Deregulation of AP-1 Proteins in Collagen Gel-induced Epithelial Cell Apoptosis Mediated by Low Substratum Rigidity*

In this study, we established that collagen gel, but not collagen gel coating, induced apoptosis exclusively in epithelial cell lines, which indicated that low substratum rigidity might trigger cell apoptosis. To confirm this, we used collagen gels with different rigidities due to cross-linking or physical disruption of collagen fibrils caused by sonication. We found that collagen gel-induced apoptosis was inversely correlated with substratum rigidity. Low substratum rigidity collagen gel-induced apoptosis was neither prevented by Bcl-2 overexpression nor preceded by mitochondrial release of cytochrome c. This suggested that the mitochondrial pathway was not involved in low substratum rigidity-induced apoptosis. Low substratum rigidity activated c-Jun N-terminal kinase (JNK) within 4 h, but it also rapidly down-regulated c-Jun within 1 h and triggered persistent aberrant expression of c-Fos for at least 24 h. Either reduced c-Jun expression or c-Fos overexpression induced apoptosis in several epithelial cells. Inhibiting low substratum rigidity-induced JNK activation prevented aberrant c-Fos expression but only partially blocked low substratum rigidity-induced apoptosis. Taking these results together, we conclude that low substratum rigidity collagen gel induced apoptosis in epithelial cells and that deregulated AP-1 proteins mediated that apoptosis, at least in part.

The physical properties of extracellular matrix influenced the locomotion (1, 2), survival (3), gene expression, and proliferation (4) of cells. Several methods have been developed to study the effects of the mechanical properties of extracellular matrix on cell behaviors. Type I collagen-coated polyacrylamide gel with different rigidities was used to study the functions of substratum flexibility on cell responsiveness, such as propul-sive forces, migration, and apoptosis (2,5). Elastic micro-patterned substrates were used to study force generation and focal adhesion assembling in cells (6). In addition, a micro-fabrication strategy to generate micrometer-scale rods of an elastomeric polymer, polydimethylsiloxane, was used to study the traction force generated by cells (7).
Fibril collagen is the most abundant extracellular matrix in normal interstitium. Due to its fibril nature, collagen gel is the most commonly used three-dimensional scaffolding material for biological studies. In the past, we used hydrated fibril collagen as a three-dimensional culture model to study cystogenesis and branching tubulogenesis (8 -11). We frequently found that epithelial cells developed apoptosis in collagen gel (8). In addition, we found that collagen gel overlay induced apoptosis only in epithelial cells (12). Subsequent findings indicated that collagen gel overlay triggered FAK proteolysis, which was associated with MDCK 3 cell apoptosis (13). Here we examined whether the physical properties of collagen gel triggers cell apoptosis. Collagen gel extracted from the tail tendons of older rats showed more cross-linking of collagen fibrils than those from younger rats (10,14). To prepare collagen gel with different physical properties, we extracted collagen from the tail tendons of rats of different ages, and then we modified the collagen gel rigidity using sonication (15,16). Collagen gel is a viscoelastic material with non-linear elasticity whose shear mechanical properties are assessable using a rheometer, a well established tool (17,18). We used modified methods of parallel plating to determine the viscoelastic modulus of collagen gel.
To elucidate the signal mechanism underlying low substratum rigidity-induced epithelial cell death, we investigated the traditional cell-death pathway. The mitochondrial pathway did not seem to be involved, however. On the other hand, we found that low substratum rigidity activated the JNK group of MAP kinases, also known as stress-activated protein kinases. Previous studies (19 -21) reported that the JNK protein kinases * This work was supported by the National Health Research Institute (Grant phosphorylated transcriptional activation domains of the AP-1 family proteins. The AP-1 family consists of several groups of bZIP domain (bZIP is basic region leucine zipper) proteins: the Jun, Fos, and ATF-2 subfamilies (22). C-Jun overexpression induced apoptosis in 3T3 fibroblasts (23). In addition, JNK/c-Jun signaling was necessary for the apoptotic response in certain neuronal cell types (24,25). On the other hand, c-Jun knock-out mouse embryonic fibroblast cells exhibited defects in proliferation and increases in UV-induced apoptosis (26,27). Consistent with cell culture studies, the lack of c-Jun resulted in massive apoptosis of hepatoblasts and erythroblasts in the developing mouse liver in vivo (28). The c-fos proto-oncogene, the other major AP-1 protein member, encoded a nuclear protein that dimerized with Jun family proteins (29 -31) to form a transcription factor complex (32). The Fos protein has been implicated as a key molecule in cell proliferation (22,32), differentiation (33,34), and transformation (35). In addition to a primary role in normal development and cellular growth, c-Fos protein has been associated with apoptotic cell death induced by anti-proliferative conditions (36,37). In this study, we explored the novel role of AP-1 deregulation in low rigidity collagen gelinduced epithelial cell apoptosis.

EXPERIMENTAL PROCEDURES
Cell Lines and Cultures-MDCK, LLC-PK1, BS-C-1, NMuMG, NRK-52E, HK-2, HEK 293, and Chang liver cells were purchased from ATCC and regularly maintained in Dulbecco's modified Eagle's medium (DMEM) supplemented with 10% fetal calf serum (FCS). BS-C-1 cells were cultured in modified Eagle's medium (MEM) supplemented with 10% FCS. HK-2 cells were cultured in MEM plus methanol-soluble factor with 5 ng/ml EGF, 40 ng/ml bovine pituitary extract and supplemented with 10% FCS. Chang liver cells were cultured in MEM supplemented with 10% FCS. Bovine aortic endothelial cells (BAECs) were a gift from Dr. Wen-Chan Chang, NIH/3T3 cells were a gift from Dr. Hsiao-Shen Liu, U-373MG were a gift from Dr. Jih-Ing Chaung, and four oral cancer cell lines (OC-2, DOK, SSC-25, and HSC-3) were gifts from Dr. Dar-Bin Shieh (all from National Cheng Kung University Medical College, Tainan, Taiwan). BAECs, NIH/3T3, and U-373MG cell lines were cultured in DMEM supplemented with 10% FCS. OC-2 cells were cultured in RPMI medium supplemented with 10% FCS. DOK cells were cultured in DMEM with 2 mM glutamine and 5 g/ml hydrocortisone and supplemented with 10% FCS.
Preparation of Hydrated Collagen Gel-Type I collagen was prepared from rat tail tendons according to the established procedure (10,38). The final concentration of type I collagen stock was 1% (wet weight) dissolved in 0.025 N acetic acid. To prepare collagen gel, 3 vol of collagen stock was mixed with 5.7 ϫ DMEM (1 vol), 2.5% NaHCO 3 (0.5 vol), 0.1 M HEPES (1 vol), 0.17 M CaCl 2 (0.1 vol), 1 N NaOH (0.1 vol), and 4.3 vol of 1ϫ culture medium (DMEM plus 10% FCS) under chilled conditions. The mixture was dispensed on culture dishes and kept at room temperature to allow gelation. After gelation, each culture was overlaid with 2 ml of culture medium that was replaced every other day. To prepare collagen gel-coated dishes, collagen gel mixtures were dispensed in culture dishes and then aspirated to allow only a thin sheet of collagen to cover the dish. The collagen gel-coated dishes were semi-air-dried in a culture hood. Every culture plate was freshly prepared before it was used in an experiment.
DNA Extraction and Electrophoresis-The method of extracting low molecular genomic DNA has been previously described (39). Briefly, cultured cells were extracted with 0.5% Triton X-100, 10 mM EDTA, and 10 mM Tris (pH 7.4) and phenol/chloroform three times. The DNA was precipitated in propanol and electrophoresed in 1.5% agarose gel. Finally the DNA was visualized using ethidium bromide staining under UV light.
Hoechst 33258 Staining and Assessing Nuclear Areas-LLC-PK1 cells cultured on culture dishes or collagen gel for 24 h were washed twice with PBS and then fixed with 2% paraformaldehyde in the gel. After they had been washed, the cells were permeabilized using buffer containing 0.1% Triton X-100 and then stained with Hoechst 33258 (5 mg/ml) for 1 h in the dark. Finally, the stained nuclei were visualized under a fluorescence microscope (BX-51, Olympus, Tokyo, Japan). To assess nuclear areas, cells were cultured under different conditions for 24 h and subjected to Hoechst 33258 staining, and then the nuclear area was evaluated under a fluorescence microscope with imaging software (Image-Pro Plus 6.0, Media Cybernetics, Inc., Silver Spring, MD).
Analysis of Apoptosis Ratio-The apoptosis ratio was assessed using flow cytometry with propidium iodide as previously described (8,12). After cells were seeded on collagen gel for different time periods, the gel was removed and treated with 0.2% collagenase at 37°C for 10 min. The cell mixtures collected were washed with PBS and fixed in 70% alcohol. After fixation, the cells were treated with RNase (100 mg/ml PBS) and stained with propidium iodide (40 mg/ml PBS). The mixed cells were incubated in the dark at room temperature for 30 min and analyzed using flow cytometry (FACScan, BD Biosciences) with excitation set at 488 nm. Data were analyzed using CellQuest software.
Western Blotting-We used the following antibodies to assess protein levels: caspase 2 (Santa Cruz Biotechnology), pro-caspase 3 (Upstate Biotechnology), Bcl-2 (DAKO), cytochrome c (BD Pharmingen), cleaved caspase 3 and caspase 9 (Cell Signaling Technology), caspase 8 (Oncogene), JNK and phosphorylated JNK (Cell Signaling), c-Jun (BD Transduction), c-Fos (Upstate Biotechnology), and horseradish peroxidaseconjugated goat anti-rabbit and anti-mouse IgG (Santa Cruz Biotechnology). Protein concentration was measured using the Lowry method (40). For Western blotting, 30 or 50 g of cell homogenate from specific samples was resolved using 10% SDS-PAGE and electrophoretically blotted onto nitrocellulose paper. The nitrocellulose paper was incubated with the specific antibody listed above, and then immunocomplexes were detected using horseradish peroxidase-conjugated IgG, and, finally, the immunocomplexes were made visible using fluorography with an enhanced chemiluminescence detection kit (ECL, Amersham Biosciences).
Treatment with Caspase or MAP Kinase Inhibitors-Cells were pretreated for 1 h and then seeded onto a culture dish, a collagen gel-coated dish, or on collagen gel with the caspase inhibitors z-VAD and DEVD-fmk (Calbiochem-Novabiochem) and JNK inhibitor SP600125 (Tocris). At various time points, cells were harvested and the apoptosis ratio was assessed using FACScan analysis.
Scanning Electron Microscopy-LLC-PK1 cells cultured for 48 h on a dish with normal culture, a collagen gel-coated dish, or collagen gel were rinsed twice with PBS solution. The cells were fixed with 2% buffered paraformaldehyde for 1 h. After fixation, the cells were rinsed twice with PBS to remove paraformaldehyde. Samples were dehydrated using incubation with gradient alcohol from 50% to 95% for 10 min under each concentration. In the final step, absolute alcohol was used to complete dehydration three times for 5 min. Samples were critical point-dried in liquid CO 2 solution and then coated with a thin layer of gold-particle film. Finally the samples were visualized under a scanning electron microscope (SEM, Hitachi S5200).
Quantification of Collagen Gel Rigidity-The detailed protocol of the quantification of collagen gel rigidity was previously described (41). In brief, the viscoelastic properties of collagen gel were analyzed using a rheometer (AR1000, TA Instruments Ltd., West Sussex, UK) using cone-plate geometry with a 2°a ngle and a 60-mm diameter cone. To avoid the fracture of collagen gel and to increase the adhesion force between the collagen fiber network and the cone-plate geometry, neutralized collagen solution was first poured onto a pre-cooled plate disk at 5°C, and then the cone was moved to approach the solution. After connecting the cone and solution, the temperature was shifted to 37°C for 25 min to allow the collagen solution to gel. While we measured the viscoelasticity of the gel, the temperature was held at 25°C. The dynamic shear storage modulus (GЈ), also called dynamic rigidity (42), and loss modulus (GЉ) of gels were acquired with angular frequency ranging from 0.1 to 10 rads/s. The maximum strain was set at 5% according to the requirements, and as suggested (43), of within a 10% linear viscoelastic range for shear measurements.
Plasmid Constructs and Transfections-The cDNA encoding c-Jun or c-Fos was generated using PCR from a 1-month-old mouse-cerebrum cDNA library. Plasmid expressing full-length c-Jun or c-Fos was constructed using standard molecular cloning techniques. The expressing sequence was constructed by inserting pEGFP-N1 (BD Biosciences Clontech). To generate the EGFP expression vector, c-Jun/EGFP, the pEGFP-N1 was digested with EcoRI and HindIII. For c-Fos/EGFP vector, the pEGFP-N1 was digested with EcoRI and SacII. Plasmid was amplified in Escherichia coli using transfection and purified using plasmid extraction kits (Plasmid Midi Kit, Qiagen). Cells were transfected with c-Jun/EGFP, or c-Fos/EGFP plasmid using Lipofectamine 2000 (Invitrogen) according to the manufacturer's instructions.
Construction of an shRNA Expression System-Plasmid expressing shRNA was constructed using standard molecular cloning techniques. The pSUPER/EGFP-expressing enhanced green fluorescence protein (EGFP) was constructed by inserting the EGFP expression cassette from pEGFP-N1 (BD Bio-sciences Clontech) into pSUPER vector (44) (kindly provided by Dr. R. Agami, The Netherlands Cancer Institute, Amsterdam, The Netherlands). To generate the shRNA expression vector pSUPER-Mmc-Jun/EGFP, the pSUPER/EGFP was digested with BglII and HindIII, and an annealed oligonucleotide duplex against the sequence of 5Ј-CGCAGCAGTTG-CAAACGTT-3Ј for murine c-Jun (45) was ligated into the vector.
Confocal Microscopy and Real-time Fluorescence Images-LLC-PK1, NMuMG, or HeLa cells were transiently transfected with EGFP-containing plasmid for the indicated time points. For a vital stain assay, cells were pre-stained with propidium iodide (200 g/ml in culture medium) for 30 min before fixation. For fluorescence images, cells were rinsed twice with PBS solution, fixed with 4% buffered paraformaldehyde for 15 min, and then blocked with super blocking buffer (Pierce) for 1 h. To detect c-Jun expression, cells were stained with mouse anti-c-Jun antibody (BD transduction) for 1 h. Cells were rinsed five times with PBS. Finally, the cells were stained with anti-mouse IgG conjugated with Alexa 594 (Molecular Probes) for 1 h. The fluorophore was excited using a laser at 488 or 594 nm and detected at emission spectrum (520 or 640 nm) using a photomultiplier with a ϫ40 water-immersion objective lens (HCX APO L ϫ40/0.90 W-U-V-1) for display as a high resolution image. Finally the stained results were visualized under a confocal microscope (SP-2, Leica Mikrosysteme Vertrieb GmbH, Bensheim, Germany). Images were taken using Leica SP-2 software. For time-lapse fluorescence recording, NMuMG cells were transiently transfected with shRNA-containing EGFP expression plasmid and recovered 4 h after transfection. The cells were moved from an incubator to a culture system on confocal microscopy (FV1000, Olympus), in which the temperature was kept at 37 Ϯ 0.5°C, and 5% CO 2 was injected into PBS to maintain the CO 2 concentration and moisture in the chamber. The EGFP was excited using an argon laser at 488 nm and detected at emission spectrum (520 nm) using a photomultiplier with a ϫ40-inverted objective lens (Uplan Apo ϫ40/1.0 oil, Olympus) to display as a high resolution image for 10-min intervals (scan speed: 3.928 s/frame). Images were taken or composed (video) using imaging software (FV10-SW, Olympus).
Statistics-All data are expressed as means Ϯ S.E. of at least three independent experiments. One-way analysis of variance was used to test for statistical differences. Statistical significance was set at p Ͻ 0.05.

Characterization of LLC-PK1 Cell Death Induced by Fibril
Collagen Gel-LLC-PK1 and HEK 293 cells were cultured for 24 h on normal culture dishes or 0.3% type I collagen gel prepared from the tail tendons of 1-month-old rats. Collagen gel did not alter the morphology of HEK 293 cells. However, LLC-PK1 cells cultured on collagen gel showed distinct morphological changes. These cells formed contracted cell islands with many disintegrated vesicles (Fig. 1A). An analysis of the DNA extracted from LLC-PK1 cells cultured on collagen gel showed a DNA ladder pattern, indicating that collagen gel-induced cell death was apoptosis (Fig. 1B). In addition, an assessment of the

AP-1 Deregulation in Low Rigidity-induced Apoptosis
integrity of cell nuclei using Hoechst 33258 staining revealed nuclear condensation as well as fragmentation in LLC-PK1 cells cultured on collagen gel (Fig. 1C). The results of the FACScan analysis of the nuclear size also showed, after 48 h, a significantly (p Ͻ 0.001) increased (56.4 Ϯ 9.8%) sub-G 0 population in LLC-PK1 cells cultured on collagen gel, but only a Ͻ7% increase in cells cultured on a normal dish (Fig. 1D). To examine whether other types of extracellular matrix had similar effects, LLC-PK1 cells were cultured on dishes pretreated with fibronectin, laminin, vitronectin, or Matrigel for 48 h, and the apoptosis ratio was assessed. None of these extracellular matrices triggered apoptosis in LLC-PK1 cells (data not shown). However, when LLC-PK1 cells were cultured on 0.3% type III collagen gel for 48 h, they displayed approximately similar levels of apoptosis (Fig. 1E). Because both type I and type III collagen belong to the fibril collagen family, these data suggested that collagen gel-induced apoptosis may be caused by the physical nature of the three-dimensional gel. To determine whether collagen gel-induced apoptosis was cell-type specific, we cultured epithelial (NMuMG, BS-C-1, MDCK, and NRK-52E), endothelial (BAECs), mesenchymal (HEK 293 and NIH-3T3), and tumor cells (HK-2, Chang liver, U-373 MG, OC-2, DOK, SSC-25, and HSC-3) on type I collagen gel for 48 h and assessed the apoptosis ratio using FACScan analysis. We found that only polarized cells, including epithelial and endothelial cells, had elevated levels of apoptosis on collagen gel (Fig. 1F). In contrast, mesenchymal as well as tumor cells showed little apoptosis when they were cultured on collagen gel.
Collagen Gel Coating Alleviated Collagen Gel-induced Apoptosis-To determine whether collagen gelinduced apoptosis in epithelial cells was caused by its physical or chemical nature, we used collagen gelcoated dishes that exerted a chemical impact similar to that of collagen gel but were physically quite rigid, as previously described (46). LLC-PK1 cells cultured on these dishes were mostly normal epithelial cell-shaped with reasonable extension. Although some of LLC-PK1 cells were elongated and became spindleshaped, they were relatively more extended than the cells with retracted morphology induced by collagen gel (Fig. 2A, panels  a-c). To investigate the cell morphology in more detail, we used an SEM. We found that cells grown on normal culture or collagen gel-coated dishes mostly spread well with little sign of apoptosis ( Fig. 2A, panels d and e). In contrast, cells grown on collagen gel exhibited more limited extension and a significant pulling of collagen fibrils. In addition, these cells also disintegrated into apoptotic bodies ( Fig. 2A, panel f). We examined the ultrastructure of collagen fibrils in collagen gel-coated dishes and collagen gel using SEM. Basically, the diameters of the collagen fibrils and the levels of cross-linking were not different under these two conditions. However, collagen gel coating showed only a few layers of collagen fibrils, whereas collagen gel showed multiple layers of collagen fibrils with greater depth (Fig. 2B). We assessed the growth characteristic as well as apoptosis ratio of LLC-PK1 cells cultured under three conditions. Although collagen gel coating reduced the growth rate of LLC-PK1 cells, collagen gel completely blocked cell growth. Cell viability was significantly higher in cells cultured on collagen gel-coated dishes than on collagen gel (Fig. 2C). More importantly, collagen gel coating significantly (p Ͻ 0.001) inhibited collagen gel-induced apoptosis (Fig. 2D). Taking these data together strongly suggested that the low substratum rigidity physical property of collagen gel caused collagen gel-induced apoptosis.
Age-augmented Cross-linking of Collagen Fibril Alleviated Collagen Gel-induced Apoptosis-We attributed the crosslinking of collagen fibril to the substratum rigidity of collagen gel. Collagen gels prepared from rat tail tendons of different ages have different mechanical properties due to the cross-links of collagen fibrils (9,14). To determine whether the age of the collagen fibril affects collagen gel-induced apoptosis, we cultured LLC-PK1 cells on 0.3% collagen gels prepared from the tail tendons of 1-, 4-, or 16-month-old rats. Cells cultured on 1-month-old collagen gel showed active pulling of collagen fibrils, indicating that cross-links of collagen fibrils might be disrupted by cell retraction. The cell pulling of collagen fibrils was clearly reduced by 4-month-old collagen gel (Fig. 3A). The morphology of cells cultured on collagen gel of different ages was further examined using SEM, which showed that LLC-PK1 cells cultured on 1-month-old collagen gel formed a cell island with little cell extension (Fig. 3B). We also observed cells actively pulling collagen fibrils. However, cells cultured on 4-month-old collagen gel showed cell extension manifested by the extensive lamellipodia and membrane raffle on the outer rim of the cell island. In addition, a FACScan analysis demonstrated that the age of the tail tendon used to make the collagen gel was inversely related to the level of collagen gel-induced apoptosis (Fig. 3C). These findings indicated that augmentation of the rigidity of collagen gel produced by an increase in the cross-links of collagen fibers significantly inhibited collagen gel-induced apoptosis.
Reduction of the Rigidity of Collagen Gel by Sonication Increased Collagen Gel-induced Apoptosis-Sonication may affect the rigidity of collagen gel by disrupting the structure of collagen fibrils (16). To test this possibility, we treated collagen stock solutions with sonication of different durations (1, 2, and 4 min) and then allowed them to gel. We found that the gel time rose in proportion to the length of sonication. We examined the ultrastructure of sonicated collagen gel using SEM and found that cross-linking levels of collagen fibrils were inversely proportional to the sonication time (Fig. 4A). This finding indicated that sonication may reduce the rigidity of collagen gel. To test whether lowering the rigidity of collagen gel alters the level of apoptosis in LLC-PK1 cells, we used collagen gel pretreated with 1, 2, and 4 min of sonication. The FACScan analysis showed that sonication dose-dependently increased collagen gel-induced apoptosis (Fig. 4B), which meant that low-  ϭ 1 m). B, SEM examination of the ultrastructures of collagen fibrils in type I collagen gel-coated dishes and collagen gel (black bars ϭ 1 m). C, the growth curve of LLC-PK1 cells cultured on normal culture dishes, type I collagen gel-coated dishes, or collagen gel for 1 and 2 days. LLC-PK1 cells (3 ϫ 10 5 ) were seeded, and cell number was assessed using a hemocytometer. D, the apoptotic ratio (%) of LLC-PK1 cells cultured on normal culture dishes, type I collagen gel-coated dishes, or collagen gel for 48 h, assessed using FACScan analysis.

AP-1 Deregulation in Low Rigidity-induced Apoptosis
ering the substratum rigidity of collagen gel increased the level of apoptosis.
Quantification of Collagen Gel Rigidity-To detect the rigidity of collagen gel, we used both a rheometer and a dynamic mechanical analyzer. The dynamic rigidity, in Pascals (Pa), of collagen gel extracted from 8-month-old rat tails was 123.43 Ϯ 7.99 Pa (N/m 2 ), from 4-month-old rat tails was 79.16 Ϯ 6.73 Pa, and from 1-month-old rat tails was 14.25 Ϯ 0.92 Pa. After sonication for 1, 2, or 4 min, the dynamic rigidity of 0.3% collagen gel from 1-month-old rats dose-dependently decreased to 7.84 Ϯ 1.01 Pa, 6.25 Ϯ 0.98 Pa, and 5.21 Ϯ 1.12 Pa, respectively. In addition, the results of the dynamic rigidity of collagen gel, assessed using a dynamic mechanical analyzer, were similar, which confirmed that the age of the tail tendon used to make the collagen gel affected the rigidity of the collagen gel (41). In summary, our findings indicated that lowering the substratum rigidity of collagen gel triggered a higher rate of apoptosis in epithelial cells.

The Rate of Epithelial Cell Apoptosis Was Inversely Proportional to the Nuclear Area of Cells and the Rigidity of Collagen
Fibrils-Cell tension regulated by cytoskeletal organization may control cellular as well as nuclear morphology. The area of the nucleus may reflect the degree of cell extension. To understand whether cell extension is associated with epithelial cell survival/apoptosis, we cultured LLC-PK1, NMuMG, NRK-52E, and BS-C-1 cells on collagen gel prepared from 1-or 8-monthold rat tails and assessed their nuclear size and apoptosis ratio. The nuclear area was smaller in all cells cultured on collagen gel (Fig. 5A). A reduction in cross-links of collagen fibrils caused a decrease in the nuclear area. On the other hand, the apoptosis ratio remained very low when cells were cultured on a dish. A reduction in the cross-links of collagen fibrils triggered apoptosis in all epithelial cells examined (Fig. 5B). The apoptosis ratio was inversely associated with the size of the nuclear area in all cells examined (Fig. 5C). Interestingly, there seemed to be a lower limit of nuclear area for each cell line to maintain survival. When the nuclear area of the cell was below this limit, the apoptosis ratio was significantly higher (p Ͻ 0.02 for all cell lines). These findings taken together strongly suggest that low substratum rigidity collagen-fibril-induced epithelial cell apoptosis might result from inhibiting cell extension, which indicated that extension is required for the survival of epithelial cells.  ϭ 1 m). The group of cells cultured on 4-month-old collagen gel had extended cell processes, or lamellipodia, whereas those on 1-month-old gel did not. C, the apoptotic ratio of LLC-PK1 cells cultured on normal culture dishes, 1-month-old collagen gel, 4-monthold collagen gel, or 16-month-old collagen gel for 2 days, assessed using FACScan analysis.  ϭ 3 m). B, effects of sonication on the collagen gel-induced apoptotic ratio in NMuMG, LLC-PK1, and NRK-52E cells. Cells were cultured for 24 h on 1-month-old collagen gel or 1-month-old collagen gel pretreated with 1, 2, or 4 min of sonication. (47). To see whether Bcl-2 blocks collagen gel-induced apoptosis, we used LLC-PK1 and MDCK cells overexpressing Bcl-2. We found that Bcl-2 overexpression did not inhibit collagen gel-induced apoptosis in these cells (Fig. 6A). To determine whether the release of cytochrome c from mitochondria was involved in collagen gel-induced apoptosis, LLC-PK1 cells were cultured on culture dishes or collagen gel for 24 h. We analyzed the mitochondrial and cytosolic fractions of cell lysates for cytochrome c levels. We found that collagen gel did not alter the distribution of cytochrome c levels in either mitochondrial or cytosolic fractions (Fig. 6B), which indicated that mitochondrial pathways were not involved in collagen gel-induced apoptosis.

Neither Mitochondria nor Caspase 3 Was Involved in Collagen Gel-induced Apoptosis-Bcl-2 is important for preventing apoptosis induced by the mitochondrial pathway
To see which caspases were involved in collagen gel-induced apoptosis, LLC-PK1 cells were cultured on culture dishes, collagen gel-coated dishes, and collagen gel. Western blotting showed that collagen gel had not activated caspase-2, -8, or -9, the initiators and mediators of apoptosis (Fig. 6C). We did find, however, that caspase 3 was activated within 8 h in cells seeded on collagen gel. Various caspase inhibitors were used to evaluate which caspase cascade was involved in collagen gel-induced apoptosis. Pan-caspase inhibitor z-VAD partially inhibited collagen gel-induced apoptosis, but caspase-3-specific inhibitor DEVD-fmk did not (Fig. 6D).
Low Substratum Rigidity Induced Down-regulation of c-Jun and Aberrant Expression of c-Fos-The activation of JNK and its downstream signals reflects the environmental stress on cells. To determine whether low substratum rigidity activates JNK and its downstream signaling, we used Western blotting to detect JNK, phosphorylated JNK, and AP-1 proteins. When LLC-PK1 cells were cultured on collagen gel, JNK was activated within 4 h and continued for 18 h (Fig. 7A). However, the down-stream signal molecule, c-Jun, was down-regulated within 1 h and gradually disappeared because of the low substratum rigidity. In contrast, c-Fos was consistently expressed over 24 h because of the low substratum rigidity (Fig. 7B). Similar results were found in NMuMG cells (Fig. 7C). We examined other members of the AP-1 protein family and found that JunD and SP-1 were down-regulated because of the low substratum rigidity but that ATF-2 levels were not significantly different in cells cultured in culture dishes, collagen gel-coated dishes, or collagen gel (supplemental Fig. S1). In contrast, HeLa cells that survived low substratum rigidity showed no c-Jun degradation (data not shown). Increasing substratum rigidity, by increasing the age of the rat tail tendons used to make the collagen gel, partially reversed the c-Jun degradation in NMuMG cells (Fig.  7D). These findings indicated that the down-regulation of c-Jun and aberrant expression of c-Fos might be involved in low substratum rigidity-induced apoptosis.
We assessed the time course changes of p-JNK, JNK, c-Jun, and c-Fos expression in BAECs cultured on dish, collagen gelcoated dish, or collagen gel (data not shown). Low rigidity of collagen gel induced JNK activation in both LLC-PK1 cells and BAECs. The activation of JNK was induced in LLC-PK1 cells within 4 h and in BAECs within 1 h. Low rigidity of collagen gel induced a decrease in c-Jun level within 1 h in LLC-PK1 cells or BAECs (data not shown). However, low rigidity-induced decrease in c-Jun was not further down-regulated in BAECs from 8 to 18, h and on the other hand the decrease was reversed (data not shown). This might be the reason why low rigidityinduced apoptosis in BAECs was markedly lower than that in epithelial cells. In addition, low rigidity-induced aberrant expression of c-Fos was found in LLC-PK1 cells but not in BAECs, which could be another reason why low rigidity induced relatively lower apoptosis in BAECs.
shRNA Inhibited c-Jun Expression-induced Cell Death in NMuMG Cells-Because low substratum rigidity degraded c-Jun, we wanted to test whether c-Jun degradation induced cell death in epithelial cells. We used shRNA to down-regulate c-Jun expression in NMuMG cells and then visualized cell viability using vital stain. Cells with green fluorescence (Fig. 8A, panel a, left panel) indicated positive transfection of control or shRNA for c-Jun; the expression of c-Jun in the same cells is shown in the right panel. Higher expression of shRNA completely inhibited the expression of c-Jun. Vital stain indicated that shRNA-transfected cells underwent cell death (Fig. 8A,  panel b). We also used a time-lapse recording method to observe the process of cell death in transfected cells under a confocal microscope (Fig. 8B and supplemental video). Cell death was seen (Fig. 8B) within 10 h of shRNA transfection, but there was no detectable cell death in cells transfected with EGFP control or in cells with low shRNA expression. The transfection rate of control EGFP in NMuMG cells was quite constant, with a maximum ratio of ϳ10%. However, the percentage of cells that expressed shRNA 8 h after transfection remained very low (Fig. 8C, panel a), possibly because cells transfected with shRNA underwent cell death because of reduced c-Jun levels. To test this possibility, we quantified the cell-death ratio of cells transfected with shRNA by counting the propidium iodide positively stained cells (Fig. 8C, panel b). We found that the cell-death ratio in cells expressing shRNA increased over time. These findings indicated that shRNA down-regulation of c-Jun triggered cell death in NMuMG cells.
Overexpression of Either c-Fos or c-Jun Induced Cell Death in Various Cells-Because low substratum rigidity also induced aberrant expression of c-Fos, we wanted to see whether c-Fos overexpression resulted in cell death. NMuMG cells were transfected with c-Jun and c-Fos. We found that overexpression of both c-Jun and c-Fos induced cell death not only in NMuMG cells (Fig. 9A) but also in LLC-PK1 and HeLa cells (Fig. 9B). Overexpression of c-Jun increased the cell-death ratio from 17% to 48% in HeLa cells, from 23% to 44% in LLC-PK1 cells, and from 20% to 50% in NMuMG cells at 12 and 24 h, respectively. Overexpression of c-Fos increased the celldeath ratio from 8% to 32% in HeLa cells, 17% to 34% in LLC-PK1 cells, and 18% to 26% in NMuMG cells at 12 and 24 h, respectively. Interestingly, we found that JNK inhibitor suppressed the aberrant expression of c-Fos (Fig. 9C, panel a). However, inhibiting JNK expression and the aberrant expression of c-Fos reduced the apoptosis ratio only ϳ20% (Fig. 9C, panel b). JNK activation might be responsible for the aberrant expression of c-Fos. According to these results, levels of c-Jun and c-Fos are tightly regulated in epithelial cells.

DISCUSSION
In this study, we demonstrated that low substratum rigidity, the physical property of fibril collagen, induced apoptosis in epithelial cells but not in fibroblasts or transformed cells. Normal epithelial cells could not extend on collagen gel due to inadequate cross-linking of collagen fibrils; their configuration was reminiscent of the morphology of cells grown on a very limited area of fibronectin (48). In both conditions, maintenance of cell extension is required for cell survival. We showed that collagen gel constructed from the tendons of older rat tails resulted in greater cell extension and significantly attenuated collagen gel-induced apoptosis. Cell extension in cells grown on collagen gel prepared from older rat tail tendons was not as great at that in cells grown on collagen gelcoated dishes. Therefore, cells grown on collagen gel extracted from 16-month-old rats had a higher rate of apoptosis than cells grown on collagen gel-coated dishes (24 Ϯ 1% versus 14 Ϯ 0.5%). Because these results are consistent with observations that cell geometry governs cellular life and death, the model presented here is useful for studying the signal transduction mechanism through which low rigidity induces apoptosis in epithelial cells.
Cells are able to sense the rigidity of their environments and change their motility through integrins and FAK (5,49,50). The tensional forces are generated by cytoskeletons that resist forces from the extracellular matrix (51,52). The force applied or generated through cell matrix interactions altered cell morphology, migration, and differentiation (53)(54)(55). Previous data also showed that extracellular matrix rigidity influenced the strengthening of integrin-cytoskeleton linkages (56). Cultured on collagen gel, epithelial cells not only displayed restricted cell extension but also disorganized actin filaments due to low substratum rigidity (data not shown), which is consistent with recent studies (5,7) showing that altering the balance of physical forces transmitted from the cell surface changed the cytoskeletal structure of the cells. Tensegrity (57) is a model that explains how a cell stabilizes itself mechanically by balancing contractive forces. In our model, we hypothesize that, because collagen fibers did not provide enough strength for cell extension, epithelial cells developed apoptosis because they lost tensegrity.
Sonication breaks triple helices into short segments but does not change their structures (15,16). In the present study, sonication prolonged gelation time and may have increased the lateral aggregation of collagen fibrils. Sonication also significantly decreased cross-linking of collagen fibrils, which contributed to the reduction in rigidity of collagen gel. Under the assumption that the nuclear area was proportionally correlated to cell extension, we assessed the nuclear area of cells. We found that substratum rigidity and the nuclear area were positively correlated and that the nuclear area and the degree of cell extension were inversely correlated. Our data are compatible with the results of the recent study by Storm et al. (18). However, results on the rigidity of collagen gel, different by about one order, were found when an electromechanical computer-controlled indenter was used in another study (58). The difference might have been caused by different detection methods. Nevertheless, our data, obtained using a rheometer, showed that collagen gel rigidity increased directly with the age of the rat tail tendons used to make the gel and decreased after sonication and that the apoptosis ratio was inversely proportional to the substratum rigidity of the gel. In general, lowering substratum rigidity below 100 Pa significantly up-regulated epithelial cell apoptosis. We have not, however, determined the threshold of substratum rigidity for triggering epithelial-cell apoptosis.
The morphology of cells cultured on low substratum rigidity collagen gel is reminiscent of that of cells deprived of cell-matrix interactions, so called "homeless cell death" or "anoikis" (59). Both low rigidity-induced cell death and homeless cell death are exclusive characteristics of epithelial cells. The anoikis signal is mediated through the activation of caspase-8, caspase-3, MEKK-1, and phospho-JNK (60,61) or Bim (62), and is blocked by Bcl-2 overexpression. This contrasts with our finding that Bcl-2 overexpression did not prevent low substratum rigidity-induced apoptosis. We also found that low rigidity down-regulated FAK (data not shown), which was not observed in the anoikis model. In addition, we found that the mitochondrial pathway was not involved in low rigidity-induced apoptosis and that pan-caspase inhibitor only partially blocked this apoptosis. These findings taken together indicate that low substratum rigidity collagen gel-induced apoptosis is not mediated by the mechanism that mediates anoikis. A novel apoptosisinducing pathway seems to be involved in low substratum rigidity collagen gel-induced apoptosis.
AP-1 proteins were activated by MAP kinases, and the phosphorylated AP-1 proteins resulted in dimerization, which in turn triggered the expression of genes associated with cell growth or survival (63). We found that low substratum rigidity activated JNK, but not p38 or ERK, within 4 h (data not shown). Under stress, caused by, for example, UV or drug treatment, the activation of JNK and downstream amplification of c-Jun have been considered cues for apoptosis. In addition, JNK activation may trigger the mitochondrial apoptosis pathway that increases Bcl-2 phosphorylation. The present study shows that low substratum rigidity activated JNK without triggering cytochrome c release from mitochondria. In addition, we demonstrated that low rigidity rapidly down-regulated c-Jun before JNK was activated. The study also showed that low substratum rigidity did not alter c-Jun mRNA levels and that low substratum rigidityinduced down-regulation of c-Jun was mediated by post-transcriptional regulation. Ample evidence has shown that the polyubiquitination of c-Jun and subsequent degradation by the 26 S proteasome may contribute to the degradation of c-Jun (64,65). Whether low substratum rigidity-induced down-regulation of c-Jun is mediated through this pathway remains to be investigated. In the present study, c-Jun down-regulation caused apoptosis in NMuMG cells. In vivo study showed that a c-Jun deficiency resulted in embryonic lethality. Although embryonic fibroblasts were cultured from c-Jun-deficient mice, they lasted for only two passages. These cells easily developed apoptosis, because they accumulated spontaneous DNA damage (66). Lowering the rigidity of collagen gel decreased the expression level of c-Jun and increased apoptosis in epithelial cells. On the other hand, transformed cells cultured on collagen gel showed persistent c-Jun expression and evaded low substratum rigidity-induced apoptosis. Taking these findings together, we speculate that the degradation of c-Jun is important in low substratum rigidity-induced apoptosis in epithelial cells.
Although experiments attempting to maintain c-Jun expression in cells cultured on collagen gel have been tried, they have been very difficult, because cells transfected with c-Jun have all developed apoptosis regardless of cell type.
Low substratum rigidity also induced persistent expression of c-Fos, which is mediated by activated JNK; inhibiting JNK activation reduced c-Fos expression. We examined whether persistent expression of c-Fos is involved in low substratum rigidity-induced epithelial cell apoptosis. In the present study, c-Fos overexpression induced cell death in many epithelial cell types, which is consistent with the findings in photoreceptor cells (67) and ganglion cells (68). Normally, cells express c-Fos only transiently. In experiments using white light-induced apoptosis, inducing persistent c-Fos expression is required for chromosome condensation and DNA fragmentation. Depletion of c-Fos can prevent light damage-induced apoptosis morphology in the retina (69). These findings suggest the importance of persistent c-Fos expression in apoptosis.
In this study, we found that only epithelial cells are more susceptible to collagen gel-induced apoptosis than mesenchymal and tumor cells. Among all extracellular matrices examined, only type I and type III collagen gel induced apoptosis in epithelial cells. Because type I and type III collagen are fibril in nature, our data indicate that collagen gelinduced apoptosis may be related to their fibril characteristics. Under physiological conditions, epithelial cells do not contact with collagen fibril directly. In fact, there is a sheet of basement membrane that separates epithelium from direct contact with the interstitial tissues. Therefore, the function of basement membrane could be considered as a natural shield, in addition to playing roles in growth and differentiation for epithelium. On the other hand, fibroblast or transformed cells do not develop apoptosis when they contact directly with fibril collagen. Fibroblast belongs to the mesenchymal cells. Under the physiological con- ditions, they grow in type I collagen-enriched environments. We showed that fibroblasts grown on collagen gel exhibited elongated shape. They may change the architectural state of the cytoskeleton to evade the low rigidity collagen gel-induced apoptosis. Fibroblast or transformed cells are also more resistant to homeless cell death than epithelial cells. Recent reports show that expression of certain oncogene in epithelial cells conferred resistance to anoikis (70,71). Ras, for example, is a common oncogene that overexpresses or mutates in many malignant cancer cells (72). It could auto-transduce the growth factor downstream signals for cell survival. Ras also governs the cytoskeleton-altering proteins in the cytoskeletal control in the cells. It is plausible that transformed cells may actively change their cytoskeletal construction upon contacting to matrix of low rigidity. Our findings are important for tumor biology, because invasive tumor must have acquired this phenotype to survive their paths in interstitium. We postulate that one of the characteristics that invasive cancer cells acquire during transformation is to escape collagen gel-induced apoptosis.