Cordycepin (3′-deoxyadenosine) suppressed HMGA2, Twist1 and ZEB1-dependent melanoma invasion and metastasis by targeting miR-33b

Malignant melanoma, the most deadly form of skin cancer, has a high propensity for metastatic spread and is notoriously chemotherapy-resistant. Cordycepin, the active component of Cordyceps spp., has been identified to have anti-metastatic effect on tumor progression and thus possesses pharmacological and therapeutic potentials. However, the mechanisms of anti-metastatic effects of cordycepin at cellular levels remain elusive. We analyzed the effect of cordycepin on human melanoma miRNA expression profiles by miRNAarray and found that miR-33b was upregulated in highly-metastatic melanoma cell lines following cordycepin exposure. Cordycepin-mediated miR-33b expression was dependent on LXR-RXR heterodimer activation. miR-33b directly binds to HMGA2, Twist1 and ZEB1 3′-UTR to suppress their expression. The negative correlations between miR-33b levels and HMGA2, Twist1 or ZEB1 expression were detected in 72 patient melanoma tissue samples. By targeting HMGA2 and Twist1, miR-33b attenuated melanoma migration and invasiveness upon cordycepin exposure. miR-33b knockdown or ZEB1 overexpression reverted cordycepin-mediated mesenchymal-epithelial transition (MET), triggering the expression of N-cadherin. In spontaneous metastasis models, cordycepin suppressed tumor metastasis without altering primary tumor growth. We showed for the first time that targeting miRNA by cordycepin indicates a new mechanism of cordycepin-induced suppression of tumor metastasis and miR-33b/HMGA2/Twist1/ZEB1 axis plays critical roles in regulating melanoma dissemination.


Computational target prediction
For the identification of miR-33b targets, three different software algorithms were used to find conserved target sites throughout transcriptomes: TargetScan (http://www.targetscan.org), PicTar (http:// pictar.mdc-berlin.de) and miRanda (http://www. microrna.org). The three lists obtained from in silico computational target predictions were compared. The intersection of resulted in a list of 10 common genes as potential targets.
cDNA clones of human ZEB1, Twist1 and HMGA2 were obtained from Origene (Rockville, MD). The cDNA constructs were inserted into the expression vector pcDNA3.1 (Invitrogen). 200 ng constructs were transfected into Lu1205 and A375 cells with Mirus TransIT (Mirus Bio LLC.) and then subjected to G418 selection for 2-3 weeks. Expression of the transfected constructs was assessed by Western blot analysis.
Mutant 3'UTR of these genes were generated by site-directed mutagenesis. For reporter assay, Lu1205 and A375 cells were plated onto 12-well plates and transfected with 100 ng of pGL3-3'-UTR of genes using Lipofectamine 2000 (Invitrogen). A Renilla luciferase vector pRL-SV50 (5 ng; Promega) was also co-transfected to normalize the differences in transfection efficiency. After transfection for 48 hr, cells were harvested and assayed with Dual-Luciferase Reporter Assay System (Promega) using a Tecan M200 luminescence reader according to the manufacturer's instructions. Transfection was repeated three times in triplicate.

Chromatin immunoprecipitation assay
Lu1205 cells were treated with cordycepin before ChIP assay was conducted using ChIP-IT kit (Active Motif) according to manufacturer's instructions. Briefly, cells were treated with 5% formaldehyde for 20 min to cross-link DNA with proteins, lysed, passed through a 26 gauge-needle to facilitate nuclei release, and sonicated. After pre-clearing the chromatin with protein G beads, the supernatants were immunoprecipitated with 1 μg anti-LXR antibody or 1 μg negative control IgG overnight. Then, the protein G-antibody-chromatin complex was washed and eluted from the protein G. DNA was purified with supplied columns and samples were analyzed by PCR.

Quantitative RT-PCR of miRNAs
For miRNAs, the isolated total RNA was polyadenylated and reverse transcribed for use in a twostep quantitative RT-PCR using the TaqMan microRNA reverse transcription kit and qRT-PCR kits (Invitrogen) according to the manufacturer's instructions. Primers specific for human, miR-200b, miR-200c, miR-205, miR-33b, and miR-211 (Qiagen) were used. The relative amounts of miRNA to small nuclear RNU6B RNA was calculated using equation 2 -ΔCt , where ΔCt = (Ct -miR -Ct -RNU6B RNA ). The PCR conditions were 2 min at 50°C and 10 min at 95°C, followed by 40 cycles of 95°C for 15 s and 60°C for 30 s.

Wound healing assay
One day before transfection, equal numbers of Lu1205 and A375 cells were seeded onto 24-well plates. Cells were then transfected with miR-33b antagomir, HMGA2 construct, Twist1 construct or controls. When the cell confluence reached about 90% after transfection, cell monolayer was treated or untreated with cordycepin at different doses for 24 hr. An artificial homogenous wound was made onto the monolayer with the use of a sterile plastic 200 μl micropipette tip. After wounding, the debris was removed by washing the cells with PBS. At different time points, cells that migrated into the wounded area were photographed under an inverted microscope (40× objective) (Nikon TE2000-U, Nikon Inc.). Wound healing effect was determined by calculating the ratio of the cell-free area at specific time point to that at time 0.

Matrigel invasion assay
RFP-Lu1205 and RFP-A375 cells were transfected with nontargeting control or miR-33b antagomir as mentioned above. Transfected cells (0.5 × 10 6 ) were reseeded into the rehydrated insert after treated with cordycepin. Medium with 10% FBS was added to the lower chamber as chemoattractant. After 24 hr incubation, non-invading cells on the upper surface of the membrane were scrubbed gently with a cotton-tipped swab. The invasive cells were photographed under an inverted light microscope (40 × objective) and quantified by manual counting in 6 randomly selected areas.

Gelatin zymography
Lu1205 and A375 cells were transfected with sh-NT or sh-miR-33b before being treated with cordycepin. After 24 hr, the medium that was collected from the incubated cells was mixed with substrate gel sample buffer (40% (v/v) glycerol, 0.25 M Tris-HCl, pH 6.8, and 0.1% bromophenol blue), and then it was loaded without boiling onto 10% SDSpolyacrylamide gel that contained gelatin (1.5 mg/ml; Sigma, St. Louis, MO). After performing electrophoresis, the gel was then soaked in 2% Triton X-100 with gentle shaking for 60 min with single change of detergent solution. The gel was rinsed and next incubated overnight in substrate buffer (50 mM Tris-HCl, pH 7.5, 5 mM CaCl 2 , and 0.02% NaN 3 ). Following the incubation, the gel was stained with 0.05% Coomassie brilliant blue G-250 and then destained in 10% acetic acid and 20% methanol. The gel was photographed and then measured by densitometry.

Rho activation assay
Rho activation assay kit (Upstate Biotechnology, Lake Placid, NY) was employed to determine the activity of RhoA in melanoma cells. Briefly, cell lysates were incubated with rhotekin Rho binding peptide coated on agarose beads for 1 hr, and the GTP-Rho bound to rhotekin-agarose beads were immunoprecipitated and subject to Western blotting detection with anti-RhoA mouse IgG antibody (Santa Cruz).
Immunofluorescence miRNA antagomir and/or overexpressing plasmidtransfected melanoma cells were grown on cover slips coated with fibronectin (1 μg/ml) before being treated with cordycepin. Cells were then washed with PBS and fixed with 5% paraformaldehyde for 10 min. Cells were permeabilized with 0.3% Triton X-100 in PBS and blocked for 30 min with 5% BSA. Subsequently, cover slips were incubated with anti-E-cadherin, anti-N-cadherin or antipaxillin for 1 hr at room temperature. This was followed by staining with Alexa 555 or Alexa 488-conjugated antirabbit IgG. To image actin filaments, rhodamine-phalloidin (1:40; Life Technologies, Carlsbad, CA) was incubated with cells. Finally, fluorescent staining was visualized with IX71 Olympus inverted microscopy (Olympus, Inc.) with 40X magnification. Colocalization of paxillin and actin was processed by Image J. To analyze the size and number of paxillin-containing focal adhesions, images were background subtracted before thresholding and segmentation were conducted to detect the edges of focal adhesions. Then, the mean size (in pixels) and number of focal adhesions in each cell were calculated.