We obtained primary antibodies against the following proteins from commercial sources: p-Smad2 (Ser245/250/255), p-Smad2 (Ser465/467), Smad2, p-Smad3, Smad3, p-p38MAPK, p-Akt, Akt, p-JNK, JNK, p-ERK, p-c-Jun and c-Jun (Cell Signaling Technology, Danvers, MA, USA); e-cadherin, fibronectin, collagen type 1 A2, p38MAPK, ERK and TβRI (Santa Cruz Biotechnology, Santa Cruz, CA, USA); N-cadherin, SNAIL, SLUG and ZEB1 (BD biosciences, Palo Alto, CA, USA); vimentin and α-SMA were obtained from Sigma (St. Louis, MO, USA); p-FGFR3, FGFR3 and TβRII (Abcam, Cambridge, MA, USA); and GAPDH (AbFrontier, Seoul, Korea). Horseradish peroxidase (HRP)-conjugated secondary antibodies were purchased from Thermo (Pittsburgh, PA, USA). For use in the in vitro invasion assay, Transwells were purchased from Corning and Matrigel was purchased from BD Biosciences. IM-412 (C₁₆H₁₄ClN₅O₂, MW: 344; ID 9073517) was isolated from a chemical library of screening compounds and was purchased from ChemBridge Corporation (San Diego, CA, USA).

The human breast cancer cell line MDA-MB-231 was purchased from the American Type Culture Collection (ATCC; Manassas, VA, USA). The cells were maintained at 37°C in RPMI-1640 supplemented with 10% fetal bovine serum (Gibco-BRL, Grand Island, NY, USA), 100 U/ml penicillin and 100 µg/ml streptomycin in a humidified atmosphere containing 5% Co₂.

Cell viability was assessed using the 3-[4,5-dimethylthiazol-2-yl]-2,5-diphenyltetrazolium bromide (MTT) assay (Sigma). MDA-MB-231 cells were plated at a density of 8×10³ cells/well into 24-well plates (SPL, Gyeonggi, Korea), and the seeded cells were stabilized for 24 h, and were subsequently treated with each indicated dose of chemicals. MTT was directly added to each well at a final concentration of 0.5 mg/ml, and after 4 h, the medium was removed, the formazan crystals in MDA-MB-231 cells were dissolved in dimethylsulfoxide and the absorbance of the formazan solution was measured using a microplate reader (Multiskan EX; Thermo LabSystems, waltham, MA, USA) equipped with a 540-nm filter. each sample was assayed in triplicate, and the experiment was repeated thrice.

Cells were seeded into 6-well plates (SPL) at 90% confluency (5×10⁵ cells/well) and incubated overnight. On the next day, the culture medium was aspirated, and the cells were scratched using a 200-µl pipette tip. After wounding, the cultures were washed with phosphate-buffered saline (PBS) and treated with IM-412. The cells were allowed to migrate for 24 h and were subsequently stained with crystal violet. Migration patterns were examined under a microscope and photographed at a magnification of ×40.

The invasiveness of the cultured cells was determined using a 24-well Transwell system (pore size, 8-µm) (Corning Incorporated, Corning, NY, USA). The upper side of the Transwell membrane was coated with 1 mg/ml Matrigel (in 10 µl) and then cells were seeded into the upper chamber at a density of 5×10⁴ cells in 100 µl of media; the complete medium was added to the lower chamber. Cells were incubated for 24 h at 37°C in 5% Co₂, and then the cells that had not penetrated the filter were wiped away using cotton swabs; the cells that had invaded the lower surface of the filter were fixed with methanol, stained with crystal violet, examined under a microscope, and photographed at a magnification of ×40. The numbers of invaded cells were determined by counting the cells in 5 randomly selected regions.

Cells were lysed with RIPA buffer (50 mM Tris-Cl, pH 7.4, 1% NP-40, 150 mM NaCl, 1 mM EDTA) supplemented with protease inhibitors (1 mM phenylmethylsulfonyl fluoride, 1 µg/ml aprotinin and 1 µg/ml leupeptin) and phosphatase inhibitors (1 mM Na₃VO₄ and 1 mM NaF). Proteins in the whole-cell lysates were separated on 8–15% SDS-polyacrylamide gels and transferred to nitrocellulose membranes (Bio-Rad, Hercules, CA, USA). The membranes were blocked (1 h) with 5% skim milk in tris-buffered saline containing 0.1% Tween-20 and then probed with primary antibodies (overnight, 4°C). After multiple washes, the membranes were incubated with HRP-conjugated secondary antibodies, and then the immunoreactive bands were detected using enhanced chemiluminescence reagents according to the manufacturer's recommendations (GE Healthcare; Little Chalfont, UK). The experiments were repeated at least thrice.

MDA-MB-231 cells (5×10⁵ cells) were grown on glass coverslips, stabilized for 24 h and then treated for 24 h with the indicated doses of IM-412. Next, the cells were washed with Hank's balanced salt solution (HBSS), fixed with 3.7% formaldehyde in HBSS for 10 min at room temperature, and then washed and permeabilized with 0.1% Triton X-100 in HBSS for 15 min at room temperature. Mouse α-SMA antibody (1:500) was used as the primary antibody and it was detected using an Alexa 546-conjugated secondary antibody. Nuclei were counterstained with 300 ng/ml 4′,6-diamidino-2-phenylindole (DAPI; Invitrogen, Carlsbad, CA, USA). Coverslips were mounted on glass slide with Gel/Mount™ (biomeda, Foster City, CA, USA), and images were obtained using a laser-scanning confocal microscope (LSM710; Carl Zeiss, oberkochen, Germany) at a magnification of ×400.

We prepared 3D cultures of cells as described with a few modifications (17). To mimic the in vivo environment, a type I collagen (Col I) gel matrix was reconstituted according to the manufacturer's specifications (Nitta Gelatin, Osaka, Japan), and then 250 µl of this mixture was plated into 12-mm polycarbonate filter chambers (3.0 µm Millicell; Millipore, Temecula, CA, USA). MDA-Mb-231 cells were seeded at a density of 2×10⁵ cells/chamber and cultured in the submerged state in culture medium for 7 days, and then in the air-liquid interface state for 24 h. During this period of air-liquid interface culture, cells were concurrently treated with or without IM-412. After the 1-day air-liquid interface culture, the 3D cultures were fixed in Carnoy's solution (ethanol:chloroform:acetic acid, 6:3:1) for 30 min at 4°C, and then the fixed samples were embedded in paraffin and sectioned (3 µm). For staining, sections were deparaffinized and the endogenous peroxidase activity was blocked. Mouse α-SMA antibody (1:500) was used as the primary antibody and the sections were developed using the Cap-Plus™ Detection kit (Invitrogen); after the development step, samples were counterstained with hematoxylin (Dako, Carpinteria, CA, USA). The stained samples were examined under a microscope and photographed at a magnification of ×200.

We performed calculations using Discover 2.98/Insight II with CVFF force field as described by Hagler et al (18). The crystal structure of FGFR3 bound to an ATP mimetic molecule (phosphomethylphosphonic acid adenylate ester) was used as the computational model (4K33.pdb). IM-412 was built from the coordinates of an adenine group, as found in the crystal structure using the Builder Module in Insight II. The computational complex model was initially minimized using 1,000 steps of steepest decent and 3,000 steps of conjugated gradient with a 14.0 Å non-bonded cut-off distance.

Data are represented as means ± SD. Statistical comparisons between groups were performed using the Student's t-test. Data were considered statistically significant at P<0.05, P<0.01 or P<0.001 as indicated in the figure legends.

TGF-β is widely recognized to exhibit pleiotropic activity in biological processes ranging from development, proliferation, differentiation, angiogenesis and immune regulation to tumor biology. Moreover, TGF-β signaling plays a prominent role in EMT (1,3,19). Since our previous results demonstrated that the novel compound IM-412 effectively inhibited TGF-β signaling in human lung fibroblasts (16), we aimed to investigate whether IM-412 prevents the migration and invasion of cancer cells. First, we examined the cytotoxic effect of IM-412 on MDA-MB-231 cells, which are highly metastatic human breast carcinoma cells. At concentrations above 2 µM, IM-412 markedly reduced the viability of the cells, but at concentrations below 2 µM, the viability levels were the same as those in untreated control cells (Fig. 1A). Next, the changes in cell motility induced by IM-412 were investigated using the wound migration (Fig. 1B) and the Transwell invasion assays (Fig. 1C) under comparatively less cytotoxic conditions. IM-412 drastically reduced the abilities of migration and invasion of MDA-MB-231 cells in a dose-dependent manner.

To evaluate the capacity of IM-412 to affect EMT-related markers during MDA-MB-231 cell invasion, we determined the expression levels of EMT-associated molecules by performing western blot analysis (Fig. 2A). IM-412 markedly reduced the expression of mesenchymal markers such as fibronectin, collagen type 1 A2 and α-SMA (ACTA2; α-smooth muscle actin). The expression level of N-cadherin was also decreased following IM-412 treatment, but increased expression of E-cadherin, an epithelial marker, was not detected in the MDA-MB-231 cells, which were previously reported to be E-cadherin-negative cells (20). The expression of vimentin, another well-known EMT marker, also remained unchanged in response to IM-412. In close agreement with these results, the expression of EMT-activating transcription factors such as SNAIL, SLUG and ZEB1 was drastically suppressed by IM-412 in a dose-dependent manner.

We also verified the effect of IM-412 on the downregulation of the representative molecule α-SMA by performing immunocytochemical and immunohistochemical staining. At 2 µM, IM-412 significantly downregulated the cellular expression of α-SMA (Fig. 2B). Immunohistochemical analysis of 3D collagen-gel-cultured MDA-MB-231 cells revealed numerous invaded cells in the collagen matrix. However, the cells treated with IM-412 did not exhibit invasive activity and expressed α-SMA at a low level (Fig. 2C). These results indicated that IM-412 effectively suppresses cell motility by inhibiting the expression of EMT-associated transcription factors and mesenchymal markers.

To investigate the molecular mechanism underlying IM-412 inhibition of cancer-cell invasion, we first examined the major TGF-β downstream signaling pathway, the Smad-dependent pathway. As expected, IM-412 significantly suppressed the phosphorylation of both Smad2 and Smad3 in a dose-dependent manner (Fig. 3A). Since the non-canonical signaling pathway of TGF-β has also been clearly shown to fully activate the EMT process (1–3), we examined how non-Smad signaling was affected by IM-412; the phosphorylation levels of p38MAPK, Akt and JNK were significantly decreased by IM-412 in a concentration-dependent manner, but the level of ERK phosphorylation was not affected (Fig. 3B). To determine whether IM-412 influences the upstream molecules of the TGF-β signaling pathways, we examined the expression of TβRII and TβRI. In contrast to our previous findings in lung fibroblasts, IM-412 did not alter the expression of TβRII and TβRI (Fig. 3C).

In addition to TGF-β-family proteins, several other growth factors can induce partial or full EMT by activating receptor tyrosine kinases and triggering downstream signaling cascades (2,6); among the receptor tyrosine kinases, FGFRs, in particular, are involved in diverse cellular processes, including metastasis and EMT (10). In our unpublished data, we demonstrating that another derivative compound of IM-412 [one fluoryl and one methyl group was added to IM-412, 3-(2-chloro-6-fluorobenzyl)-1,6,7-tri-methyl-1H-imidazo[2,1-f]purine-2,4(3H,8H)-dione] exhibited the highest inhibitory activity on FGFR3 among 65 kinases and 2-fold higher inhibitory activity against FGFR3 compared with FGFR1 in an in vitro kinase assay. Therefore, we focused on FGFR3 as it appeared to be the most probable target molecule of IM-412. We employed computational prediction to identify the binding site of FGFR3 for IM-412 (Fig. 4A). The molecular-modeling results obtained for the FGFR3/IM-412 complex showed the two strong hydrogen bonds between the N-H imidazo group of IM-412 and the C=O backbone of Ala558 (2.0 Å), the nitrogen of the purine dione group, and the N-H backbone of Ala558 (2.14 Å) at the hinge backbone of FGFR3. In the structure, the 2-chlorobenzyl is directed toward the gatekeeper residue Val555. Hydrophobic interactions were observed between the benzyl ring and the gatekeeper residue Val555 in addition to the several hydrophobic interactions observed with the 2-chloro substitutions (Lys508 and Ala634). Thus, we considered the imidazopurine structure of IM-412 to be highly active in blocking the ATP-binding-pocket region of FGFR3.

To confirm the potential ability of IM-412 to block the activation of FGFR3, we examined the levels of FGFR3 and phosphorylated FGFR3. The phosphorylation of FGFR3 was significantly inhibited by IM-412, while the expression of FGFR3 protein was not changed (Fig. 4B).