Dickkopf-related protein 1 (DKK1) has been reported involved in metastasis and invasion in several tumors. This study sought to investigate the prognostic value of DKK1 in intrahepatic cholangiocarcinoma (ICC) and its role in promoting ICC metastasis.
Tissue microarrays of 138 ICC patient samples were employed to detect DKK1, vascular endothelial growth factor C (VEGF-C), and matrix metalloproteinase 9 (MMP9) expression using immunohistochemistry. The prognostic significances were assessed by Kaplan-Meier survival estimates. DKK1 expression was measured in an ICC cell line (HCCC-9810) and ICC tissues by immunofluorescence assay, quantitative real-time polymerase chain reaction, and western blot. Serum levels of DKK1 from 37 ICC patients were tested by enzyme-linked immunosorbent assay. The role of DKK1 in proliferation, migration, invasion, and gene expression regulation was assessed by DKK1 depletion using small interfering RNA.
Multivariate analyses revealed that DKK1 was an unfavorable predictor for overall survival and time to recurrence. The prognostic significance was retained in ICC patients with low recurrence risk (P < .05). DKK1 expression was elevated in an ICC cell line, tumor samples, and patient sera. High levels of DKK1 in ICC tissues correlated with elevated MMP9, VEGF-C, and metastasis of hepatic hilar lymph nodes. DKK1 depletion caused a decrease in cell migration and invasiveness, and down-regulation of MMP9 and VEGF-C expression.
Intrahepatic cholangiocarcinoma (ICC), a primary adenocarcinoma arising from the epithelial cells of intrahepatic bile ducts, is the second most common primary liver cancer, accounting for 4% to 6% of all cases.1,2 The morbidity and mortality of ICC are rising in recent years,3 and surgery has been regarded as the only effective treatment for this disease.4 The resectability of ICC remains very low, and the outcome after resection has not improved over the past decade.5 The poor outcome of ICC is due mainly to late diagnosis, which prevents effective therapeutic options, leading to high rates of recurrence and metastasis after operations.6 Although several diagnostic and prognostic biomarkers for ICC have been described recently,7–9 there still remains a lack of ideal biomarkers that can be widely accepted for use in clinical settings.
Dickkopf-related protein 1 (DKK1) is a secretory protein that was discovered in embryonic development.10,11 The clinical and prognostic significance of DKK1 has been reported in many malignant tumors, such as breast cancer, lung cancer, esophageal carcinoma, and myeloma.12–19 However, the expression of DKK1 in ICC has not been reported and its clinical significance in ICC is still unclear.
In this study, we found DKK1 was an independent prognostic factor for ICC after surgery. High expression of DKK1 in ICC tissues was associated with elevated matrix metalloproteinase 9 (MMP9), vascular endothelial growth factor C (VEGF-C) expression, and high lymph node metastasis. DKK1 might enhance tumor cell invasion and promotes lymph node metastasis of ICC through the induction of MMP9 and VEGF-C.
MATERIALS AND METHODS
Patients and Follow-Up
Tumor specimens used in tissue microarray (TMA) assays were obtained from 138 consecutive ICC patients who underwent resection in the Liver Cancer Institute, Zhongshan Hospital, Fudan University, China, between February 2001 and November 2006. Fresh tumor tissues and corresponding peritumoral tissues used in quantitative real-time polymerase chain reaction (qRT-PCR) and western blot analyses were randomly chosen from the tissue bank of Zhongshan Hospital. All specimens were taken and preserved as described.8
Serum samples were collected from 37 patients with ICC and 50 healthy volunteers in 2009. Serum was obtained before each operation and stored at −80°C.
The enrollment criteria for all patients in this study were the same as in the previous report.8 The seventh edition of the tumor-node-metastasis (TNM) classification system was used.20 Ethical approval was obtained from the Zhongshan Hospital Research Ethics Committee, and informed consent was obtained from each patient. The detailed clinicopathological characteristics for 138 patients with ICC are listed in Table 1. The follow-up procedures were described in our previous article,21 and all data were summarized at the end of March 15, 2011. The median follow-up was 28 months (range, 1–109 months).
Table 1. Correlation Between DKK1 and Clinical Characteristics in 138 ICC Patients
TMAs were constructed and immunohistochemistry was performed as described.21 Polyclonal rabbit anti-human DKK1 (sc-25516, 1:50 dilution; Santa Cruz Biotechnology, Santa Cruz, Calif), polyclonal rabbit anti-human VEGF-C antibody (AP2042d, 1:50 dilution; Abgent, San Diego, Calif), and polyclonal rabbit anti-human MMP9 antibody (AP6214a, 1:50 dilution; Abgent) were used to detect the expression of DKK1, VEGF-C, and MMP9, respectively.
Evaluation of Immunohistochemical Variables
Immunohistochemical staining was assessed as described.21 The intensity of DKK1 staining was evaluated as follows: strong positive samples (scored as 3+) had dark brown staining in > 50% of tumor cells, completely obscuring the cytoplasm; moderate positive (scored as 2+) had dark brown staining in 25% to 50% of tumor cells obscuring the cytoplasm; weak positive (1+) showed a lesser degree of brown staining in the tumor cell cytoplasm; and, absent (scored as 0) had no appreciable staining in tumor cells. Strong and moderate scores were regarded as positive results.17 The intensity of VEGF-C staining was scored according to a previous report with intensities of 0% to 50% being considered negative and those > 50% as positive.22 The intensity of MMP9 staining was scored as described,23 with 0 as negative, 1 as focally positive (1%-10% positive cells in the lesion), 2 as moderately positive (11%-50%), and 3 as markedly positive (more than 50%). Cases with scores 2 or 3 were considered positive cases in this study.
A human ICC cell line, HCCC-9810 (purchased from the Chinese Academy of Sciences Shanghai Branch Cell Bank, Shanghai, China) was maintained as described.8 The intrahepatic biliary epithelial cell line IBEpiC (ScienCell Research Laboratories, Carlsbad, Calif) was maintained in Epithelial Cell Medium (EpiCM) (ScienCell) at 37°C in a humidified incubator under 5% CO2 conditions.
qRT-PCR, Immunofluorescence, and Western Blot Analyses
Thirty-six ICC samples and matched peritumoral tissues were randomly chosen and analyzed by qRT-PCR as described.21 The primers were as follows: DKK1 forward primer: 5′-TAGAGTCTAGAACGCAAGGATCTC-3′ and reverse primer: 5′-CAAAAACTATCACAGCCTA AAGGG-3′; glyceraldehyde 3-phosphate dehydrogenase (GAPDH) forward primer: 5′-GGGAGCCAAAAGG GTCATCATCTC-3′ and reverse primer: 5′-CCATGC CAGTGAGCTTCCCGTTC-3′; VEGF-C forward primer: 5′-CAAGGCCCCAAACCAGTAACAAT-3′ and reverse primer: 5′-GCTGGCAGGGAACGTCTAATAA TG-3′; MMP9 forward primer: 5′-TCTGCCCGGAC CAAGGATACA-3′ and reverse primer: 5′-AGGCCGT GGCTCAGGTTCAG-3′.
DKK1 expression in HCCC-9810 and IBEpiC cells was detected by immunofluorescence assay as described.24 The antibodies used were monoclonal mouse anti-human DKK1 (H00022943-M11, 1:50 dilution; Novus Biologicals, Littleton, Colo).
Thirty-three ICC tumor and 29 peritumoral tissues were further used for western blot analysis as described.21 Monoclonal mouse anti-human DKK1 (H00022943-M11, 1:200 dilution; Novus Biologicals) antibody was used. GAPDH (KC-5G4, 1:5000 dilution; Kangchen Pharmaceuticals, China) was used as an internal control. The bands were quantified by Image J 1.44p software (from Wayne Rasband, National Institutes of Health, Bethesda, Md).
Enzyme-Linked Immunosorbent Assays
The concentrations of protein in DKK1 culture medium or in human serum were measured using standard enzyme-linked immunosorbent assay (ELISA) kits (R&D Systems, Minneapolis, Minn) according to the manufacturer's instructions. Cell culture medium experiments were performed as follows. Twenty-four hours after the cells were cultured in 25-cm2 flasks, they were aspirated, washed with phosphate-buffered saline twice, and dispensed into 0.5 mL fresh culture medium. At 48 and 72 hours after cell culturing, the medium was aspirated for ELISA and cell numbers were counted in each flask for accurate evaluation of DKK1 expression. The final result was calculated as follows, with OD indicating optical density at 450 nm wavelength: [OD (sample cell culture medium) − OD (empty culture medium)]/sample cell number. ELISAs were performed as described.25
Cell Proliferation, Migration, and Matrigel Invasion Assays
The functional role of DKK1 in ICC cells was assessed using small interfering RNA (siRNA). The target siRNA sequences were siRNA-DKK1-1: CCAATACTACAAG AGACAGAT, siRNA-DKK1-2: CCUUGAACUCGGU UCUCAA, Negative control: UUCUCCGAACGUGU CACGU. The transfection of siRNA into HCCC-9810 was done using Lipofectamine 2000 (Invitrogen) according to the manufacturer's instructions.
Cell proliferation, cell migration, and Matrigel invasion assays were done as described.21,24
Microarray-Based Gene Expression Profile
The 35 K oligonucleotide microarrays were constructed by Capital Bio Corp (Beijing, China). Gene expression profile changes between the DKK1 siRNA group and control group were analyzed as reported.24
Statistical analyses were performed with the SPSS software package (version 19.0; IBM, Armonk, NY). The receiver operating characteristic (ROC) analysis was used to determine the diagnostic potential of DKK1. The chi-square test and student t test were used for comparison between groups. Overall survival (OS) and time to recurrence (TTR) were defined as described.24 The prognostic significance was assessed by log-rank test. The correlation between expression of DKK1, VEGF-C, and MMP9 was analyzed using Kendall's tau-b correlation. Univariate and multivariate analyses were based on the Cox proportional hazards regression model. P < .05 was considered statistically significant.
High Expression of DKK1 in ICC Tissues Indicates Poor Prognosis of ICC Patients
DKK1 was mainly observed in the cytoplasm of tumor cells, where it exhibited a variety of patterns with respect to staining intensity and percentage of positive cells. A total of 38.4% (53 of 138) of the ICC patients had positive scores for DKK1 in tumor tissues (Fig. 1A).
For the whole study population, the OS and tumor recurrence rates were 54.73% and 55.82% at 1 year, 42.00% and 68.22% at 3 years, and 27.06% and 77.00% at 5 years, respectively. Child-Pugh score, tumor encapsulation, serum γ-glutamyl transferase levels, serum carcinoembryonic antigen levels, TNM stages, and hepatic hilar lymph node metastasis were prognostic factors for OS and/or TTR, based on univariate analysis, (Table 2).
Table 2. Multivariate Analyses of Factors Associated With Survival and Recurrencea
Hazard Ratio (95% CI)
Hazard Ratio (95% CI)
Abbreviations: CA19-9, carbohydrate antigen 19-9; CEA, carcinoembryonic antigen; CI, confidence interval; DKK1, Dickkopf-related protein 1; GGT, serum γ-glutamyl transferase; NA, not adopted; OS, overall survival; TTR, time to recurrence.
Multivariate analysis, Cox proportional hazards regression model. The clinicopathological variables were adopted for their prognostic significance by univariate analyses.
Six cases without CEA and CA19-9 information were not included.
DKK1 was prognostic for OS (P = .003) and TTR (P = .005) in univariate analysis. The 5-year OS and recurrence rates of the DKK1-negative group were significantly higher than those of the DKK1-positive group (31.11% and 73.83% vs 20.95% and 82.36%, respectively; Fig. 2A, B). Because TNM stage and Child-Pugh score were associated with several clinical indices, they were not entered into multiple analyses with these indices to avoid potential bias.3
Using multivariate analysis, DKK1 expression was an independent prognostic factor for both OS (P = .038) and TTR (P = .004). DKK1-positive patients were more than 2 times likely than DKK1-negative patients to suffer from recurrence (hazard ratio = 1.900; 95% confidence interval = 1.225–2.947; Table 2).
Predictive value of DKK1 in the low recurrent risk subgroup of ICC,2,26,27 such as in those with TNM stage I+II (early stage) disease, without hepatic hilar lymph node metastasis, no vascular invasion, and well-differentiated subgroups were further investigated. DKK1 also showed prognostic significance in those subgroups (Fig. 2C, D and Fig. 3A-F).
Correlation of DKK1 Expression With Clinicopathological Features and Recurrence Patterns
Patients with high DKK1 expression were more likely to have hepatic hilar lymph node metastasis (P = .041), high γ-glutamyl transferase levels (P = .026), and be in Child-Pugh class B (P = .044) (Table 1).
For postoperative recurrence patterns, DKK1-positive groups in patients with TNM stage I+II ICC indicated a much higher tendency to metastasis to peritoneal lymph nodes (Table 3).
Table 3. Relationship Between DKK1, VEGF-C, and MMP-9 Expression With Postoperative Recurrence Patterns of TNM Stage I+II ICC Patients
Negative (n = 45)
Positive (n = 30)
Negative (n = 25)
Positive (n = 50)
Negative (n = 24)
Positive (n = 51)
Abbreviations: DKK1, Dickkopf-related protein 1; ICC, intrahepatic cholangiocarcinoma; MMP9, matrix metalloproteinase 9; TNM, tumor-node-metastasis; VEGF-C, vascular endothelial growth factor C.
Lung, brain, bone, or other remote organ
Peritoneal lymph nodes
DKK1 Was Highly Expressed in ICC Cells and Tissues
The DKK1 messenger RNA (mRNA) expression in HCCC-9810 cells was higher than in IBEpiC cells (2−ΔCt values: 0.0011 ± 0.00004 vs 0.00029 ± 0.000008, P < .001). Similar results were observed in western blots and immunofluorescence assays (Fig. 4A).
The immunohistochemistry results in TMAs indicated that there were only 18 of 138 cases (13%) with DKK1-positive expression in the peritumoral tissues, whereas 53 of 138 cases (38.4%) were positive in tumor tissues (Fig. 1A). Both the results of qRT-PCR and western blot also indicated that DKK1 was highly expressed in ICC tissues (Fig. 4B-E).
Patients suffering from ICC recurrence (20 of 36 patients) had higher expression levels of DKK1 mRNA than those without recurrence (16 of 36 patients; 2−ΔCt values: 0.206 ± 0.043 vs 0.107 ± 0.028, P = .036; Fig. 4C). The results of western blot analyses were consistent with those observed in qRT-PCR analyses (Fig. 4D, E).
DKK1 Secreted in ICC Patients' Sera and Cell Supernatant Fractions
High concentrations of DKK1 were detected in the culture medium of HCCC-9810 cells, and low concentrations in the medium of IBEpiC cells (180.03 ± 17.00 ng/mL vs 36.82 ± 17.0 ng/mL per 1 × 106cells in 24-hour cultures and 241.19 ng/mL vs 89.49 ng/mL per 1 × 106cells in 48-hour cultures, all P values < .01, Fig. 4A).
The DKK1 levels in serum samples were determined, and DKK1 concentrations were significantly higher in ICC patients (4.15 ± 2.36 ng/mL vs 1.61 ± 0.41 ng/mL, P < .001; Fig. 4F) than in healthy volunteers. ROC analyses resulted in an optimal cutoff point at a serum DKK1 level of 2.49 ng/mL, which had the highest area under the curve (area = 0.872; sensitivity = 75.7%; specificity = 100%; Fig. 4F)
Inhibition of DKK1 by siRNA Attenuated Migration and Invasiveness of ICC Cells
Successful knockdown of DKK1 expression was confirmed by qRT-PCR, western blot, and ELISA analyses (Fig. 5A, B). Matrigel invasion assays revealed that decreased DKK1 expression was accompanied by impairment in the invasiveness of HCCC-9810 cells (12.96 vs 85.29, P < .001; Fig. 5C). A wound-healing assay revealed an evident delay in the wound closure rate of DKK1 siRNA–transfected HCCC-9810 cells at 48 hours compared with negative control–transfected HCCC-9810 cells (P = .001; Fig. 5D). However, the down-regulation of DKK1 showed no significant influence on cell proliferation (P > .05; Fig. 5E).
Microarray Analysis of Gene Expression
According to microarray analysis, 138 genes were differentially expressed between the DKK1 siRNA–treated HCCC-9810 cells and the scramble siRNA–treated HCCC-9810 cells. Of these, 42 genes were up-regulated (ratio > 1.5) and 96 genes were down-regulated (ratio < 0.67). Among these genes, VEGF-C and MMP9 expression were decreased significantly (ratio = 0.22 and 0.26, respectively). The results of qRT-PCR analysis also confirmed that mRNA of VEGF-C and MMP9 were significantly down-regulated after DKK1 knockdown (P < .05; Fig. 5F).
Correlation of DKK1 Expression With VEGF-C and MMP9
VEGF-C and MMP9 were further stained in TMAs. VEGF-C cytoplasmic staining was observed in cancer cells with low staining in fibroblastic cells (Fig. 1B). A total of 72.5% cases showed positive VEGF-C staining in ICC tissues (100 of 138 cases). MMP9 was expressed in the cytoplasm of ICC cells as a granular pattern and was positively observed in 45.6% of cases (63 of 138 cases) (Fig. 1B). DKK1 expression was positively associated with VEGF-C and MMP9 expression (P < .001; Table 1, Fig. 1B).
We also examined mRNA expression of MMP9 and VEGF-C in 36 ICC tissues. Levels of DKK1 positively correlated with MMP9 (r = 0.477, P = .017) and VEGF-C (r = 0.604, P = .017) mRNA levels in ICC tissue samples (Fig. 4G).
Accumulating evidence has demonstrated that high expression of DKK1 might be a key alteration contributing to the invasion and metastasis of tumor cells.16,17 Recently, DKK1 was also found to be associated with prognosis in several kinds of tumors.13,17 Until now, systematic investigation of the prognostic significance of DKK1 in ICC has not been reported, especially with long-term follow-up and a large number of patients. Thus, the role of DKK1 in ICC progression has not been clearly defined.
Cancer is a heterogeneous disease, and patients at the same clinical stage of disease, with similar histopathological tumor features, and similar treatment strategies (such as surgical resection) can have different clinical outcomes. DKK1 was highly expressed in ICC patients with postoperative recurrence and was related to poor prognosis of ICC patients after surgery. Further analysis of the prognostic significance of DKK1 in clinical subgroups indicated that the OS and TTR of the DKK1-positive patients who had TNM stage I or II disease were significantly worse than DKK1-negative patients in the same stage. This suggests that DKK1 might be a promising prognostic biomarker in early-stage ICC patients. Even after surgical resection, the recurrence rate of early stage ICC was still very high (recurrence rate within 2 years in TNM I+II is 50% in this study), and no strategy could effectively identify these patients. The predictive value of DKK1 in this subgroup can help clinicians to distinguish patients with high recurrent risk and enable clinicians to administer rational adjuvant therapy in a timely manner.
Clinicopathological correlation has shown that high DKK1 expression was associated with lymph node metastasis and advanced tumor stage. ICC is relatively hypovascular and has a tendency for lymphatic metastasis. Early lymphatic spread limits the efficacy of surgery.2 The presence of lymph node metastasis, recognized as the most common metastatic lesion, is regarded as an important prognostic factor for patients with ICC,26 which was consistent with our study. Further investigations indicated that DKK1 depletion by siRNA significantly inhibited the migration and invasion of ICC cells and decreased their MMP9 and VEGF-C expression, which are widely recognized as key molecules in promoting tumor cell invasive potential and lymph node metastasis.22,23,28,29 Consistent with the in vitro data, high DKK1 expression in tumor samples also correlated with high expression of MMP9 and VEGF-C in fresh tissues and TMAs of ICC. All data suggest that DKK1 improves tumor cell invasive potential by MMP9 and promotes lymphatic spread through VEGF-C, which causes the poor prognosis and high recurrence rate of ICC patients after surgery.
The potential for targeting DKK1 in cancer therapy seems promising, because DKK1 is always overexpressed in many human cancers including ICC, but is barely detectable in their adjacent normal tissues.13,16,17 Several recent reports showed that down-regulation of DKK1 using RNA interference or anti-DKK1 monoclonal antibodies has achieved significant inhibition of tumor development in vitro and in vivo.17–19 Our in vitro study using RNA interference also showed that DKK1 might be a promising target protein for preventing ICC metastasis and recurrence.
In conclusion, we have provided the first evidence that ICC patients exhibited up-regulation of DKK1 in tumor tissues and increased serum levels of DKK1. DKK1 enhances the invasive abilities of tumor cells and promotes lymph node metastasis of ICC through the induction of MMP9 and VEGF-C. DKK1 may be a novel prognostic biomarker and a potential therapeutic target for ICC.
We thank Yuan Ji, MD, PhD, and Hai-Ying Zeng for expertise in immunohistochemistry.
This study was supported by grants from the Major Program of National Natural Science Foundation of China (No. 81030038), National Key Sci-Tech Project (2012ZX10002-019-002), the National Natural Science Foundation of China (No. 81000927), and the Research Fund for the Doctoral Program of Higher Education of China (No. 20100071120064).