Abstract
Increasing evidence suggests that there is a unique cell subpopulation in melanoma that can form nonadherent melanospheres in serum-free stem cell medium, mimicking aggressive malignancy. Using melanospheres as a model to investigate progression mechanisms, we found that miR-519d overexpression was sufficient to promote cell proliferation, migration, invasion, and adhesion in vitro and lung metastatic capability in vivo. The cell adhesion receptor EphA4 was determined to be a direct target of miR-519d. Forced expression of EphA4 reversed the effects of miR-519d overexpression, whereas silencing of EphA4 phenocopied the effect of miR-519d. Malignant progression phenotypes were also affected at the level of epithelial-to-mesenchymal transition and the ERK1/2 signaling pathway inversely affected by miR-519d or EphA4 expression. In clinical specimens of metastatic melanoma, we observed significant upregulation of miR-519d and downregulation of EphA4, in the latter case correlated inversely with overall survival. Taken together, our results suggest a significant functional role for miR-519d in determining EphA4 expression and melanoma progression.
Significance: These results suggest a significant role for miR-519d in determining expression of a pivotal cell adhesion molecule that may impact risks of malignant progression in many cancers. Cancer Res; 78(1); 216–29. ©2017 AACR.
Introduction
Malignant melanoma, which is associated with high morbidity and mortality, is a tumor resulting from transformed melanocytes or nevocytes that are derived from the neural crest. Melanoma cells and neural crest stem cells were also found to exploit common molecular mechanisms (1). Historically, neural stem cells are typically cultured in vitro under nonadherent conditions as neurospheres (2). Sphere-forming assays are also widely used in diverse cancer tissues and the cancer stem cell research (3, 4). It has been found that a subpopulation of cells in human melanomas can form nonadherent melanospheres in serum-free stem cell medium (5). In comparison with the adherent melanoma cells, these spheroid cells were demonstrated to have the stem-like properties of self-renewing and multipotency to differentiate under appropriate conditions into multiple cell lineages, such as melanocytes, adipocytes, osteocytes, and chondrocytes (5). The spheroid cells were highly tumorigenic that intradermal injections in SCID mice of as few as 100 cells generated tumors that maintained tumorigenic potential into subsequent recipients (6–8). Although recent evidence has challenged the notion of only a selected subpopulation of cancer cells having the ability to form a melanoma tumor (9), melanospheres still represent an enrichment of heterogeneous cell groups with enhanced clonogenic potential (6). Moreover, melanospheres presented a more aggressive phenotype with higher invasiveness and increased antitumor immune escape (10, 11). Taking advantage of this in vitro spheroid culture model, we exploited this system to further characterize the mechanism of melanoma progression and aggressiveness.
miRNAs are a class of single-stranded, small noncoding RNAs (∼22 nucleotides) that act through complementary interaction with the target mRNAs, resulting in translational repression or cleavage of the target mRNAs (12, 13). Studies have shown that dysregulation of miRNAs are implicated in melanoma formation and progression. As melanospheres represent a cell model of stemness as well as aggressiveness, we searched the literature for a possible miRNA candidate and found miR-519d, one of the members in chromosome 19 miRNA cluster (C19MC), which is the largest primate-specific miRNA cluster (14) preferentially expressed in stem cells (15–17), was shown to be differentially expressed between primary melanomas and metastases (18). In this study, we are interested in investigating whether miR-519d contributed to melanoma progression by studying three-dimensional spheroid system.
EphA4 belongs to the large Eph (erythropoietin-producing hepatocellular carcinoma) family of receptor tyrosine kinases and has been demonstrated to have different, even paradoxical, effects in various human tumor context. For example, upregulation of EphA4 correlated with liver metastasis in colorectal cancer (19), and overexpression of EphA4 could also enhance cell adhesion to facilitate metastasis of hepatocellular carcinoma (20). Through the EphA4–FGFR1 signaling pathway, EphA4 promoted cell proliferation and migration in the human glioma U251 cell line (21). In contrast, EphA4 overexpression in lung adenocarcinoma reduced cell migration and invasion and was associated with an improved patient outcome (22). The roles of EphA4 receptor in melanoma, however, have not been extensively characterized.
Originating from the melanosphere model, the current study focuses on the biologic effect of miR-519d on tumor progression in cutaneous melanoma. We have identified EphA4 as the direct and functional target gene of miR-519d. Collectively, these results highlight the importance of miRNA-regulating mechanisms for EphA4 in modulating the metastatic properties of melanoma and may provide new insights into the pathogenesis of cutaneous melanoma.
Materials and Methods
Melanosphere culture and sphere formation assay
To generate melanospheres, melanoma cells were seeded at a 1,000 cell/cm2 density per well in 6-well ultralow dish (CORNING 3471) in stem cell medium (SCM) consisting of DMEM/F12, supplemented with EGF 20 ng/mL, basic FGF (bFGF) 20 ng/mL (BD Biosciences), B27 (1:50), 0.4% BSA, and insulin 4 μg/mL (Thermo Fisher Scientific). Twice a week, the medium was changed. Every week, single-cell suspensions were generated by enzymatic digestion of melanospheres. For the sphere formation assay, the size of melanospheres was measured, and melanospheres >20 μm in diameter were counted after 6- to 14-day continuous culture. Cells greater than three passages of spheroid culture were used for the sphere experiments.
While culturing melanospheres from murine xenografts, tumor fragments were minced using scissors and incubated in HBSS supplemented with 3 mmol/L calcium chloride and 1 mg/mL collagenase IV (Sigma-Aldrich) for 3 hours at 37°C. DNase I (10 mg/mL; Sigma-Aldrich) was added to reduce clumping, and cells were filtered through a 70-μm pore size filter to remove undigested tissue aggregates. Isolated cells were cultured for 1 day in complete medium and subjected to SCM for sphere culture.
miRNA target prediction
The analysis of miR-519d–predicted targets was performed using two online computational algorithms: TargetScan 5.0-5.2 version and MiRanda 2008.
Luciferase reporter assay
For 3′UTR reporter assays, the 3′UTR region of human EphA4 gene was amplified by PCR and cloned into pMIR-REPORT Luciferase vector (Ambion). Mutant plasmids were generated using a QuickChange Site-Directed Mutagenesis Kit (Stratagene). There were two predicted sites on 3′UTR EphA4 for miR-519d binding, which were located between nucleotide 49-57 (Site 1) and 54-76 (Site 2) of 3′UTR EphA4, respectively. Site 1 was further mutated to 5′- atgcccggt -3′ and 5′- atcggcttt -3′, as mutant 1 and mutant 2, respectively. Site 2 was mutated as mutant 3. Control and miR-519d–overexpressing stable A2058 melanoma cell clones were cotransfected with the indicated firefly luciferase reporter plasmid and a pTK-Renilla luciferase plasmid with a ratio of 10:1. Lysates were collected 48 hours after transfection. Firefly and Renilla luciferase activities were measured with a Dual-Luciferase Reporter System (Promega) according to the manufacturer's instructions.
In vivo tumorigenicity and lung metastasis assay
Animal studies were approved by Institutional Animal Care and Use Committees of National Taiwan University College of Medicine and College of Public Health (Taipei, Taiwan; approval no: 20130473). Male 8-week-old C.B-17/Icr-scid mice were purchased from National Laboratory Animal Center and bred at Laboratory Animal Center of National Taiwan University College of Medicine. For subcutaneous tumorigenicity assay, cells were trypsinized, washed with PBS, and adjusted to a concentration of 102, 103, or 4 × 105 in 0.1 mL PBS depending on the purpose of experiments. The cell suspension was then mixed with an equal volume of Matrigel (Becton Dickinson) before subcutaneous injection into mice. At the indicated time, the mice were killed and the tumors were removed, sized, and weighed. For lung metastasis assay, 5 × 105 cells were resuspended in 0.1 mL PBS and injected into the lateral tail vein. The mice were sacrificed 7 weeks after intravenous injection. The lungs were weighted, fixed, photographed, and the number of visible grayish nodules on the lung surface was counted with a dissecting microscope. The lung tissues were then embedded in paraffin and sectioned at three different levels; the sections were stained with hematoxylin and eosin for light microscopy.
Patient samples and IHC
RNAs from melanoma samples and adjacent noncancerous skin tissues were obtained from surgeries at the National Taiwan University Hospital. Patients' RNAs were reverse transcribed using the SV Total RNA Isolation System (Promega) and underwent qRT-PCR. Overall survival data were collected from chart reviews and the Taiwan Cancer Registry. The study protocol was approved by the Research Ethics Committee of National Taiwan University Hospital (no. 201212073RIND) and was in compliance with the ethical principles of the Declaration of Helsinki.
For IHC, formalin-fixed, paraffin-embedded 5-μm tissue sections were stained according to the procedures described previously (23) with the primary rabbit anti-human EphA4 antibody 1:200 (GTX111602, GeneTex Inc.). 3-Amino-9-ethylcarbazole (Dako) was applied for colorization, and the slides were counterstained for nuclei with hematoxylin.
Standard molecular biology methods as well as statistical analysis can be found in the Supplementary Data.
Results
Melanospheres possess more clonogenic and tumorigenic capacities than adherent cells
A range of diverse melanoma cell lines, including A2058, SK-Mel-28, A375 (three BRAF V600E positive cell lines), SK-Mel-90 (a NRAS Q61K-mutant cell line), and MeWo (a cell line wild type for BRAF and NRAS but with focal NF deletion), were investigated for their ability to form spheres. Melanospheres were successfully grown from these cell lines in stem cell media except MeWo cell line (Fig. 1A), which was the least invasive melanoma cell line among them (5, 23). These melanospheres could be serially propagated for more than 20 passages. When melanospheres were dissociated and then plated in culture medium with serum, the majority of cells became attached to form a monolayer within 2 days. These cells resumed the ability to form spheres when cultured in stem cell medium. In agreement with the previous results (6), we did not detect any specific enrichment of CD20+, CD133+, ABCB5+, or ABCG2+ cells in A2058 melanospheres compared with the adherent monolayer cells by flow cytometry (data not shown). To obtain sphere-forming cells with higher tumorigenic potential, we also cultured melanospheres derived from SCID mouse xenografts, which was developed from subcutaneous injection of the 12th passage of sphere-forming cells. The self-renewing or transformation ability of A2058 adherent, sphere-forming cells, and sphere-forming cells coming from mouse xenografts was assessed by anchorage-independent growth in soft agar. The sphere-forming cells especially those from mouse xenografts had significantly stronger ability to form colonies, compared with the adherent cells (Fig. 1B). We next evaluated the capacity to form tumors by subcutaneous injection of melanoma cells into SCID mice. At a dose of 5,000 and 1,000 cells, both the adherent (A) and sphere-forming (S) cells generated tumors in all of the injected mice. Notably, only sphere-forming cells could develop tumors when as few as 100 cells were injected per mouse. The size and weight of the S100 tumors were even significantly larger than A1000 tumors (Fig. 1C and D). Furthermore, we found that sphere-forming cells were more resistant to cisplatin-induced cytotoxicity in a dose-dependent manner. Cisplatin treatment at the concentration of 30 μmol/L for 24 hours resulted in a significant reduction of cell death in sphere-forming cells compared with the adherent cells. Sphere-forming cells derived from mouse xenografts had even stronger chemoresistance than sphere-forming cells continuously kept in cultures (Fig. 1E). Taken together, our results demonstrated that, compared with adherent melanoma cells, the sphere-forming capacities contributed to the establishment of tumors in vivo and a more drug-resistant potential.
Melanospheres have higher tumorigenic capacities and chemoresistance compared with adherent melanoma cells. A, Sphere formation in A2058, A375, and SK-Mel-90 human melanoma cell lines. Only cell aggregation but no true spheres were observed in MeWo cells. Bar, 100 μm. B, The clonogenic ability as revealed by soft agar assay was enhanced in A2058 sphere-forming cells, especially those cultured from SCID xenografts, compared with the adherent cells. Bar, 1 mm. C, Tumor volume of subcutaneous xenografts derived from 100 or 1,000 A2058 adherent cells (A) and sphere-forming cells (S) in SCID mice. D, The mice were sacrificed 7 weeks after subcutaneous inoculation of adherent or sphere-forming cells in the left and right sides of flank, respectively (n = 5 in each group). Tumors of adherent cells weighed significantly less than those of sphere-forming cells. Adherent cells could not form tumors when as few as 100 cells were injected. Bar, 1 cm. E, Chemosensitivity to cisplatin is decreased in sphere-forming cells. Cells were treated with the indicated concentration of cisplatin for 24 hours, and cell viability was measured by MTT assay. Data are expressed as mean values ± SD; *, P < 0.05; **, P < 0.01.
Melanospheres have higher tumorigenic capacities and chemoresistance compared with adherent melanoma cells. A, Sphere formation in A2058, A375, and SK-Mel-90 human melanoma cell lines. Only cell aggregation but no true spheres were observed in MeWo cells. Bar, 100 μm. B, The clonogenic ability as revealed by soft agar assay was enhanced in A2058 sphere-forming cells, especially those cultured from SCID xenografts, compared with the adherent cells. Bar, 1 mm. C, Tumor volume of subcutaneous xenografts derived from 100 or 1,000 A2058 adherent cells (A) and sphere-forming cells (S) in SCID mice. D, The mice were sacrificed 7 weeks after subcutaneous inoculation of adherent or sphere-forming cells in the left and right sides of flank, respectively (n = 5 in each group). Tumors of adherent cells weighed significantly less than those of sphere-forming cells. Adherent cells could not form tumors when as few as 100 cells were injected. Bar, 1 cm. E, Chemosensitivity to cisplatin is decreased in sphere-forming cells. Cells were treated with the indicated concentration of cisplatin for 24 hours, and cell viability was measured by MTT assay. Data are expressed as mean values ± SD; *, P < 0.05; **, P < 0.01.
miR-519d is overexpressed in melanospheres
Dysregulation of miRNAs has been identified to cause melanoma formation and progression (24). As melanospheres had more tumorigenic potential and performed more aggressive behavior than adherent melanoma cells, we exploited the unique model system to exploring miRNA regulatory mechanisms, with primary importance given to melanoma tumorigenesis and progression. In the study performed by Mueller and colleagues (18), miR-526b was found to be upregulated in primary melanoma cell lines compared with normal human epidermal melanocytes (malignant transformation or early progression), and miR-519d was demonstrated to be overexpressed in metastatic melanomas compared with primary melanomas (metastasis). Both miR-519d and miR-526b belong to the chromosome 19 miRNA cluster (C19MC), the largest human miRNA cluster (16, 25). To determine the expression status of miR-519d and miR-526b, qRT-PCR analysis was performed, which proofed miR-519d and miR-526b were significantly overexpressed in melanospheres by 11.4 ± 1.5-fold and 18.4 ± 4.4-fold, respectively, compared with adherent A2058 cells (Fig. 2A).
miR-519d promotes cell proliferation and sphere formation. A, qRT-PCR analyses of miR-519d and miR-526b expression in A2058 adherent and sphere-forming cells. B, qRT-PCR confirmed successful construction of miR-519d- and miR-526b–overexpressing stable cell clones in A2058 melanoma cells. C, Cell proliferation. Cells were counted by Trypan blue exclusion assay after culturing for the indicated duration. Cells overexpressing miR-519d exhibited significant increased cell proliferation. D, Sphere-forming assay showed significant increased sphere size in the stable cell clone with miR-519d overexpression on day 14 cultures. The sphere diameter was 98.3 ± 20.6 μm for cells with miR-519d overexpression, which was significantly larger than the control (49.7 ± 13.3 μm) and miR-526b cells (59.1 ± 2.6 μm). Bar, 100 μm. E, Sphere-forming assay showed significant increased sphere number in the miR-519d–overexpressing stable cell clone (left). The addition of anti–miR-519d in miR-519d–overexpressing cells reduced the sphere number (right). F, Inhibition of miR-519d function with anti–miR-519d reduced the sphere size on day 9 cultures (51.3 ± 14.5 mm vs. 84.9 ± 26.6 mm in the control). Bar, 100 μm. *, P < 0.05; **, P < 0.01.
miR-519d promotes cell proliferation and sphere formation. A, qRT-PCR analyses of miR-519d and miR-526b expression in A2058 adherent and sphere-forming cells. B, qRT-PCR confirmed successful construction of miR-519d- and miR-526b–overexpressing stable cell clones in A2058 melanoma cells. C, Cell proliferation. Cells were counted by Trypan blue exclusion assay after culturing for the indicated duration. Cells overexpressing miR-519d exhibited significant increased cell proliferation. D, Sphere-forming assay showed significant increased sphere size in the stable cell clone with miR-519d overexpression on day 14 cultures. The sphere diameter was 98.3 ± 20.6 μm for cells with miR-519d overexpression, which was significantly larger than the control (49.7 ± 13.3 μm) and miR-526b cells (59.1 ± 2.6 μm). Bar, 100 μm. E, Sphere-forming assay showed significant increased sphere number in the miR-519d–overexpressing stable cell clone (left). The addition of anti–miR-519d in miR-519d–overexpressing cells reduced the sphere number (right). F, Inhibition of miR-519d function with anti–miR-519d reduced the sphere size on day 9 cultures (51.3 ± 14.5 mm vs. 84.9 ± 26.6 mm in the control). Bar, 100 μm. *, P < 0.05; **, P < 0.01.
miR-519d promotes melanoma cell proliferation and sphere formation
To determine whether miR-519d and miR-526b had an effect on the malignant phenotype of melanoma cells, we stably overexpressed pre–miR-control, pre–miR-519d, or pre–miR-526b in A2058 cells. Quantitative RT-PCR showed that the establishment of the stable clones was successful, with a 13.2-fold increase of miR-519d and a 14.2-fold increase of miR-526b, similar to their respective mRNA levels in melanospheres (Fig. 2B). The growth property of the three stable clones was first assessed by MTT assay. We found that the proliferation rate was significantly increased in miR-519d–overexpressing cells compared with the control-transfected cells, but there was no statistically significant difference between miR-526b–overexpressing and control cells (Fig. 2C). Moreover, overexpression of miR-519d in A2058 cells enhanced sphere-forming ability, including the sphere size and number, compared with the vector control (Fig. 2D and E). However, no significant increase of sphere formation was found in miR-526b clones. Therefore, we did not investigate miR-526b in the following study because no sphere-enhancing ability was found. To investigate whether miR-519d could regulate sphere formation under endogenous conditions, we blocked the function of miR-519d by the respective antagomiR (anti–miR-519d) in miR-519d–overexpressing cells and found that miR-519d was sufficient to significantly reduce the sphere number and size (Fig. 2E and F).
miR-519d promotes melanoma cell migration, invasion, adhesion, and in vivo lung metastasis
To investigate the role of miR-519d in melanoma cell migration, A2058 stable cell clones with or without miR-519d overexpression were used for the wound scratch and transwell migration assays. We found that miR-519d overexpression significantly increased cell migratory velocity as revealed by the wound scratch assay (Fig. 3A) and transwell migration (Fig. 3B). Adhesion to extracellular matrix is an essential process for metastatic planting of cancer cells in distant locations. We demonstrated that overexpression of miR-519d significantly enhanced cell adhesion by 1.7 ± 0.07-, 1.7 ± 0.12-, and 1.6 ± 0.42-fold on fibronectin, laminin, and poly-l-lysine coating wells, respectively, compared with the controls (Fig. 3C). Overexpression of miR-519d also increased invasion capacity of A2058 as evidenced by Boyden chamber invasion assay (Fig. 3D).
miR-519d enhances cell migration, adhesion, invasion in vitro, and lung metastasis in vivo. A and B, The wound scratch assay (A) and transwell migration (B) showed a significant increased migratory velocity in A2058 melanoma cells with miR-519d overexpression. C, Adhesion assay on fibronectin, laminin, or poly-l-lysine showed enhanced adhesion of cells with miR-519d overexpression. D, Using the Boyden chamber invasion assay, representative images with the cresyl violet stain (top, magnification, ×200) and the bar graph demonstrated the status of cellular invasion between cells with or without miR-519d overexpression. E, The expression levels of mesenchymal proteins, including fibronectin, vimentin, N-cadherin, Snail, Slug, and Twist, were increased when miR-519d was overexpressed. F, miR-519d enhanced lung metastasis in vivo. The number of lung metastatic melanoma nodules was significantly increased in mice receiving intravenous injection of melanoma cells overexpressing miR-519d in comparison with the control. There were 27.0 ± 2.0 nodules in the miR-519d–overexpressing group versus 3.8 ± 1.8 nodules in the control group (n = 6 in each group). Tumor nodules are shown in representative grossly dissected lungs (top). **, P < 0.01.
miR-519d enhances cell migration, adhesion, invasion in vitro, and lung metastasis in vivo. A and B, The wound scratch assay (A) and transwell migration (B) showed a significant increased migratory velocity in A2058 melanoma cells with miR-519d overexpression. C, Adhesion assay on fibronectin, laminin, or poly-l-lysine showed enhanced adhesion of cells with miR-519d overexpression. D, Using the Boyden chamber invasion assay, representative images with the cresyl violet stain (top, magnification, ×200) and the bar graph demonstrated the status of cellular invasion between cells with or without miR-519d overexpression. E, The expression levels of mesenchymal proteins, including fibronectin, vimentin, N-cadherin, Snail, Slug, and Twist, were increased when miR-519d was overexpressed. F, miR-519d enhanced lung metastasis in vivo. The number of lung metastatic melanoma nodules was significantly increased in mice receiving intravenous injection of melanoma cells overexpressing miR-519d in comparison with the control. There were 27.0 ± 2.0 nodules in the miR-519d–overexpressing group versus 3.8 ± 1.8 nodules in the control group (n = 6 in each group). Tumor nodules are shown in representative grossly dissected lungs (top). **, P < 0.01.
Because epithelial-to-mesenchymal transition (EMT), characterized by loss of intercellular adhesions and acquisition of a migratory phenotype, plays a major role in cancer cell metastasis, we examined whether EMT markers were changed under miR-519d overexpression. We found that the protein levels of mesenchymal markers, including fibronectin, vimentin, N-cadherin, Snail, Slug, and Twist, were upregulated following the overexpression of miR-519d (Fig. 3E). Interestingly, E-cadherin protein was not detected in A2058 cells (Supplementary Fig. S1). It has been reported that certain melanoma cell lines, like WM115 and WC62, also did not express detectable E-cadherin mRNA (26). To further investigate the effect of miR-519d on metastasis of melanoma in vivo, we injected the control or miR-519d–overexpressing stable cell clones into the lateral tail vein of SCID mice and detected lung metastatic nodules under 400× field dissecting microcopy 5 weeks after injection. The number of lung metastatic nodules was dramatically increased when miR-519d was overexpressed (Fig. 3F). These findings suggested that miR-519d played a tumor-promoting role in melanoma.
miR-519d targets the 3′UTR of EphA4 transcripts and downregulates EphA4 expression
miR-519d's ability to coordinate melanoma progression can be attributed to multiple direct and indirect regulation of target genes. To identify the direct targets of miR-519d, we performed the 3′UTR binding site prediction by TargetScan 6.2 (27) and MiRanda (28). EphA4 was chosen as our candidate molecule because it was differently expressed in a significant degree in both prediction tools. Although MiRanda and TargetScan algorithms identified CDKN1A/p21, PTEN, PPAR1, and AKT3 as hypothetical targets of miR-519d, we found no correlation between miR-519d expression and protein levels of CDKN1A/p21, PTEN, and PPAR1 (Supplementary Fig. S2). We examined the effect of miR-519d on EphA4 expression on mRNA and also on protein levels. Nevertheless, qRT-PCR and Western blot analysis demonstrated that overexpression of miR-519d significantly suppressed the endogenous mRNA and protein level of EphA4 by 67% (Fig. 4A) and 50%, respectively, whereas inhibition of miR-519d by the addition of anti–miR-519d resumed the expression of EphA4 in miR-519d–overexpressing A2058 cells (Fig. 4B). In consistence with the above results, the staining intensity of membranous EphA4 was decreased in cells with miR-519d overexpression, while comparing with the control, under confocal microscopy (Supplementary Fig. S3). We initially found miR-519d was upregulated in the sphere-forming cells. In support of the above findings that EphA4 was downregulated by miR-519d, we demonstrated that the expression of EphA4 proteins was reduced in spheroid cells, compared with the adherent A2058 cells (Fig. 4C). Upregulation of miR-519d and reciprocal downregulation of EphA4 was also found in melanospheres derived from SK-Mel-28 and SK-Mel-90 melanoma cells (Supplementary Fig. S4).
EphA4 is a direct target of miR-519d. A, Downregulation of EphA4 mRNA after lentiviral overexpression of miR-519d in A2058 melanoma cells. shR-EphA4 transfection was used as a positive control. Expression of EphA4 mRNA relative to actin mRNA was calculated and normalized to the scramble-treated control. B, Overexpression of miR-519d in A2058 cells significantly repressed the endogenous protein level of EphA4, which was recovered after addition of anti–miR-519d. HepG2 cell lysate was served as a positive control. The α-tubulin expression level was used for normalization. C, Decreased EphA4 expression level in sphere-forming cells, compared with the adherent cells. D, miR-519d modulated the expression of EphA4 via direct interaction with the 3′UTR as detected by luciferase reporter and mutagenesis assays. Sequences of the human EphA4 3′UTR (nucleotide 37-58 and 52-77, respectively) showing the two putative miR-519d-3p–binding sites. Matching regions are highlighted by lines. Alteration of either binding site by mutation (Mut 1, 2, or 3) is shown. E, Forced expression of miR-519d could significantly repress luciferase activity of the reporter gene bearing the 3′UTR of EphA4, whereas mutation at either one of the predicted target sites in the 3′UTR abrogated the repression. A Renilla luciferase control was included in all transfections to compensate for different transfection efficiencies. The normalized luciferase activities were given as the firefly:Renilla ratios normalized to the respective control-transfected setting. F, Addition of anti–miR-519d abrogated the repression of luciferase activity. **, P < 0.01. NS, nonsignificant.
EphA4 is a direct target of miR-519d. A, Downregulation of EphA4 mRNA after lentiviral overexpression of miR-519d in A2058 melanoma cells. shR-EphA4 transfection was used as a positive control. Expression of EphA4 mRNA relative to actin mRNA was calculated and normalized to the scramble-treated control. B, Overexpression of miR-519d in A2058 cells significantly repressed the endogenous protein level of EphA4, which was recovered after addition of anti–miR-519d. HepG2 cell lysate was served as a positive control. The α-tubulin expression level was used for normalization. C, Decreased EphA4 expression level in sphere-forming cells, compared with the adherent cells. D, miR-519d modulated the expression of EphA4 via direct interaction with the 3′UTR as detected by luciferase reporter and mutagenesis assays. Sequences of the human EphA4 3′UTR (nucleotide 37-58 and 52-77, respectively) showing the two putative miR-519d-3p–binding sites. Matching regions are highlighted by lines. Alteration of either binding site by mutation (Mut 1, 2, or 3) is shown. E, Forced expression of miR-519d could significantly repress luciferase activity of the reporter gene bearing the 3′UTR of EphA4, whereas mutation at either one of the predicted target sites in the 3′UTR abrogated the repression. A Renilla luciferase control was included in all transfections to compensate for different transfection efficiencies. The normalized luciferase activities were given as the firefly:Renilla ratios normalized to the respective control-transfected setting. F, Addition of anti–miR-519d abrogated the repression of luciferase activity. **, P < 0.01. NS, nonsignificant.
Having proven that miR-519d regulates EphA4 expression, we next sought to identify whether this regulation is mediated via direct binding of miR-519d to the EphA4 3′UTR by the luciferase reporter assay. We found that overexpression of miR-519d, but not the control, significantly suppressed the luciferase activity of EphA4 3′UTR in A2058 cells. To further validate the specific binding of miR-519d to its pairing sites in the 3′UTR of EphA4, we identified two potential binding sites within the EphA4 3′UTR by miRNA target prediction algorithms. Although two mature sequences, has-miR-519d-3p and has-miR-519d-5p, were derived from the stem-loop miR-519d, the two predicted sites on 3′UTR EphA4 for miR-519d binding were solely located on miR-519d-3p. We therefore generated three mutant forms of 3′UTR EphA4 harboring the mutated sequences of the predicted binding sites on miR-519d-3p, which included mutant (Mut)-1, -2 for binding site 1, and Mut-3 for binding site 2 (Fig. 4D). We found that all three mutant constructs restored the luciferase signal intensity in miR-519d–overexpressing cells (Fig. 4E), indicating that miR-519d suppressed EphA4 expression through the predicted sequences in the 3′UTR of EphA4. Moreover, blockage of miR-519d overexpression using anti–miR-519d resulted in a significant increase of luciferase activity of EphA4 3′UTR (Fig. 4F). Our luciferase reporter and mutagenesis assays showed that miR-519d targeted the 3′UTR of EphA4 directly in melanoma.
Forced expression of EphA4 reverses the effects of miR-519d in melanoma cell migration, invasion, and adhesion
Because EphA4 was identified as a direct target of miR-519d, we next investigated whether EphA4 was a functionally important target gene of miR-519d in melanoma cells. We hypothesized that reintroduction of miRNA-resistant full-length EphA4 (sequence without the 3′UTR of EphA4) in miR-519d–overexpressing A2058 cells could inhibit its highly metastatic phenotype, or on the other hand, downregulation of EphA4 by RNA interference in the control A2058 stable cell clone could mimic the effect of miR-519d overexpression. We exploited two specific shRNA(shR)s (#1 and #2) designed for targeting toward EphA4 in control A2058 cells. Both shRNAs have reduced EphA4 expression to the level less than 30% of that in scramble-infected cells (Fig. 5A, left), and the control/shR-EphA4#1 cells were used in the subsequent experiments. We also successfully established stable cell lines by transfecting EphA4-pWPT–expressing vectors or control vectors into miR-519d–overexpressing A2058 cells (Fig. 5A, right). As expected, in a manner opposite to miR-519d, forced expression of EphA4 alone in miR-519d–overexpressing cells (miR-519d/pWPT-EphA4) could significantly decrease cell migration (Fig. 5B), invasion (Fig. 5C), and adhesion (Fig. 5D), whereas knockdown of EphA4 in the control A2058 cells (control/shR-EphA4) behaved like miR-519d overexpression that cell migration (Fig. 5B), invasion (Fig. 5C), and adhesion (Fig. 5E) were enhanced. Moreover, forced expression of EphA4 could significantly suppress the formation of melanosphere in A2058 cells (Fig. 5F).
EphA4 is a functionally important target gene of miR-519d in melanoma progression. A, Immunoblotting for EphA4 protein showed effective knockdown of endogenous EphA4 in control A2058 cells (control/shR-EphA4#1 and #2) and successful restoration of EphA4 in miR-519d–overexpressing A2058 cells (miR-519d/pWPT-EphA4). B, Migration by scratch wound assay. C, Transwell invasion assay. D and E, Adhesion assays in miR-519d–overexpressing A2058 cells with or without EphA4 restoration (D) and in A2058 control cells with or without EphA4 knockdown (E). F, The effect of EphA4 overexpression on sphere formation in miR-519d–overexpressing A2058 cells. The sphere diameter on day 6 in culture was 48.1 ± 14.8 μm in the control/pWPT cells, 71.7 ± 22.1 μm in the miR-519d–overexpressing cells (miR-519d/pWPT), but significantly decreased to 40.3 ± 11.5 μm after forced expression of EphA4 (miR-519d/pWPT-EphA4). The representative phase-contrast photomicrographs are shown at the top. Bar, 100 μm. Data, mean values ± SD; *, P < 0.05, **; P < 0.01.
EphA4 is a functionally important target gene of miR-519d in melanoma progression. A, Immunoblotting for EphA4 protein showed effective knockdown of endogenous EphA4 in control A2058 cells (control/shR-EphA4#1 and #2) and successful restoration of EphA4 in miR-519d–overexpressing A2058 cells (miR-519d/pWPT-EphA4). B, Migration by scratch wound assay. C, Transwell invasion assay. D and E, Adhesion assays in miR-519d–overexpressing A2058 cells with or without EphA4 restoration (D) and in A2058 control cells with or without EphA4 knockdown (E). F, The effect of EphA4 overexpression on sphere formation in miR-519d–overexpressing A2058 cells. The sphere diameter on day 6 in culture was 48.1 ± 14.8 μm in the control/pWPT cells, 71.7 ± 22.1 μm in the miR-519d–overexpressing cells (miR-519d/pWPT), but significantly decreased to 40.3 ± 11.5 μm after forced expression of EphA4 (miR-519d/pWPT-EphA4). The representative phase-contrast photomicrographs are shown at the top. Bar, 100 μm. Data, mean values ± SD; *, P < 0.05, **; P < 0.01.
miR-519d/EphA4 pathway enhances the downstream ERK signaling and EMT protein expression
We showed previously that miR-519d was involved in EMT, which enhanced melanoma cell migration, invasion, and metastasis (Fig. 3E). To explore the mechanism how EphA4 acted as a suppressor in melanoma progression, we first examined various EMT markers and EMT-inducing transcription factors (EMT-TF) following the manipulation of EphA4 expression. Analyses with Western blotting revealed that forced expression of EphA4 (miR-519d/pWPT-EphA4) suppressed EMT by reducing the expression of the mesenchymal proteins N-cadherin, fibronectin, and EMT-TF Snail. In agreement with the above findings, knockdown of EphA4 (control/shR-EphA4) led to an increase of N-cadherin, fibronectin, and Snail. There was no significant change of the protein levels of vimentin and Twist. Interestingly, the amount of Slug was more abundant in miR-519d/pWPT-EphA4 cells compared with that in miR-519d/pWPT cells, and decreased when EphA4 was knocked down (Fig. 6A).
Restoration of EphA4 inhibits miR-519d–mediated ERK activation, epidermal-to-mesenchymal transition and in vivo lung metastasis. A, Overexpression of EphA4 decreased protein levels of mesenchymal markers fibronectin, N-cadherin, and Snail. The expression levels of vimentin and Twist were not altered. B, Overexpression of EphA4 resulted in decreased ERK1/2 phosphorylation. The protein expression level of phospho-AKT remained unchanged. The reverse effect was observed following depletion of EphA4. C and D, Measurement of the lung metastasis capability after tail vein injection of miR-519d/pWPT-EphA4 cells, control/shR-EphA4 cells, and their respective controls. Tumor nodules are shown in representative grossly dissected mouse lungs (C) and tissue sections stained with hematoxylin and eosin (D, original magnification, ×200) from one animal of each group (n = 6 in each group). E, The number of lung metastatic melanoma nodules was inversely correlated with the EphA4 expression level. F, In accordance with E, the lung weight was inversely correlated with the EphA4 expression level. Data, mean values ± SD; *, P < 0.05; **, P < 0.01.
Restoration of EphA4 inhibits miR-519d–mediated ERK activation, epidermal-to-mesenchymal transition and in vivo lung metastasis. A, Overexpression of EphA4 decreased protein levels of mesenchymal markers fibronectin, N-cadherin, and Snail. The expression levels of vimentin and Twist were not altered. B, Overexpression of EphA4 resulted in decreased ERK1/2 phosphorylation. The protein expression level of phospho-AKT remained unchanged. The reverse effect was observed following depletion of EphA4. C and D, Measurement of the lung metastasis capability after tail vein injection of miR-519d/pWPT-EphA4 cells, control/shR-EphA4 cells, and their respective controls. Tumor nodules are shown in representative grossly dissected mouse lungs (C) and tissue sections stained with hematoxylin and eosin (D, original magnification, ×200) from one animal of each group (n = 6 in each group). E, The number of lung metastatic melanoma nodules was inversely correlated with the EphA4 expression level. F, In accordance with E, the lung weight was inversely correlated with the EphA4 expression level. Data, mean values ± SD; *, P < 0.05; **, P < 0.01.
We next looked at the effects of EphA4 on signaling pathways known to affect melanoma progression. Significantly, we found that miR-519d–overexpressing stable cell line had an increase of ERK1/2 activation, which was reversed while EphA4 was overexpressed. Similar to the condition of miR-519d overexpression, EphA4 knockdown also led to an increased activation of ERK1/2 (Fig. 6B). We have also noticed that constitutive activation of AKT occurred in A2058 cell line and that the amount of phospho-Akt was not changed when we manipulated the expression levels of EphA4 (Fig. 6B). Taken together, these data indicated that EphA4 could interfere with the activation status of ERK1/2, which represents a major pathway mediating cell migration, invasion, and adhesion in melanoma (29).
Forced expression of EphA4 reverses the effects of miR-519d in mouse lung metastasis
To further investigate the role of EphA4 associated with miR-519d in an in vivo model, we injected stable transfected cell lines miR-519d/pWPT, miR-519d/pWPT-EphA4, control/scramble, and control/shR-EphA4 into the lateral tail vein of SCID mice. The animals were assessed for lung metastases 7 weeks after injection (n = 8 in each group). We found that mice injected with miR-519d/pWPT and control/shR-EphA4 cells had more severe lung metastases than those injected with miR-519d/pWPT-EphA4 and control/scramble cells, respectively (Fig. 6C), which was confirmed by the histopathologic analysis (Fig. 6D). Taking account into the expression levels of EphA4 in control and miR-519d cells, respectively, the number of lung metastatic melanoma nodules as well as the lung weight of the injected mice was inversely correlated with EphA4 expression, with statistical significance (Fig. 6E and F).
Reciprocal miR-519d-EphA4 expression correlates with clinical metastasis in melanoma patients
To determine the clinical significance of miR-519d expression in patients with melanoma, the mRNA levels of miR-519d and EphA4 were measured in surgical specimens from normal skin obtained from peritumoral tissues, primary cutaneous melanoma, and lymph node or distant metastatic melanoma using qRT-PCR. There was no significant difference in miR-519d expression between normal skin (n = 19) and primary melanoma (n = 21), but miR-519d expression was significantly increased in metastatic melanoma (n = 21) in comparison with that in normal skin and primary melanoma (Fig. 7A). We also found that compared with primary melanoma (n = 13), EphA4 mRNA was significantly downregulated in metastatic melanoma (n = 17, Fig. 7B), and this downregulation was correlated with the upregulation of miR-519d (R = −0.6; Pearson correlation, P = 0.034). In addition, the protein levels of EphA4 were analyzed in 10 cases of nevocellular nevus, 105 cases of primary cutaneous melanoma, and 37 cases of metastatic melanoma. Among the 105 cases of primary cutaneous melanoma, there were 75 with acral lentiginous melanoma, 16 with superficial spreading melanoma, 12 with nodular melanoma, and 2 with lentigo malugna melanoma. Pancreatic ductal adenocarcinoma tissues were served as the positive control (Supplementary Fig. S5A; ref. 30). The intensity of EphA4 immunostain was examined and scored as negative, weakly positive, moderately positive, or strongly positive. Negative or weak EphA4 immunoreactivity was seen in the normal skin (Supplementary Fig. S5B) and nevocellular nevus (Supplementary Fig. S5C). Strong immunoreactivity was found in the radial growth phase of primary cutaneous melanoma, and the immunointensity was decreased in the vertical growth tumor parts in the dermis (Supplementary Fig. S5D). Positive EphA4 immunoreactivity was observed in 49 of the 105 cases of primary malignant melanoma (46.7%) and 9 of the 37 cases metastatic melanoma (24.3%), which showed both membranous and cytoplasmic staining patterns (Fig. 7C and D). Statistical analysis with χ2 test showed significantly decreased EphA4 expression in the metastatic tumors compared with that in primary melanoma (P = 0.017). Furthermore, comparing the immunointensity of EphA4 between 28 sets of paired primary melanoma and metastatic melanoma (Supplementary Fig. S5E), melanoma patients with higher EphA4 expression levels in metastatic tumors than its primary counterpart had longer overall survival than those with the same degree (including both negative staining in paired primary and metastatic melanomas) or decreased EphA4 expression in their paired melanoma metastases (93.6 months vs. 37.2 months; log-rank test, P = 0.034; Fig. 7E). The HR for patients with decreased EphA4 expression in their metastatic tumors was 3.0 (95% confidence interval, 1.08–7.25; P < 0.05). Taken together, these results indicated that melanoma patients with increased miR-519d levels and reciprocally reduced EphA4 expression was less beneficial to the overall survival.
Reciprocal miR-519d upregulation and EphA4 downregulation correlates with metastasis of melanoma patients. A, qRT-PCR analysis for miR-519d expression was conducted on melanoma surgical samples showing upregulation of miR-519d in metastatic melanoma (n = 21) compared with primary melanoma (n = 21) and normal noncancerous skin tissues (n = 19). The miR-519d abundance was normalized to RNU6B (U6). B, qRT-PCR analysis for EphA4 expression showed that compared with primary melanoma (n = 13), EphA4 mRNA was significantly downregulated in metastatic melanoma (n = 17). C, IHC analysis of EphA4 in two representative paired malignant melanocytic lesions. Positive EphA4 immunoreactivity was found in both cases of primary cutaneous melanoma (top left, ×100; bottom left, ×400), but the immunoreactivity was significantly decreased in their counterpart metastatic melanoma of regional lymph nodes (top right, ×200; bottom right, × 100). Nevertheless, focal moderate staining could be observed in part of the metastatic lymph nodes (bottom right). D, Bar graph summary of the grading of EphA4 immunoreactivity in 105 cases of primary cutaneous melanoma and 37 cases of metastatic tumors. E, Kaplan–Meier survival analysis between changes of EphA4 immunointensity (metastasis vs. primary in 28 paired samples) and the overall survival in melanoma patients. Patients with higher EphA4 staining levels in metastatic tumors than its primary counterpart had significantly better overall survival than those with unchanged or decreased EphA4 expression (P < 0.05, log-rank test).
Reciprocal miR-519d upregulation and EphA4 downregulation correlates with metastasis of melanoma patients. A, qRT-PCR analysis for miR-519d expression was conducted on melanoma surgical samples showing upregulation of miR-519d in metastatic melanoma (n = 21) compared with primary melanoma (n = 21) and normal noncancerous skin tissues (n = 19). The miR-519d abundance was normalized to RNU6B (U6). B, qRT-PCR analysis for EphA4 expression showed that compared with primary melanoma (n = 13), EphA4 mRNA was significantly downregulated in metastatic melanoma (n = 17). C, IHC analysis of EphA4 in two representative paired malignant melanocytic lesions. Positive EphA4 immunoreactivity was found in both cases of primary cutaneous melanoma (top left, ×100; bottom left, ×400), but the immunoreactivity was significantly decreased in their counterpart metastatic melanoma of regional lymph nodes (top right, ×200; bottom right, × 100). Nevertheless, focal moderate staining could be observed in part of the metastatic lymph nodes (bottom right). D, Bar graph summary of the grading of EphA4 immunoreactivity in 105 cases of primary cutaneous melanoma and 37 cases of metastatic tumors. E, Kaplan–Meier survival analysis between changes of EphA4 immunointensity (metastasis vs. primary in 28 paired samples) and the overall survival in melanoma patients. Patients with higher EphA4 staining levels in metastatic tumors than its primary counterpart had significantly better overall survival than those with unchanged or decreased EphA4 expression (P < 0.05, log-rank test).
Discussion
Besides the characteristics of stem-like properties and strong tumorigenicity (5–7), previous studies also showed that multicellular spheroids represented an appropriate three-dimensional in vitro system that was similar to a tumor micromilieu or a micrometastasis in vivo and could be used to understand the molecular pathogenesis more closely mimicking tumor metastasis in the clinical setting (31, 32). Therefore, characterization of the nature of melanospheres can greatly facilitate the new understanding of melanoma progression than studying cells grown in an adherent state. In this study, A2058 human melanoma cell line passaged as adherent cultures or formed melanospheres in stem cell media have been studied to compare their cellular characteristics and miRNA regulation. In accordance with the previous study (6), we found the melanospheres obtained from our sphere culture system represented an enrichment of heterogeneous cell subpopulations with enhanced clonogenic ability and were highly tumorigenic. These melanospheres also showed higher intrinsic chemoresistance. Furthermore, a propensity for EMT identified in melanospheres contributed to melanoma metastasis. These unique features of melanospheres can be regulated by specific miRNAs.
In this study, we demonstrated increased expression of miR-519d and miR-526b in melanospheres, compared with adherent monolayers. In addition, we found that the amount of the other two miRNAs, miR-519a and miR-519c, was elevated in melanospheres by qRT-PCR (unpublished data). miR-519a, miR-519c, miR-519d, and miR-526b were all members of chromosome 19 miRNA gene cluster (C19MC), the largest primate-specific miRNA cluster comprising 46 pre-miRNAs, located on human chromosome 19 q13.41 region (33, 14). Previous studies have indicated that C19MC represents a stem cell miRNA gene cluster that is preferentially expressed in the fetal brain, placenta tissues, and human embryonic stem cells but has low expression in adult tissues (15–17). In addition to maintaining stem cell function, the polycistronic miRNA locus C19MC also involves in carcinogenesis. Li and colleagues have demonstrated that amplification of C19MC, which results in extensive expression of miR-517c and miR-520g, is linked to tumorigenesis and clinical aggressiveness of primitive neuroectodermal tumors in the central nervous system (34). However, previous reports have not found amplification of C19MC in cutaneous melanoma (35, 36). Alterations in DNA methylation on their nearby CpG islands can also regulate miRNA expression (37–39). Hypomethylation of the CpG islands located 17.6 kb upstream the C19MC combined with upregulation of miR-519d was found in human gastric cancer AGS cells (40) and hepatocellular carcinoma (41). It is plausible to explore in the future if other members of C19MC in addition to miR-519d are reexpressed in melanospheres to provide the synergistic effects from multiple miRNAs of the C19MC in melanoma development and progression.
From the melanosphere model, we have demonstrated that miR-519d functioned as a critical tumor oncogene in melanoma. miR-519d promoted melanoma tumorigenesis and progression by targeting the crucial oncogenic pathways, including Eph, ERK, as well as facilitating EMT. In cervical cancer, miR-519d was found to promote progression and metastasis through direct targeting Smad7 (41). In hepatocellular carcinoma, miR-519d has been shown to play an important role in promoting cell proliferation, invasion, and apoptosis resistance through the direct targeting of cyclin-dependent kinase inhibitor (CDKN) 1A/p21, PTEN, AKT3, and tissue inhibitor of metalloproteinase 2 (TIMP2), at least in part (42). However, when we overexpressed miR-519d in A2058 melanoma cells, there was no significant alteration of protein levels of CDKN1A/p21, PTEN, AKT3, and peroxisome proliferator activated receptor 1 (PPAR1), the potential targets predicted by computational algorithms. Instead, we found the mRNA and protein levels of another predicted target EphA4 exhibited a significant decrease under miR-519d overexpression in melanoma cells, which suggested a tissue-specific miRNA regulation at the transcriptional or posttranscriptional stages. Furthermore, we found miR-519d targeted the 3′UTR of EphA4 transcripts and downregulated EphA4 expression. Forced expression of EphA4 in miR-519d–overexpressing cells inhibited melanoma progression including in vitro cell migration, invasion, adhesion, as well as lung metastasis in vivo. Our in vivo lung metastasis experiments were achieved by directly injecting the melanoma cells into mouse circulation for mimicking circulating or disseminating tumor cells and, thus, did not recapitulate the entire metastatic process of melanoma cells from the primary sites. Orthotopic melanoma models using GFP-labeled cells could be used to determine spontaneous metastasis in future studies (43–45). Nevertheless, in accordance with the above findings, upregulation of miR-519d and downregulation of EphA4 were found in a subset of clinical metastatic melanoma surgical samples, which was associated with a worse clinical outcome. Taken together, these results contributed to the characterization of a novel molecular mechanism of melanoma progression and metastasis.
Using genetic modulation of EphA4 expression in melanoma cells, we found that EphA4 effectively suppressed the migratory, invasive, adhesion, and metastatic capabilities of melanoma cells. Moreover, our results demonstrated the clinical significance of EphA4, as its expression significantly decreased in metastatic melanoma and correlated with patient survival. Thus, both in vitro and in vivo evidence indicated that downregulation of EphA4 played a crucial role in melanoma progression. Although the percentage of positive EphA4 immunostaining was higher in primary cutaneous melanoma than that in metastatic melanoma, however, when the immunoreactivity of EphA4 was compared between primary and metastatic melanoma, no statistical significance was observed. We found interestingly that the EphA4 immunostain of some melanoma cases was accentuated in the radial growth phase of primary cutaneous melanoma and the immunointensity was decreased in the vertical growth tumor parts in the dermis (Supplementary Fig. S4E). These observations suggested that EphA4 overexpression might be involved in the initial process of melanoma tumorigenesis but was downregulated, in part by miR-519d, to facilitate local tumor invasiveness and further metastasis.
Aberrant Eph receptor expression has been implicated in a variety of malignancies as critical regulators in adhesion, migration, and invasion of cancer cells, including melanoma (46–48). Both tumor-suppressive or promoting functions for Ephs have been documented, depending on mutual interactions within the Eph/ephrin system as well as on cell type–specific context from the tumor microenvironment (49). The complexity of Eph signaling is also signified between tumor stages (50). For instance, in the early stages of malignant transformation in the colon, EphA1, EphA2, EphB1, EphB2, and EphB4 are upregulated and may play an oncogenic role. Although in the late stage of colon cancer, subsequent gene silencing most likely through methylation may play a role in promoting tumor progression (50). Here, we proposed a novel miR-519d–mediated mechanism by modulating the expression of EphA4 during melanoma progression, which might provide new potential biomarkers for the advanced status of melanoma.
Disclosure of Potential Conflicts of Interest
No potential conflicts of interest were disclosed.
Authors' Contributions
Conception and design: Y.-H. Liao
Development of methodology: Y.-L. Huang, Y.-H. Liao
Acquisition of data (provided animals, acquired and managed patients, provided facilities, etc.): J.-B. Hong, H.-Y. Huang, J.-S. Chen, Y.-H. Liao
Analysis and interpretation of data (e.g., statistical analysis, biostatistics, computational analysis): K.-T. Hua, Y.-H. Liao
Writing, review, and/or revision of the manuscript: K.-T. Hua, Y.-H. Liao
Administrative, technical, or material support (i.e., reporting or organizing data, constructing databases): Y.-S. Sheen, Y.-L. Huang
Study supervision: Y.-H. Liao
Acknowledgments
The authors gratefully acknowledge the excellent technical assistance by Chun-Ru Lin, Jhu-Yun Deng, Ching-Yi Lin, Jia-Fang Tsai, and Pei-Jung Lin. This work was supported by Ministry of Science and Technology (MOST) under grant no. NSC-102-2314-B-002-162 and MOST 103-2314-B-002-067-MY3 (Y-H. Liao).
The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked advertisement in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.