Human NSCLC cell lines, A549, Lu99 and EBC-1, (Riken BioResearch Center, Tsukuba, Japan) were cultured in 100-mm dishes (Becton Dickinson Labware) in RPMI-1640 medium supplemented with 10% fetal bovine serum (FBS) (Moregate Biotech), penicillin (5 µg/ml), streptomycin (5 µg/ml) and neomycin (10 µg/ml). Human umbilical vein endothelial cells (HUVECs) isolated from human umbilical cord were purchased from Lonza Walkersville, Inc. and cultured as described previously (8). Incubation was carried out at 37°C in 95% air and 5% CO₂. Phase contrast imaging with a light microscope (Nikon, Tokyo, Japan) was performed at the indicated time points. Representative images of phase contrast were obtained. Type I collagen solution (Atelocollagen Bovine Dermis, IPC-30) was purchased from Koken, Co., Ltd. Other chemicals were purchased from Sigma-Aldrich; Merck KGaA unless otherwise stated. All cells used in this study were authenticated by short tandem repeat analysis and confirmed to be mycoplasma-negative.

Serum-free culture was performed as described previously (8) with slight modifications. The NSCLC cell lines were trypsinised, spun down at 4°C and washed twice with cold serum-free MCDB-104GK medium (Nihon Pharmaceutical Co., Ltd.). The cell lines were then serum-deprived, seeded at a density of 0.2 or 2×10⁵ cells/cm² in 35-mm dishes (Asahi Techno Glass), 60-mm dishes (Asahi Techno Glass) or 100-mm dishes (Becton Dickinson Labware) in serum-free MCDB-104GK medium supplemented with penicillin (5 µg/ml), streptomycin (5 µg/ml) and neomycin (10 µg/ml) and incubated for 24 h. The serum-free culture supernatant derived from EBC-1 cells (EBC-1 supernatant) and the cells were spun down, and each supernatant was filtrated through a 0.45-µm polyvinylidene difluoride membrane filter (Merck Millipore, Italy) and concentrated by ultrafiltration (Amicon Ultra 3K, Merck Millipore). Protein concentrations in EBC-1 supernatants were determined using the Bradford method.

Cell death analyses in EBC-1 cells were performed using the Muse™ Cell Analyzer (Merck Millipore) according to the manufacturer's instructions as described previously (8) with slight modifications. Briefly, EBC-1 cells were harvested after 24-h cultures with or without 10% FBS and suspended at 3×10⁵ cells/ml in phosphate-buffered saline (PBS) containing 1% FBS. Each 100-µl cell suspension was then labelled for 20 min in the dark with the same volume of Annexin-V/7-amino-actinomycin D (7-AAD) reagent (Muse™ Annexin-V & Dead Cell kit, Merck Millipore). Quantitative detection of Annexin V/7-AAD-positive cells was performed using the Muse™ Cell Analyzer.

Tube formation in sandwich culture was performed as described previously (8) with slight modifications. Briefly, HUVECs (1.1×10⁵ cells/cm²) were sandwiched between two layers of collagen gel (0.258% type I collagen) with tube-induction medium composed of MCDB-104GK medium and 199 medium at a 13:7 ratio, supplemented with 2% FBS, l-ascorbic acid (25 µg/ml), penicillin (5 µg/ml), streptomycin (5 µg/ml) and neomycin (10 µg/ml) in 24-well culture plates (Becton Dickinson Labware) for 24 h. Tube formation was induced in the presence of tube-induction medium containing or stratifying culture supernatants derived from NSCLC cell lines, recombinant human VEGF-A (rhVEGF-A; HumanZyme), rhHDGF (Abnova) or rhFGF-2 (Fuji Film Wako Pure Chemical Corp.). Inhibitory analysis of tube formation was performed using tube-induction medium containing or stratifying the supernatants or the rhGFs together with neutralising antibodies listed in Table SI. Tube formation was quantified as described previously (8). Tube areas were quantified as the ratio of the area of the formed tubes to that of the imaged field using the Scion Image 4.0.3 program (Scion Corp.), and the ratio of tube areas of vehicle treatment or control was regarded as 1.0.

Tube formation of HUVECs in 3D culture was performed as described previously (8) with slight modifications. Briefly, HUVECs were trypsinised and spun down. The culture supernatants were aspirated, and the cell pellets were mixed with 0.258% type I collagen gels as described above. HUVECs mixed with collagen gel were added to each well (2.86×10⁶ cells/ml) of 96-well culture plates (Becton Dickinson Labware) for cell viability analysis as described below. The plates were incubated at 37°C for 1 h to allow the collagen gel to solidify. HUVECs were then incubated for 24 h to induce tube formation in the abovementioned tube-induction medium containing EBC-1 supernatants at 2–50 µg/ml. Cell viabilities were determined using the Cell Counting Kit-8 (Dojindo) as described previously (8). Cell viability assays in 3D culture were performed in triplicate.

Cell proliferation was analysed as described previously (8) with slight modifications. Briefly, HUVECs (5.0×10³ cells/cm²) in collagen-coated 24-well culture plates were treated with the abovementioned tube-induction medium containing EBC-1 supernatants (2–50 µg/ml) at the indicated time points. The cells were harvested by trypsinisation, and the cell counts were determined with the Coulter Counter Z1 (Coulter Japan). Cell proliferation assays were performed in duplicate.

Western blotting was performed as described previously (8) with slight modifications. Briefly, HUVECs (1×10⁵ cells/cm²) were seeded in collagen-coated 60-mm culture dishes and incubated in the culture medium. The culture medium was aspirated after 24 h and the dishes were washed with PBS. HUVECs were then incubated in tube-induction medium without FBS for 3 h before stimulation. Stimulation was achieved by supplementing the serum-free tube-induction medium containing EBC-1 supernatant (50 µg/ml), 3D culture supernatant derived from Lu99 cells described below or rhVEGF-A (30 ng/ml) with or without a mouse monoclonal anti-IgG2B antibody (10 µg/ml) or an anti-VEGF-A neutralising antibody (10 µg/ml) at the indicated time points. After the stimulations, proteins were extracted from HUVECs. Culture supernatants were collected from serum-free cultures of EBC-1 cells and 3D culture of Lu99 cells treated with small interfering RNA (siRNA) described below. Protein concentration of each supernatant was determined using the Bradford method. Proteins extracted from HUVECs, siRNA-treated EBC-1 supernatants (50 µg) and siRNA-treated Lu99 cells were used for western blotting as described previously (8). The antibodies used for western blotting are described in Table SI.

Stealth RNAi Negative Control Duplexes (#12935-113) were purchased from Thermo Fisher Scientific, Inc. and used as a control. Each Stealth siRNA duplex oligoribonucleotide against VEGF-A, midkine (MK), HDGF, granulin (GRN) and FGF-2 (GenBank™ accession nos. NM_003376, NM_00101233, NM_004494, NM_002087 and NM_002006, respectively) were synthesised by Thermo Fisher Scientific, Inc. (sequences shown in Table SII). The duplex oligoribonucleotides were dissolved in diethyl pyrocarbonate-treated water to 20 µM. EBC-1 and Lu99 cells were transfected with siRNAs using Lipofectamine RNAiMAX (Thermo Fisher Scientific, Inc.) in accordance with the manufacturer's instructions. Stealth RNAi compounds were transfected at a final concentration of 5 nM in culture medium as described previously (8) with slight modifications. Briefly, 1×10⁶ cells were incubated overnight in 100-mm dishes containing 10 ml of RPMI-1640 medium supplemented with 10% FBS without penicillin, streptomycin and neomycin. Lipofectamine RNAiMAX and siRNA were each diluted in 1 ml of RPMI-1640 medium for 5 min at room temperature, and then they were combined and incubated for 15 min at room temperature to form complexes. Two millilitre of the mixture was added to each dish and the cells were further incubated. The old RPMI-1640 medium containing the mixture was aspirated after a 48-h incubation, the dishes were washed with PBS, and 10 ml of fresh RPMI-1640 medium supplemented with 10% FBS was added and incubated for another 24 h (a total of 72-h incubations after starting the siRNA transfections). The cells were harvested by trypsinisation. EBC-1 cells were then seeded at a density of 2×10⁵ cells/cm² in 35-mm, 60-mm or 100-mm dishes in serum-free MCDB-104GK medium supplemented with penicillin, streptomycin and neomycin and incubated further for 24 h as described above. The serum-free culture supernatants from siRNA-treated EBC-1 cells were collected after the 24-h incubations (a total of 96-h incubations after starting the siRNA transfection). Following trypsinisation, Lu99 cells were incubated in 3D cultures as described below.

The levels of VEGF-A protein in serum-free culture supernatants (50 µg/ml) from EBC-1 cells treated with or without siRNAs were measured using the Human VEGF-A ELISA Kit (RayBiotech, Inc.) according to the manufacturer's instructions as described previously (8).

After EBC-1 supernatant was collected as described above, 20 µl (≥1 mg/ml) of each supernatant was precipitated using trichloroacetic acid. The precipitates derived from each supernatant were dissolved in Tris buffer (2 mM EDTA/250 mM Tris-HCl) at pH 8.5, reduced with 0.67 M dithiothreitol, alkylated with 1.4 M iodoacetamide and digested with trypsin. The recovered peptides were analysed using a Q Exactive Plus MS instrument (Thermo Fisher Scientific, Inc.) coupled with a capillary high-performance liquid chromatography system (EASY-nLC 1200, Thermo Fisher Scientific, Inc.) to acquire MS/MS spectra. A 0.075×150 mm-EASY-Spray column (3-µm particle diameter, 100-Å pore size, Thermo Fisher Scientific, Inc.) was used with mobile phases of 0.1% formic acid and 0.1% formic acid/80% acetonitrile. Data derived from the MS/MS spectra were searched in the SWISS-Prot database using the MASCOT Server (http://www. matrixscience. com) and proteins were identified using the Scaffold viewer program (http://www.proteomesoftware. com/products/scaffold). Protein identification by MS was performed in two independent experiments, and proteins identified in both experiments are shown in Table SIII.

The levels of HDGF protein in EBC-1 supernatants (50 µg/ml) and those in Lu99 supernatants were measured using the Human HDGF ELISA Kit (Arigo biolaboratories Corp., Hsinchu, Taiwan) according to the manufacturer's instructions. Briefly, each microtitre plate was pre-coated with a polyclonal anti-human HDGF antibody (Arigo Biolaboratories Corp.). After collecting the supernatants, EBC-1 supernatants (50 µg/ml), Lu99 supernatants or rhHDGF (Arigo biolaboratories Corp.) were added to each well for 2 h at 37°C. After 3 washes with wash buffer, 100 µl of peroxidase-linked anti-HDGF antibody (ARG81356, Arigo biolaboratories Corp.) was added to each well for 1 h at 37°C. Detection was performed with tetramethyl-benzidine, dihydrochloride, dihydrate and hydrogen peroxide. The measurements were repeated in duplicate. The lowest concentration of HDGF detected by this system was 30 pg/ml. The colour intensity of each solution after development was quantified using a SpectraMax PLUS 384 microplate reader (Molecular Devices, Sunnyvale, CA).

Isolation of total RNA, synthesis of first-strand cDNA and PCR were performed as described previously (8) with slight modifications. Briefly, total RNAs from EBC-1, A549 and Lu99 cells were isolated using the TRIzol reagent (Enzo Life Sciences Inc.). First-strand cDNA was synthesised from total RNA (1.25 µg) using the PrimeScript RT Reagent Kit (Takara Bio, Inc.). PCR was performed with the synthesised cDNA products using TaqMan Gene Expression Master Mix (Thermo Fisher Scientific, Inc.) for each target gene. All reactions were carried out in triplicate. The sequences of the PCR primer pairs and fluorogenic probes used for HDGF, glyceraldehyde-3-phosphate dehydrogenase (GAPDH), 18S-rRNA, VEGF-A, brain-derived neurotrophic factor (BDNF), FGF-2 and FGF-5 are available on the Thermo Fisher Scientific website (http://www.thermofisher.com, HDGF assay ID: Hs00610314_m1; GAPDH assay ID: Hs99999905_m1; 18S-rRNA assay ID: Hs99999901_s1; VEGF-A assay ID: Hs00900054_m1; BDNF assay ID: Hs00380947_m1; FGF-2 assay ID: Hs00266645_m1; FGF-5 assay ID: Hs00170454_m1). The PCR products were analysed using the ABI 7500 real-time PCR system (Thermo Fisher Scientific. Inc.) and quantified by employing the 2^−ΔΔCq quantification method (9). As internal controls, the expression levels of HDGF mRNA in EBC-1 cells cultured with or without FBS were normalised to corresponding expression levels of GAPDH mRNA, and those of HDGF, VEGF-A, BDNF, FGF-2 and FGF-5 mRNAs in the NSCLC cell lines were normalised to the corresponding expression levels of 18S-rRNA.

Isolation of total RNA from EBC-1 and Lu99 cells was performed as described above. The degrees of RNA cross-linking and RNA degradation were analysed by electrophoresis using the Agilent 2100 Bioanalyzer (Agilent Technologies). Microarray analysis was performed with the 3D-Gene Human Oligo Chip 25k (Toray Industries Inc.). For efficient hybridisation, this microarray was designed with a columnar structure to stabilise spot morphologies and to enable micro-bead agitation. Total RNA (0.5 µg) isolated from EBC-1 and Lu99 cells was amplified using the Amino Allyl MessageAMP II aRNA Amplification Kit (Applied Biosystems). Amplified RNAs derived from EBC-1 and Lu99 cells (10 µg) were labeled with Cyanine 5 (Cy5) and Cy3, respectively. Purified Cy5- and Cy3-labeled aRNA pools (each 1 µg) were individually mixed with hybridisation buffer, and hybridised at 37°C for 16 h. The hybridisation was performed according to the manufacturer's instructions (www.3d-gene.com). The hybridisation signals were obtained using the 3D-Gene Scanner (Toray Industries Inc.) and processed using the 3D-Gene Extraction software (Toray Industries Inc.). The signals detected for each gene were normalised using global normalisation method.

Lu99 cells were trypsinised and spun down, the culture supernatants were aspirated, and the cell pellets were mixed with 0.258% type I collagen gels as described above. Collagen gel (42 or 350 µl) was added to each well of 96- or 24-well culture plates, respectively, and the plates were incubated at 37°C for 3 h to allow the collagen gel to solidify. The abovementioned tube-induction medium was added to each well for a total of 120 µl (96-well plates) or 1 ml (24-well plates) at cell densities of 0–5×10⁶ cells/ml for cell viability analysis or collecting 3D culture supernatants derived from Lu99 cells. Lu99 cells in 3D culture were incubated in 24-well culture plates for 24 h, each supernatant was then collected, and 500 µl of each supernatant was stratified in tube-induction medium to induce tube formation in sandwich culture described above. Cell viability was determined using the Cell Counting Kit-8 (Dojindo, Japan) as described previously (8). Cell viability assays in 3D culture were performed in triplicate. Protein concentrations in 3D culture supernatant derived from Lu99 cells cultured at a cell density of 2×10⁶ cells/ml (Lu99 supernatant) were determined using the Bradford method.

All data are presented as means ± standard errors of the means (SEMs, n=3) of three independent experiments. Differences in mean values among groups for multiple comparisons were subjected to one-way factorial ANOVA and subsequent Dunnett's test or Tukey's test, and were considered significant when P-values were <0.05 (*P<0.05, **P<0.01, ***P<0.005 and ****P<0.001) as indicated in the figures and legends by the relevant symbol. Statistical analyses were performed using JMP Pro 13 (SAS Institute Inc., Japan).

To identify angiogenic factors in NSCLC, we performed serum-free culture in this study, based on our previous study where angiogenic factors were identified on human mesothelioma cells (8). As shown in Fig. 1A, spheroid-like aggregation and cell shrinkage-like cell death were observed at 24 h after serum-free culture in human NSCLC cell lines, A549 and Lu99. Meanwhile, the serum-free culture induced neither the aggregation nor cell death in another NSCLC cell line, EBC-1 (Fig. 1A and B). These results indicate that EBC-1 cells can adapt to serum-free culture.

We examined the effects of serum-free culture supernatant derived from EBC-1 cells (EBC-1 supernatant) on angiogenesis. EBC-1 supernatant induced the tube formation of HUVECs (Fig. 1C) and increased the cell viability of HUVECs in 3D cultures (Fig. S1A) for 24 h in concentration-dependent manners. The numbers of HUVECs in monolayer cultures were significantly increased by EBC-1 supernatant after 48-h or 72-h incubations, but not after a 24-h incubation (Fig. S1B). These results indicate that EBC-1 supernatant induces tube formation and suggest that the increase in cell viability of HUVECs by EBC-1 supernatant is due to suppressed cell death rather than to promotion of cell proliferation.

We then examined VEGFR2 phosphorylation in HUVECs treated with EBC-1 supernatant. EBC-1 supernatant transiently phosphorylated VEGFR2 and ERK1/2 (Fig. 2A). An anti-VEGF-A antibody (10 µg/ml) significantly, but not completely, suppressed EBC-1 supernatant-induced tube formation (Fig. 2B). In addition, the antibody markedly suppressed the phosphorylation of VEGFR2, but not completely that of ERK1/2 (Fig. 2C). We also performed RNAi using a siRNA targeting VEGF-A (siVEGF-A) in EBC-1 cells. Treatment of EBC-1 cells with siVEGF-A did not induce cell death more significantly than those with vehicle or control siRNA (siControl, Fig. S2). ELISA for VEGF-A showed that the mean VEGF-A concentration in the supernatant (50 µg/ml) was 16.47±2.65 ng/ml (Fig. 3A). We previously reported that the anti-VEGF-A antibody (10 µg/ml) completely abrogated tube formation and VEGFR2 phosphorylation induced by VEGF-A (30 ng/ml) (8). The result of ELISA for VEGF-A and our previous findings indicate that the anti-VEGF-A antibody completely inhibited VEGF-dependent tube formation. ELISA also showed that siVEGF-A significantly abrogated VEGF-A secretion (Fig. 3A). Furthermore, the serum-free culture supernatant derived from EBC-1 cells transfected with siVEGF-A significantly, but not completely, suppressed tube formation (Fig. 3B). These results suggest that EBC-1 supernatant-induced tube formation is regulated by both VEGF-dependent and -independent pathways.

To explore humoral factors that regulate EBC-1 supernatant-induced tube formation independently of VEGF-A, we analysed proteins in the supernatant by using mass spectrometry (MS) and identified 1,007 proteins (Table SIII), most related to adhesion, cytoskeletal and nuclear proteins, indicating that there were microvesicles such as exosomes in the EBC-1 supernatant (10,11). Among the proteins, interleukin-8 (IL-8) (12), macrophage migration inhibitory factor (MIF) (13), galectin-1 (14), MK (15), IL-18 (16), galectin-3 (17), HDGF (18), osteopontin (OPN) (19), connective tissue growth factor (CTGF) (20) and GRN (21) are known to be involved in angiogenesis besides VEGF-A. We examined the involvements of IL-8, MIF, galectin-1, IL-18, galectin-3, OPN and CTGF in EBC-1 supernatant-induced tube formation using commercially available neutralising antibodies corresponding to each of the above proteins or their receptors. None of the antibodies suppressed EBC-1 supernatant-induced tube formation (Fig. S3), suggesting that IL-8, MIF, galectin-1, IL-18, galectin-3, OPN and CTGF are not directly involved in EBC-1 supernatant-induced tube formation.

We then performed RNAi using siRNAs targeting MK (siMK), HDGF (siHDGF) and GRN (siGRN) in EBC-1 cells and investigated the involvements of these three proteins in EBC-1 supernatant-induced tube formation. Each siRNA and siControl did not induce cell death in EBC-1 cells compared to vehicle in the 24-h serum-free cultures (Fig. S4A). We also confirmed that each siRNA abrogated secretion of the protein corresponding to the target gene (Fig. 4A). The supernatant of EBC-1 cells transfected with HDGF siRNA (siHDGF) significantly, but not completely, suppressed tube formation compared to those treated with vehicle or other siRNAs (Fig. 4B).

We further investigated the involvement of HDGF in EBC-1 supernatant-induced tube formation using another siHDGF (#761). siHDGF#761 did not induce cell death in EBC-1 cells compared to vehicle and siControl (Fig. S4B) and abrogated HDGF secretion in 24-h serum-free cultures (Fig. 5A). Similar to the result shown in Fig. 4B, the supernatant of EBC-1 cells transfected with siHDGF#761 significantly, but not completely, suppressed tube formation compared to those treated with vehicle or siControl (Fig. 5B). We also investigated the effect of HDGF on tube formation using rhHDGF. rhHDGF (10 ng/ml) significantly induced tube formation, but much less potently than did the EBC-1 supernatant (Fig. 5C).

ELISA for HDGF showed that the mean HDGF concentration in EBC-1 supernatant (50 µg/ml) was 1.67±0.26 ng/ml (Table SIV), while HDGF mRNA was expressed in the EBC-1 cells independently of the presence of FBS (Fig. 6A). In addition, the expression levels of HDGF mRNA in A549 and Lu99 cells were significantly higher than that in the EBC-1 cells (Fig. 6B). Furthermore, additions of the anti-VEGF-A antibody to the supernatants of EBC-1 cells transfected with siHDGFs (#724 or #761) suppressed tube formation more significantly than those of the antibody to the supernatants of the cells treated with vehicle or siControl (Fig. 6C). These results indicate that VEGF mainly regulates EBC-1 supernatant-induced tube formation and HDGF enhances VEGF-dependent tube formation.

To determine whether HDGF is involved in angiogenesis induced by the other NSCLC cell lines, we established a novel culture method; NSCLC cells were embedded in collagen gel and cultured three-dimensionally in tube-induction medium containing FBS (Fig. 7A). The 3D culture supernatants derived from Lu99 cells incubated at 0–5×10⁶ cells/ml for 24 h significantly induced tube formation in a cell density-dependent manner, whereas no significant differences were noted in tube formation when using different cell densities (2-5×10⁶ cells/ml; Fig. 7B). Thus, we used the supernatant derived from Lu99 cells cultured at a cell density of 2×10⁶ cells/ml (Lu99 supernatant) in subsequent experiments.

We performed RNAi in Lu99 cells using siHDGFs (#724 or #761) and investigated the involvement of HDGF in Lu99 supernatant-induced tube formation. Both siHDGFs (#724 or #761) had little effects on cell viability of Lu99 cells in 3D cultures (Fig. 7C) and on protein concentrations in Lu99 supernatants (Fig. 7D), and markedly abrogated HDGF secretion (Fig. 7E) compared to treatments with vehicle and siControl. However, the supernatants derived from Lu99 cells transfected with siHDGFs did not suppress tube formation (Fig. 7F), although the HDGF concentrations in EBC-1 and Lu99 supernatants were almost the same (Table SIV and Fig. 7E). The reason why HDGF was not directly involved in Lu99 supernatant-induced tube formation was presumed to be due to a decrease in HDGF concentration by stratification of the Lu99 supernatant in the tube-induction medium (Fig. 7E). These results suggest an involvement of more potent angiogenic factors than HDGF in Lu99 supernatant-induced tube formation.

We also investigated the expression levels of VEGF-A mRNA in EBC-1, A549 and Lu99 cells. Intriguingly, VEGF-A mRNA expression in Lu99 cells was significantly weaker than those in EBC-1 and A549 cells (Fig. 8A). The Lu99 supernatant did not induce VEGFR2 phosphorylation (Fig. 8B). In addition, the anti-VEGF-A antibody failed to suppress Lu99 supernatant-induced tube formation (Fig. 8C). These results indicate that the Lu99 supernatant induces HDGF- and VEGF-independent tube formation.

To explore the humoral factor that regulates Lu99 supernatant-induced tube formation, we comprehensively analysed mRNA expression in Lu99 and EBC-1 cells by using gene microarray, and thereby identified 61 mRNAs expressed in Lu99 cells, but not in EBC-1 cells (Table SV). Among these mRNA-encoded proteins, brain-derived neurotrophic factor (BDNF) (22), FGF-2 (23) and FGF-5 (24) are secreted extracellularly and are known to be involved in angiogenesis. The expression levels of BDNF, FGF-2 and FGF-5 mRNAs were significantly higher in Lu99 cells than in EBC-1 and A549 cells (Fig. 9A). We then examined the involvements of BDNF, FGF-2 and FGF-5 in Lu99 supernatant-induced tube formation using commercially available neutralising antibodies corresponding to each of the above proteins. Consequently, Lu99 supernatant-induced tube formation was not suppressed by antibodies to BDNF and FGF-5, but was markedly inhibited by that to FGF-2 (Fig. 9B). As expected, rhFGF-2 (≥3 ng/ml) induced tube formation as potently as did Lu99 supernatant (Fig. S5). We further confirmed the involvement of FGF-2 in Lu99 supernatant-induced tube formation using another monoclonal anti-FGF-2 neutralising antibody. Similar to the result shown in Fig. 9B, the antibody significantly inhibited Lu99 supernatant-induced tube formation (Fig. 9C). These results indicate that FGF-2 regulates HDGF- and VEGF-independent tube formation induced by Lu99 supernatant.