Specimens were collected at Sappasitthiprasong Hospital, Ubon Ratchathani, Thailand. The definitive diagnosis of cholangiocarcinoma and hepatocellular carcinoma was based on the histopathological examination of biopsy or surgical specimens. Detailed information of patients (n=7; 4 cholangiocarcinoma and 3 hepatocellular carcinoma) including sex, age, the definitive diagnosis and staging are shown in Table I. Blood specimens were collected at the first patient visit when there was clinical suspicion of cholangiocarcinoma, because of signs and symptoms of obstructive jaundice and presence in a population at risk in the high prevalence area. Control blood group (n=5; 3 males, 2 females, with average age of 53.8±2.7) was defined as healthy individuals who presented for an annual checkup. Plasma-EDTA samples were retrieved as left-over specimens within 30 min after routine complete blood counting and then kept aliquots at −80°C until used. The study was approved by the local ethics committee of Faculty of Medicine Ramathibodi Hospital, Mahidol University and Sappasitthiprasong Hospital (protocol ID 03-58-68) and written informed consent was waived due to use of left-over specimens.

Human cholangiocarcinoma cell line (HuCCA-1) originating from a Thai patient (19) was cultured in Ham's F-12 (Gibco; Thermo Fisher Scientific, Inc., Waltham, MA, USA) with 10% fetal bovine serum (FBS, Hyclone; GE Healthcare Life Sciences, Chicago, IL, USA) and 1% antibiotic-antimycotic (Gibco). Hepatocellular carcinoma cell line (HCC-S102) established from a Thai patient (20) was grown in RPMI-1640 (Gibco; Thermo Fisher Scientific, Inc.) containing 10% FBS (Hyclone) and 1% antibiotic-antimycotic. After the cells reached ~70% confluence, culture medium was replaced with serum-free medium. The culture was further incubated for 24 h before the entire conditioned medium was harvested. For experiments comparing MNC and HFB systems, HuCCA-1 cells were cultured in the serum-free media with a CDM-HD serum replacement (FiberCell^® Systems Inc., Frederick, MD, USA) as described in our previous study (12). The conditioned media were collected daily and replaced with fresh media. The collected media were harvested by sequential centrifugation at 480 × g at 4°C for 5 min and 2,000 × g for 10 min to remove contaminated cells and cell debris, respectively. After centrifugation, the conditioned media were kept at −80°C until use.

Approximately 1×10⁸ CDM-adapted HuCCA-1 cells were inoculated into the hollow fiber bioreactor with a 5 kDa MWCO cartridge (#C2008, FiberCell Systems Inc.) according to the manufacturer's instructions. The entire conditioned media were collected daily and the fresh new media were replaced. The collected media were harvested using the same protocol performed in the MNC system as described previously.

The growth of cells cultured in the HFB and MNC systems was indirectly monitored by glucose consumption. The depletion rate of glucose in the media was determined using a blood glucose meter (Medisafe-mini GR-102, Terumo Co., Tokyo, Japan) as previously described (12).

The lactate dehydrogenase (LDH) cytotoxicity detection kit (Takara Bio USA, Inc., Mountain View, CA, USA) was used to determine the extent of cell leakage. Briefly, 100 µl of conditioned media was mixed with an equal volume of the assay mixture (catalyst, dye, and reaction mixture). The reaction mixtures were incubated at room temperature for 30 min in the dark and stopped by adding 1/10 volume of 1 N HCl. End product was measured by a microplate reader (Molecular Devices, LLC., Sunnyvale, CA, USA) with the absorbance at 450 and 690 nm.

The conditioned media were concentrated by an Amicon^® Ultra Centrifugal Filter Unit with 3 kDa MWCO (Merck KGaA, Darmstadt, Germany). Then trace contaminants were removed using a 2-D Clean-Up kit (GE Healthcare Life Sciences). The sample was then resuspended in 2D lysis buffer containing 7 M urea, 2 M thiourea, 130 mM DTT, 4% CHAPS, 2% ampholine pH 3.0–10.0 Iso-Dalt (Servalyt), 30 µM Tris buffer, and cOmplete™ Mini protease inhibitor cocktail (Roche Molecular Systems, Inc., Pleasanton, CA, USA) and incubated at room temperature for 30 min. The protein concentration was determined using Bio-Rad protein assay (Bio-Rad Laboratories, Hercules, CA, USA). Protein samples (100 µg) were adjusted to 50 µl with 2D lysis buffer and mixed with 75 µl of rehydration buffer containing 8 M urea, 18 mM DTT, 2% CHAPS, and 2% Ipg buffer pH 3.0–10.0 NL (GE Healthcare Life Sciences). The sample was left at room temperature for another 30 min before applied to Ipg strips, 7 cm, pH 3.0–10.0 non-linear (GE Healthcare Life Sciences) for 16 h. The first dimension separation was performed at 20°C with 50 µA per strip with Ettan IPGphor 3 (GE Healthcare Life Sciences) according to the manufacturer's instructions. The IPG strip was equilibrated and separated by 12.5% SDS polyacrylamide gel electrophoresis (SDS-PAGE) using a SE600 Ruby apparatus (GE Healthcare Life Sciences) at a current of 25 mA per gel, as previously described (21). The gel was stained using Sypro^® Ruby (Molecular Probes; Thermo Fisher Scientific, Inc.) and scanned using an Ettan DIGE Imager (GE Healthcare Life Sciences). The gel image was analyzed by ImageMaster 2D Platinum 7.0 software (GE Healthcare Life Sciences).

The protein spots were excised and subjected for in-gel digestion as described (10). The digested peptides were identified by a nanoflow liquid chromatography system using a C18 Easy-nLC™ (75 µm id × 100 mm) column (Thermo Fisher Scientific, Inc.) coupled with the amaZon speed ion trap mass spectrometer (Bruker, Billerica, MA, USA). Protein identification was interpreted using a MASCOT search engine (www.matrixscience.com) with the SwissProt database, limited to Homo sapiens only, with default settings except for the following parameters: 0.3 Da for peptide tolerance and MS/MS tolerance; 1+, 2+, and 3+ for peptide charge; and ESI-TRAP for Instrument.

The STRING software version 9.1 (http://string-db.org/) was used to predict the potential protein-protein interactions and the clusters of the function were grouped. The proteins with 2-fold higher expression in HFB in comparison to MNC with statistical difference (p<0.05) were selected.

Protein samples (15 µg) were separated by 12.5% SDS-PAGE and transferred onto Immobilon-P membrane (Millipore, Merck KGaA). The membrane was blocked with 0.1% Tween-20 in Tris-buffer saline (TBS/T) containing 5% BSA for 1 h at room temperature, followed by probing overnight at 4°C with monoclonal antibodies against human neutrophil gelatinase-associated lipocalin (NGAL) (1:200, Abcam Cambridge, UK), L-plastin (LCP1) (1:400, Abcam), protein deglycase DJ-1 (DJ-1) (1:2,000, Abcam), α-enolase (ENO1) (1:2,000, Abcam), topoisomerase (TPI) (1:4,000, Abcam), proteasome subunit α type-3 (PSMA3) (1:5,000, Cell Signaling Technology, Inc., Danvers, MA, USA), proteasome subunit α type-6 (PSMA6) (1:5,000, Abcam), and proteasome subunit β type-6 (PSMB6) (1:2,000, Abcam). The membrane was incubated with the corresponding secondary antibodies (Dako; Agilent Technologies, Santa Clara, CA, USA). Protein signal was detected by a WesternBright ECL (Advansta, Inc., Menlo Park, CA, USA) using an ImageQuant LAS 4000 mini digital imaging system (GE Healthcare Life Sciences).

High abundance proteins were removed from human plasma samples using the Pierce™ Top 2 Abundant Protein Depletion Spin Columns (Thermo Fisher Scientific, Inc.) according to the manufacturer's instructions. Briefly, 10 µl of each sample was applied to the column and rotated gently end over end at room temperature for 30 min. The columns were spun down at 1,000 × g for 2 min. The depleted samples were collected and then concentrated using an Amicon Ultra-0.5 Centrifugal Filter, with 3 kDa MWCO (Merck KGaA). Samples were centrifuged at 14,000 × g for 60 min at 4°C and the filtrate was discarded. The filter devices were then placed upside down in a new tube and spun down at 1,000 × g for 2 min in order to collect the remaining concentrated sample. Total protein concentrations of samples were determined by Bradford protein assay (Bio-Rad Laboratories). Protein samples (20 µg) were separated and immunoblotted with an antibody against PSMA3as described above.

Secreted proteins from each condition were buffer-exchanged into 1X PBS using Bio-Spin 6 columns (Bio-Rad Laboratories), followed by determination of proteasomal activity using a Proteasome Activity Fluorometric assay kit (BioVision, Inc., Milpitas, CA, USA) according to the manufacturer's instructions. Briefly, 10 µg of each sample was adjusted to 50 µl with PBS, mixed with 50 µl of assay buffer, 1 µl of substrate, and 1 µl of inhibitor or assay buffer. The reaction was monitored every 10 min using a fluorescentmicroplate reader with the excitation at 350 nm and the emission at 440 nm. The specific activity of proteasome was then calculated as the difference of relative fluorescence units divided time per micrograms of protein used.

The difference between mean values between groups was analyzed by using one-way ANOVA with post hoc Tukey HSD analysis. Values were considered statistically significant at p<0.05.

Since the viability of cells cultured in HFB could not be directly determined, the glucose consumption and the cellular release of lactate dehydrogenase were measured to observe the cell growth and cellular cytotoxicity, respectively. Glucose consumption of the cells grown in HFB was increased until the end of the observation period at day 9, while cell leakage as detected by lactate dehydrogenase release showed a slight decrease with the observation period (Fig. 1A). On the other hand, the glucose consumption of the cells cultured in MNC increased continuously until day 4 before entering into a stationary phase on day 5 and 6 and then decreasing afterwards, while the lactate dehydrogenase release was stable during the first 4 days and increased from day 5 onwards (Fig. 1B). These results suggested that the cells cultured in HFB were in good condition and kept growing throughout the experiment, while the growth of cells in MNC was limited to the first 5 days. Therefore, the conditioned media were collected at day 4 for each experiment for further analysis.

The conditioned media of cholangiocarcinoma cells from HFB and MNC systems were collected separately and analyzed using 2DE (Fig. 2). The ImageMaster software version 7.0 was employed to analyze the protein spots. The results showed that 1,003 spots were detected in HFB, while 977 spots were found in MNC. Significantly different expression of 138 protein spots were identified by LC/MS/MS. Ninety-five secretory proteins in conditioned media of HFB culture system showed a significant 2-fold increase as compared to those of MNC system. These proteins could be mainly categorized into 6 groups according to their function namely cell cycle, cytoskeleton structure, immune system, metabolic pathway, signaling transduction and ubiquitin-proteasome pathway (Table II). Among these proteins, neutrophil gelatinase-associated lipocalin (NGAL or lipocalin-2) (spot no. 6), pterin-4-α-carbinolamine dehydratase (spot no. 55), proteasome subunit β type-7 (spot no. 108) and triose phosphate isomerase (spot no. 109) were expressed only in conditioned media of HFB system. Eight interesting proteins showed increased expression by >10-fold, while 27 proteins were expressed between 5- and 10-fold higher and 20 proteins were expressed between 2- and 5-fold higher. NGAL or lipocalin-2 (spot nos. 11 and 66) were found to show the greatest increase in expression by 19.93-fold and 18.79-fold, and 2 more spots also belong to the same protein (spot nos. 6 and 15) at different pI. Ferritin light chain, ferritin heavy chain, proteasome subunit α type-3, Annexin A3, triose phosphate isomerase and mitochondrial enoyl-CoA Δ isomerase 1 showed >10-fold higher expression. Proteins with 5-10-fold increased expression include macrophage migration inhibitory factor, calmodulin, 60S ribosomal protein L12, proteasome subunit β type-5, superoxide dismutase (Mn), proteasome subunit β type-1, proteasome subunit β type-7, NGAL, proteasome subunit β type-4, NAD(P)H-hydrate epimerase, proteasome subunit β type-7, proteasome subunit α type-5, complement component 1 Q, N(G),N(G)-dimethylarginine dimethylaminohydrolase 1. Interestingly, the expression levels of all 20S proteasome subunits including proteasome subunit α type-1 to type-7 and β type-1 to type-7 were all increased in HFB compared to those in the MNC, with the-fold changes ranging between 2.92 and 12.13 (except for spot no. 108 which was only present in the HFB).

The potential protein interactions were predicted by a bioinformatics program (STRING). The analysis was performed on 95 proteins with 2-fold higher expression in the HFB system when compared to those in the MNC system. Upregulated proteins were selected for this analysis because they should be more suitable as for potential biomarkers (22). The upregulated proteins could form at least three clusters of the interaction networks as shown in Fig. 3. The first cluster (purple background) belongs to the 20S proteasome subunits including proteasome subunit α type-1 to type-7 (PSMA1-7) and β type-1 to type-7 (PSMB1-7) which, according to the KEGG pathway (23), is involved in the regulation of various functions. The proteins associated in cell cycle pathway are the 14-3-3 protein family including 14-3-3 protein ε (YWHAE), 14-3-3 protein σ (SFN), 14-3-3 protein β/α (YWHAB), 14-3-3 protein γ (YWHAG), 14-3-3 protein η (YWHAH), 14-3-3 protein θ (YWHAQ) are shown in the second cluster (red background). The third cluster (green background) are proteins involved in metabolism including glycolysis/gluconeogenesis, citrate cycle (TCA cycle), pyruvate metabolism, 2-oxocarboxylic acid metabolism, cysteine and methionine metabolism, tyrosine metabolism, glycine, serine and threonine metabolism, and phenylalanine metabolism.

According to the results from 2DE and prediction of protein-protein interactions, 95 secretory proteins from HFB system were significantly increased and can be clustered as a functional relationship into 3 groups as compared with MNC system. Immunodetection was further used to validate some distinct secreted proteins that may be useful for cancer biomarker study including NGAL, PSMA6, PSMB6, PSMA3, LCP1, DJ-1, TPI, and ENO1 as shown in Fig. 4. In agreement with 2DE results, western blot analyses revealed that the expression levels of these proteins were significantly increased in the HFB system compared to those in the MNC system. Interestingly, among of these secreted proteins, only 20S proteasome has never been proposed as a potential biomarker for early detection in cholangiocarcinoma.

Proteasomes function mainly inside the cells, therefore, we next studied whether the identified 20S proteasome subunits present in the extracellular compartment had active functions. The activity of the 20S proteasome was therefore studied in conditioned media from both HFB and MNC systems. Cumulative fluorescent signals from the proteasomal-related proteolytic activity were detected in conditioned media from both HFB and MNC systems (Fig. 5A). This indicated the 20S proteasome subunits secreted from cells were still active in function. Consistent with the immunodetection results, higher proteasome activity was observed in conditioned media of HFB system in comparison to that of MNC system. Proteasomal specific activity of HFB and MNC samples were 38.63 and 17.22 U/µg, respectively (Fig. 5A). In addition, the proteasome activities were also investigated in the conditioned media from HuCCA-1 and HCC-S102, with the result that the proteasome activity of HuCCA-1 secretome was higher than that of HCC-S102 secretome (42.63 and 30.78 U/µg, respectively) (Fig. 5B).

To validate 20S proteasome as potential biomarker for CCA, the level of plasma PSMA3 was measured using western blot analysis in normal donor (n=5), in CCA patients with stage I (n=1), stage III (n=2), stage IV (n=1), in HCC patients with stage I (n=1) and stage III (n=2). To enrich low-abundance proteins, human plasma albumin and IgG was removed prior to immunodetection. PSMA3 were detected in blood plasma from normal subjects, CCA and HCC patients (Fig. 6A). Interestingly, plasma PSMA3 of CCA patients were significantly increased as compared to normal subjects. The levels of PSMA3 were obviously increased in the plasma of CCA patients as compared to HCC patients (Fig. 6B). Moreover, the levels of plasma PSMA3 were significantly elevated in cancer patients with stage III and IV compared to normal subjects (Fig. 6B).