Introduction

Cholangiocarcinoma (CCA) is a cancer originating from the bile duct epithelium (cholangiocytes). It is rare in Western countries, but it is highly prevalent in the Northeast Thailand particularly in Khon Kaen province, where the prevalence is highest in the world. [1]. The main risk factor for CCA in this area is an infection with carcinogenic liver flukes, Opisthorchis viverrini (OV) due to dietary habits of eating raw or undercooked freshwater fish and poor sanitation practice. Liver fluke infestation induces chronic inflammation leading to oxidative DNA damage and CCA transformation of the infected bile ducts [1, 2]. At present, opisthorchiasis-associated CCA is a serious public health problem in this region.

Surgery to remove the cancer completely is the only possible curative treatment for CCA. However, most of CCA patients are diagnosed with advanced unresectable tumors. For postoperative treatment or the treatment of unresectable cases, chemotherapy in combination with or without radiotherapy has been used even with poor prognosis. Although several attempts have been developed for screening, diagnosis and treatment of CCA, practically none of those attempts brought about apparent improvement of survival of patients. However, targeted cancer therapies are still required to increase the efficacy of cancer treatment and reduce the side effects [3].

Mitochondrial proteins play key roles in pathogenesis, cellular patterns [4], and carcinogenesis of various cancers e.g., renal [5], gastric [6], breast [7], hepatocellular [8], and colon cancer [9]. Thus, mitochondria are considered as target organelles for cancer chemotherapy. In particular, the expression levels of the mitochondrial proteins such as serine hydroxymethyltransferase 2, complement 1q-binding protein, ornithine aminotransferase, mitochondrial uncoupling protein 2, mortalin, succinyl-CoA/3, and ketoacid coenzyme A transferase1 were found to correlate with the progression of various cancers [7–12]. For CCA therapy, combinations of mitochondrial protein induction and chemotherapy or radiotherapy were reported to enhance apoptosis of CCA cells [13–15]. Also, high expression of a mitochondrial protein, superoxide dismutase 2 (SOD2), was observed during CCA genesis in hamster model [16]. Recent study of whole mitochondrial DNA (mtDNA) genome sequence analysis, however, revealed that there is no direct association between mtDNA mutations and CCA [17]. Therefore, differential expression of mitochondrial proteins in CCA and adjacent non-cancerous tissues may provide better understanding of the roles of mitochondrial proteins in carcinogenesis and pave us the way to discover new molecular targets for diagnosis and treatment of CCA. Here, we report that using proteomic analysis of mitochondrial proteins in CCA and adjacent non-cancerous tissues, AIFM3 was overexpressed only in CCA regardless of the tumor types but not in non-cancerous tissues, suggesting its potential target for chemotherapy.

Materials and methods

Tumor samples

A total of 25 CCA samples (11 non-papillary and 14 papillary types) were used in this study. Among those samples, three, each of non-papillary and papillary CCA containing cancerous and adjacent non-cancerous tissues were used for the isolations of mitochondria and subsequently for proteomic analyses. Paraffin-embedded sections of all 25 samples were used for immunohistochemistry.

All samples were kindly provided from the Liver Flukes and Cholangiocarcinoma Research Center (LFCRC), Faculty of Medicine, Khon Kaen University, Thailand. They were immediately snap-frozen in liquid nitrogen and stored at −80 °C until use. For histopathological typing of CCA, frozen sections were collected from each tissue and a couple of sections from each tissue were fixed in a 4 % buffered formalin and stained with hematoxylin and eosin. This project was approved by the Khon Kaen University Ethics Committee for Human Research (HE571416).

Isolation of mitochondrial protein fractions

The methods for isolation of mitochondrial protein fractions followed to those described previously [18–20]. Approximately, 100–200 mg of frozen human CCA and its adjacent non-cancerous tissues were washed with ice-cold phosphate buffered saline (PBS, 0.1 M Sodium phosphate, 0.15 M Sodium chloride, pH 7.2) to remove any blood clots and connective tissues. Mitochondria were isolated using a Mitochondria Isolation Kit for Tissue (Pierce, Rockford, IL, USA) according to the manufacturer’s instructions with some modifications. In brief, human CCA and adjacent non-cancerous tissues were separately minced in each of the 800 μl of the Mitochondria Isolation Reagent 1, supplemented with protease inhibitor cocktail (Roche, Indianapolis, IN) and carefully homogenized about 40–80 strokes on ice [19]. Each crude homogenate was then returned to the original tube, and 800 μl of the Mitochondria Isolation Reagent 2 supplemented with protease inhibitor cocktail (Roche) was added for further cell lysis. The tube was centrifuged at 700×g for 5 min at 4 °C to spin down unbroken tissues, cell debris, and nuclei. The supernatant was transferred into a new tube and centrifuged at 12,000×g for 10 min at 4 °C to spin down mitochondria [20]. The supernatant was transferred into a new tube as cytosolic lysates. Then, 500 μl of the Mitochondria Isolation Reagent 2 was added to the pellet, re-suspended, and the suspension was centrifuged at 12,000×g for 5 min at 4 °C to reduce contamination of other subcellular components. The pellet was lysed in lysis buffer containing 0.5 % SDS and protease inhibitors (Roche). Protein concentrations of both mitochondrial and cytosolic lysates were determined by Lowry’s method [21]. Voltage-dependent anion channel (VDAC) protein, a mitochondrial marker, was investigated using western blot analysis.

Electrophoresis and tryptic digestion

Mitochondrial proteins from each non-papillary and papillary CCA and their corresponding adjacent non-cancerous tissues were separately pooled (50 μg aliquots for each sample). Protein samples (50 μg each) were separated on 12.5 % SDS-PAGE (ATTO AE-6530 system, Tokyo, Japan). After Coomassie Brilliant Blue (CBB) staining, the gel lanes were cut into 12 ranges. Each gel range was chopped into fine pieces (1 mm3/piece), and transferred to a 96-well plate (12–16 pieces/well). The gel pieces were subjected to in-gel digestion according to the method developed by the Proteomics Research Laboratory, National Center for Genetic Engineering and Biotechnology (BIOTEC) and the National Science and Technology Development Agency (NSTDA), Thailand. In brief, the gel pieces were dehydrated twice in 100 % acetonitrile (ACN) for 5 min each with agitation and dried at room temperature for 15 min. Subsequently, disulfide bonds were reduced with 10 mM dithiothreitol (DTT)/10 mM ammonium bicarbonate (NH4HCO3) for 1 h at room temperature and alkylated with 100 mM iodoacetamide (IAA)/10 mM NH4HCO3 for 1 h at room temperature in the dark. Then, 20 μl of 10 ng/μl trypsin in 50 % ACN/10 mM NH4HCO3 was added and incubated for 20 min before adding 20 μl of 30 % ACN. The reaction mixture was incubated overnight at 37 °C. The tryptic peptides were extracted from the gel for three times with 30 μl of 50 % ACN/0.1 % formic acid. Finally, the tryptic peptide mixtures were dried and kept at −80 °C until LC-MS/MS analysis.

LC-MS/MS analysis

The extracted peptides were dissolved in 15 μl of 0.1 % formic acid, centrifuged at 12,000×g for 10 min, and separated using an Ultimate 3000 LC System (Dionex Ltd., U.K.) on a nanocolumn PepSwift monolithic column 100 mm-i.d.650 mm, with a flow rate of 300 nl/min using multi-steps gradient of a linear concentration increase from 10 to 90 % of 80 % acetonitrile in 0.1 % formic acid within 20 min. The nanoLC system was connected with an electrospray interface with ESI-Ion Trap MS (Bruker Daltonik GmbH, Bremen, Germany). The collected LC-MS raw data were directly used for quantification based on MS signal intensities of individual analysis using DeCyder MS differential analysis software (DeCyderMS, GE Healthcare, Piscataway). All differential peptide data were submitted to database search against the NCBI human mitochondria database using MASCOT software version 2.2 (Matrix Science, London, UK). Search parameters were as follows: Trypsin was selected as the enzyme, with three potential missed cleavage, carbamido-methylated cysteine as fixed modification, and oxidation of methionine residues as a variable modification. Peptide mass tolerance was 1.2 Da, while fragment mass tolerance was 0.6 Da, and ESI ion trap was selected for the instrument type. Proteins were selected with statistical significance at P < 0.05 as shown in maps of protein levels using MultiExperiment Viewer (MeV, version 4.6.1) software. Then, all significantly expressed proteins were analyzed for their intersections among the different sample groups using jvenn (http://bioinfo.genotoul.fr/jvenn/example.html). Finally, the mitochondrial proteins that exclusively expressed in both types of CCA tissues, but not in adjacent non-cancerous tissues, were chosen as candidate proteins.

Western blot analysis

To validate the integrity of harvested mitochondria, immune-blot distribution patterns of human voltage-dependent anion channel (VDAC) protein in the isolated mitochondrial and cytosolic fractions were examined using anti-human VDAC antibody (Biorbyt, UK). Fifty micrograms of proteins were size-fractionated by SDS-PAGE using 12.5 % polyacrylamide gels and electrophoretically transferred onto a polyvinylidene difluoride (PVDF) membrane (GE Healthcare, U.K.) for 1 h, and the membrane was blocked with 5 % skim milk in Tris-buffer saline with 0.1 % Tween-20 (TBS-T, pH 7.4) for 1 h at room temperature. Then, the membrane was incubated overnight at 4 °C with rabbit polyclonal antibody against human VDAC. After washing with TBS-T, the membrane was incubated for 1 h at room temperature with horseradish peroxidase (HRP)-conjugated secondary antibody and goat anti-rabbbit IgG (GE Healthcare) diluted in 5 % skim milk in TBS-T. The chemiluminescence was detected using the enhanced chemiluminescence ECL plus system (GE Healthcare) and then visualized using the Image Quant LASmini400.

To validate the candidate mitochondrial protein, AIFM3, its expression level in the individual sample was investigated by western blotting using AIFM3 antibody (Biorbyt, UK). The membrane was incubated overnight at 4 °C with rabbit polyclonal antibody against human AIFM3 (diluted 1:500 in TBS-T) and treated with the secondary antibody labeled with horseradish peroxidase. The antigen-antibody interaction on the membrane was detected as described above.

Potential interaction analysis

Potential interaction of the identified proteins was analyzed using public-domain software, the search tool for interacting chemicals known as STITCH 4.0 [22]. In brief, a set of input page icons of the software was shown at the start page. The page icon namely “multiple names” was selected. The list of proteins of interest was keyed in e.g. AIFM3 and/or other specified proteins and an item “Homo sapiens” was selected as the organism, and then clicked “Go”. The pop-up new page with the list of the protein names of interest for the example, AIFM3 and other proteins were seen and then clicked “Continue”. The other pop-up new page showed the confidence view. To interpret the confidence view, stronger associations are represented by the thicker lines. Protein-protein interactions are shown in solid lines, chemical-protein interactions in dashed lines and interaction between chemicals in dotted lines. In our study, highly confident interaction was focused with a score greater than 0.7.

Immunohistochemistry

To evaluate the AIFM3 expression in CCA tissues, paraffin-embedded tissue sections of 25 CCA cases (including 3 papillomatous and 3 non-papillomatous CCA cases used for mitochondrial proteomic analysis as previously mentioned above) were examined using standard immunohistochemistry protocols. The tissue sections were incubated with anti-AIFM3 antibody at 1:300 in antibody dilution buffer (PBS-T, 0.05 % sodium azide) (Biorbyt, UK) at 4 °C overnight. Then, the sections were incubated with peroxidase-conjugated secondary antibody (Dako Cytomation, Inc., California, USA). Peroxidase activity was visualized using the 3, 3′-Diaminobenzidine (DAB) substrate solution, with hematoxylin counterstaining. The staining was assessed using H-score by recording both the intensity of staining (0 = no staining; 1 + = weak staining; 2 + = moderate staining; and 3 + = strong staining), and percentage of stained tumor cells (0–100 %). The H-score was calculated by the following formula: H-score = (% of positively stained tumor cells at weak intensity × 1) + (% of positively stained tumor cells at moderate intensity × 2) + (% of positively stained tumor cells at strong intensity × 3) as previously described [23].

Statistical analysis

Statistical analysis was performed using SPSS 17.0. The H-scores between matched pairs of non-papillary, papillary CCA, and their corresponding adjacent non-cancerous tissues were compared using Student’s t test. P value < 0.05 was considered statistically significant.

Results

Mitochondria isolation

Mitochondrial and cytosolic fractions were prepared from CCA and adjacent non-cancerous tissues of 3 cases each of non-papillary and papillary CCA patients. Successful separation of both fractions was validated by western blot analysis using antibody against mitochondria-specific marker protein. VDAC (Fig. 1a) shows strongly positive reaction in mitochondrial fractions but negative in cytosolic fractions.

Fig. 1
figure 1

Proteomic analysis of mitochondrial proteome. a Validation of the mitochondrial protein extracts; N, adjacent non-cancerous tissue; C, cancerous tissue of CCA; Mito, mitochondrial protein fraction; Cyto, cytosolic fraction. b Mitochondrial proteins (50 μg each) were pooled and 50 μg proteins were loaded onto each well of 12.5 % SDS-PAGE. Protein fractions were visualized by CBB staining; Lane 1: three matched-pairs of adjacent non-cancerous tissue of papillary CCA, Lane 2: three papillary CCA, Lane 3: three matched-pairs of adjacent non-cancerous tissue of non-papillary CCA, Lane 4: three non-papillary CCA

Full size image

Proteomic analysis of mitochondrial proteins

Mitochondrial protein fraction from the three samples each of non-papillary and papillary CCA and the corresponding adjacent non-cancerous tissues were pooled by tissue type and electrophoresed on 12.5 % SDS-PAGE (Fig. 1b). After dividing into 12 ranges according to the molecular size, each gel plug was digested in gel. The extracted tryptic peptides were subjected to LC-MS/MS analysis. The differentially expressed peptides data obtained from DecyderMS software were searched against the NCBI human mitochondria database for protein identification. In total, 281 mitochondrial proteins were identified. Using the MultiExperiment Viewer (MeV, version 4.6.1) software, 105 proteins were selected based on statistically significant (P < 0.05) and protein expression levels as shown in the heatmap (Fig. 2).

Fig. 2
figure 2

Heatmap of protein expression profile pattern of total 105 proteins. Each row represents an individual protein and each column represents matched-pairs of (adj-pap) adjacent non-cancerous tissue of papillary type, (Cancer-pap) CCA papillary type, (adj-nonpap) matched-pairs of adjacent non-cancerous tissue of non-papillary type, and (Cancer-nonpap) CCA non-papillary type, respectively (bar chart represents the signal intensity level as shown in the corresponding shades)

Full size image

Identification of AIFM3 as a candidate protein

A total of 105 significantly different in mitochondrial protein expressions in CCA and the adjacent non-cancerous tissues were analyzed using the “jvenn”, interactive Venn diagram viewer software. The results show that five and seven proteins are over-expressed in papillary and non-papillary CCA, respectively. However, only one protein was commonly found in both papillary and non-papillary type CCA tissues, but not in the adjacent tissues (Fig. 3a). This protein was identified as AIFM3. Western blot analysis of the mitochondrial solutions of the 3 cases each of papillary and non-papillary CCA cases revealed that AIFM3 protein was uniquely expressed in all 3 CCA tissues, but not in the corresponding adjacent non-cancerous tissues (Fig. 3b).

Fig. 3
figure 3

Identification of candidate protein. a Venn diagram presents the number of proteins in each sample and degree of individual overlap between cancerous tissue and adjacent non-cancerous tissue. The dot line is total identified proteins in adjacent tissue of papillary CCA and solid line is total identified proteins in papillary CCA. The dash line is total identified proteins in adjacent tissue of non-papillary CCA and dash-dot line is total identified proteins in papillary CCA. b Detection of the AIFM3 protein in mitochondrial fractions (n = 6) using western blot analysis; N, adjacent non-cancerous tissue; C, cancerous tissue of CCA

Full size image

To investigate the expression of AIFM3 in clinical specimens, apart from 3 cases each of papillary and non-papillary CCA used for mitochondrial extraction, 25 more human CCA and their corresponding adjacent non-cancerous tissues were examined by immunohistochemistry. The results show that H-scores of the AIFM3 staining were significantly higher in CCA than in the corresponding adjacent non-cancerous tissues (Fig. 4a-d).

Fig. 4
figure 4

Immunohistochemical detection of AIFM3 in cancerous and non-cancerous tissues from CCA patients. a Expression of AIFM3 in human tissues was examined by immunohistochemistry (magnification, ×400). b–d Comparison of AIFM3 levels in adjacent and cancerous tissues (n = 25), a level of P < 0.05 was considered statistically significant using matched-pairs t test in papillary CCA, non-papillary CCA, and overall CCA; Legend and symbol: Adjacent non-cancerous tissues (squares), CCA tissues (triangles)

Full size image

AIFM3 protein interaction

To speculate the potential role of AIFM3 in CCA-genesis, possible interaction of AIFM3 and mitochondrial carcinogenesis-related proteins were predicted by STITCH version 4.0 (Fig. 5). The results show that superoxide dismutase 2 (SOD2), NADPH oxidase 1 (NOX1), serine hydroxymethyltransferase 2 (SHMT2), complement component 1q-binding protein (C1QBP), ornithine aminotransferase (OAT), 3-oxoacid CoA transferase 1 (OXCT1), anamorsin (CIAPIN1), heat shock 70 kDa protein 9 (mortalin or HSPA9), and protein kinase C alpha (PRKCA) were identified as proteins interacting with AIFM3.

Fig. 5
figure 5

Protein-chemical interaction network of the AIFM3. Filtering interactions network was according to apoptosis-inducing factor (in bold circle) and carcinogenesis processes associated proteins (in bold box) predicted by STITCH Version 4.0 (http://stitch.embl.de). AIFM3 was found to be associated with superoxide dismutase 2 (SOD2), NADPH oxidase 1 (NOX1), serine hydroxymethyltransferase 2 (SHMT2), complement component 1q-binding protein (C1QBP), ornithine aminotransferase (OAT), 3-oxoacid CoA transferase 1 (OXCT1), anamorsin (CIAPIN1), heat shock 70 kDa protein 9 (mortalin or HSPA9), and protein kinase C alpha (PRKCA). Stronger associations are represented by the thicker lines. Weak associations are represented by thin lines. Protein-protein interactions are shown in solid lines, chemical-protein interactions in dashed lines

Full size image

Discussion

Proteomic analysis revealed that the expression patterns of mitochondrial proteins are different between cancerous and their adjacent non-cancerous tissues of both types of CCA. The significant differences of the expression patterns of mitochondrial proteins are also seen between both non-papillary and papillary types of CCA. The concordant findings the difference between mitochondrial protein expression in cancerous and non-cancerous tissues have also been reported in renal carcinoma, breast cancer, and gastric cancer [5, 6, 24].

In this study, five and seven proteins are over-expressed in papillary and non-papillary CCA, respectively. However, only AIFM3, which is also known as AIF-like protein (AIFL), was commonly over-expressed in both papillary and non-papillary types of CCA. AIFM3 consists of 598 amino acids. It is a mitochondrial flavoenzyme that arises from alternative splicing of human AIF-like gene during transcription [25]. NADH oxidase-dependent complex I function of AIF is important for oxidative phosphorylation and tumorigenicity [26]. AIF has been reported to act as a free radical scavenger that suppresses oxidative stress in neuronal cells and play a role in anti-apoptogenic function [27]. In addition, AIFM3 is a direct target of miR-210 that is related to proliferation of human hepatoma cells [28]. In fact, AIFM3 expression level was increased in miR-210 knocked-down hepatoma xenograft [29]. Some studies demonstrated that AIF induced apoptosis [25, 30]. Thus, the precise involvement of AIF in regulating life-death balances may be rather specific for the cell type and death inducer [26]. Overexpression of AIFM3 was found in human CCA tissues, which represented alteration of mitochondrial proteins in CCA. This high expression of AIFM3 in CCA was consistent with previous finding that showed high expression of SOD2 during CCA genesis in hamster model [16]. SOD2 is antioxidant enzyme that regulates ROS level occurring during OV infection. SOD2 expression seemed to be increased during cholangiocarcinogenesis, although decreased expression when the tumor has fully developed. Thus, high expression of SOD2 was found in CCA-induced hamsters, when tumors were developed and an expression of this enzyme was impaired. Taken together, AIFM3 was classified into AIF family members that were considered as free radical scavenger, apoptotic protein, and non-apoptotic protein. In the present study, we speculated that AIFM3 may be a potential mitochondria protein in CCA.

Interactions of AIFM3 and mitochondrial carcinogenesis-related proteins have been reported not only in CCA but also in several other cancers such as breast cancer, hepatocellular carcinoma, gastric cancer, and colon cancer [6, 9]. As the predicted counterparts to interact with AIFM3, the following molecules have been listed; superoxide dismutase 2 (SOD2), NADPH oxidase 1 (NOX1), serine hydroxymethyltransferase 2 (SHMT2), complement component 1q-binding protein (C1QBP), ornithine aminotransferase (OAT), 3-oxoacid CoA transferase 1 (OXCT1), anamorsin (CIAPIN1), heat shock 70 kDa protein 9 (mortalin or HSPA9), and protein kinase C alpha (PRKCA) [7, 8, 10–12, 16, 31–33]. Interestingly, 2 out of 4 mitochondrial proteins, namely PRKCA and CIAPIN1, were uniquely expressed in papillary CCA in this study. Their roles in association with AIFM3 should be explored in the future.

In this study, protein-chemical interaction networks analysis revealed that AIFM3 linked to various key molecules in cancer progression through z-VAD-fmk, which is a pan-caspase inhibitor having been widely used to block apoptosis [34]. In addition, z-VAD-fmk can also inhibit cysteine proteases such as calpain and cathepsin B in vitro [35]. This AIFM3 mediated pathway may also be associated with carcinogenesis of CCA. Since the molecular structure and function of AIFM3 have not been fully understood, further studies are required to elucidate its role in carcinogenesis and metastasis.

In terms of technical point of view, we have isolated mitochondria from cancerous and adjacent noncancerous tissues at the first step of proteomic analysis to avoid the complexity of the proteins of the starting material [36, 37]. Although a special reagent kit is commercially available and ready to use, however, few steps for optimization were required to obtain the optimum conditions for each tissue type because of biological variance of human tissues also affected the yield of protein samples [38]. So far, our methods in this study were suitable for proteomic analysis of mitochondrial proteins of CCA.

In conclusion, among mitochondrial proteins, AIFM3 was aberrantly expressed in CCA tissues and could be a potential target molecule for CCA chemotherapy.