Overexpression of FoxM1 predicts poor prognosis of intrahepatic cholangiocarcinoma

FoxM1 is upregulated in ICC

qRT-PCR and Western blotting analysis showed that the mRNA and protein expression levels of FoxM1 were significantly higher in ICC cancer tissues than in the matched peritumoral tissues (Fig. 1A, B). Then, we used immunohistochemical staining to detect the protein expression of FoxM1 in 184 ICC samples. The results showed that FoxM1 was mainly expressed in the cell membrane and cytoplasm of tumor cells. Based on the results of immunohistochemical staining, the entire cohort of 184 ICC cases was dichotomized into the FoxM1^high expression group (> 4 points, n = 111) and the FoxM1^low expression group (≤ 4 points, n = 73) (Fig. 1C). The details of these patients were described in a previous study [19]. The clinicopathological features of all ICC patients are listed in Table 1.

Figure 1. FoxM1 was upregulation in ICC and correlated with survival. The mRNA expression (A) and protein expression (B) of FoxM1 in ICC tumor tissues compared with the expression in paired peritumoral tissues. (C) Representative immunohistochemical staining images of FoxM1 in ICC. (D) The bar graph shows the statistics for the staining intensity of FoxM1 in 184 ICC tumor samples and paired peritumoral tissues. (E) Kaplan-Meier curves for the overall survival and disease-free survival of 184 patients with ICC according to the expression of FoxM1. Abbreviations: ICC, intrahepatic cholangiocarcinoma; P, peritumoral tissue; T, tumor tissue. ^*P < 0.05.

Overexpression of FoxM1 is correlated with ICC progression and adverse prognosis

A higher expression level of FoxM1 was correlated with the presence of multiple tumors (χ² = 17.030, P < 0.001), tumor diameter > 5 cm (χ² = 19.503, P < 0.001), lymph node metastasis (χ² = 13.576, P < 0.001), and advanced TNM stage (χ² = 6.947, P = 0.008) (Table 1). However, there were no significant relationships between FoxM1 expression and clinicopathological parameters such as age, gender, cirrhosis, hepatitis B virus infection, tumor encapsulation, tumor differentiation, vascular invasion, resected margin, serum CEA, and serum CA19-9 (all P > 0.05, Table 1). The results suggested that overexpression of FoxM1 was markedly associated with adverse clinicopathological factors and may function as a biomarker of poor prognosis in patients with ICC.

The OS and DFS of the FoxM1^high expression group were significantly lower than those of the FoxM1^low expression group (all P < 0.001, Fig. 1E). The median OS of the FoxM1^high expression group was significantly lower than that of the FoxM1^low expression group (7 months vs. 22 months), and the median DFS of the FoxM1^high expression group was also markedly shorter than that of the FoxM1^low expression group (3 months vs. 9 months). In addition, the 1-, 3-, and 5-year OS rates of the FoxM1^high expression group were 18.9%, 1.2%, and 1.2%, respectively, which were significantly lower than those of the FoxM1^low expression group (64.4%, 32.1%, and 28.5%, respectively). The same trend was seen in the 1-, 3-, and 5-year DFS rates of the FoxM1^high and FoxM1^low expression groups (6.3%, 0.9%, and 0.9% vs. 46.6%, 28.1%, and 21.8%, respectively). The Cox regression proportional hazards model for multivariate analysis showed that tumor differentiation, resected margins, and the FoxM1 expression level were independent prognostic predictors for the OS of patients with ICC (all P < 0.05, Table 2). Tumor size, resected margin, and the FoxM1 expression level were independent prognostic indicators for DFS (all P < 0.05, Table 2). The above results indicated that high expression levels of FoxM1 were significantly correlated with ICC progression and adverse prognosis.

FoxM1 promoted the proliferation, migration, and invasion of ICC in vitro

To verify the exact biological effect of FoxM1 on ICC, we first established cell lines that stably up/downregulated FoxM1 expression. The FoxM1 mRNA and protein expression in 3 ICC cell lines were respectively detected by qRT-PCR and Western blotting. It was found that the expression of FoxM1 was highest in the SSP-25 cell line and lowest in HCCC-9810 (P < 0.0001, Fig. 2A). Thus, HCCC-9810 was selected for targeted upregulation of FoxM1 (HCCC-9810-FoxM1, compared with its control HCCC-9810-Control), and SSP-25 was selected for targeted downregulation of FoxM1 (SSP-25-shFoxM1, compared with its control SSP-25-Control) by lentiviral transfection. The transfection efficiencies of both cell lines reached greater than 90%, and the cells were further screened using RPMI-1640 mediums supplemented with puromycin reagent (1μg/mL). The results showed that the transfected cells (HCCC-9810-FoxM1, SSP-25-shFoxM1) could stably up/downregulate the protein and mRNA expression of FoxM1 compared with their respective control cells (all P < 0.001, Fig. 2B, C). In addition, both cell lines were able to consistently express the labeled green fluorescence and grew well. Therefore, the selected cell lines were used for further formal experiments.

Figure 2. Selection and establishment of stably transfected ICC cell lines with FoxM1. (A) The protein and mRNA expression of FoxM1 in ICC cell lines (HCCC-9810, RBE, and SSP-25). Western blotting and qRT-PCR analysis showed successful overexpression (B) and knockdown (C) of FoxM1 in ICC cells (HCCC-9810 and SSP-25, respectively). ^***P < 0.001, ^****P < 0.0001.

Next, BrdU incorporation assay, MTT assays and plate cloning experiments were performed to detect the effect of FoxM1 on ICC cell proliferation. The results showed that the cell line with upregulated FoxM1 (HCCC-9810-FoxM1) had a significantly higher proliferative capacity than its control cell line (HCCC-9810-Control) (all P < 0.01, Fig. 3A-C, E, and F), while the proliferative ability of the cell line with downregulated FoxM1 (SSP-25-shFoxM1) was markedly lower than that of its control cell line (SSP-25-Control) (all P < 0.05, Fig. 3A, B, D, E, and G).In addition, it was indicated that plating efficiency of HCCC-9810-Control and HCCC-9810-FoxM1 were 13.8% (69/500, cells/well) and 34.4% (172/500, cells/well), while SSP-25-Control and SSP-25-shFoxM1 were 62.6% (313/500, cells/well) and 23.2% (116/500, cells/well) respectively (all shown with mean from at least three independent assays). Furthermore, the cell scratch assay was used to compare the horizontal migration ability of cells with up- and downregulated FoxM1. The results revealed that 48 hours after the scratch wound was inflicted, the percentage of the scratch repaired in the FoxM1 overexpression group (HCCC-9810-FoxM1) was significantly higher than that in the control group (HCCC-9810-Control) (P < 0.01, Fig. 4A), while the percentage of the wound closed in the FoxM1 downregulated group (SSP-25-shFoxM1) was markedly lower than that of its control group (SSP-25-Control) (P < 0.0001, Fig. 4A). We further used Transwell migration assays to compare the vertical motility of the transfected cells. The results showed that 48 hours after the start of the experiment, the number of penetrating cells in the HCCC-9810-FoxM1 group was notably greater than that of the HCCC-9810-Control group (P < 0.0001, Fig. 4B), while the number of penetrating cells in the SSP-25-shFoxM1 group was significantly lower than that in the SSP-25-Control group (P < 0.01, Fig. 4B). This tendency also existed in the subsequent Transwell invasion assay. The number of cells on the lower surface of the membrane in the HCCC-9810-FoxM1 group was markedly greater than that in the HCCC-9810-Control group (P < 0.01, Fig. 4C), and the number of penetrating cells in the SSP-25-shFoxM1 group was significantly lower than that of the SSP-25-Control group (P < 0.001, Fig. 4C). Together, the above results indicated that FoxM1 enhanced the proliferative, migrative and invasive capacities of ICC cells in vitro.

Figure 3. FoxM1 promoted the proliferation of ICC in vitro. The proliferative abilities of the indicated ICC cells were determined by Brdu incorporation assay (A, B), and cell number counting (C, D), MTT assays (E) and plate cloning tests (F, G, statistics are shown with a bar graph). ^*P < 0.05, ^**P < 0.01, ^***P < 0.001.

Figure 4. FoxM1 promoted the migration and invasion of ICC cells in vitro. The motility of ICC cells was detected by scratch assays (A) and Transwell migration assays (B). The invasiveness of ICC cells was determined by Transwell invasion assays (C). Statistics are shown with a bar graph. ^**P < 0.01, ^***P < 0.001, ^****P < 0.0001.

FoxM1 promoted ICC progression in vivo

We then examined the effects of FoxM1 on ICC progression in vivo with a nude mouse liver tumorigenicity model (Fig. 5A, B, C, and D). The number of HCCC-9810-FoxM1-derived tumor nodules (including metastasis) was significantly greater than that in the HCCC-9810-Control group (P < 0.01, Fig. 5E). In contrast, the number of SSP-25-shFoxM1-derived tumor nodules (including metastasis) was markedly lower than that in the SSP-25-Control group (P < 0.01, Fig. 5G). Similarly, the weight of the liver including the HCCC-9810-FoxM1-derived tumors was significantly heavier than that of its control group (P < 0.0001, Fig. 5F), while the weight of liver including SSP-25-shFoxM1-derived tumors was markedly lighter than that of its control group (P < 0.001, Fig. 5H). These results show that FoxM1 promotes ICC progression in vivo.

Figure 5. Overexpression of FoxM1 promoted ICC progression in vivo. Representative images of nude mouse ICC hepatic xenograft models (red arrow head show primary xenografted tumor, while the yellow arrow indicates metastasis in (B-a, B-b and C-a, C-b) and HE-stained sections (A-c, B-c, C-c, and D-c: magnification × 40; A-d, B-d, C-d, and D-d: magnification × 400). The total number of tumor nodules/animal (E, G) and liver weight (g)/animal (F, H) of the hepatic xenograft tumors showed marked differences between the experimental and control groups (n = 5 per group). Statistics are shown with a bar graph. ^**P < 0.01, ^***P < 0.001, ^****P < 0.0001.

Overexpression of FoxM1 upregulates the activity of the c-Myc signaling pathway

Currently, bioinformatics is a powerful tool for data mining in the microarray RNA database. In this study, GSEA analysis was performed to find potential differentially expressed genes and the related signaling pathways based on the high/low expression of FoxM1 mRNA in ICC. The results of the GSEA gene signaling pathway enrichment analysis are shown in Table 3, including the c-Myc, P53, WNT canonical, c-MET, and ATM pathways, which are involved in the proliferation, invasion, metastasis, and apoptosis of ICC. The results indicated that FoxM1 may play a wide range of roles in different disease stages of ICC.

The GSEA analysis showed that the MYC_ACTIV_PATHWAY (Fig. 6A) and the MYC_PATHWAY (Fig. 6B) had higher enrichment values, and both FDR q-values were < 0.250 when FoxM1 was highly expressed, which indicated that the expression levels of FoxM1 affect the activity of the MYC_PATHWAY. Western blotting then confirmed that the expression levels of the c-Myc protein in HCCC-9810-FoxM1 were significantly higher than those in HCCC-9810-Control, while expression levels of the c-Myc protein in SSP-25-shFoxM1 were markedly lower than those in SSP-25-Control (all P < 0.0001, Fig. 6C). Those results showed that the overexpression of FoxM1 in ICC upregulates the activity of the c-Myc signaling pathway.

Figure 6. GSEA (Gene Set Enrichment Analysis) of the c-Myc signaling pathway based on the high and low expression of FoxM1 in ICC. The results indicated marked correlations of FoxM1 expression in ICC with the c-Myc signaling pathway (A and B). Subsequently, Western blotting verified that the overexpression of FoxM1 upregulates the activity of c-Myc signaling (C) and promotes MMP-9 and MMP-2 (D) protein expression in ICC. Statistics are shown with a bar graph. ^**P < 0.01, ^***P < 0.001, ^****P < 0.0001.

Overexpression of FoxM1 promotes MMP-9 and MMP-2 protein expression in ICC

Furthermore, to investigate the underlying mechanisms by which FoxM1 promotes ICC invasion and metastasis, Western blotting was used to detect the expression of the MMP-9 and MMP-2 proteins in the transfected cells. The results showed that the expression levels of the MMP-9 and MMP-2 proteins in HCCC-9810-FoxM1 were significantly higher than those in HCCC-9810-Control, while the expression levels of the MMP-9 and MMP-2 proteins in SSP-25-shFoxM1 were markedly lower than those in SSP-25-Control (all P < 0.01, Fig. 6D). This result indicated that high expression levels of FoxM1 in ICC cells enhance MMP-9 and MMP-2 protein expression; FoxM1 may promote the invasion and metastasis of ICC cells by regulating the expression of the MMP-9 and MMP-2 proteins.

Clinical ICC samples

In this study, 184 patients diagnosed for the first time with ICC who underwent hepatectomy at the First Affiliated Hospital of Sun Yat-sen University in Guangzhou from April 2004 to September 2015 were consecutively selected. All cases were pathologically diagnosed as ICC. All patients were older than 18 years old and had complete clinical follow-up and pathological data. No other anticancer treatment measures were performed before surgery, including radiochemotherapy, immunotherapy, and percutaneous ablation. Paired paraneoplastic tissues were defined as intrahepatic tissues with no infiltrating tumor cells within 2 cm of the ICC tumor tissues. The degree of tumor differentiation was determined based on the 2010 version of the World Health Organization (WHO) classification of tumors of the digestive system [40]. The TNM stage was assigned according to the seventh edition of the American Joint Committee on Cancer (AJCC) Cancer Staging Manual [41]. This study was approved by the ethics committee of Sun Yat-sen University, and all patients provided written consent. Routine postoperative follow-up was performed with all patients. The disease-free survival (DFS) was defined as the period from the day of surgery to the time of tumor relapse, and the overall survival (OS) was defined as the period from the day of the surgery to the time of ICC-related death. Data were censored for patients without recurrence or cancer-related death at the time of the last follow-up (August 2016), as previously reported [19].

Archived paraffin-embedded tumors and paired peritumoral tissues derived from all 184 patients with ICC were used for immunohistochemistry. Another 24 paired samples from patients with ICC were randomly collected between October 2015 and September 2016 for the analysis of FoxM1 mRNA expression, from which 8 tumors and the paired peritumoral tissues were also used for the analysis of the protein expression levels of FoxM1 (details for those patients are listed in Supplementary Table S1).

Quantitative Real-Time PCR (qRT-PCR)

Following the manufacturer's instructions, total RNA was extracted using TRIzol reagent (Invitrogen, USA) from the tissue samples frozen in liquid nitrogen or from the ICC cell lines. cDNA was derived from RNA by reverse transcription using the PrimeScript^TM RT Master Mix (Takara, Japan). Subsequently, the qRT-PCR analysis was performed using SYBR^®Premix Ex Taq^TM II (Takara, Japan) on the Bio-Rad platform. The sequences of the primers used for human FoxM1 were 5’-CGTCGGCCACTGATTCTCAAA-3’ (forward) and 5’-GGCAGGGGATCTCTTAGGTTC-3’ (reverse). The relative mRNA expression was assessed by comparative cycle threshold methods and normalized to the internal control GAPDH.

Western blotting

Western blotting was performed as described previously [42]. Briefly, the proteins extracted from the tissues or cells were subjected to SDS-PAGE and then transferred onto polyvinylidene difluoride (PVDF) membranes (Millipore, USA). The PVDF membranes were blocked with 5% bovine serum albumin (BSA) for 1 h at room temperature and then incubated with the corresponding primary antibodies overnight at 4˚C. GAPDH was used as an internal loading control. Subsequently, the PVDF membranes were washed adequately with TBST and incubated with the corresponding secondary antibodies. Finally, the PVDF membranes were detected by enhanced chemiluminescence solution (Millipore, USA). The primary antibodies used for Western blotting are listed in Supplementary Table S2.

Immunohistochemistry (IHC)

Rabbit polyclonal anti-human FoxM1 antibodies (1:100, Abcam, USA) were used for IHC according to the manufacturer's instructions. Antigen heat retrieval was performed with 0.01 mol/L sodium citrate buffer (pH 6.0) before beginning the IHC staining protocol. The specific procedures for the IHC were performed as previously described [43]. Five random fields under the microscope for each slice were used for IHC scoring, with the mean as the final result. The results of IHC staining were judged by two investigators independently based on the following criteria: 1) The extent of positivity according to the proportion of positively stained cells as follows: 0, no positive staining cells; 1, 0-25%; 2, 26-50%; 3, 51-75%; and 4, > 75%; 2) Staining intensity was scored as follows: 0, negative staining; 1, weak staining (light yellow); 2, moderate staining (yellow-brown); 3, strong staining (brown); and 3) The immunoreactivity scores (IRSs) were determined by adding the above scores, yielding a range from 0 to 7. Samples with an IRS > 4 and those with an IRS ≤ 4 were dichotomized as the high and low expression groups, respectively.

Cell lines and lentiviral transfection

Three human ICC cell lines, HCCC-9810, RBE and SSP-25, were purchased from the Cell Bank of Chinese Academy of Sciences (Shanghai, China). The cells were cultured in RPMI-1640 medium (Gibco, USA) contained 10% fetal bovine serum (FBS) (Gibco, USA), 100 U/ml penicillin (Gibco, USA), and 100 μg/ml streptomycin (Gibco, USA) in a humidified incubator at 37°C with 5% CO₂. The short hairpin FoxM1 lentiviral vectors (GeneChem Co., Ltd. Shanghai, China) designed to downregulate the expression of FoxM1 were transfected into SSP-25 cells, which were then designated SSP-25-shFoxM1. The lentiviral vectors with FoxM1 were transfected into HCCC-9810 cells, which were then designated HCCC-9810-FoxM1. Meanwhile, the empty lentiviral vectors were transfected into SSP-25 and HCCC-9810 cells as controls, which were designated SSP-25-control and HCCC-9810-control, respectively. The lentiviral vectors carried puromycin resistance and green fluorescent sequences. Subsequently, puromycin (1 µg/ml) was used for screening the stable cell lines. The transfection procedure was performed based strictly on the manufacturer’s instructions. The short hairpin RNA (shRNA) and cDNA clone sequences are listed in Supplementary Table S3.

Cell counting and Bromodeoxyuridine (BrdU) incorporation and immunofluorescence assay for cell proliferation

BrdU analysis were performed according to described previously [44]. For proliferation assay, cells were placed in a 6-well plate at a concentration of 5×10⁴ cells per well. After incubation for 24–72 hours, triplicate wells were harvested and cells were counted with a hemocytometer after trypan blue staining. For Brdu incorporation assay, cells were plated in 12-well plate. 48 hours later, cells were incubated with BrdU at final concentration of 10uM for 4hrs. Cells were then fixed by 4% paraformaldehyde at RT for 10mins, followed by 0.1% Triton X-100 permeabilization and 1.5N HCl DNA hydrolysis. Cells were then probed by O/N incubation with BrdU antibody (1:100, Abcam, ab221240). Samples were then subjected to DAPI nuclear counterstain. BrdU positive cells were then quantified and data were presented as percentage of BrdU positive cells.

MTT [3-(4,5-Dimethyl-2-thiazolyl)-2,5-diphenyl-2H-tetrazolium bromide] assays

MTT assays were performed to investigate the cell proliferative ability of HCCC-9810-FoxM1 and SSP-25-shFoxM1 compared with their respective controls. These four cell lines were added to 96-well plates at a concentration of 2 × 10³ cells/well and were incubated for 24 hours, 48 hours, and 72 hours in an incubator with 37°C and 5% CO₂. Then, the MTT solution (20 µl) (KeyGEN BioTECH, Nanjing, China) was added to each well at the indicated time point and further incubated for 4 h. Dimethyl sulfoxide reagent (150 µl) was added to replace with original culture fluid, after which the plates were shaken at 37°C for 10 minutes. The absorbance at 490 nm was measured to detect the number of living cells in each well.

Plate colony-forming assays

The effects of FoxM1 up/downregulation on cell proliferation in vitro were also investigated by plate colony-forming assays. HCCC-9810-FoxM1, HCCC-9810- Control, SSP-25-shFoxM1 and SSP-25-Control in the logarithmic phase were digested by trypsin, counted, seeded into six-well plates at 500 cells/well, and then incubated for 14 days under conventional cell culture conditions. After three washes with PBS, the clone points were fixed with methanol at room temperature and then stained with 0.1% crystal violet. Subsequently, the cell colonies were carefully washed again, photographed, and counted.

Scratch, transwell migration and invasion assays

Scratch, Transwell migration and invasion assays were used to detect the motility and invasiveness of the 4 tumor cell lines. The scratch experiment was performed to evaluate the horizontal motor ability of tumor cells. Cells in the logarithmic phase were grown to confluence in 6-well plates, and a wound line was scratched into the cell monolayer using a sterile 200 µL pipette tip. The detached cells were removed by washing 2~3 times with sterile PBS, which was then replaced by the serum-free RPMI-1640 medium. The percent wound closure was observed by a Leica inverted microscope at 0 and 48 hours after the scratch was generated.

In addition, to assess the vertical motion capability and invasiveness of tumor cells, Transwell migration and invasion assays were performed, respectively. A total of 5 × 10⁴ cells suspended in 50 μl of serum-free RPMI-1640 medium was added to the upper chamber of the insert with/without Matrigel (Transwell migration/invasion assays, respectively), and 600 μl of RPMI-1640 containing 10% FBS was placed into the lower chamber. After a 48-h incubation at 37°C and 5% CO₂, the cells on the lower surface of the insert were counted after fixation in 4% paraformaldehyde and stained with 0.1% crystal violet for 30 min. The number of cells in 9 random microscopic fields (magnification, × 100) of each Transwell chamber were counted and photographed. All results are presented as the mean ± standard error (SE) of 3 independent experiments conducted in triplicate.

Animal studies and HE staining

The animal studies were approved by the Ethics Committee of the First Affiliated Hospital of Sun Yat-sen University (Guangzhou, China) (Grant number: 2014-12). Male athymic BALB/C nude mice (4 weeks old) were purchased from the model animal research center of Nanjing University (Nanjing, China). Tumor cells (1×10⁷) in the logarithmic phase that were suspended in 200 μL of serum-free RPMI-1640 medium and Matrigel (BD Biosciences, USA; dilution, 1:1) were subcutaneously implanted into the flanks of the mice (2 mice in each group). Three weeks later, the xenografts were observed in all treated nude mice, with the largest diameter ranging from 0.5 cm to 1.2 cm. Then, those mice were sacrificed, and the xenografts were excised cut into pieces (1 × 1 × 1 mm³ per piece). The remaining nude mice (4 weeks old, n = 5 per group) were anesthetized by intraperitoneal injection of 2% sodium pentobarbital (45 mg/kg). A transverse upper abdominal incision into the abdominal cavity was used to expose the liver, and then a tumor mass was implanted in the largest hepatic lobe. All surgical operations were carried out on a sterile, clean bench in the SPF operating room and were completed within 2 hours without antibiotics. Eight weeks after the operation, all nude mice were euthanized and examined by exploratory surgery. The liver and lungs were quickly removed, and the livers were weighed and photographed. HE staining was performed using paraffin-embedded sections derived from 10% formalin-fixed tissue.

Bioinformatics analysis (Gene Set Enrichment Analysis, GSEA)

The mRNA data of 36 patients with ICC were downloaded from The Cancer Genome Atlas (TCGA) (https://cancergenome.nih.gov/) database. The 36 ICC samples were divided into a high expression group and a low expression group according to the median total mRNA expression of FoxM1. Subsequently, a bioinformatics analysis was performed by the GSEA software (version 3.0) to find the differentially expressed genes and related signaling pathways derived from the high/low expression of FoxM1 in ICC [45,46]. In the GSEA analysis, we used the Molecular Signature Database (MsigDB) provided by the GSEA as the signaling pathways annotation source, as it contains common pathways annotation data from KEGG and REACTOME. Significantly enriched signaling pathways were screened through a permutation of 1000 cycles.

Statistical analysis

The statistical analysis were performed with SPSS version 20.0 software for Windows (IBM, Chicago, IL, USA). Data are presented as the mean ± SE from at least three independent assays. Numerical variables were compared using Student's t-test or one-way ANOVA. The χ² test or Fisher's exact test were performed to evaluate the differences in categorical data, and the correlations among variables were compared by Pearson’s bivariate correlate analysis. The survival differences were assessed using the Kaplan-Meier method with a log-rank test. The Cox regression proportional hazards model (backward stepwise) was performed to determine the independent prognostic predictors. P < 0.05 (two-tailed) was considered statistically significant.

Ethics approval

This study was approved by the ethics committee of Sun Yat-sen University, and all patients provided written consent. All animal procedures were performed according to national guidelines and approved by the Ethics Committee of the First Affiliated Hospital of Sun Yat-sen University (Guangzhou, China) (Grant number: 2014-12).