Human cholangiocarcinoma cells (HuCCA-1), derived from a Thai patient, were grown in Ham’s F-12 media (Gibco^®, Invitrogen, USA) (14). Hepatocellular carcinoma cells HCC-S102 established from a Thai patient were grown in RPMI-1640 (Gibco) (15) while HepG2 and SK-HEP-1 cell lines were purchased from American Type Culture Collection (ATCC) and grown in DMEM (Gibco). All cell culture media contained 10% fetal bovine serum (FBS, Hyclone Laboratories, USA), 100 U/ml penicillin, 100 mg/ml streptomycin, and 125 ng/ml amphotericin B (Gibco) and were maintained at 37°C in a humidified atmosphere, 95% air, 5% CO₂. In serum-free culture, FBS gradually decreased and was simultaneously replaced with chemically defined medium for high density cell culture serum replacement (CDM-HD; FiberCells^® System, USA) (16).

Scaffold-based 3D culture was established using natural collagen type I fibrils named Lyostypt^® (Braun, Germany) (17). Briefly, cells were seeded onto sterilized collagen scaffold, followed by incubation for 2 h to allow cell attachment. Media was gently added to completely submerge each scaffold piece then incubated overnight and media was subsequently changed every alternate day throughout the experiment.

Cell-seeded scaffolds were incubated with MTT in a humidified incubator with 5% CO₂ at 37°C for 2 h and the number of cells was qualitatively determined under light microscope. For Hoechst staining, cell-seeded scaffolds were incubated in the dark with 1 μg/ml Hoechst dye for 2 min and stained nuclei were detected by a fluorescence microscope. Quantitative numbers of cells were determined by PicoGreen^® assay (Invitrogen) that binds to DNA and measured by fluorescence microplate reader at wavelength of 485 nm and emission at 520 nm. Metabolic activity was also determined to confirm cell viability by assessing remaining glucose concentration in the conditioned medium. The culture medium (2 μl) was assayed daily using Medisafe-mini GR-102 blood glucose meter (Terumo, Japan). The amount of depleted glucose in the medium each day was calculated and reported as cumulative glucose consumption by viable cells.

Cells were fixed in 4% paraformaldehyde. Specimens were stained with H&E and mucins for microscopic analysis. For morphological evaluation, cells from both culture systems were fixed with 2.5% glutaraldehyde (pH 7.4). Dried samples were sputter-coated with platinum-palladium and observed under scanning electron microscope (SEM) (Hitachi SEM S-2500, Japan).

Conditioned medium was collected and cell debris was removed by centrifugation. Supernatant was concentrated by lyophilization and precipitated in 10% TCA at 4°C overnight. Samples were centrifuged at 12,000 rpm, 4°C for 10 min, washed with 25% v/v acetone, and dried by speed vacuum. Specimens were resuspended in lysis buffer containing 9 M urea, 2% CHAPS, 2% DTT, 2% ampholine pH 3.5–10.0, and 1:500 protease inhibitor (Sigma-Aldrich, USA). Protein concentration was measured by Bradford assay. IPG strips, 7 cm, non-linear, pH 3.0–10.0 gradient (GE Healthcare, USA) were rehydrated with 150 mg protein overnight. IEF was performed at 7,000 Vh, 55 mA per gel strip using an Ettan IPGphor 3 (GE Healthcare). The IPG strips were equilibrated as previously described (8) and separated in 12.5% SDS-PAGE followed by CBB R-250 staining. Gels were analyzed by ImageMaster 2D Platinum 7.0 software (GE Healthcare).

Protein spots having a volume ratio change of >1.5-fold were excised and subjected to in-gel digestion according to the protocol by Srisomsap et al (9). Briefly, gels were destained with 50% ACN in 0.1 M NH₄HCO₃, reduced with 10 mM DTT, and alkylated with 100 mM iodoacetamide, respectively. Gel pieces were dried then 0.3 μg of trypsin (Promega, USA) was added, followed by incubation at 37°C overnight and digested peptides were collected for protein identification.

Nanoflow liquid chromatography coupled with the amaZon speed ion trap mass spectrometry (Bruker, USA) was utilized to identify protein spots. A 75 μm id × 100 mm C18 EASY-nLC™ column (Thermo Scientific, USA) was used. Gradient separation was performed using 0.1% formic acid in water (solution A) and 0.1% formic acid in ACN (solution B) followed by MS/MS equipped with CaptiveSpray™ source. Parent mass peaks with a range from 50 to 3,000 m/z were selected for MS/MS analysis in which collision energy was fixed at 1,300 V. MS/MS data were processed by Bruker Compass 1.4 software and proteins were then identified using MASCOT with similar search parameters as in a previous study (18). Proteins with molecular weight and pI consistent to gel spot with MASCOT score >25 using p-value ≤0.05 were considered positively identified.

Proteins were resolved in 10% SDS-PAGE and electrophoretically transferred to PVDF membranes (Millipore, USA). The membranes were probed with antibody against human TPI, SFN (1:2,000, Abcam, USA), PGAM1, ENO1, LCP1 (1:1,000, Abcam), β-tubulin (1:2,000, Cell Signaling Technology, USA), and actin (1:5,000, Sigma, USA) at 4°C overnight. Membranes were washed and incubated with corresponding secondary antibody conjugated with HRP (DakoCytomation, Denmark) at room temperature for 1 h. Membranes were probed with ECL (GE Healthcare) and detected by ImageQuant™ LAS 4000 (GE Healthcare). Then, membranes were stained with CBB R-250 and band intensity was determined to show equal protein loadings.

The differences between monolayer and 3D cell culture were analyzed with STATA 10.1 using unpaired t-test. p<0.05 was considered statistically significant.

The serum-free media cell culture was completely switched to CDM-HD and morphology observed under a light microscope. There was no significant difference in percentage of viable cells between serum-containing and serum-free cultures. Cell viability was consistently maintained at >95%.

The serum-free cells were then grown in natural collagen-based scaffold to mimic the dense 3D microenvironment of the in vivo tumor. Growth and localization of cells were studied (Fig. 1A). At the initial phase (day 2), single cells were dispersed throughout the scaffold. Then, the cells aggregated into colonies and expanded between the spaces of the spun fibers (days 4 and 14). When cell-seeded scaffolds were stained with MTT (left column), black spots were visible representing cells or small colonies of cells (day 2). Hoechst staining of dsDNA (middle column) showed increasing numbers of blue spots under a fluorescence microscope, indicating increased number of nuclei of cells at increasing time-points (days 2, 4, and 14). H&E staining (right column) indicated cells with deep blue nuclei and faint pink cytoplasm throughout the pink collagen fibers. These three different methods confirmed progressive growth of HuCCA-1 cells inside collagen scaffolds.

Morphological study under SEM (Fig. 1B) showed alteration of cell shape in 3D culture when compared to monolayer culture. The appearance of cells in the scaffold was round, growing into a dense cluster, whereas the attached cells were flat and expanded as a single thin layer during the 14 days of culture. Cell-cell contacts as well as cell-collagen scaffold contacts were observed in our 3D culture system.

The growth pattern of cells in different culture systems was determined by PicoGreen assay. Results showed alterations in growth pattern between 3D and monolayer cultures (Fig. 1C). Cells in monolayer culture showed rapid increase in cell proliferation but decreased after reaching the maximum growth in day 4, while in the 3D culture system, cells showed slower proliferation but maintained a longer growth phase covering a period of over 14 days.

The levels of glucose were monitored and cumulative glucose consumption was calculated compared to starting glucose level in the culture media (Fig. 1D). The pattern of cumulative glucose consumption of cells in 3D culture continuously increased until day 7, while cumulative glucose consumption of cells in monolayer culture became constant after day 4, reflecting the cessation of growth in monolayer culture after 4 days (Fig. 1C). Thus, cells at day 4 of both culture systems were selected and used for further comparative analyses.

Histological analysis revealed that 3D culture produced increased level of secreted ECM when compared to monolayer culture. Compared to plain collagen scaffold, a positive result (pink) of mucicarmine staining represented accumulation of mucins in 3D culture, which was not found in monolayer culture (Fig. 1E).

The protein concentrations extracted from 3D culture were significantly increased (>5-fold, p<0.05) compared to monolayer culture (Fig. 2A). Thus, the secreted proteins from 3D culture were more extensively enriched and sufficient for further studies.

Contamination of intracellular protein leakage from damaged or dead cells is also a major concern in secretome analysis. To evaluate interference by cell lysis, western blotting of two major cytosolic proteins (actin and β-tubulin) was performed in conditioned media from both culture systems at day 4. The data indicated significant levels of actin and β-tubulin in conditioned media obtained from monolayer culture, but decreased presence of these proteins in conditioned media prepared from 3D culture (Fig. 2B). Thus, the lack of these cytoplasmic protein contaminations confirmed the high quality of secreted proteins obtained from 3D culture, making samples from 3D culture ideal for secretome analysis.

The secreted proteins from cells grown under the two culture conditions were compared by 2DE (Fig. 3A). Overall, 431±14 protein spots were found in 3D culture while 410±20 spots were found in monolayer culture. Image analysis showed that 35 secreted protein spots were differentially expressed in 3D culture compared to monolayer culture. Of these, 25 spots were significantly increased in 3D culture and 10 spots were found to be expressed only in 3D culture system (Table I). These proteins were categorized according to their functions, namely: signal transduction, metabolic enzymes, chaperone and stress response, protein synthesis and degradation, cytoskeleton and related, and miscellaneous function proteins (Fig. 3B).

Western blot analyses were used to verify the expression of selected secreted proteins which differed considerably between two culture conditions including 14-3-3 σ (SFN), triosephosphate isomerase (TPI), α-enolase (ENO1), phosphoglycerate mutase 1 (PGAM1), and L-plastin (LCP1). These results were consistent with the 2DE data in which the expressions of these secreted proteins were higher in 3D than monolayer culture (Fig. 4A). Moreover, the expression of LCP1 in secretomes of CCA cells was used to compare with three liver cancer cell lines (HepG2, SK-HEP-1, and HCC-S102). The result showed that LCP1 was found only in the conditioned media of CCA, but not in the conditioned media of the other liver cell lines (Fig. 4B).