Runx2 Isoform I Controls a Panel of Proinvasive Genes Driving Aggressiveness of Papillary Thyroid Carcinomas

Abstract

Context:

The ability of tumor cells to invade adjacent tissues is governed by a complicated network of molecular signals, most of which have not yet been identified. In a recent work, we reported that the transcriptional regulator Id1 contributes to thyroid cancer progression by powering the invasion capacity of tumor cells.

Objective:

The intent of this work was to further investigate the biology of invasive thyroid tumors, through the analysis of the molecular interactions existing between Id1 and some of its target genes and through the characterization of the function of these factors in the progression of thyroid tumors.

Results:

We showed that Id1 controls the expression of the Runx2 isoform I and that this transcription factor plays a central role in mediating the Id1 proinvasive function in thyroid tumor cells. We demonstrated that Runx2 regulates proliferation, migration, and invasiveness by activating a panel of genes involved in matrix degradation and cellular invasion, which we previously identified as Id1 target genes in thyroid tumor cells. Finally, we show that Runx2 is strongly expressed in metastatic human thyroid tumors both at the primary site and in metastases.

Conclusion:

Overall, our experiments demonstrate the existence of a previously unknown molecular axis that controls thyroid tumor invasiveness by altering the ability of tumor cells to interact with the surrounding microenvironment. These factors could prove to be valuable markers that permit early diagnosis of aggressive thyroid tumors.

Papillary thyroid carcinomas (PTC) are considered indolent lesions with slow growth rate and a generally favorable outcome. About 20–50% of PTC develop lymph node metastases, whereas only 6–20% of these lesions progress as distantly metastatic disease (1–3). The molecular mechanisms determining the metastatic potential of PTC remain largely unknown.

We have recently reported that the transcription factor Id1 (inhibitor of DNA binding 1) controls progression of thyroid carcinomas by powering the invasion capacity of tumor cells. The aggressive behavior induced by Id1 in thyroid tumor cells is accompanied by the deregulation of more than 400 genes, most of which encode for proteins already identified as determinants of aggressiveness in other types of epithelial tumors. The Runt-related transcription factor 2 (Runx2) and some of its target genes, including the matrix metalloproteinase (MMP) 13 and the glycoprotein osteopontin (OPN), are significantly induced by Id1 in thyroid tumor cells (4). Runx2 [also known as core binding factor α1 (Cbfa1)] is a transcription factor belonging to the Runt-related family and known mainly for its role in controlling development and homeostasis of the skeletal tissue (5). Id1 and Runx2 are common downstream targets of several signaling pathways [e.g. TGF, bone morphogenetic protein (BMP), and wingless related MMTV integration site (Wnt)] (6, 7) and are commonly involved in the same biological processes, including bone homeostasis and cell fate commitment.

A number of works have recently identified Runx2 as a crucial mediator of aggressiveness and metastasization in different epithelial tumors, in particular in breast (8–10) and prostate cancer (11, 12). The ability of Runx2 to power the metastatic potential of tumor cells has been linked to its ability to regulate genes crucial to tumor progression including VEGF, MMP9, MMP13, and OPN (9, 13–15). Furthermore, genomic occupancy analysis in osteosarcoma cells has revealed that Runx2 binds the promoter regions of a cluster of genes involved in cell adhesion and motility (16).

Based on these findings, we hypothesized that Runx2 contributed to the aggressive phenotype induced by Id1 in thyroid tumor cells.

In this work, we demonstrate that Runx2 is a mediator of aggressive features, controlling migration and invasiveness of thyroid tumor cells. We show that Runx2 controls the expression of an entire panel of genes that we previously identified as targets of Id1 in thyroid tumor cells and that are involved in matrix degradation and loosening of the cell-cell interactions. Finally, we show that Runx2 is strongly expressed in metastatic PTC both at the primary and metastatic site, leading us to suggest that it may be a new marker of aggressiveness of this type of tumor.

Materials and Methods

Cell cultures and Western blot

B-CPAP, TPC1, and WRO human cell lines were obtained from Dr. Massimo Santoro, University of Naples (Naples, Italy). B-CPAP and TPC1 were derived from PTC samples, whereas WRO was derived from a follicular thyroid carcinoma. Id1A, Id1B, ct3, and ct4 lines were clonally derived from B-CPAP cells as previously described (4). All cell lines were grown at 37 C and 5% CO₂ in DMEM supplemented with 10% fetal bovine serum. Id1- and Runx2-overexpressing clones as well as control clones were grown in presence of 400 μg/ml geneticin (Invitrogen, Monza, Italy).

Western blot analysis was performed as described elsewhere (17). Cells were lysed either in 1× sodium dodecyl sulfate sample buffer or in RIPA buffer. Protein extracts were analyzed by SDS-PAGE using the Bio-Rad (Hercules, CA) Mini-Protean apparatus. Staining was performed with the ECL Western blot detection reagent (GE, Healthcare, Piscataway, NJ). Antibodies used were anti-Runx2 (R&D Systems, Rovereto, Italy), anti-p21 (Santa Cruz Biotechnology, Santa Cruz, CA), anti-actin (Sigma-Aldrich, Milan, Italy), antigoat (Santa Cruz Biotechnology), and antimouse (GE Healthcare, Milan, Italy).

Small interfering RNA (siRNA) transfection

Stealth RNA interference oligos against Runx2 and control were purchased from Invitrogen (Monza, Italy), and 30 nm of either anti-Runx2 or control siRNA was transfected using the RNAiMax Lipofectamine reagent (Invitrogen) using the reverse transfection protocol.

Scratch wound healing assay

Each cell line was seeded at 90% confluence in a six-well plate. Scratches were applied after cell adhesion by using a cell scraper. Healing areas were pictured at time 0, 12, and 24 h using a Nikon Ti-E inverted microscope.

Invasion chamber assay

This assay was conducted using the Matrigel Invasion Chamber (BD Biosciences, San Jose, CA), and 5 × 10⁴ cell/ml of each line was tested in triplicate. Twenty-two hours after seeding, invading cells were fixed in 100% methanol, stained with crystal violet, pictured using a Nikon Ti-E inverted microscope, and counted. Three fields for each well, captured with a magnification of ×100, were analyzed.

Cell proliferation assay

A total of 1 × 10⁴ cells for each cell line were seeded in triplicate (unless otherwise specified) in a 24-well plate in regular growth medium. Viable cells were counted every 24 h for 3 d using trypan blue (Sigma-Aldrich, Milan, Italy) staining and manual counting in a hemocytometer.

Colony-forming assay

One hundred cells for each clone were seeded as a single-cell suspension in triplicate in a 10-cm plate in regular medium. Three weeks later, colonies were fixed in 100% methanol and stained using crystal violet. Formed colonies were pictured and manually counted.

Immunohistochemistry

All PTC samples were retrieved from the archive of the Pathology Unit of Arcispedale S. Maria Nuova. Description of the clinicopathological features of the analyzed PTC patients is provided in Supplemental Table 1 (published on The Endocrine Society's Journals Online web site at http://jcem.endojournals.org). Tissue specimens were fixed in 10% formalin, paraffin embedded, and processed for light microscopy. For immunohistochemistry 4-μm sections were cut and stained with a mouse monoclonal anti-Runx2 antibody (AbFrontier; Histo-Line Laboratories, Milan, Italy) and then counterstained with hematoxylin. Images were captured using a Nikon Eclipse E80 microscope. This project has been approved by the local ethical committee.

Additional methods are described in Supplemental Methods.

Results

Id1 controls expression of Runx2 isoform I

We analyzed Runx2 mRNA and protein levels in Id1-overexpressing cells. This analysis was conducted in the Id1A and Id1B lines that overexpress Id1 and were clonally derived from B-CPAP papillary thyroid tumor cell line (18), as previously described (4). Figure 1A shows a quantitative RT-PCR (qRT-PCR) analysis of Runx2 and Id1 mRNA levels in Id1A and Id1B cells as compared with control cells. Runx2 was strongly induced by Id1 in both lines in a dose-dependent manner. Western blot analysis confirmed that Runx2 protein was strongly induced in Id1-overexpressing cells as compared with control cells (Fig. 1B). Immunofluorescence confirmed that Runx2 levels were higher in Id1A cells and mainly localized in the nucleus (Fig. 1C). Furthermore, Runx2 levels were higher in TPC1 and WRO cells compared with B-CPAP cells in accordance with that observed for Id1 (4) (Supplemental Fig. 1A).

Fig. 1.

Id1 controls Runx2 isoform I in thyroid tumor cells. A, qRT-PCR of Id1 (white bars) and Runx2 (black bars) mRNA levels in Id1A and Id1B cells. The bars represent the averaged fold change of Id1 and Runx2 in Id1A and Id1B cells compared with control cells, normalized to the GAPDH levels. The control represents the average expression in three different stable clones that do not overexpress Id1 (ct2, ct3, and ct4). ***, P ≤ 0.001; **, P ≤ 0.01. B, Western blot analysis of Runx2 and actin in Id1A, Id1B, and control cells. C, Immunofluorescent staining of Id1A and ct3 cells with anti-Runx2 antibody. Magnification, ×400. D, Schematic representation of the Runx2 gene. Boxes 1-8 represent the exons of the Runx2 genes. Reverse common primers (white arrow), isoform II forward specific primer (black arrow), and isoform I forward specific primer (gray arrow) are indicated. E, Semiquantitative RT-PCR of Runx2 isoforms expression in Id1A, Id1B, and control cells (ct3). GAPDH expression was used as control. Expected molecular weights: isoform II, 142 bp; isoform I, 122 bp. M, Molecular weight marker.

The Runx2 gene encodes two major isoforms using two different promoters (Fig. 1D) (19). To determine which isoform is controlled by Id1 in thyroid tumor cells, we designed specific primers for each variant and analyzed the mRNA levels in Id1-overexpressing cells by semiquantitative RT-PCR. Intriguingly, the P2 promoter-dependent isoform I was strongly induced in Id1-overexpressing cells compared with control cells (Fig. 1E). By contrast, we did not detect expression of the isoform II in any of the cell lines.

Runx2 is a necessary mediator of aggressive phenotype in thyroid tumor cells

Because a Runx2 proinvasive function has been reported in other types of epithelial tumors, we investigated whether Runx2 was a mediator of the increased migration and invasion properties of Id1-overexpressing thyroid tumor cells.

Id1A cells were transfected with siRNA oligonucleotides against Runx2 or with control siRNA, and down-regulation of Runx2 was analyzed at the mRNA and protein levels by qRT-PCR and Western blot, respectively (Fig. 2, A and B). Specific siRNA transfection resulted in a down-regulation of Runx2 of about 80% as compared with control siRNA. To analyze how Runx2 reduction affected the migration capacity of Id1-overexpressing cells, we performed a scratch wound healing assay in Id1A cells transfected with siRNA against Runx2 or control oligos. Transfected cells were seeded as a monolayer, and scratches of comparable dimensions were applied to the plates (Fig. 2C). After 24 h, Id1A cells transfected with control siRNA had completely healed the scratches. By contrast, Id1A cells transfected with siRNA against Runx2 had reduced the empty spaces only slightly, proving that reduction of Runx2 levels decreases migration of Id1-overexpressing cells. The same experiment was conducted in the parental B-CPAP cells, with comparable results (Supplemental Fig. 1, B–D). We then investigated the effect of Runx2 reduction on invasiveness by using an invasion chamber assay. Id1A cells transfected with Runx2 or control siRNA were seeded in triplicate in the upper part of an invasion chamber. Invasive cells were counted 22 h after seeding. Reduction of Runx2 levels profoundly impaired the Id1A cell invasion properties (Fig. 2D). Taken together, these experiments demonstrate that Runx2 is a crucial mediator of the Id1 proinvasive function in these tumors. Runx2 is a transcription factor and controls biological processes by regulating the expression of specific target genes. We hypothesized that some of the genes that we had previously identified as Id1 targets may indeed be controlled by Runx2. We focused on targets that might be involved in controlling migration and invasiveness. Several Id1 target genes encode proteins known to control cell-cell interaction or matrix degradation. In particular, three metalloproteinases (MMP2, MMP13, and MMP14), the MMP inhibitor tissue inhibitor of metalloproteinases 3 (TIMP), and the MMP interactor OPN were induced by Id1. We transfected the Id1A cells with siRNA against Runx2 (white bars) or control siRNA (black bars) and analyzed how this affected the expression of MMP2, MMP13, MMP14, OPN, and TIMP3 in Id1A cells. The expression of all three MMP was reduced upon Runx2 down-regulation (Fig. 2E), showing that Runx2 mediates the Id1 control over these genes. Interestingly, the expression of TIMP3 and OPN was induced upon down-regulation of Runx2. This observation implies that the positive effect of Id1 on TIMP3 and OPN expression must be mediated by a different molecular mechanism. Interestingly, siRNA-mediated down-regulation of Runx2 causes decreased expression of MMP2, MMP13, MMP14, and OPN in B-CPAP cells, while inducing expression of TIMP3, further confirming the ability of Runx2 to control these genes (Supplemental Fig. 1E).

Fig. 2.

Down-regulation of Runx2 expression impairs migration and invasiveness of Id1-overexpressing thyroid tumor cells. A, qRT-PCR analysis of Runx2 mRNA levels in Id1A cells after Runx2 and control siRNA transfection; 50 nm control siRNA were used. Results were normalized to the GAPDH levels. The bars represent the averaged fold change of Runx2 in cells transfected with siRNA against Runx2 compared with the cells transfected with control siRNA. B, Western blot analysis of Runx2 and actin in Id1A cells after Runx2 and control siRNA transfection. C, Scratch wound healing assay. Light microscopy images are shown of Id1A cells transfected with Runx2 or control siRNA during a representative scratch wound healing assay at 0 and 24 h after scratch application. Magnification, ×100. Six scratches for each well were analyzed in this experiment, all obtaining the same results. The experiment was replicated twice obtaining comparable results. D, Results of the invasion chamber assay on the Id1A cells transfected with Runx2 and control siRNA. The histograms represent the averaged number of invading cells per field obtained in two separate experiments. For both siRNA transfections, the mean values ± sem are indicated. Insets show pictures of representative fields. Magnification, ×100. E, qRT-PCR analysis of the MMP2, MMP13, MMP14, TIMP3, and OPN in Id1A cells transfected with siRNA against Runx2 (white bars) or control siRNA (black bars). Results were normalized to the GAPDH levels. The bars represent the averaged fold change of indicated mRNA in cells transfected with siRNA against Runx2 compared with the cells transfected with control siRNA. ***, P ≤ 0.001; **, P ≤ 0.01. Ctrl, Control.

Runx2 overexpression increases motility and invasiveness in thyroid tumor cells

To strengthen these observations and to further explore the function of Runx2 in controlling invasiveness, we stably overexpressed Runx2 in thyroid tumor cells. To this end, B-CPAP cells were transfected with a Runx2 isoform I-expressing vector or with an empty vector, and single clones were derived by means of antibiotic selection. Runx2 mRNA and protein levels were analyzed in different clones (Fig. 3, A and B), and the Rx12 and Rx21 lines were chosen for additional experiments (arrows). To test migration capacity of Runx2-overexpressing cells, we performed a scratch wound healing assay. As shown in Fig. 3C, at 12 h, the healing was more pronounced in the Rx12 and Rx21 than in any of the control lines. After 24 h, the scratches were completely healed in Runx2-overexpressing cells while still open in all the control lines. Next, we tested the invasion properties of Runx2-overexpressing cells in an invasion chamber assay. As shown in Fig. 3D, the number of invasive cells was significantly higher in both Rx12 and Rx21 cells than in control cells. These experiments demonstrated that Runx2 sustains both migration and invasiveness of thyroid tumor cells, even in an Id1-independent manner.

Fig. 3.

Runx2 overexpression increases migration and invasion in thyroid tumor cells. A, qRT-PCR analysis of Runx2 mRNA levels in a representative set of B-CPAP-derived clones. The histograms represent the fold induction of Runx2 levels in the indicated Runx2-overexpressing stable lines compared with a control line (Ctrl4N). B, Western blot analysis of Runx2 and actin in the indicated Runx2-overexpressing and control stable lines. Arrows indicate the levels of Runx2 expression in Rx12 and Rx21 cells that were used in the following experiments. C, Scratch wound healing assay. Light microscopy images are shown of Rx12, Rx21, and three control lines (Ctrl4N, -5N, and -6N) during a representative scratch wound healing assay at 0, 12, and 24 h after scratch application. Magnification, ×100. Four scratches for each line were analyzed in this experiment, all obtaining the same results. The experiment was replicated twice, obtaining comparable results. D, Results of the invasion chamber assay. Rx12, Rx21, and Ctrl5N were tested in triplicate. The histograms represent the averaged number of invading cells per field obtained in two separate experiments. For all lines, the mean values ± sem are indicated. E, qRT-PCR analysis of Runx2, MMP2, MMP13, MMP14, TIMP3, and OPN in Rx12 (white bars), Rx21 (black bars), and Id1A (gray bars) and control lines Ctrl3N, -4N, and -5N. Results were normalized to the GAPDH levels. The bars represent the averaged fold change of indicated mRNA in Rx12, Rx21, and Id1A lines compared with the averaged value of the indicated genes in the three control lines and corresponding to baseline. ***, P ≤ 0.001; **, P ≤ 0.01; *, P ≤ 0.05.

Finally, we tested the expression of MMP2, MMP13, MMP14, TIMP3, and OPN in Runx2-overexpressing cells. Figure 3E shows the results of a qRT-PCR analysis of the expression levels of the aforementioned mRNAs in Rx12 (black bars), Rx21 (white bars), and Id1A cells (gray bars) expressed as a fold change relative to the averaged levels of four different control lines (baseline). MMP2, MMP13, MMP14, and OPN were induced in both Runx2-overexpressing lines compared with control cells, whereas TIMP3 was severely down-regulated. These results confirm that Runx2 is a positive regulator of an entire panel of MMP and inhibits the expression of their inhibitor TIMP3. These observations, together with the increased migration and invasiveness displayed by the Runx2-overexpressing cells, strongly support the hypothesis that Runx2 is an important determinant of the metastatic potential of thyroid tumors.

Runx2 stimulates cell growth in thyroid tumor cells

We analyzed the proliferation rate of Runx2-overexpressing cells through the construction of a growth curve (Fig. 4A). Noticeably, the Rx12 and Rx21 lines showed a significantly higher proliferation rate at each time point than did the control lines and B-CPAP cells. After 72 h, the number of Runx2-overexpressing cells increased by about 40% compared with the numbers of control cells. To confirm this observation, we investigated growth properties of B-CPAP cells after Runx2 silencing by siRNA transfection. Figure 4B shows that reduction of Runx2 levels significantly reduced the proliferation of B-CPAP cells. Comparable results were obtained in TPC1 cells (Supplemental Fig. 2, A–C). A colony-forming assay was also performed. After 3 wk, both Rx12 and Rx21 clones had formed a significantly higher number of colonies than controls (Fig. 4C). Interestingly, the size of formed colonies was considerably larger in the Runx2-overexpressing lines than in the control, confirming a positive effect of Runx2 in supporting cell growth. It has been reported that Runx2 is able to control the cell cycle inhibitor p21 (20). Indeed, p21 levels were lower in both Rx12 and Rx21 lines than in controls (Fig. 4D and Supplemental Fig. 2, D and E). No differences were observed in p21 cellular localization in these cells (Supplemental Fig. 2F).

Fig. 4.

Runx2 controls cell proliferation in thyroid tumor cells. A, Proliferation curves of Rx12, Rx21, Ctrl4N, Ctrl5N, and B-CPAP cells. Curves represent the averaged number of cells per well for each line at each time point ± sem. B, Proliferation curves of B-CPAP cells transfected with Runx2 or control siRNA. Curves represent the averaged number of cells per well for each transfection at each time point ± sem. C, Colony formation assay. Pictures of formed colonies in a representative plate for the indicated cell lines (Rx12, Rx21 and Ctrl4N) are shown. The insets show magnification of single colonies. The histograms represent the averaged number of formed colonies ± sem in the three replicates analyzed for each line. The experiments were performed twice, obtaining comparable results. D, qRT-PCR analysis of p21 mRNA levels. The histograms represent the fold induction of p21 levels in the Rx12 and Rx21 compared with the averaged value obtained for two different control lines (Ctrl4N and Ctrl5N). ***, P ≤ 0.001; **, P ≤ 0.01. Ctrl, Control.

Runx2 is strongly expressed in human metastatic PTC

To validate our results in vivo, we investigated the expression of Runx2 in human PTC. Thirty-five PTC from patients of the Arcispedale S. Maria Nuova Hospital were analyzed for Runx2 expression by immunohistochemistry. Clinicopathological features of the analyzed cases are described in the Supplemental Table 1. Control of the antibody specificity was performed in Rx21 cells and Ctrl4N (Supplemental Fig. 2G). All tumors analyzed showed a diffuse expression of Runx2 (Fig. 5, A–D). Twenty of the analyzed PTC had developed metastases. From 18 of the 20 metastatic PTC, we could retrieve matched lymph-node metastases, and we have analyzed Runx2 by immunohistochemistry. Noticeably, Runx2 was expressed in all metastatic tissues analyzed. Figure 5, E–L, shows Runx2 staining in three representative primary PTC and matched metastases.

Fig. 5.

Runx2 expression in human metastatic PTC. A–D, Immunohistochemistry staining of Runx2 (brown) in four representative metastatic PTC, with hematoxylin (blue). Magnification, ×40. The insets (a–d) show higher magnification of the same field. Magnification, ×200. Runx2 staining is confined to nuclei and clearly visible in tumor cells. E–L, Immunohistochemistry staining of Runx2 (brown) in three representative metastatic PTC (E, F, and G) and corresponding lymph node metastases (H, I, and L), with hematoxylin (blue). Magnification, ×200. Runx2 is strongly expressed in human thyroid carcinoma cells both at the primary and metastatic site. M, Semiquantitative RT-PCR of Runx2 isoform expression in three human metastatic PTC (PTC1, PTC2, and PTC3). GAPDH expression was used as control for the amount of starting RNA used for the analysis. M, Molecular weight marker. N, Deoxyribonuclease sensitivity assay of the P1 and P2 Runx2 promoters in Id1A, Id1B, and control cells (ct3). Chromatin accessibility is expressed as a percentage on a scale ranging from a constitutively expressed gene (GAPDH) to a constitutively repressed gene (rhodopsin).

Because we have observed that in thyroid tumor cells only Runx2 isoform I is expressed, we have investigated the expression of Runx2 isoforms in human metastatic PTC samples by means of semiquantitative RT-PCR (Fig. 5M). In accordance with our in vitro data, isoform II was barely detectable. By contrast, isoform I was readily observed in all tumor tissues analyzed, and its levels were comparable to those of total Runx2. This observation demonstrates that Runx2 expression in human PTC is dependent mainly on the activity of the P2 promoter. To further investigate the differential activity of the P1 and P2 promoters in thyroid tumor cells, we have analyzed the chromatin accessibility of P1 and P2 promoters in Id1A, Id1B, and control cells by performing a deoxyribonuclease sensitivity assay (Fig. 5N). In accordance with the expression levels of the two Runx2 isoforms, the P2 DNA structure was largely accessible (more than 85%), whereas the P1 promoter region was extremely compact and poorly accessible (about 20–30%).

As previously reported (21), a weak staining for Runx2 was detected also in normal thyroid (Fig. 5, A–D, and Supplemental Fig. 2H). Therefore, we decided to perform a quantitative analysis of Runx2 expression in thyroid tumors. We collected total RNA from normal tissue, primary tumor, and matched lymph-node metastasis of eight of the metastatic PTC patients, and we analyzed Runx2 levels by qRT-PCR (Fig. 6A). Noticeably, Runx2 expression was significantly higher in all the tumor samples and metastases analyzed compared with the normal tissue. This observation confirms the idea that Runx2 sustains thyroid tumor growth and progression. Next, we have investigated the expression of Runx2 target genes in primary tumor and normal counterpart in the same eight PTC samples in which Runx2 expression was analyzed (Fig. 6B and Supplemental Fig. 3). MMP13, MMP14, and OPN were significantly up-regulated in tumor compared with normal tissue in the majority of the PTC samples. By contrast, TIMP3 expression was consistently down-regulated in all tumor samples compared with the normal tissue. MMP2 showed a less consistent variation in tumor vs. normal tissue. Noticeably, this evidence is in agreement with our in vitro observations.

Fig. 6.

Expression of Runx2 and its target genes in human primary PTC and normal tissues. A, qRT-PCR analysis of Runx2 levels in normal tissue (white bars), primary tumor (black bars), and lymph-node metastasis (gray bars) from eight patients with PTC. For PTC4 and PTC6, it was not possible to measure Runx2 expression in the metastatic tissue due to insufficient RNA amount for the analysis. For each PTC patient, the histograms represent the fold induction of Runx2 levels in primary tumor and metastasis compared with normal tissue. ***, P ≤ 0.001; **, P ≤ 0.01; *, P ≤ 0.05. B, Scatter plot representation of the relative fold expression of MMP2, MMP13, MMP14, OPN, and TIMP3 obtained by qRT-PCR in primary tumor compared with respective normal tissue from eight patients with PTC. Detailed histograms for each of the indicated mRNA in each PTC patient are provided in Supplemental Fig. 3. C, Schematic representation of the Runx2 proinvasive molecular axis.

Discussion

Although PTC are the most common endocrine malignancies, the molecular mechanisms controlling aggressiveness and metastatic potential of these lesions are still largely unknown. We recently identified Id1 as a major determinant of aggressiveness of thyroid tumor cells (4). The present work characterizes for the first time the existence of a molecular axis controlled by Id1 that, through Runx2, potentiates migration and invasion capacity of thyroid tumor cells. The Runx2 gene encodes two major isoforms: isoform I, controlled by the P2 proximal promoter, and isoform II, transcribed from the P1 distal promoter. The two isoforms differ only in terms of a few amino acids in the N-terminal region, even though major differences in the distribution and transcriptional function of the two proteins have been reported (19). Isoform II is expressed almost exclusively in skeletal tissue, whereas isoform I has a broader expression pattern. Interestingly, the vast majority of signaling pathways able to control Runx2 expression [including bone morphogenetic protein (BMP), wingless related MMTV integration site (Wnt), and TNFα] alters P1 activity but does not affect the P2 promoter (22–24). Noticeably, our data show that the P2-dependent isoform I is the major Runx2 variant expressed in thyroid tumor cells both in vitro (Fig. 1E) and in vivo (Fig. 5M) and that Id1 induces selectively the up-regulation of Runx2 isoform I (Fig. 1E). Our data demonstrate that Runx2 is a necessary mediator of the Id1 proinvasive function in thyroid tumor cells. Down-regulation of Runx2 expression by siRNA strongly impaired the migration and invasive capacity of Id1-overexpressing cells (Fig. 2, C and D). Simultaneously, ectopic overexpression of Runx2 can power migration and invasiveness of thyroid tumor cells (Fig. 3, C and D). Solid evidence has linked Runx2 to the development and the metastatic spreading of breast (10, 25–27), prostate (9, 11, 28, 29), and bone malignancies (16, 30). In particular, Runx2 seems to play a crucial role in determining the interaction between metastatic cells and osteoblasts, creating a favorable environment for bone colonization and metastasis development (8, 28). In addition, Runx2 has been shown to sustain tumor progression by supporting invasiveness and proliferation of tumor cells (31). Expression of Runx2 in thyroid tumor cells and in human tumors has been also reported (32, 33). In particular, Runx2 has been implicated in calcification processes (32), a common feature of PTC, or has been shown to be highly expressed at the invasion front of PTC and to be associated with signs of mesenchymal transdifferentiation (33). However, direct proof of the proinvasive function of Runx2 in thyroid tumors has never been reported before. Using several in vitro approaches, we have demonstrated that Runx2 induces the expression of three different MMP (MMP2, MMP13, and MMP14) in thyroid tumor cells although inhibiting the MMP inhibitor TIMP3. As previously shown, Runx2 also positively controls OPN, which is a glycoprotein able to bind MMP by docking them to the cellular membrane (15). MMP2, MMP13, MMP14, and OPN are all genes we previously identified as Id1 target genes in thyroid tumor cells (4). The MMP are a large family of proteases involved in cellular interaction with the surrounding microenvironment (34). Molecular features and functions of the 23 known MMP are extremely heterogeneous and not always associated with tumor progression. Noticeably, the three MMP we have described as Runx2 targets are all known for their ability to degrade the extracellular matrix and to sustain tumor progression, thereby favoring vascular and tissue invasion (34). In addition, the membrane-anchored MMP14 has recently been described as a tumor growth factor that, through proteolytic activity, allows tumor cells to overcome extracellular matrix constraint and to grow in a tridimensional structure (35, 36). Remarkably, the ability to grow tridimensionally in the absence of physical anchorage is one of the features of metastatic cells. The negative control of Runx2 over TIMP3 indicates that, besides its positive effect at the transcriptional level, Runx2 endorses MMP function by promoting their active state. Finally, we showed that also in thyroid tumor cells, Runx2 controls OPN expression. Overexpression of Runx2 in stable clones induced OPN expression, whereas siRNA-mediated Runx2 down-regulation caused decreased levels of OPN in B-CPAP cells. Intriguingly, down-regulation of Runx2 in Id1A cells resulted in a significant up-regulation of the OPN expression. It is likely that Id1 controls OPN expression through a Runx2-independent mechanism. Therefore, high levels of Runx2 in Id1-overexpressing cells may interfere with this alternative activation pathway. It is worth noticing that we have found several transcription factors beside Runx2 being up-regulated upon Id1 overexpression in thyroid tumor cells (4). Noticeably, OPN has already been shown to be induced in thyroid tumor and to be significantly correlated with aggressiveness and lymph-node metastases in these lesions (37). Overall, our data demonstrate the existence of a new molecular axis able to modify how thyroid tumor cells interact with the surrounding microenvironment and to potentiate their invasion capacity. Our data also demonstrate that Runx2 supports thyroid tumor cell proliferation. The role of Runx2 in controlling cell growth is quite controversial. It is known that in osteoblasts Runx2 inhibits cell proliferation to favor differentiation. Runx2 expression varies across the cell cycle, with high levels during M phase and early G1 and low levels in late G1 or S phase (38, 39). Moreover, serum starvation of osteoblast cell lines induces Runx2 expression, which is promptly reduced upon serum reintegration. By contrast, numerous works have reported a positive effect of Runx2 on proliferation of different types of tumor cells as well as other normal cell types, including endothelial cells (12, 39–41). It is reasonable to suppose that the diverse effect of Runx2 in controlling proliferation is affected by the cellular context, including different posttranslation controlling mechanisms or the presence of different specific transcriptional partners. Finally, to strengthen the validity of our observation, we have proven that Runx2 is strongly and consistently expressed in human metastatic PTC both at the primary sites and in metastases. Noticeably, we have proven that isoform I is the major Runx2 variant expressed in vivo in tumor cells, whereas only small traces of isoform II could be detected. In our opinion, the possibility that isoform I is indeed the Runx2 variant responsible for the invasiveness and metastatic potential in epithelial tumors is extremely fascinating. Unfortunately, no work published on this topic so far has addressed the diverse function of the two Runx2 isoforms in this context.

Our immunohistochemistry data demonstrate that Runx2 is expressed also in normal thyroid tissue. This observation is not new; Endo and Kobayashi (21) have observed that Runx2 is expressed in mouse thyrocytes, controlling thyroglobulin expression and that its ablation causes hypothyroidism in mice. However, we show that the expression of Runx2 is strongly induced in primary tumors and lymph-node metastases compared with normal thyroid. This observation is in accordance with what is observed in breast tumors, where Runx2 is expressed and has a role in normal epithelial cells but whose increased expression confers metastatic potential to tumor cells. In accordance with what is observed for Runx2, we showed that the expression of MMP13, MMP14, and OPN is consistently higher in human thyroid tumors than in normal tissues, whereas TIMP3 is significantly down-regulated (Fig. 6B and Supplemental Fig. 3), strongly supporting the validity of our model.

In conclusion, we have described a completely new molecular program that, through gene expression regulation, controls invasiveness and aggressiveness of PTC (Fig. 6C). We have shown that Runx2 is a major determinant of aggressiveness and invasiveness in vitro and is highly expressed in human thyroid tumors. We have identified five Runx2 target genes whose function may be a determinant in mediating its aggressive function. Retrospective and prospective analyses of the expression of these factors in aggressive PTC will reveal whether they prove to be valuable molecular markers for the early diagnosis and prognosis of thyroid tumors.