(A–B) qRT-PCR reveals upregulation of miR-33b, miR-200b, miR-200c, miR-205 and miR-211 in Lu1205 and A375 melanoma cells which received 200 μg/ml cordycepin compared with their matched untreated controls. (C) Relative expression of miR-33b in Lu1205 and A375 melanoma cells which received 0, 50, 100 and 200 μg/ml cordycepin treatment. Data are expressed as mean±SEM from three independent experiments. *p < 0.05, **p < 0.01 compared with control. (D) LXR and RXR binding sites overlap and cluster around target genes. All target genes are shown in their native chromosomal location according to Human Genome Assembly (hg18) in UCSC Genome Browser. Vertical lines in target genomes represent exons, and horizontal lines represent introns. (E) Sequences of the consensus LXRE. (F–G) 200 μg/ml cordycepin-mediated binding of LXR/RXR to LXRE as measured by chromatin immunoprecipitation assays. The chromatin was extracted from Lu1205 cells treated with or without cordycepin. Crosslinked lysates were immunoprecipitated with control Ig or anti-LXR antibody or not (input). DNA precipitates were isolated and subjected to semiquantitative PCR (F) or quantitative PCR (G) using specific primers covering LXRE. Quantitative PCR data are represented normalized and expressed relative to the levels detected in Ig-immunoprecipitated lysates. Data represent mean ± SEM of 3 independent assays. (H) Lu1205 cells were transfected with LXRE luciferase reporter constructs. Cells were treated or untreated with 200 μg/ml cordycepin and transfected with scrambled, siLXRβ, siRXRα and siLXRβ/RXRα before the luciferase activities were measured. Results are relative to the levels in untreated cells. Data represent mean±SEM of 3 independent assays. *p < 0.05 compared with control. (I) Relative expression of miR-33b in Lu1205 and A375 melanoma cells which received scrambled or siLXRβ/RXRα followed by 200 μg/ml cordycepin treatment. Data are expressed as mean±SEM from three independent experiments. *p < 0.05 compared with control.

(A) The effect of 200 μg/ml cordycepin treatment on ADMA9, ZEB1, ZEB2, CTNNB1, HMGA2, Rac1, SALL4, Twist1 and Yes1 mRNA expression levels in Lu1205 and A375 were measured by qRT-PCR. mRNA levels were normalized against GAPDH reference genes. (B) Cordycepin treatment or transient overexpression of miR-33b mimic in Lu1205 reduced luciferase activities of ZEB1, HMGA2 and Twist1 3′UTR reporter. (C) miR-33b inhibition reverted cordycepin-mediated suppression of luciferase activities of HMGA2, Twist1 and ZEB1 3′UTR reporter in Lu1205 cells. Lu1205 were transfected with sh-NT, sh-pre-miR-33b or sh-miR-33b for 24 hr before being treated or untreated with 200 μg/ml cordycepin for 24 hr. The relative luciferase activities were measured. (D–F) Mutation of miR-33b target sites in HMGA2, Twist1 and ZEB1 3′UTR constructs completely rescued the down-regulatory effect of corycepin on luciferase activity. (D) Top, HMGA2 3′-UTR-wt (HMGA2-1 and HMGA2-2 wt) and two 3′-UTR mutants (HMGA2-1 mut and HMGA2-2 mut) in which the sequence in red was mutagenized to abolish miR-33b binding. Bottom, luciferase assays showing decreased activity after co-transfection of HMGA2-wt with miR-33b in Lu1205 and A375 cells. The HMGA2-1 mut and HMGA2-2 mut were partially refractory to cordycepin. (E) Top, Twist1 3′-UTR-wt (Twist1 wt) and 3′-UTR mutant (Twist1 mut) in which the sequence in red was mutagenized to abolish miR-33b binding. Bottom, luciferase assays showing decreased activity after co-transfection of Twist1-wt with miR-33b in Lu1205 and A375 cells. The Twist1 mut were refractory to cordycepin. (F) Top, ZEB1 3′-UTR-wt (ZEB1 wt) and 3′-UTR mutant (ZEB1 mut) in which the sequence in red was mutagenized to abolish miR-33b binding. Bottom, luciferase assays showing decreased activity after co-transfection of ZEB1-wt with miR-33b in Lu1205 and A375 cells. The ZEB1 mut were refractory to cordycepin. Luciferase activity was assayed and renilla luciferase was used as internal control. Each data are expressed as mean±SEM from three independent experiments. *p < 0.05 compared with control.

(A) Relationship of miR-33b, HMGA2, Twist1 and ZEB1 relative expression and melanoma clinicopathological features. miR-33b expression levels in FFPE melanoma tissue samples from 72 patients with clinical TNM stage 0–IV were detected with in situ hybridization and quantified by ImageJ for grey value. HMGA2, Twist1 and ZEB1 expression levels in these FFPE melanoma tissue samples were detected with IHC and quantified by ImageJ for grey value. Relative expression levels of target molecules were compared among different TNM stages with ANOVA. P value is shown. Color circles represent outliers in each group. (B–D) The correlation between miR-33b relative expression levels (log transformed) and HMGA2 (B), Twist1 (C) or ZEB1 (D) relative expression levels (log transformed) in FFPE melanoma tissue samples from 72 patients with clinical TNM stage 0–IV. Pearson correlation coefficients and R²-values were calculated as indicated.

(A–C) sh-miR-33b reverted cordycepin-mediated suppression of Lu1205 melanoma migration capacities as measured by wound healing assay. (A) Representative images photographed at 0 and 24 hr post-wounding were shown at magnification of 20X. (B) Increasing cordycepin concentration reduced Lu1205 and A375 cell motility. (C) sh-miR-33b but not sh-miR-200b, sh-miR-200c, sh-miR-205 and sh-miR-211 reverted cordycepin-mediated suppression of Lu1205 and A375 motility. The cell free area were quantified at 24 hr and normalized against that at 0 hr. Data are expressed as mean ± SEM from three independent experiments. *p < 0.05 compared with control. ^†p < 0.05 compared with sh-NT. (D) siLXRβ/RXRα cordycepin-mediated suppression of Lu1205 and A375 motility. The cell free area were quantified at 24 hr and normalized against that at 0 hr. Data are expressed as mean ± SEM from three independent experiments. *p < 0.05 compared with control. (E) sh-miR-33b reverted cordycepin-mediated suppression of phosphorylation of FAK, Src and MLC and GTP-coupling of RhoA in Lu1205 and A375 cells. Cells were untransfected or transfected with sh-NT or sh-miR-33b for 24 hr before being treated or untreated with 200 μg/ml cordycepin for 24 hr. Cells were lysed and subjected to Western blotting analysis of phosphorylation of FAK, Src and MLC and GTP-coupling of RhoA. Total FAK, Src, MLC and RhoA were used as loading controls. (F) HMGA2 and Twist1 overexpression reverted cordycepin-mediated suppression of Lu1205 migration as measured by wound healing assay. Lu1205 monolayer were transfected with either mock, HMGA2 or Twist1 construct before being treated with 200 μg/ml cordycepin for 24 hr. The cell free area were quantified at 24 hr and normalized against that at 0 hr. Data are expressed as mean ± SEM from three independent experiments. *p < 0.05 compared with cordycepin + mock. (G) Cordycepin inhibited Lu1205 melanoma stress fiber and focal adhesion formation via miR-33b. Lu1205 cells were transfected with a variety of miRNA mimic, sh-miRNA, and/or overexpressing constructs before being treated or untreated with cordycepin. Lu1205 cells were stained with rhodamine-phalloidin and paxillin antibody (Geen = paxillin, red = F-actin). Bar = 10 μm. (H–I) Quantification of the average number and size (μm2) of paxillin-containing focal adhesions in Lu1205 cells using ImageJ software.12 cells were analyzed per condition in each experiment. Data are expressed as mean ± SEM of three independent experiments. *p < 0.05, **p < 0.01 compared with ctrl + scramble.

(A) Matrigel-coated transwell invasion assay revealed that the miR-33b knockdown rescued cordycepin-mediated inhibition of Lu1205 and A375 invasion. Top, representative images show migrated RFP-Lu1205 cells underneath membrane. Bar = 100 μm. Bottom, quantification of number of migrated Lu1205 cells which were transfected with sh-NT or sh-miR-33b before being treated or untreated with cordycepin. Data are expressed as mean ± SEM from three independent experiments. **p < 0.01 compared with control+sh-NT. (B) Cordycepin downregulated MMP-2 and MMP-9 activities through miR-33b in Lu1205 cells as shown by gelatin zymograph assay. (C) qRT-PCR showed that cordycepin inhibited the mRNA expressions of invasion-related genes, MMP-2, MMP-9, and CD44, and increased the expression of TIMP-1 in Lu1205 and A375 cells. (D) Twist1 and HMGA2 restoration rescued cordycepin-mediated downregulation of mRNA expressions of invasion-related genes, MMP-2, MMP-9, and CD44, and decreased the expression of TIMP-1 in Lu1205 and A375 cells. Data are expressed as mean ± SEM from three independent experiments. *p < 0.05, **p < 0.01 compared with control+sh-NT or cordycepin+mock.

(A) sh-miR-33b transfection or (B) ZEB1 overexpression reverted cordycepin-mediate MET in both Lu1205 and A375 cells as measured with Western blotting. E-cadherin is epithelial marker, while N-cadherin and vimentin are mesenchymal markers. β-tubulin serves as loading controls. (C) Immunofluorescence staining of E-cadherin (red) and N-cadherin (green) in untreated or cordycepin-treated Lu1205 cells which were transfected with sh-NT or sh-miR-33b and mock or ZEB1. Nuclei were counterstained with DAPI. Bar = 10 μm.

(A) Bioluminescence images of primary tumor, lung, liver and bone from sh-NT or sh-miR-33b stable Lu1205 cells following 28 days post implantation. (B) Quantification of primary tumor weight (g) and metastasis to the lung, liver and bone as measured by bioluminescent intensity (represented by photons/sec). Cordycepin (3′-deoxyadenosine) was administrated i.p. daily to the nude mice in a dosage of 2 mg/kg body weight after the tumor inoculation. Ten mice were injected under each condition with each point representing a measurement from an individual mouse with the quartiles and median values indicated by a boxplot. P-values were calculated using a two-tailed Mann-Whitney U-test. (C) H&E (upper) staining of liver tissue and IHC (lower) staining of N-catenin in primary tumor. Arrow: metastatic tumor. (D) Prominent lung lesions of injected metastatic GFP-Lu1205 and GFP-A375 cells transfected with sh-NT and sh-miR-33b. Vesicle (ctrl) or cordycepin was administrated one week before tumor inoculation. Images were representative of six independent experiments. Number of lung lesions was quantified. **p < 0.01 compared with control. (E) Kaplan–Meier analysis comparing survival in sh-NT- and sh-miR-33b-transfected mice (left) and vector, HMGA2, Twist1 and ZEB1-transfected mice (right). Statistic significance between groups was analyzed by log-rank test. *p < 0.05 among groups. (F) Model of melanoma cancer metastasis suppressed via cordycepin-initiated signaling pathways.