Fig 2: TGF-β dependent CDH6 induction is mediated by RUNX2.A) qRT-PCR analysis of CDH6 levels in non-treated (NT) and TGF-β treated (TGF) Nthy.ori cells, in presence of DMSO (mock, white bars), Actinomycin D (grey bars), cycloheximide (black bars). The bars represent the fold expression of the CDH6 mRNA in the indicated samples normalized to the GAPDH levels. The results are the average of two different replicates. B) qRT-PCR analysis of CDH6 levels in Nthy.ori 3.1 cells non-treated (black bars) or treated with TGF-β (white bars) after transfection with RUNX2 siRNA (right) or control siRNA (left). For both control siRNA and RUNX2 siRNA-treated samples, the bars represent the relative fold change of CDH6 upon TGF-β treatment as compared to non-treated cells. The graphs show one representative experiment. The experiment was replicated three times, obtaining comparable results. C and D) qRT-PCR analysis of RUNX2 and CDH6 mRNA levels in B-CPAP (C) and TPC1 (D) cells non-treated (NT, black bars) or treated with 10mM SB-431542 for 24h (grey bars) or 48h (white bars). Expression levels of CDH6 and RUNX2 were normalized to the geometric mean of GAPDH and CYPA. p-value was calculated by two-tailed Student’s t-test. *** p≤ 0.001; ** p≤ 0.01; * p≤ 0.05.

Fig 3: Tumor-derived cell lines display a constitutive EMT-like phenotype.qRT-PCR analysis of EMT markers (A) (E-CAD; N-CAD; CDH16; TNC; VIM; FN 1) and CDH6 (B) in non-treated thyroid-derived cell lines. The bars represent the average fold change of the indicated genes in tumor cells (B-CPAP, TPC1, WRO, and FTC-133) as compared to thyrocytes (Nthy.ori 3.1), normalized to the geometric mean of levels of three reference genes: GAPDH, CYPA, GUSB. C) Western Blot analysis of E-CAD, N-CAD, FN1 and Actin in non-treated Nthy.ori 3.1; B-CPAP and TPC1 cells. D) Western Blot analysis of phosphorylated ERK, phosphorylated AKT, and Actin in Nthy.ori 3.1 (Left panels); B-CPAP (middle panels) and TPC1 (right panels) cells untreated (NT) or after TGF-ß exposure for the indicated times. E) Immunofluorescence staining of SMAD2/3 proteins (green) in Nthy.ori 3.1 (upper panels); B-CPAP (middle panels) and TPC1 (lower panels), non-treated (NT) or after TGF-ß exposure for the indicated times. DAPI (Blue) stains the nuclei. Magnification 200X. F-H) qRT-PCR analysis of transcription factors known to partake in the EMT program (SNAI1, SNAI2, ZEB1, TWIST, Id1, and RUNX2) in Nthy.ori 3.1 (E), B-CPAP (F) and TPC1 (G) cells, non-treated (NT) or treated with TGF-ß for the indicated times. The bars represent the fold change of the indicated genes in TGF-ß treated cells as compared to the non-treated cell levels, normalized to the geometric mean of GAPDH, CYPA, GUSB levels. p-value was calculated by two-tailed Student’s t-test. *** p= 0.001; ** p= 0.01; * p= 0.05. Error bars represent s.e.m. (n=3).

Fig 4: Analysis of the expression levels of CDH6 splicing variants in thyroid-derived cells.A) Representative amplification curves of CDH6-L (long isoform; black line) and CDH6-S (short isoform, gray line) in non-treated Nthy.ori 3.1 (left panel), B-CPAP (middle panel) and TPC1 (right panel) obtained by qRT-PCR amplification. Numbers represent the average Ct value +/- s.e.m. of a triplicate amplification. B) qRT-PCR analysis of CDH6-L (left) and CDH6-S (right) in non-treated thyroid-derived cell lines. The bars represent the average fold change of the indicated genes in B-CPAP and TPC1 as compared to Nthy.ori 3.1. Target genes were normalized to the geometric mean of levels of three reference genes: GAPDH, CYPA, GUSB. C-E) qRT-PCR analysis of CDH6-L and CDH6-S in non-treated (NT; black bars) or TGF-β treated (white bars) Nthy.ori 3.1 (C), B-CPAP (D), and TPC1 cells (E). The bars represent the average fold change of the CDH6 variants in TGF-β treated cells as compared to non-treated cells, normalized to the geometric mean of GAPDH, CYPA, GUSB. p-value was calculated by two-tailed Student’s t-test. *** p≤ 0.001; ** p≤ 0.01; * p≤ 0.05.

Fig 5: CDH6 is a new RUNX2 target in thyroid tumor cells.A) qRT-PCR analysis of RUNX2 (black bars) and CDH6 (white bars) mRNA levels in Id1-overexpressing cells (Id1A) upon RUNX2 siRNA transfection (24 hours and 72 hours after siRNA transfection). The bars represent the average fold change of indicated genes in cells transfected with RUNX2 siRNA as compared to cells transfected with control siRNA, normalized to the GAPDH levels. B) qRT-PCR analysis of CDH6 levels in cells stably overexpressing RUNX2 (Rx12 and Rx21) and control cells (CT). The bars represent the average fold change of CDH6 in RUNX2 overexpressing cells as compared to control cells, normalized to the GAPDH levels. Control represents the average CDH6 expression in two different stable clones transfected with an empty vector (Ctrl 4N and Ctrl 5N). Id1A, Rx12, Rx21, Ctrl4N, and Ctrl5N were previously described (26). p-value was calculated by two-tailed Student’s t-test. *** p≤ 0.001; ** p≤ 0.01; * p≤ 0.05.C) Linear regression analysis of the relative fold expression of CDH6 (Y-axis) and RUNX2 (X-axis) in primary tumors (n=11) as compared to the respective normal tissue from human PTC patients. CDH6 and RUNX2 levels were normalized to the geometric mean of GAPDH, CYPA and GUSB. Correlation was statistically significant (R2=0.4985; p-value= 0.0033). T/N= fold change Tumor vs Normal. D) Schematic representation of the TGF-β pathway in controlling CDH6 expression. TGF-β signaling induces activation of RUNX2 that in turn control CDH6 expression. Our data seem to suggest that other TGF-β dependent mechanisms, involving other transcription factors (EMT-TF) cooperate with RUNX2 in controlling CDH6 expression. We showed that CDH6 is a target of Id1 in thyroid tumor cells with aggressive phenotype. In this work we showed that the Id1 dependent CDH6 induction is mediated by RUNX2. However, since both RUNX2 and Id1 are early target genes of the TGF-β signaling we hypothesize that TGF-β controls RUNX2 expression independently by Id1.