Original Paper | Published:

HOXD3 enhances motility and invasiveness through the TGF-β-dependent and -independent pathways in A549 cells

Abstract

Homeobox genes regulate sets of genes that determine cellular fates in embryonic morphogenesis and maintenance of adult tissue architecture by regulating cellular motility and cell-cell interactions. Our previous studies showed that a specific member, HOXD3, when overexpressed, upregulates integrin β3 expression in human erythroleukemia HEL cells and lung cancer A549 cells, and enhances their motility and invasiveness. We performed a microarray study of over 7075 genes to determine the mechanisms underlying the HOXD3-enhanced motility and invasiveness in A549 cells. RT–PCR-based tracking gene analyses highlighted a set of TGF-β-upregulated genes, which included matrix metalloproteinase-2, syndecan-1, CD44, and TGF-β-induced 68 kDa protein. Exogenous TGF-β also caused this pattern of upregulation in A549 cells and enhanced their migratory and invasive activity, confirming the involvement of TGF-β signaling. However, HOXD3 reduced the expression of TGF-β-independent genes coding for desmosomal components such as desmoglein, desmoplakin and plakoglobin which are known to suppress tumor invasion and metastasis. These results suggest that HOXD3 enhances the invasive and metastatic potential of cancer cells through the TGF-β-dependent and -independent pathways.

Introduction

Homeobox-containing genes are the master regulators of cell differentiation and morphogenesis in animals (Gehring and Hiromi, 1986). They contain a common sequence element of 183 bp, the homeobox, which encodes a highly conserved 61-amino-acid homeodomain. The homeodomain is responsible for recognition and binding of sequence-specific DNA motifs, and cis-regulates the transcription of genes relevant to the formation of specific segmental architecture (McGinnis and Krumlauf, 1992). Homeobox-containing genes are subdivided into more than 20 classes according to their primary sequences. Class I homeobox-containing genes were the first to be discovered and have been most extensively studied. In mammals, 39 class I homeobox genes are clustered in a similar arrangement of 13 paralog groups on four different chromosomal/genomic regions, HOXA, B, C, and D (Graham et al., 1989; Apiou et al., 1996; Mark et al., 1997). They are expressed in a spatiotemporal manner during embryonic morphogenesis, each regulating a group of genes involved in modeling a specific segmental architecture. Class I homeobox-containing genes have also been demonstrated in normal adult tissues with characteristic patterns, suggesting their possible role in the maintenance of tissue-specific architecture (Cillo et al., 1992; Cillo, 1994–1995).

The deregulated expressions of HOX genes have been observed in certain cancers. In acute myeloid leukemia, a chromosomal translocation results in the fusion of the nuclear pore complex protein NUP98 and HOXA9 protein, which seems to promote leukemogenesis through inhibition of HOXA9-mediated differentiation (Borrow et al., 1996; Nakamura et al., 1996a). Proviral activation of Hoxa9 and Hoxa7 by retroviruses has been shown to be involved in leukemia development in a mouse myeloid leukemia model (Nakamura et al., 1996b). In solid tumors, HOX genes exhibit altered expression patterns in human kidney, colon and lung cancers, compared to those in normal organs (Cillo et al., 1992, 1999; De Vita et al., 1993; Tiberio et al., 1994). Altered HOX gene expression is also noted in metastatic lesions of lung and colon cancers, compared to those in their primary lesions (Cillo, 1994–1995; De Vita et al., 1993).

We previously showed that overexpression of the HOXD3 gene enhanced integrin β3 expression in both human erythroleukemia HEL cells and lung carcinoma A549 cells, and that these cells acquired strong ability to adhere to and migrate toward the integrin β3 ligands. However, this finding was not observed in the control cells unexpressing HOXD3 gene (Taniguchi et al., 1995; Hamada et al., 2001). The HOXD3-overexpressing A549 cells acquired ability to produce large amounts of extracellular matrix-degrading enzymes including urokinase-type plasminogen activator (uPA) and matrix metalloproteinase-2 (MMP-2), resulting in the enhancement of in vitro cell invasion of Matrigel (Hamada et al., 2001). Boudreau et al. (1997) noted the HOXD3-mediated conversion of endothelium from resting to an activated angiogenic state. They showed that stimulation of endothelial cells with basic fibroblast growth factor (bFGF) caused increased expression of HOXD3, and enhanced expression of both the integrin αvβ3 and uPA in endothelial cells. These lines of evidence suggest that HOXD3 plays a pivotal role in the regulation of genes related to invasion and metastasis. However, these studies remain incomplete in our attempt of understanding downstream genes of the HOXD3 gene. To better understand the mechanisms of HOXD3-mediated enhancement of invasive and metastatic potential, we monitored effects of HOXD3-overwxpression on global gene expression by using a human cDNA microarray of 7,075 genes. This analysis highlighted the involvement of many effectors, especially molecules associated with cell-cell and cell-extracellular matrix interactions and the activation of a TGF-β-regulated pathway in HOXD3-overexpressing A549 cells.

Results

Identification of genes responsive to HOXD3 transduction by cDNA microarray analysis

We analysed the downstream effects of HOXD3 in A549 cells, using 7075 human cDNA microarrays. A HOXD3-transfected A549 clone (HOX+2) and a control vector-transfected clone (Neo1) were investigated as representative cell populations. Of the 7075 genes analysed, 6438 (91.0%) satisfied the examination criterion (see Materials and methods). In HOX+2 cells, the signal intensities (fluorescence units) ranged from 127 (Human PDGF-associated protein mRNA, complete cds, U41745) to 27 785 (solute carrier family 24 member 1, AF062921). In Neo1 cells, the fluorescence signal ranged from 86 (H sapiens mRNA for RP3 gene, AI885178) to 10 638 (solute carrier family 24 member 1, AF062921). The ratios of the relative signal of HOX+2/Neo1 varied from 5.6 (thrombospondin 1, X14787) to 1/3.4 (cystein-rich protein 1, AI433969). Our previous study had revealed that HOXD3-overexpression enhanced the expression of the integrin β3 in HEL cells and A549 cells (Taniguchi et al., 1995; Hamada et al., 2001). As expression of the integrin β3 was upregulated 1.6-fold in the microarray analysis, we regarded the increases in the ratio by more than 1.6-fold as upregulation and the decreases to less than 1/1.6 as downregulation by HOXD3-overexpression. We identified 74 genes (70 cDNAs and four ESTs) (1.0%) as upregulated (more than 1.6-fold) and 383 genes (167 cDNAs and 216 ESTs) (5.4%) as downregulated (less than 1/1.6) by HOXD3-overexpression. The expression of 6185 genes (87.4%) did not differ between HOX+2 and Neo1 cells. The remaining 433 genes (6.1%) were not expressed in either samples.

To verify the microarray results, we performed semi-quantitative RT–PCR on RNA extracted from the two HOXD3-overexpressing clones (HOX+1 and HOX+2) and two control clones transfected with the empty vector (Neo1 and Neo2). PCR products separated by agarose gel electrophoresis were stained with ethidium bromide, and observed under UV light (Figure 1). The intensity of each band relative to control was analysed by densitometry. As shown in Table 1, the HOXD3-responsive genes indicated by the microarray were classified into five groups: (1) extracellular matrix (ECM) components; (2) cell adhesion molecules; (3) molecules associated with ECM-degradation; (4) cytoskeletal system-associated molecules; and (5) growth factors, cytokines and their related molecules. Most genes profiled differentially by the microarray analysis were likewise characterized by RT–PCR. However, in seven gene expressions (Developmental endothelial locus 1, tissue transglutaminase 2, galectin 3, annexin VIII, interferon γ-inducible protein 16, bone morphogenetic protein 5 and neurotensin), the ratios of the relative signal of HOX+2/Neo1 by RT–PCR were different from those by the microarray analysis. Nonetheless, none of them showed an inverse profile.

Figure 1
figure1

Expression patterns of representative genes responding to HOXD3-overexpression or TGF-β stimulation. RNA was isolated from HOXD3-overexpressing cells (HOX+1 and HOX+2) and the control transfectant cells (Neo1 and Neo2) which had been treated with TGF-β (0, 0.4, 2 or 10 ng/ml) for 24 h. (a) The expression of TGF-β-induced 68 kDa protein (βig-h3) was upregulated by HOXD3 and TGF-β; (b) vitronectin was down-regulated by HOXD3 and TGF-β; (c) HOXD3 was not affected by TGF-β, and integrin β3 was upregulated by HOXD3 but not by TGF-β; (d) plakoglobin was down-regulated by HOXD3 whereas it was not affected by TGF-β

Table 1 Genes that were upregulated or downregulated in A549 cells overexpressing HOXD3 gene

Production of active TGF-β by HOXD3-overexpressing A549 cells

The genes of which expression were altered by HOXD3 included thrombospondin-1 (5.6-fold), TGF-β-induced 68 kDa protein (βig-h3) (5.4-fold), PAI-1 (2.9-fold), tissue transglutaminase (2.8-fold), matrix metalloproteinase 2 (2.7-fold) and CD44 (2.1-fold). These expressions were previously documented to be upregulated by TGF-β (Negoescu et al., 1995; Skonier et al., 1992; Keski-Oja et al., 1988; George et al., 1990; Overall et al., 1991; Nakashio et al., 1997), therefore we investigated whether the TGF-β signaling pathway was activated in these cells. First, we measured TGF-β in the medium conditioned with HOXD3-overexpressing cells (HOX+1 and HOX+2) and control transfectant cells (Neo1 and Neo2), using a growth inhibition assay with Mv1Lu cells (Figure 2a). The bioassay revealed that the medium conditioned with HOXD3-overexpressing cells contained approximately 3 ng/ml of the active form of TGF-β (the concentration of total-latent and active form-TGF-β was 6–9 ng/ml). Although the media conditioned with the parent, Neo1 and Neo2 cells, respectively, contained an equivalent total TGF-β to that of the HOXD3-overexpressing cell media, they contained less amounts of active TGF-β. The growth of Mv1Lu cells, which was inhibited by the media conditioned with HOX+1 and HOX+2 cells, was completely recovered by an addition of neutralizing anti-TGF-β antibody (Figure 2e). We also determined the amounts of TGF-β in the conditioned media by an enzyme-linked immunosorbent assay. Although there were no differences in production of the latent form of TGF-β1, the media conditioned with HOX+1 and HOX+2 cells contained higher levels of active TGF-β1 protein than those with the parent, Neo1 and Neo2 cells (Figure 2b). The immunoblot analysis using an anti-TGF-β antibody demonstrated a 25 kDa band which represented the active form of TGF-β in the media conditioned with HOXD3-overexpressing cells, but not in the media with parent cells or neo-transfected cells (Figure 2c). To examine whether the HOXD3-overexpressing cells cleaved latent form of TGF-β to produce active TGF-β, we checked the latency protein by immunoblotting with the use of an anti-TGF-β1-latency-associated peptide (β1-LAP) antibody. As shown in Figure 2d, three bands (Mr >200 kDa, ≈110 kDa and ≈85 kDa) were detected in the conditioned media from every cell clone. The media conditioned with HOX+1 and HOX+2 cells contained a more amount of ≈85 kDa protein which represented a β1-LAP homodimer (Miyazono et al., 1991) than those with Neo1 and Neo2 cells. On the contrary, the media conditioned with Neo1 and Neo2 cells contained a greater amount of ≈110 kDa protein which represented a TGF-β1/β1-LAP complex (Miyazono et al., 1991) than those with HOX+1 and HOX+2 cells.

Figure 2
figure2

Production of TGF-β by A549 cells transfected with HOXD3. (a) TGF-β activity in the medium conditioned with parental A549 cells and the transfectant cells. TGF-β activity was assessed by growth inhibition assay using mink lung epithelial Mv1Lu cells, and expressed as the concentration (ng/ml) of recombinant human TGF-β1 which showed equivalent growth inhibition on the standard curve. A half of the conditioned medium was acid-treated for activation of latent form of TGF-β (total TGF-β, open columns). Growth inhibition by non-treated conditioned medium indicates active TGF-β (solid columns). The columns represent mean±standard deviation for quadruplicate samples. *P<0.001 compared to the parental or neomycin-resistant gene transfectant cells by one-way ANOVA followed by Fisher's probable least-squares difference analysis as a post hoc test. (b) The amounts of TGF-β protein in these samples were assayed with the use of a TGF-β1-ELISA kit. The amount of TGF-β1 protein in acid-treated conditioned media (total TGF-β) is shown by open columns, and that in acid-non-treated conditioned media (active TGF-β) by solid columns. The columns represent mean±standard deviation for quadruplicate samples. *P<0.001 compared to the parental or neomycin-resistant gene transfectant cells by one-way ANOVA followed by Fisher's probable least-squares difference analysis as a post hoc test. (c) Immunoblot analysis of active form of TGF-β in medium conditioned with each cell line. Samples pulled down with heparin-Sepharose beads from the medium conditioned with each cell line were resolved with SDS–PAGE, and electrotransferred to Immobilon P membrane. The membrane was probed with a mouse monoclonal antibody to TGF-β (active form), and visualized with Enhanced Chemiluminescent Detection System. As a positive control, 4 ng of recombinant human TGF-β (rhTGF-β) was used. (d) Immunoblot analysis of TGF-β1-latency-associated peptide (β1-LAP) in medium conditioned with each cell clone. The membrane was probed with mouse monoclonal antibody to β1-LAP. (e,f) The amounts of active TGF-β in media conditioned with HOXD3-overexpressing cells treated with proteinase inhibitors or a blocking peptide to bind TSP-1 to CD36. The conditioned media were collected from the cells cultured in the presence of 50 nM TIMP-1 (), 50 nM TIMP-2 (□), 20 μM GGWSHW (white dot on black box), 20 μM GGYSHW as a control peptide (horizontal lined square), 10 μg/ml aprotinin (light dotted square), 10 μg/ml leupeptin () or nothing (▪). (e) TGF-β activity without acid-treatment was measured by the growth inhibition assay using Mv1Lu cells. Mv1Lu cells were incubated with the medium conditioned with HOX+1 or HOX+2 cells in the presence of either a neutralizing antibody to TGF-β ([square with thick diagonal from top left to bottom right]) or normal rabbit IgG (dark dotted square) as control. *P<0.001 compared to the medium conditioned with the cells treated with normal rabbit IgG or the non-treated cells by one-way ANOVA followed by Fisher's probable least-squares difference analysis as a post hoc test. (f) The amount of TGF-β1 protein in acid-non-treated conditioned media was measured by TGF-β1-ELISA

MMP-2 and TSP-1 which were upregulated by HOXD3-overexpression (Table 1) are known as activators of TGF-β (Schultz-Cherry et al., 1995; Yu and Stamenkovic, 2000). Our previous study demonstrated that HOX+1 and HOX+2 cells expressed a higher level of urokinase-type plasminogen activator (uPA) than parental, Neo1 and Neo2 cells (Hamada et al., 2001). uPA converts plasminogen into plasmin which enzymatically activates TGF-β (Khalil, 1999). Therefore, we collected conditioned media from HOX+1 and HOX+2 cells which had been cultured in the presence of TIMP-2 (a natural inhibitor of MMP-2), a GGWSHW peptide (a peptide to block the binding of TSP-1 to CD36) or aprotinin and leupeptin (as inhibitors of uPA and plasmin). The amount of active TGF-β in the conditioned media was measured by the growth inhibition assay and ELISA. As shown in Figure 1e,f, there was no difference in the amount of active TGF-β between the conditioned media from the cells treated with any of the inhibitors and those from the control cells.

Responses of endogenous genes to active TGF-β in control transfectant (Neo1 and Neo2) cells

To identify the genes which were upregulated or downregulated by TGF-β signaling in HOXD3-overexpressing clones (HOX+1 and HOX+2), we treated the control transfectant (Neo1 and Neo2) cells with active TGF-β at concentrations of 0, 0.4, 2 or 10 ng/ml for 24 h. Isolated total RNA was used for semi-quantitative RT–PCR on 35 genes that were up- or down-regulated by HOXD3 in the microarray analysis. As represented by βig-h3 in Figure 1a, thrombospondin-1, syndecan 1, PAI-1, MMP-2, and transgelin were upregulated by TGF-β-treatment in a dose-dependent manner. These genes were also upregulated in HOXD3-overexpressing clones (Table 1 and Figure 1a). The expression of vitronectin, which was downregulated in the HOXD3-overexpressing clones, was reduced by TGF-β (Figure 1b). TGF-β treatment did not affect the expressions of HOXD3, integrin β3, developmental endothelial locus-1 (Del-1), tissue transglutaminase, CD24, CD44 and quiescin Q6; except HOXD3, they were upregulated by HOXD3-overexpression (Table 1 and Figure 1c). The expression of desmosomal components such as desmoglein 1, desmoplakin and plakoglobin was repressed by HOXD3-overexpression but not by TGF-β-treatment (Table 1 and Figure 1d).

TGF-β enhancement of in vitro cell motility and invasiveness

To examine whether TGF-β influences metastasis-related properties of A549 cells, we assessed the effects of TGF-β on the in vitro invasion and motility of parent, HOXD3-overexpressing (HOX+1 and HOX+2) and control neo-transfected (Neo 1 and Neo2) cells. HOX+1 and HOX+2 cells showed aggressive invasion of Matrigel (a reconstituted basement membrane) and type I collagen gel compared to the parent, Neo1 or Neo2 cells (Figure 3a). When the parent, Neo1 or Neo2 cells were treated with TGF-β, the invasion of type I collagen gels by these cells was enhanced, whereas their invasion of Matrigel was unchanged (Figure 3b). To confirm that endogenous active TGF-β was responsible for the increased invasion of type I collagen gel, we tested the effects of neutralizing antibodies against TGF-β and TGF-β receptor type II and a human recombinant soluble TGF-β receptor type II (sTGFRII) on the invasion of type I collagen gel by HOX+1 and HOX+2 cells. As shown in Figure 3c, these two antibodies and sTGF-RII reduced the invasion.

Figure 3
figure3

In vitro invasiveness of parental A549 cells and the transfectant cells when treated with TGF-β. In vitro invasion assay was performed by using Transwell chambers with Matrigel-coated membranes (a) or with membranes embedded in type I collagen gel (b). The cells which had been treated with TGF-β for 24 h at the indicated concentrations (square;, no TGF-β; dark dotted square, 0.4 ng/ml; white dots on black square, 2 ng/ml; ▪, 10 ng/ml) were placed in the upper compartment of the chamber. The same concentrations of TGF-β were added to the upper compartment during the invasion assay. After 24 h-incubation, invaded cells were counted under a microscope. The columns represent mean±standard deviation for triplicate assays. #P<0.001 compared to the parental or neomycin-resistant gene transfectant cells, *P<0.001 compared to the cells untreated with TGF-β, **P<0.001 compared to the cells treated with 0.4 ng/ml of TGF-β by one-way ANOVA followed by Fisher's probable least-squares difference analysis as a post hoc test. (c) Invasion of type I collagen gel by HOXD3-overexpressing cells (HOX+1 and HOX+2) was inhibited by treatment with antibody to TGF-β or TGF-β receptor II, or soluble TGF-b receptor type II. The cells were incubated with DME/F12 containing 0.1% BSA (□) with anti-TGF-β antibody (squf;), anti-TGF-β receptor type II (white dot on black square), normal rabbit IgG (), normal goat IgG (), and recombinant human soluble TGF-β receptor type II (dark dotted square) for 30 min at 4°C. The cells (2×104) treated with each antibody or soluble TGF-β receptor type II were placed in the upper compartment of the chamber. After 24 h-incubation, invaded cells were counted under a microscope. The columns represent mean±standard deviation for triplicate assays. *P<0.001 compared to the non-treated or normal IgG treated cells by one-way ANOVA followed by Fisher's probable least-squares difference analysis as a post hoc test

HOXD3-overexpressing cells showed a highly haptotactic response to vitronectin, fibronectin and type I collagen compared to parent, Neo1 or Neo2 cells (Figure 4a–c). TGF-β stimulated the haptotaxis of the parent and vector-transfectants to type I collagen but not to vitronectin or fibronectin, whereas it did not affect the haptotaxis of HOXD3-overexpressing cells to any of the extracellular matrix components (Figure 4a–c). To confirm that endogenous active TGF-β was responsible for the increased haptotaxis to type I collagen, we tested the effects of neutralizing antibodies against TGF-β and TGF-β receptor type II and sTGFRII on haptotaxis of HOX+1 and HOX+2 cells to type I collagen. As shown in Figure 4d, these two antibodies and sTGFRII reduced the haptotactic migration.

Figure 4
figure4

Haptotactic activities of parental A549 cells and the transfectant cells to fibronectin (a), vitronectin (b) and type I collagen (c,d) when treated with TGF-β. Haptotaxis assay was performed by using Transwell chambers. The lower surface of the membranes of transwell chambers was coated with 10 μg of fibronectin, vitronectin or type I collagen. The cells (2×104) which had been treated with TGF-β for 24 h at the indicated concentrations (a, b and c: □, no TGF-β; dark dotted square, 0.4 ng/ml; white dots on black square, 2 ng/ml; ▪, 10 ng/ml) were placed in the upper compartment of the chamber. The same concentrations of TGF-β were added to the upper compartment during the haptotaxis assay. After 6 h-incubation, migrated cells were counted under a microscope. The columns represent mean±standard deviation in randomly selected 20 fields per well at ×200 magnification. #P<0.001 compared to the parental, Neo1 or Neo2 cells, §P<0.001 compared to the parental or Neo1 cells, *P<0.001 compared to the cells untreated with TGF-β, ¶P<0.001 compared to the cells treated with 0.4 ng/ml of TGF-β. (d) Inhibition of haptotaxis of HOXD3-overexpressing cells (HOX+1 and HOX+2) to type I collagen by treatment with antibody to TGF-β or TGF-β receptor type II, or soluble TGF-β receptor type II. The cells were incubated with DME/F12 containing 0.1% BSA (□) with anti-TGF-β antibody (▪), anti-TGF-β receptor type II (white dots on black square), normal rabbit IgG (), normal goat IgG (), and recombinant human soluble TGF-β receptor type II (dark dotted square) for 30 min at 4°C. The cells (2×104) treated with each antibody or soluble TGF-β receptor type II were placed in the upper compartment of the chamber. After 6 h-incubation, migrated cells were counted by using a microscope. *P<0.001 compared to the non-treated or normal IgG treated cells by one-way ANOVA followed by Fisher's probable least-squares difference analysis as a post hoc test

Discussion

We revealed by the cDNA microarray analysis that gene expression patterns in HOXD3-overexpressing cells related to alterations in (1) the cell-cell interactions, (2) the cell-extracellular matrix (ECM) interactions, (3) the cytoskeletal system, and that the HOXD3-overexpression activated a specific signaling pathway (TGF-β-regulated pathway).

Interestingly, HOXD3-overexpression downregulated the gene expression of desmosomal components including desmoglein, desmoplakin and plakoglobin. Immunohistochemical studies have shown that a loss of staining for desmosomal components correlates with invasive and metastatic potential in both transitional cell carcinoma (Conn et al., 1990) and squamous cell carcinoma (Harada et al., 1992; Hiraki et al., 1996; Shinohara et al., 1998). Transfection of the cDNA encoding desmosomal components into highly invasive cells inhibits their in vitro invasion of collagen gels (Tselepis et al., 1998). These reports indicate that desmosomes have a role in suppression of tumor spreading, and therefore a reduction in the expression of desmosomal components in HOXD3-overexpressing cells may aid their dissociation or migration from the primary tumor mass.

Interaction of cells with the extracellular matrix is also important in tumor invasion and metastasis as well as in embryonic morphogenesis. Our microarray analysis and subsequent semi-quantitative RT–PCR analysis revealed that the HOXD3-overexpression upregulated the gene expression of (1) adhesion molecules (integrin β3, CD44), (2) ECM components (TSP-1, βI-GH3, Del-1) and (3) the molecules associated with ECM-degradation (PAI-1, MMP-2). Vitronectin, one of the ECM components, was downregulated by HOXD3-overexpression. We have already shown that the integrin β3 subunit induced by HOXD3-overexpression forms the integrin αvβ3 heterodimer and enhances migration in the presence of their ligands, vitronectin and fibrinogen. Furthermore, phagokinetic track assay in the absence of αvβ3 ligands shows that HOXD3-overexpressing cells have a motile activity higher than those of the parental or control cells (Hamada et al., 2001). The present study indicated a possible missing link. Namely, Del-1, one of the ECM components, upregulated by HOXD3-overexpression, has recently been identified as a novel ligand for the integrin αvβ3 (Hidai et al., 1998). And the Del-1 protein has been shown to promote αvβ3-dependent endothelial cell attachment and migration, which suggests that Del-1 functions as a motility-stimulating factor for endothelial cells during angiogenesis (Penta et al., 1999). We therefore infer that the cell motility of HOXD3-overexpressing A549 cells is stimulated by Del-1 through the integrin αvβ3 in an autocrine manner.

We found the participation of the TGF-β-regulated pathway in the conversion of the cells to a more motile and invasive phenotype during HOXD3-overexpression. Of the genes differentially expressed between the HOXD3-overexpressing cells and the control cells, some were documented as being TGF-β-regulated genes. These included TSP-1, βig-h3, PAI-1 and MMP-2 (Negoescu et al., 1995; Skonier et al., 1992; Keski-Oja et al., 1988; Overall et al., 1991). We confirmed by RT–PCR analysis that the expression patterns of these genes in the control cells stimulated with exogenous TGF-β were similar to those in the HOXD3-overexpressing cells. The bioassay and ELISA for TGF-β and the immunoblot analysis analysis using antibodies to TGFβ1 and β1-LAP demonstrated that this activation of the TGF-β pathway was due to an accelerated conversion of latent TGF-β to an active form in the HOXD3-overexpressing cells. It has been proposed that TGF-β activation occurs in vivo through the pathways of plasminogen/plasmin, TSP-1/CD36, integrin αvβ3/MMP-2 (or MMP-9) or integrin αvβ6 (Khalil, 1999; Lyons et al., 1988; Schultz-Cherry et al., 1995; Yehualaeshet et al., 1999; Munger et al., 1999; Yu and Stamenkovic, 2000). To examine whether the pathways of MMP-2, TSP-1 or plasminogen/plasmin were involved in the activation of TGF-β, we collected conditioned media in the presence of tissue inhibitor of metalloproteinase-2, a natural inhibitor of MMP-2, a peptide (GGWSHW) which disturbs the binding of TSP-1 with CD36, or aprotinin and leupeptin which are inhibitors of plasmin. However, none of the inhibitors prevented the activation of TGF-β achieved by HOX+1 or HOX+2 cells. Further, the involvement of integrin αvβ6 is unlikely as expression of αvβ6 was not observed in any of the parent cell line and the transfected clones (J Hamada, unpublished data). Thus, we have no evidence to demonstrate that MMP-2, TSP-1, plasminogen/plasmin or avb6 was involved in HOXD3-mediated TGF-β activation. We need further experiments to elucidate the mechanism responsible for the HOXD3-mediated TGF-β activation. However, it is of interest that HOXD3 is implicated in the regulation of the conversion from latent to active TGF-β, since TGF-β is known to be an important regulator in embryonic development (Moses and Serra, 1996; Massague and Chen, 2000; Kimelman and Griffin, 2000).

Finally we have shown that the activation of TGF-β signaling by HOXD3-overexpression alters the cellular properties of migration and invasion. Exogenous TGF-β promoted haptotaxis to type I collagen (COL-I) in the parent and control transfectant cells, consistent with a previous report that TGF-β stimulated the invasion of COL-I gels by A549 cells in a dose-dependent manner (Mooradian et al., 1992). Haptotactic activity of HOX+1 and HOX+2 cells to type I collagen was suppressed by an addition of blocking antibodies or soluble TGF-β receptor type II; however, their activity was still higher than that of parent, Neo1 and Neo2 cells. This phenomenon suggests that haptotaxis of HOXD3-overexpressing cells is stimulated through not only a TGF-β-mediated pathway but also other signaling pathway(s). Like haptotaxis to type I collagen, invasion of type I collagen gel by HOX+1 and HOX+2 cells was also partially suppressed by the blocking antibodies and sTGFRII. Further, exogeneous active TGF-β stimulated the invasion of parent, Neo1 and Neo2 cells, but did not reach the same level as those of HOX+1 and HOX+2 cells. It seems that involvement of TGF-β signaling in the invasion of type I collagen gel is relatively low compared to that in haptotaxis to the collagen. Unlike haptotaxis, the cells invading type I collagen gel need to exert more biological ability, for example ability to digest the collagen. In not only migration to type I collagen but also degradation of type I collagen, the HOXD3-overexpressing cells may use TGF-β-independent pathway(s) as well as TGF-β-dependent one. We previously demonstrated that the HOXD3-overexpressing A549 cells showed high haptotactic activity to vitronectin and fibrinogen and a strong invasiveness to Matrigel compared to the parent and control transfectant cells (Hamada et al., 2001). In the present study, however, exogenous TGF-β did not affect the haptotaxis of the control cells to vitronectin and fibronectin or their invasion of Matrigel, which is composed of laminin-1, collagen type IV and heparan sulfate proteoglycan (Kleinman et al., 1982). In addition, none of the expressions of integrin β3, Del-1, or desmosomal components (desmoglein, desmoplakin and plakoglobin) was affected by TGF-β-stimulation. Taken together, the TGF-β signaling activated by HOXD3-overexpression seems to mainly facilitate ligand (COL-I)-dependent cell migration and partly to participate in gain of more invasive and metastatic potential of cells. Type I collagen is the major component in tumor stroma. TGF-β-mediated haptotactic and invasive responses may facilitate invasion of tumor cells from the in vivo primary tumor tissues.

In conclusion, the phenotypic and transcriptional alteration of A549 cells by HOXD3-overexpression may be caused by signals through multiple pathways including the TGF-β-regulated pathway. Our study provides the first evidence that HOXD3 homeobox gene regulates many downstream effectors, and alters the invasive and metastatic potential of cancer cells.

Materials and methods

Antibodies and reagents

A rabbit pan-specific antibody to TGF-β, a mouse monoclonal antibody to human TGF-β1-latency-associated peptide (β1-LAP) and recombinant human TGF-β1 were purchased from R&D (Minneapolis, MN, USA). A goat antibody to human TGF-β receptor type II, a rabbit antibody to TGF-β and recombinant human TGF-β soluble receptor type II were from Genzyme (Minneapolis, MN, USA). A mouse monoclonal antibody to thrombospondin was from Oncogene Research Products (Cambridge, MA, USA). Procine type I collagen was from Nitta Gelatin (Osaka, Japan). Normal rabbit IgG was from Chemicon (Temecula, CA, USA). Normal goat IgG was from Cappel (Durham, NC, USA). Bovine vitronectin was from Yagai (Yamagata, Japan). Bovine serum albumin was from Boehringer Mannheim (Mannheim, Germany). Matrigel was from Becton Dickinson (Bedford, MA, USA). Transwell chamber was from Costar (Cambridge, MA, USA). Geneticin® (G418 sulfate), Molony murine leukemia virus reverse transcriptase, random primer and Trizol, human tissue inhibitor of metalloproteinase (TIMP)-1 and TIMP-2 were from Life Technologies (Rockville, MD, USA). Taq polymerase was from Nippon Gene (Tokyo, Japan). Glutaraldehyde, aprotinin and leupeptin were from Wako (Tokyo, Japan). Crystal violet was from Kanto Chemical (Tokyo, Japan). Heparin-Sepharose, CL-6B was from Amersham Pharmacia Biotech (Uppsala, Sweden). Oligopeptide, GGWSHW and GGYSHW were synthesized by Sigma Genosys Japan (Ishikari, Hokkaido, Japan).

Cells and cell culture

Human lung cancer A549 cells and mink lung epithelial Mv1Lu cells were obtained from the Japanese Cancer Research Resources Bank (JCRB, Tokyo, Japan). These cells were grown on tissue culture dishes in a 1 : 1 (v/v) mixture of Dulbecco's modified minimum essential medium and Ham's F12 medium (DME/F12) supplemented with 5% fetal bovine serum (FBS). Cloned A549 cell lines expressing HOXD3 gene (HOX+1 and HOX+2) or neomycin-resistant gene (Neo1 and Neo2) were established by the use of lipofection with a mammalian expression vector, pMAMneo/HOX4A(HOXD3) or empty vector pMAMneo (Hamada et al., 2001). These cells were grown on tissue culture dishes in DME/F12 supplemented with 5% FBS, Geneticin (400 μg/ml). The cell lines were cultured at 37°C in a humidified 5% CO2 atmosphere.

cDNA microarray analysis

Target cDNA was generated from 1 μg polyadenylated mRNA, which was reverse-transcribed and labeled either with Cy5 (HOX+2) or Cy3 (Neo1) dUTP. The average intensity of the Cy5 fluorescence divided by the average intensity of the Cy3 fluorescence equaled 0.97 (balance coefficient), indicating similar labeling efficiency for each set of target cDNAs. Target cDNA was hybridized on IncyteGEM, microarrays containing 7075 probes with sequences complementary to 4107 human genes and 2968 human expressed sequence tags (ESTs) (UniGem V, Genome Systems, St. Louis, MO, USA). Following the hybridization and washing, the relative expression levels of both cDNA populations were measured and compared by obtaining the Cy5/Cy3 fluorescence ratio for each target cDNA to satisfy the examination criterion. The inclusion criterion was based on an image recognition algorithm for each cDNA in the analysis and included a fluorescence signal from the cDNA exceeding a signal to background ratio of 2.5 and the cDNA covering its grid location on the microarray for >40%. In this study, genes were considered differentially expressed if an increase was more than 1.6-fold or a decrease was to less than 1/1.6.

Semiquantitative duplex RT–PCR analysis

For RT–PCR analysis, total RNA was extracted from monolayer cultures of HOX+1, HOX+2, Neo1, Neo2 cells and TGF-β (0.4, 2 or 10 ng/ml)-treated-Neo1 and -Neo2 cells with Trizol, according to the manufacturer's instruction. Two μg of total RNA sample was subjected to cDNA synthesis for 2 h at 37°C in 50 μl of reaction mixture containing 4 U/ml of Molony murine leukemia virus reverse transcriptase, 7.5 mM dithiothreitol, 0.5 mM MgCl2, 0.5 μM dNTP and 2 μM random primer. PCR amplification of cDNA was performed in 50 μl of reaction mixture containing 1 μl of cDNA sample, 10 mM Tris-HCl (pH 8.3), 50 mM KCl, 2 mM MgCl2, 0.125 U/ml of Taq polymerase and different primer sets (10 nM each). Co-amplification of the specific gene and human β-actin gene, as an internal control, was achieved by using two primer sets in a single reaction mixture. Each primer was designed to encompass an exon junction for prevention of templating possibly contaminated genomic DNA. PCR products were electrophoresed in a 2.5% agarose gel and intensity of the bands observed under a UV illuminator was analysed by Scion Image (Scion, Frederick, MD, USA).

A549 cell-conditioned medium

Tumor cells (2×105/dish) were plated on 60-mm tissue culture dish in DME/F12 supplemented with 5% FBS. After 24 h, the cultures were washed twice with DME/F12 medium without FBS and then incubated with DME/F12 medium without FBS for 24 h. The medium was collected and centrifuged at 800 g for 10 min, and the supernatants were recentrifuged at 20 000 g for 10 min. All processes were done in sterile condition.

Bioassay of TGF-β activity

In vitro bioassay to measure TGF-β activity was based on the method described by Lucas et al. (1991) with some modification. Briefly, mink lung epithelial Mv1Lu cells were plated on flat-bottomed 96-well tissue culture plates in 50 μl of DME/F12 supplemented with 5% FBS at 3.2×103 cells/well. Prior to addition of samples (conditioned media from each cell line and DME/F12) to the wells, a half of each sample was acid-treated for activation of latent form of TGF-β by incubating each 200 μl sample with 1.5 μl of 6N HCl for 30 min at room temperature and neutralized with 3 μl of neutralizing solution (1 : 1=6 M NaOH : 1 M HEPES). Fifty μl of the conditioned media, acid-treated-CM and DME/F12 were added to each quadruplicate well. Recombinant human TGF-β1 (rhTGF-β1) diluted with DME/F12 at serial concentrations (0–20 ng/ml) was added to the wells (50 μl/well) to establish a standard curve. After 3 days of incubation, the assay was terminated and cell numbers were quantified by using a colorimetric crystal violet-staining procedure (Kueng et al., 1989). To determine cell number, the absorbance of each well was read at 595 nm on a Microplate reader (BIO-RAD, Model 550, Hercules, CA, USA). Data of absorbance at 595 nm were converted to cell numbers according to the pre-tested linear correlation between the cell number and the absorbance. Percentage of growth inhibition was calculated as %=100×{1−(no. of cells in sample well with CM or rhTGF-β)/(no. of cells in control well)}. For each experiment, a standard curve for per cent growth inhibitions to rhTGF-β1 concentrations was established. The concentration of TGF-β in CM was deduced from the rhTGF-β1 concentration (ng/ml) of the equivalent growth inhibiting activity on the standard curve.

Enzyme-linked immunosorbent assay (ELISA) for TGF-β

The amounts of TGF-β1 protein in conditioned media were assayed with the use of a TGF-β1-ELISA kit (AN' ALYZA human TGF-β1, Genzyme Techne, Minneapolis, MN, USA). To activate latent TGF-β1 to be immunoreactive, the conditioned media were treated with 1 N HCl and then neutralized with 1.2 N NaOH/0.5 M HEPES, according to the manufacturer's instruction.

Immunoblot analysis

For the detection of active form of TGF-β, the media conditioned with each cell line (1.2 ml) were mixed with 50 μl of heparin-Sepharose CL-6B beads and rotated at 4°C for 18 h (McCaffrey et al., 1992). The heparin-Sepharose beads were washed five times with DME/F12; bound TGF-β was eluted with SDS–PAGE sample buffer and boiled for 4 min. Each sample was resolved with 12.5% SDS–PAGE and electrotransferred to a polyvinylidene difluoride membrane (Immobilon P, Millipore, Bedford, MA, USA). For the detection of latent form of TGF-β, the media conditioned with each cell line were concentrated to 2 mg/ml of protein using a Centricon-10 (Amicon, Millipore, Bedford, MA, USA). Ten μl each of the concentrated conditioned media was subjected to SDS–PAGE in a 7.5% gel under non-reducing conditions and electrotransferred to the membrane. The membranes were blocked in TBS-T (20 mM Tris-HCl, pH 7.5, 137 mM NaCl, 0.1% Tween 20) with 5% skim milk overnight at 4°C, and then incubated with primary antibodies for 1 h at room temperature. The membranes were then incubated with horseradish peroxidase-conjugated antibodies for 1 h at room temperature, and developed with reagents of the Enhanced Chemiluminescent Detection System (Amersham Pharmacia, Little Chalfont, UK).

Flow cytometry

The cultured cells were harvested by trypsinization and washed twice with PBS containing 0.065% sodium azide. The cells were incubated with primary antibodies for 1 h at 4°C, and then with secondary antibodies conjugated with FITC. After incubation, the cells were washed three times with PBS containing 0.065% sodium azide. The stained cells were resuspended in 1 ml of PBS containing 0.065% sodium azide and analysed with the use of FACSCalibur (Becton Dickinson, San Jose, CA, USA).

Haptotaxis assay

Haptotaxis assay was performed by using Transwell chambers. The lower surface of the membranes with 8-μm pores of Transwell chambers was coated with 10 μg each of FN, VN, or FB, and dried up in a hood overnight. Before the assay, the membranes coated with the matrix components had been washed twice with DME/F12. Six hundreds μl of DME/F12 containing 0.1% BSA was placed into the lower compartment of the Transwell chambers, and then 100 μl of the cell suspension (2×105 cells/ml in DME/F12 containing 0.1% BSA) was placed into the upper compartment. After 6 h-incubation, each membrane was fixed with 10% neutral-buffered formalin and stained in Giemsa solution. After the cells attached to the upper side of the membrane were removed by wiping with a cotton swab, those attached to the lower side of the membrane were counted under a microscope. Haptotactic activity was evaluated by the number of cells per field at ×200 magnification (mean±s.d., n=20). In some experiments, to examine the involvement of TGF-β in the haptotaxis to type I collagen, the cells were incubated in DME/F12 containing 0.1% BSA and recombinant human soluble TGF-β receptor type II (10 μg/ml), anti-TGF-β antibody (200 μg/ml), anti-TGF-β receptor type II antibody (100 μg/ml), normal rabbit IgG (200 μg/ml) or normal goat IgG (100 μg/ml) for 30 min at 4°C. The cell suspension (100 μl) was placed into the upper compartment of Transwell as described above.

In vitro invasion of reconstituted basement membrane, Matrigel, by A549 cells

In vitro invasion was assayed by the method reported by Albini et al. (1987) with some modification (Nagayasu et al., 1998). Briefly, membranes with 8-μm pores of Transwell chambers were coated with 100 μl of 20-times diluted Matrigel (Becton Dickinson, Bedford, MA, USA) in cold sterile deionized-distilled water. The Matrigel-coated Transwell chambers were air-dried in a hood overnight. The Matrigel-coated membranes were then washed twice with 100 μl of DME/F12 and incubated with DME/F12 for 1 h at room temperature. Before the assay, the medium in the upper compartment of the Transwell chamber was replaced with 100 μl of A549 cell suspension (2×105 cells/ml) in DME/F12 supplemented with 1% FBS. Into the lower compartment of the chamber, 600 μl of medium conditioned with human lung fibroblast MRC-5 cells was placed. After 24 h of incubation, membranes was fixed with 10% neutral-buffered formalin and stained in Giemsa solution. The cells attached to the upper side of the membrane were wiped out, and those attached to the lower side of the membrane were counted under a microscope. Invasiveness was evaluated by the number of cells invaded per membrane (mean±s.d., n=3).

Invasion of type I collagen gel by A549 cells

Type I collagen gel (Cellmatrix type I-P, Nitta Gelatin, Yao, Japan), 10-fold concentrated DME/F12 and NaHCO3/HEPES were mixed in volume at 9 : 1 : 1, respectively. The mixed collagen solution (600 μl) was placed in each well of 24-well plates, and a Transwell was placed onto each well. The Transwells were pressed down to assure their tight contact to the collagen solution. The 24-well plates with collagen sol and Transwells were incubated at 37°C for 30 min for gel-formation of the collagen solution. After gelation, 90 μl of cell suspension (2.2×105/ml in DME/F12 containing 0.1% BSA) was placed in the upper compartment of each Transwell. Immediately after the cell placement, 10 μl of recombinant human TGF-β diluted at a series of concentrations (0, 4, 20, 100 ng/ml) with DME/F12 containing 0.1% BSA was added into the upper compartment. After 24 h of incubation at 37°C in a CO2 incubator, the Transwell was removed and the collagen gel was washed twice with 2 ml of DME/F12. The cells in the collagen gel were observed under an inverted phase-contrast microscope (Eclipse TE300, Nikon, Tokyo, Japan). Invasiveness was evaluated by the number of cells in the collagen gel in triplicate assay.

References

  1. Albini A, Iwamoto Y, Kleinman HK, Martin GR, Aaronson SA, Kozlowski JM, McEwan RN . 1987 Cancer Res. 47: 3239–3245

  2. Apiou F, Flagiello D, Cillo C, Malfoy B, Poupon MF, Dutrillaux B . 1996 Cytogenet. Cell Genet. 73: 114–115

  3. Borrow J, Shearman AM, Stanton Jr VP, Becher R, Collins T, Williams AJ, Dube I, Katz F, Kwong YL, Morris C, Ohyashiki K, Toyama K, Rowley J, Housman DE . 1996 Nat. Genet. 12: 159–167

  4. Boudreau N, Andrews C, Srebrow A, Ravanpay A, Cheresh DA . 1997 J. Cell Biol. 139: 257–264

  5. Cillo C, Barba P, Freschi G, Bucciarelli G, Magli MC, Boncinelli E . 1992 Int. J. Cancer 51: 892–897

  6. Cillo C . 1994–1995 Invasion Metastasis 14: 38–49

  7. Cillo C, Faiella A, Cantile M, Boncinelli E . 1999 Exp. Cell. Res. 248: 1–9

  8. Conn IG, Vilela MJ, Garrod DR, Crocker J, Wallace DM . 1990 Br. J. Urol. 65: 176–180

  9. De Vita G, Barba P, Odartchenko N, Givel JC, Freschi G, Bucciarelli G, Magli MC, Boncinelli E, Cillo C . 1993 Eur. J. Cancer 29A: 887–893

  10. Gehring WJ, Hiromi Y . 1986 Annu. Rev. Genet. 20: 147–173

  11. George MD, Vollberg TM, Floyd EE, Stein JP, Jetten AM . 1990 J. Biol. Chem. 265: 11098–11104

  12. Graham A, Papalopulu N, Krumlauf R . 1989 Cell 57: 367–378

  13. Hamada J, Omatsu T, Okada F, Furuuchi K, Okubo Y, Takahashi Y, Tada M, Miyazaki YJ, Taniguchi Y, Shirato H, Miyasaka K, Moriuchi T . 2001 Int. J. Cancer 93: 516–525

  14. Harada T, Shinohara M, Nakamura S, Shimada M, Oka M . 1992 Int. J. Oral. Maxillofac. Surg. 21: 346–349

  15. Hidai C, Zupancic T, Penta K, Mikhail A, Kawana M, Quertermous EE, Aoka Y, Fukagawa M, Matsui Y, Platika D, Auerbach R, Hogan BL, Snodgrass R, Quertermous T . 1998 Genes Dev. 12: 21–33

  16. Hiraki A, Shinohara M, Ikebe T, Nakamura S, Kurahara S, Garrod DR . 1996 Br. J. Cancer 73: 1491–1497

  17. Keski-Oja J, Raghow R, Sawdey M, Loskutoff DJ, Postlethwaite AE, Kang AH, Moses HL . 1988 J. Biol. Chem. 263: 3111–3115

  18. Khalil N . 1999 Microbes Infect. 1: 1255–1263

  19. Kimelman D, Griffin KJ . 2000 Curr. Opin. Genet. Dev. 10: 350–356

  20. Kleinman HK, McGarvey ML, Liotta LA, Robey PG, Tryggvason K, Martin GR . 1982 Biochemistry 21: 6188–6193

  21. Kueng W, Silber E, Eppenberger U . 1989 Anal. Biochem. 182: 16–19

  22. Lucas C, Fendly BM, Mukku VR, Wong WL, Palladino MA . 1991 Meth. Enzymol. 198: 303–316

  23. Lyons RM, Keski-Oja J, Moses HL . 1988 J. Cell Biol. 106: 1659–1665

  24. Mark M, Rijli FM, Chambon P . 1997 Pediatr. Res. 42: 421–429

  25. Massague J, Chen Y-G . 2000 Genes Dev. 14: 627–644

  26. McCaffrey TA, Falcone DJ, Du B . 1992 J. Cell. Physiol. 152: 430–440

  27. McGinnis W, Krumlauf R . 1992 Cell 68: 283–302

  28. Miyazono K, Olofsson A, Colosetti P, Heldin CH . 1991 EMBO J. 10: 1091–1101

  29. Mooradian DL, McCarthy JB, Komanduri KV, Furcht LT . 1992 J. Natl. Cancer Inst. 84: 523–527

  30. Moses HL, Serra R . 1996 Curr. Opin. Genet. Dev. 6: 581–586

  31. Munger JS, Huang X, Kawakatsu H, Griffiths MJ, Dalton SL, Wu J, Pittet JF, Kaminski N, Garat C, Matthay MA, Rifkin DB, Sheppard D . 1999 Cell 96: 319–328

  32. Nagayasu H, Hamada J, Kawano T, Konaka S, Nakata D, Shibata T, Arisue M, Hosokawa M, Takeichi N, Moriuchi T . 1998 Br. J. Cancer 77: 1371–1377

  33. Nakamura T, Largaespada DA, Lee MP, Johnson LA, Ohyashiki K, Toyama K, Chen SJ, Willman CL, Chen IM, Feinberg AP, Jenkins NA, Copeland NG, Shaughnessy Jr JD . 1996a Nat. Genet. 12: 154–158

  34. Nakamura T, Largaespada DA, Shaughnessy Jr JD, Jenkins NA, Copeland NG . 1996b Nat. Genet. 12: 149–153

  35. Nakashio T, Narita T, Akiyama S, Kasai Y, Kondo K, Ito K, Takagi H, Kannagi R . 1997 Int. J. Cancer 70: 612–618

  36. Negoescu A, Lafeuillade B, Pellerin S, Chambaz EM, Feige JJ . 1995 Exp. Cell Res. 217: 404–409

  37. Overall CM, Wrana JL, Sodek J . 1991 J. Biol. Chem. 266: 14064–14071

  38. Penta K, Varner JA, Liaw L, Hidai C, Schatzman R, Quertermous T . 1999 J. Biol. Chem. 274: 11101–11109

  39. Schultz-Cherry S, Chen H, Mosher DF, Misenheimer TM, Krutzsch HC, Roberts DD, Murphy-Ullrich JE . 1995 J. Biol. Chem. 270: 7304–7310

  40. Shinohara M, Hiraki A, Ikebe T, Nakamura S, Kurahara S, Shirasuna K, Garrod DR . 1998 J. Pathol. 184: 369–381

  41. Skonier J, Neubauer M, Madisen L, Bennett K, Plowman GD, Purchio AF . 1992 DNA Cell Biol. 11: 511–522

  42. Taniguchi Y, Komatsu N, Moriuchi T . 1995 Blood 85: 2786–2794

  43. Tiberio C, Barba P, Magli MC, Arvelo F, Le Chevalier T, Poupon MF, Cillo C . 1994 Int. J. Cancer 58: 608–615

  44. Tselepis C, Chidgey M, North A, Garrod D . 1998 Proc. Natl. Acad. Sci. USA 95: 8064–8069

  45. Yehualaeshet T, O'Connor R, Green-Johnson J, Mai S, Silverstein R, Murphy-Ullrich JE, Khalil N . 1999 Am. J. Pathol. 155: 841–851

  46. Yu Q, Stamenkovic I . 2000 Genes Dev. 14: 163–176

Download references

Acknowledgements

The authors wish to thank Ms. Masako Yanome for help in preparing the manuscript. This work was supported by a Grant-in-Aid for Scientific Research (B) and a Grant-in-Aid for Scientific Research on Priority Areas (C) from the Ministry of Education, Culture, Sports, Science and Technology of Japan.

Author information

Correspondence to Jun-ichi Hamada.

Rights and permissions

Reprints and Permissions

About this article

Keywords

  • homeobox gene
  • HOXD3
  • microarray
  • lung cancer cells
  • TGF-β
  • invasion

Further reading