsuyoshi Akagi (KAN Research Institute, Japan) with the permission of Michinari Hamaguchi (Nagoya University, Japan). Integrin-β1-null GE11 cells (Gimond et al., 1999) and GE11 cells expressing integrin β1 were obtained from Kiyotoshi Sekiguchi with the permission of Reinhard Fässler (Max-Planck Institute for Biochemistry, Germany). All cells, except for MCF-10A cells, were maintained in DMEM supplemented with 10% FCS, penicillin and streptomycin. MCF-10A cells obtained from ATCC were maintained with mammary epithelial cell basal medium (MEBM; Clonetics), containing bovine pituitary extract (13 μg/ml), insulin (5 μg/ml), EGF (5 ng/ml), hydrocortisone (2 μg/ml) and cholera toxin (100 ng/ml). FCS-free cultures were grown in DMEM supplemented with insulin (5 μg/ml), transferrin (5 μg/ml), sodium selenite (5 ng/ml), penicillin and streptomycin.

The retrovirus vectors pCX4pur and pCX4bsr (Akagi et al., 2003) were gifts from T. Akagi (KAN Research Institute, Japan). The pCX4hyg and pCX4zeo vectors were constructed by substituting the puromycin-resistance gene of pCX4pur with hygromycin-resistance and zeocin-resistance genes, respectively. Wild-type and uncleavable mutant human HB-EGF cDNAs were cloned into pCX4pur as previously described (Wang et al., 2006). The mutant HB-EGF was characterized previously (Miyamoto et al., 2004). The cDNAs for constitutively active Raf-1 and MEK1 were described previously (Umata et al., 2001). The cDNAs for wild-type and constitutively active Akt1 were gifts from Michiyuki Matsuda (Kyoto University, Japan) with the permission of Yukiko Gotoh (The University of Tokyo, Japan). The cDNAs for wild-type Raf-1 and MEK1, which were obtained from Makoto Tsuneoka (Takasaki University of Health and Welfare, Japan), were fused with an HA tag using an HA-tagging vector (Wang et al., 2006), and then subcloned into pCX4zeo and pCX4hyg, respectively. The cDNA of human integrin β1A was obtained from Junichi Takagi (Osaka University, Japan) with the permission of Yoshikazu Takada (University of California, Davis, CA). A FAK cDNA derived from BRL cells was obtained by reverse transcription-PCR and cloned into the BamHI-NotI sites of the pCX4bsr vector using the following primers: 5′-CAGGATCCACCATGGCAGCTGCTTATCTTG-3′ (forward) and 5′-CTGCGGCCGCTCAGTGTGGCCGTGTCTGCC-3′ (reverse). Retrovirus infection was performed as described previously (Wang et al., 2006). A mixed population of cells was used to avoid clonal effects.

Cells grown in culture dishes were lysed with 50 mM Tris-HCl (pH 7.4), 150 mM NaCl, 1% (w/v) NP-40, 0.1 mM Na₃VO₄, 1 mM NaF, 10 mM sodium pyrophosphate, 10 mM β-glycerophosphate, 10 mM o-phenanthroline and protease inhibitor cocktail (Nacalai Tesque). Cells grown in collagen gels were lysed with a 2× concentration of lysis buffer. The cell lysates were centrifuged and concentrated with ice-cold trichloroacetic acid. Tumors were snap frozen with liquid nitrogen and lysed with the lysis buffer using a Polytron homogenizer (Kinematica).

Tumor-formation assays were performed as described previously (Wang et al., 2006). The handling of animals was performed in accordance with the guidelines prescribed by Osaka University (Osaka, Japan).

Ice-cold bovine type I collagen (Nitta Gelatin; 3 mg/ml), reconstitution buffer comprising 2.2% (w/v) NaHCO₃, 0.2 M HEPES and 50 mM NaOH, 10× DMEM-F12 (1:1) medium and cells suspended in serum-free DMEM (5×10⁶ cells/ml) were mixed in an 8:1:1:0.1 ratio and poured into 24-well dishes (0.5 ml/well). After a 30-minute incubation at 37°C, the gels were overlaid with 1 ml DMEM containing 15% FCS and incubated for 1 week. The collagen gels were incubated with 0.5 ml of 0.5% (w/v) bacterial collagenase (Gibco) in HBS (+) (comprising 10 mM HEPES pH 7.3, 140 mM NaCl, 4 mM KCl, 1.8 mM CaCl₂ and 1 mM MgCl₂) at 37°C until they dissolved, and the cells were then harvested by centrifugation. The cells were resuspended in PBS (–) containing 0.1% (w/v) BSA and 20 mM EDTA, and counted under a microscope. All experiments were carried out in triplicate, and the data were presented as the mean ± s.d.

Cells were mixed with ice-cold 12.3 mg/ml Matrigel (Becton Dickinson; 1.7×10⁵ cells/ml) and poured into 24-well dishes (0.3 ml/well). After a 30-minute incubation at 37°C, the gels were overlaid with 1 ml DMEM containing 10% FCS and incubated for 4 days. MCF-10A cells were cultured for 1 week in the presence of hydrocortisone (1 μg/ml), insulin (5 μg/ml) and cholera toxin (100 ng/ml). The gels were incubated with 0.5 ml of 2000 U/ml dispase (Godo Shusei) in HBS (+) at 37°C until they dissolved, and the cells were then harvested by centrifugation. All experiments were carried out in triplicate, and the data were presented as the mean ± s.d.

Bottom-layer agarose (0.5%; 4 ml) was overlaid with top-layer agarose (0.33%; 3 ml) containing 1×10⁴ cells. The numbers of colonies with diameters greater than 0.2 mm were counted under a dissecting microscope after 4 weeks of culture. For the cell-counting experiments, the cells were cultured with DMEM containing 1.5% (w/v) methylcellulose, and then recovered by centrifugation after diluting the medium with ice-cold PBS (–). All experiments were carried out in triplicate, and the data were presented as the mean ± s.d. The semisolid media cultures performed in this study were carried out in the presence of 10% FCS.

Cells (1×10⁴) suspended in DMEM containing 3% FCS were plated on NCPs (Scivax), and cultured for 3 days. All experiments were carried out in triplicate, and the data were presented as the mean ± s.d.

Upregulation ratio, which represents HB-EGF dependency for cell growth, was calculated using the following equation: doubling time of BRL-HB cells/doubling time of BRL-mock cells.

Statistical significance was determined by Student's t-test and ANOVA for a single pair of conditions and multiple pair of conditions, respectively, with a value of P<0.02 considered statistically significant.