Prognostic and Predictive Values of MAD2L1 in Cholangiocarcinoma

Background: This study aimed to investigate the expression of Mitotic arrest decient 2-like protein 1 (MAD2L1) in cholangiocarcinoma (CCA) and its biological function. Methods: Our study performed bioinformatics to analyze the microarray data from the Gene Expression Omnibus (GEO) database and obtained the differentially expressed genes (DEGs). Enrichment assay and protein-protein interaction networks analysis were conducted to extract hub genes. Mitotic arrest decient 2-like protein 1 (MAD2L1) was investigated in tumor and adjacent nonneoplastic biliary ducts in 42 samples by immunohistochemistry. Subsequently, MAD2L1 was manipulated, and its function was examined in CCA cell lines and in vivo model. Results: 297 DEGs were extracted from the microarray data. And seven hub genes were dened through the enrichment assay and protein-protein interaction networks analysis. MAD2L1 was picked up as a novel biomarker, according to hierarchical cluster analysis and Kaplan-Meier survival analysis. MAD2L1 was specically expressed in cancer tissues but not in the surrounding normal tissue, and 31 (73.81%) of 42 CCAs were MAD2L1 positive by immunohistochemistry. MAD2L1 expression was signicantly correlated with tumor size, pathological grade, and clinical stage. Kaplan-Meier analysis demonstrated an inverse correlation between MAD2L1 expression and overall survival. Real-time polymerase chain reaction (RT-PCR) and Immunoblotting results further conrmed the results of immunohistochemistry and bioinformatic analysis from the database. In vitro and in vivo models, decreasing MAD2L1 could signicantly suppress tumor growth. Conclusion: High MAD2L1 expression predicts the advanced stage of CCA. Targeting MAD2L1 may be a potential tumor suppressor and may provide the biological basis for a new therapeutic strategy.


Background
Cholangiocarcinoma (CCA) is the second most prevalent liver cancer and accounts for approximately 15% of all liver cancer in adults [1]. Depending on the anatomical location of tumor initiation, CCAs are classically subtyped into three groups: intrahepatic CCA (iCCA), perihilar CCA (pCCA), and distal CCA (dCCA) [2]. But the common pathological characteristic of these three CCAs is initiating from the malignant bile duct epithelial cells [3]. Because of the improvement in the diagnosis of CCA, incidence, and mortality are both increasing all over the world, especially the east Asian [4][5][6][7]. Unfortunately, to date, there has been still no satis cing therapeutic strategy except for surgical resection and liver transplant [7,8]. Data from the past 30 years suggests that the 5-year overall survival rate has shown no improvement [9]. The dominating reasons are approximately 65% of patients are diagnosed with unresectable, and over 90% of patients recurrent early after surgical treatment [10,11]. Developing novel biomarkers for improving early diagnosis, monitoring recurrence, and exploring new mechanisms underlying CCA progression is urgent.
In the past two decades, microarray technology and bioinformation have been widely applied to analyze the gene expression in CCA [12,13]. More and more microarray data has been uploaded to the Gene Expression Omnibus (GEO) database for sharing. Screening genetic alterations from GEO and identifying the differentially expressed genes (DEGs) and pathways involved in carcinogenesis is an emerging strategy for cancer research [14,15].
In our study, we searched from GEO and got three mRNA microarray datasets. After standardization, we analyzed to obtain DEGs between carcinoma tissues and non-carcinoma tissues. Whereafter, online bioinformatics analysis including Gene Ontology (GO), Kyoto Encyclopedia of Genes and Genomes (KEGG) pathway enrichment analysis, and protein-protein interaction (PPI) network were performed to explore the molecular mechanisms underlying tumorigenesis and progression. MAD2L1, one of the DEGs, was validated high expression in tumors and correlated with poor survival in CCA from the TCGA database. Then, we investigated the expression of MAD2L1 in tumor tissues, adjacent nonneoplastic biliary duct from CCA patients by immunohistochemistry (IHC) in 42 surgically resected primary CCAs from a single institute. We also investigated the correlation of MAD2L1 expression with histologic tumor grade, progression, metastasis, clinical stage, and postoperative survival rate. Functional exploration, we altered MAD2L1 in CCA cell lines and assayed cell proliferation, and results demonstrated that MAD2L2 could promote cell growth. And the in vivo study results also con rmed the in vitro study. In conclusion, we found MAD2L1 could be a novel biomarker in CCA, and targeting MAD2L1 may be a potential therapeutic strategy for CCA.
Bioinformatic analysis. The Database for Annotation, Visualization, and Integrated Discovery (DAVID; http://david.ncifcrf.gov) (version 6.7) was used to conduct biological analyses [19]. P<0.05 was considered statistically signi cant. Search Tool for the Retrieval of Interacting Genes (STRING; http://string-db.org) (version 11.0) was performed to predict the functional interactions between proteins and with a combined score >0.4 was considered statistically signi cant [20]. The plug-in Molecular Complex Detection (MCODE) (version 1.4.2) of Cytoscape was used to nd densely connected regions with the criteria for selection as following: MCODE scores >5, degree cut-off=2, node score cut-off=0.2, Max depth=100 and k-score=2, and the hub genes were selected with degrees ≥15 [21]. Hierarchical clustering and the overall survival analyses of hub genes were constructed using the UCSC Cancer Genomics Browser (https:// genome-cancer.ucsc.edu) [22]. Patient selection. Forty-two surgically resected primary CCAs were collected from 2004 to 2014 at our institute. The study was approved by the ethics committee and conducted following the regulations. All patients signed informed consent for the use of their CCA samples and clinical information. This study was also performed according to the ethical guidelines of the Declaration of Helsinki 2008 [23]. The specimens were anonymized and analyzed in a blinded fashion. Follow-up time for survivors ranged from 1 to 36 months (median, 15.3 months). All 42 patients were followed up for at least one year after surgery.
Histology and Immunohistochemistry. Formalin-xed and para n-embedded tissue slides (5 μm) were depara nized and rehydrated. Antigen retrieval was performed by incubating the tissue slides in 0.01 M citric acid buffer at 100°C for 10 minutes. After blocking with 3% H 2 O 2 and 5% fetal bovine serum, the slides were incubated with a monoclonal antibody against MAD2L1 (1:100, ab10691; Abcam, Burlingame, CA) at 4°C overnight. The slides were then reacted with polymer-horseradish peroxidase reagent. The peroxidase activity was visualized with diaminobenzidine tetrahydroxychloride solution. The sections were counterstained with hematoxylin. Dark brown cytoplasmic staining of at least 1% tumor cells was de ned as positive, and no staining or less than 1% cells stained was de ned as negative. As a negative control, we replaced the primary antibody with 5% fetal bovine serum.
Cell culture. All CCA cell lines, QBC939 was donated by Dr. Li lab [24], HuCCT1 cells were purchased from the Japanese Collection of Research Bioresources Cell Bank, and HEK293T cell lines were obtained from the American Type Culture Collection. Cells were maintained in Dulbecco's modi ed Eagle's medium (Invitrogen; Thermo Fisher Scienti c Inc.) supplemented with 10% fetal bovine serum (Thermo Fisher Scienti c, Inc.), 1% glutamine, and 1% penicillin/streptomycin (Invitrogen; Thermo Fisher Scienti c, Inc.).
Cells were humidi ed in a 5% (v/v) CO 2 environment at 37˚C. Tests for mycoplasma contamination yielded negative results by direct PCR.
Cell count. Trypan blue dye exclusion test by staining cells with 0.2% Trypan blue solution (Sigma-Aldrich Chemical Co, MO, USA) was performed to reveal necrotic cells after each experimental condition, according to the manufacturer's instructions. Moreover, unstained viable cells were manually counted using a hemocytometer. Counts were performed by triplicate under a 10X objective according to standard methodology. All measurements were performed in triplicate.
Animal experiments. Eight 6-week-old nude mice were purchased from the National cancer institute (NCI) and divided into two groups for injection of QBC939 cells transfected/engineered with (1) shNC, (2) shMAD2L1. The prepared stable QBC939 cells (mixed with Matrigel, 1:1) were injected at 1 × 10 6 into the Subcutaneous tissue of the mice. The mice were sacri ced after 8 weeks. Tumors were removed for study. Animal protocols were approved by the Animal Care and Use Committee of Nanjing medical university. And all animal experiments followed guidelines for ethical review of laboratory animal welfare of Nanjing medical university (http://iacuc.njmu.edu.cn).
Statistics. Statistical analysis was performed using SPSS software, version 23.0 (IBM Inc.). Mann-Whitney U test was applied to test the two groups of continuous data. Correlation between MAD2L1 expression and clinicopathological parameters was evaluated by the χ2 test and Fisher exact test. Survival rates were calculated using the unadjusted Kaplan-Meier method, and the log-rank analysis analyzed the difference in survival curves. All cell experiments were conducted in three technical replicates. Quantitation results are shown as mean ± SD. Statistical signi cance for cell experiments was determined using the independent-sample t-test. P<0.05 was considered to indicate a statistically signi cant difference.

Results
Expression pro les and integrated screening of DEGs in human CCA tissues from bioinformatic analysis Three microarray datasets containing cancerous and noncancerous tissues from GEO were analyzed to obtain DEGs. We rstly standardized the microarray results, then identi ed all DEGs (4,702 in GSE26566, 1,092 in GSE32225, and 2447 in GSE77984); lastly, overlapped the three datasets and got 297 genes as shown in the Venn diagram (Fig. 1A). Functional and pathway enrichment analyses of 297 DEGs were performed using DAVID, and the top 20 GO pathways were showed in gure 1B. The results of GO analysis demonstrated that changes in biological processes (BP) of DEGs were signi cantly enriched in protein binding, negative regulation of the apoptotic process. Changes in molecular function (MF) of DEGs were mainly enriched in protein binding, poly(A) RNA binding, and protein homodimerization activity. Changes in cell component (CC) of DEGs were mainly enriched in the cytoplasm, nucleus, and perinuclear region of cytoplasm (Fig. 1B). KEGG pathway analysis revealed that the 297 DEGs were mainly enriched in pathways in cancer (Fig. 1C). To further extract the hub genes from the DEGs, STRING and Cytoscape were used to predict the PPI network and get the most signi cant modules (Fig. 1D). A total of 7 genes were identi ed as hub genes with degrees ≥15. Hierarchical clustering showed that the hub genes could differentiate the cholangiocarcinoma samples from the noncancerous samples (Fig.   1E). Subsequently, the Kaplan-Meier survival analysis was performed to demonstrate the overall survival of the hub genes. MAD2L1 revealed a survival difference between high expression and low expression signi cantly (Fig. 1F). However, no difference in overall survival was shown among the other six hub genes (Fig. S1).
Together, the results from Fig. 1A-F suggest that MAD2L1 could play important roles in the carcinogenesis or progression of CCA.
Clinical data revealed MAD2L1 expression correlated with poor outcome in CCA To evaluate MAD2L1 expression in CCA, we applied IHC to detect MAD2L1 in 42 CCA cases. Thirty-one (73.81%) of 42 cases were MAD2L1 positive, whereas 11(26.19%) were negative. All the par-carcinoma tissues were negative, compared with tumor tissue, and intriguingly, for 30 cases (71.43%) with lymphovascular invasion, only 8(26.67%) of them had MAD2L1 expression in the metastatic lymph nodes. Representative positive staining versus negative of MAD2L1 in CCA tissue was shown in Fig. 2A.
Next, we analyzed the correlation between MAD2L1 expression and a variety of clinicopathological features. Interestingly, MAD2L1 expression was more frequently associated with larger tumor size (OR,6.231; P=0.023), higher pathological grade (OR,5.6; P=0.048), and higher stage (OR,6.000; P=0.035). However, there was no association between MAD2L1 and metastasis (p=1.000) (table 1). In 42 patients with at least 12 months of follow-up, Kaplan-Meier analysis showed that patients with MAD2L1 positivity had a shorter overall postoperative survival time than did MAD2L1-negative cases (Fig 2.B; P=0.031). To con rm the expression of MAD2L1 in CCAs, we isolated fresh tumors, adjacent nonneoplastic biliary duct from 20 CCA cases and applied real-time PCR and immunoblot to detect MAD2L1 mRNA and protein expression. As shown in Fig. 2C and D, higher MAD2L1 mRNA was detected in larger tumors than that in small tumors. Compared with the para-carcinoma tissues, MAD2L1 protein was exclusively detected in tumors.
Taken together, Fig.2A-D indicated that MAD2L1 expression could be used as a biomarker to predict prognosis in CCA.
Preclinical study using in vitro and invivo models to test the role of MAD2L1 in CCA Previous clinical data showed that positive MAD2L1 exhibited a signi cant difference with tumor size but not with metastasis (table 1). To further validate the MAD2L1 roles in the CCA progression, we conducted the gain or loss of MAD2L1 function assay through the lentiviral system in CCA cells. MAD2L1 was knocked down in QBC939 cells with MAD2L1-shRNA (Fig. 3A) and over-expressed in HuCCT1 cells with MAD2L1-cDNA (Fig. 3B). After altering MAD2L1 in QBC939 and HuCCT1 cell lines, cell count and colony formation assay were conducted to check cell growth. The results revealed that knocking down MAD2L1 with shMAD2L1 signi cantly suppresses cell proliferation (Fig. 3C). Consistently, in HuCCT1 cells, overexpression of MAD2L1 increased cell growth (Fig. 3D). We performed subcutaneous tumor formation in a total of 2 groups ((1: QBC939-shNC; 2: QBC939-shMAD2L1; four nude mice for each group) as an In vivo model.. Four weeks later, the tumor volume of the shMAD2L1 group signi cantly reduced than the control group ( Fig. 3G and H) due to low MAD2L1 protein level (Fig. 3I) In summary, the results from Fig. 3A-I suggest that MAD2L1 may promote tumor progression by increasing cell proliferation.

Discussion
In our study, we applied bioinformatic analysis to nd hub genes that were aberrant expression between CCA tissues and para-carcinoma tissues from GEO datasets. Firstly, we obtained 294 DEGs and employed DAVID to conduct the GO and KEGG analyses. GO enrichments demonstrated these DEGs enriched in protein binding, cytoplasm, and signal transduction, while the results of KEGG analysis revealed that DEGs enriched in pathways in cancer. Set with degrees ≥ 15, we obtained seven hub genes. Among these hub genes, Enhancer of zeste homolog 2 (EZH2) has been reported that its abnormal expression correlated with tumorigenesis and progression in many malignancies, such as cholangiocarcinogenesis, prostate cancer, and non-small cell lung carcinoma [25][26][27]. As a lysine methyltransferase, EZH2 catalyzes methylation of histone H3 at lysine 27, establishes and maintains H3K27 tri-methylation repressive marks [28,29]. In physiological conditions, EZH2 plays an important role in cellular differentiation; however, in pathological conditions, it contributes to cancer cell proliferation [30]. Mechanism study in hepatocellular carcinoma (HCC) revealed that EZH2 suppressed miR-30d which could degrade Karyopherin Subunit Beta 1 (KPNB1) mRNA, and eventually increased KPNB1 expression. Ectopic expression of KPNB1 promoted HCC cell growth [31]. In the present study, EZH2 also displayed higher expression in tumor tissues compared with adjacent tissues; however, the overall survival assay showed no difference between positive and negative-EZH2. A bioinformatic study on HCC disclosed overexpression of The cell-division cycle protein 20 (CDC20), MAD2L1, and DNA replication licensing factor (MCM3) could predict poor prognosis [32,33]. In our study, partially due to limited clinical samples, CDC20 and MCM3 showed no difference in overall survival analysis. According to previous published functional study, these 7 hub genes correlated with the cell-division cycle in carcinoma and could be involved in cell growth [31,34,35].
TCGA database suggests that even all the seven hub genes displayed high expression in tumor tissues, only MAD2L1 showed a signi cant difference in over survival (Fig. 1F). Our clinical data con rmed that positive MAD2L1 correlated with poor survival in CCA. Positive MAD2L1 related to bigger tumor size, higher stage, and higher grade, but intriguingly no correlation was displayed between metastasis and MAD2L1-positive ( Table 1). The IHC results of weak positive MAD2L1 in the metastatic lymph nodes backed up that MAD2L1 occupied demanding positive function in promoting tumor cell growth. To test our hypothesis, we arti cially altered MAD2L1 expression in CCA cell lines and examined cell growth. The results of cell proliferation demonstrated that MAD2L1 promoted CCA cell growth. Oppositely, knocking down MAD2L1 could suppress cell growth. Previously, abundant studies illustrated that MAD2L1 plays very important roles, either in physical or pathological conditions. Physiologically, MAD2L1 exhibits synergistic action with SCF plus GM-CSF to promote the immature hematopoietic progenitor cells growth [36]. Under the pathological conditions, many upstream genes or miRNA functioned through MAD2L1 to promote cell proliferation, such as down-regulation of miR-200c-5p increased MAD2L1 expression in HCC and elevated cell growth [37,38]. However, to date, no data showed that MAD2L1 expression in CCA. Our study has been the rst study on MAD2L1 in CCA on the level of not only vitro but also vivo models. Our data demonstrated that targeting MAD2L1 could signi cantly suppress CCA growth, and MAD2L1 may bee a valid therapeutic target for developing more effective novel therapeutic strategies to treat CCA.

Conclusion
MAD2L1 is an oncogene. Although its function is not fully understood, the present study showed that MAD2L1 is highly expressed in cancer cells which can increase cell proliferation. MAD2L1 may be involved in multiple signaling pathways in cancer development. Therefore, the application of manipulating MAD2L1 in CCA therapy may have potential breakthroughs and become a new therapeutic target worthy of further study.

Consent for publication
All included patients provided informed consent.

Availability of data and materials
All data generated or analyzed during this study are included in this published article.

Competing interests
The authors declare that they have no competing interests. Authors' contributions YG and XQ conceived and designed the study. LF TL XO and CZ made substantial contributions to the design of the current study, acquisition of data, interpretation of data and revising the manuscript. All authors read and approved the nal manuscript.